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)
298 * This is true for native and DDF, IMSM allows 16.
300 #define MD_NAME_MAX 32
302 extern const char Name
[];
305 unsigned long long sector
;
312 struct md_bb_entry
*entries
;
315 /* general information that might be extracted from a superblock */
317 mdu_array_info_t array
;
318 mdu_disk_info_t disk
;
322 unsigned long long data_offset
;
323 unsigned long long new_data_offset
;
324 unsigned long long component_size
; /* same as array.size, except in
325 * sectors and up to 64bits.
327 unsigned long long custom_array_size
; /* size for non-default sized
328 * arrays (in sectors)
331 #define VOLUME_RESHAPE 1
332 #define CONTAINER_RESHAPE 2
333 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
335 unsigned long long reshape_progress
;
336 int recovery_blocked
; /* for external metadata it
337 * indicates that there is
338 * reshape in progress in
340 * for native metadata it is
341 * reshape_active field mirror
343 int journal_device_required
;
347 CONSISTENCY_POLICY_UNKNOWN
,
348 CONSISTENCY_POLICY_NONE
,
349 CONSISTENCY_POLICY_RESYNC
,
350 CONSISTENCY_POLICY_BITMAP
,
351 CONSISTENCY_POLICY_JOURNAL
,
352 CONSISTENCY_POLICY_PPL
,
353 } consistency_policy
;
355 /* During reshape we can sometimes change the data_offset to avoid
356 * over-writing still-valid data. We need to know if there is space.
357 * So getinfo_super will fill in space_before and space_after in sectors.
358 * data_offset can be increased or decreased by this amount.
360 unsigned long long space_before
, space_after
;
362 unsigned long long resync_start
; /* per-array resync position */
363 unsigned long long recovery_start
; /* per-device rebuild position */
364 #define MaxSector (~0ULL) /* resync/recovery complete position */
366 long bitmap_offset
; /* 0 == none, 1 == a file */
367 unsigned int ppl_size
;
369 unsigned long long ppl_sector
;
370 unsigned long safe_mode_delay
; /* ms delay to mark clean */
371 int new_level
, delta_disks
, new_layout
, new_chunk
;
373 unsigned long cache_size
; /* size of raid456 stripe cache*/
375 char text_version
[50];
377 int container_member
; /* for assembling external-metatdata arrays
378 * This is to be used internally by metadata
384 /* Device info for mdmon: */
391 #define DS_WRITE_MOSTLY 4
393 #define DS_BLOCKED 16
394 #define DS_REMOVE 1024
395 #define DS_UNBLOCK 2048
396 int prev_state
, curr_state
, next_state
;
398 /* info read from sysfs */
422 struct supertype
*supertype
;
425 struct spare_criteria
{
426 unsigned long long min_size
;
427 unsigned int sector_size
;
430 typedef enum mdadm_status
{
431 MDADM_STATUS_SUCCESS
= 0,
449 extern char short_options
[];
450 extern char short_monitor_options
[];
451 extern char short_bitmap_options
[];
452 extern char short_bitmap_auto_options
[];
453 extern struct option long_options
[];
454 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
456 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
458 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
460 /* for option that don't have short equivilents, we assign arbitrary
461 * numbers later than any 'short' character option.
463 enum special_options
{
549 UOPT_LAYOUT_ORIGINAL
,
550 UOPT_LAYOUT_ALTERNATE
,
551 UOPT_LAYOUT_UNSPECIFIED
,
556 * Code specific options, cannot be set by the user
559 UOPT_SPEC_FORCE_ARRAY
,
561 UOPT_SPEC_LINEAR_GROW_NEW
,
562 UOPT_SPEC_LINEAR_GROW_UPDATE
,
563 UOPT_SPEC__RESHAPE_PROGRESS
,
564 UOPT_SPEC_WRITEMOSTLY
,
567 UOPT_SPEC_NOFAILFAST
,
568 UOPT_SPEC_REVERT_RESHAPE_NOBACKUP
,
571 extern void fprint_update_options(FILE *outf
, enum update_opt update_mode
);
573 enum prefix_standard
{
585 FlagDefault
, FlagSet
, FlagClear
,
588 /* structures read from config file */
589 /* List of mddevice names and identifiers
590 * Identifiers can be:
592 * super-minor=decimal-minor-number-from-superblock
593 * devices=comma,separated,list,of,device,names,with,wildcards
595 * If multiple fields are present, the intersection of all matching
596 * devices is considered
598 #define UnSet (0xfffe)
604 char name
[MD_NAME_MAX
+ 1];
608 char *devices
; /* comma separated list of device
609 * names with wild cards
614 struct supertype
*st
;
615 int autof
; /* 1 for normal, 2 for partitioned */
620 char *container
; /* /dev/whatever name of container, or
621 * uuid of container. You would expect
622 * this to be the 'devname' or UUID
623 * of some other entry.
625 char *member
; /* subarray within a container */
627 struct mddev_ident
*next
;
629 /* fields needed by different users of this structure */
630 int assembled
; /* set when assembly succeeds */
642 int require_homehost
;
647 enum update_opt update
;
673 unsigned long long size
;
674 unsigned long long data_offset
;
675 int consistency_policy
;
678 /* List of device names - wildcards expanded */
681 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail,
683 * Not set for names read from .config
685 enum flag_mode writemostly
;
686 enum flag_mode failfast
;
687 int used
; /* set when used */
688 long long data_offset
;
689 struct mddev_dev
*next
;
692 typedef struct mapping
{
701 char *pattern
; /* U for up, _ for down */
702 int percent
; /* -1 if no resync */
703 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
706 char * metadata_version
;
709 struct dev_member
*next
;
711 struct mdstat_ent
*next
;
714 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
715 extern void mdstat_close(void);
716 extern void free_mdstat(struct mdstat_ent
*ms
);
717 extern int mdstat_wait(int seconds
);
718 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
719 extern int mddev_busy(char *devnm
);
720 extern struct mdstat_ent
*mdstat_by_component(char *name
);
721 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, char *container
);
724 struct map_ent
*next
;
731 extern int map_update(struct map_ent
**mpp
, char *devnm
, char *metadata
,
732 int uuid
[4], char *path
);
733 extern void map_remove(struct map_ent
**map
, char *devnm
);
734 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
735 extern struct map_ent
*map_by_devnm(struct map_ent
**map
, char *devnm
);
736 extern void map_free(struct map_ent
*map
);
737 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
738 extern void map_read(struct map_ent
**melp
);
739 extern int map_write(struct map_ent
*mel
);
740 extern void map_delete(struct map_ent
**mapp
, char *devnm
);
741 extern void map_add(struct map_ent
**melp
,
742 char *devnm
, char *metadata
, int uuid
[4], char *path
);
743 extern int map_lock(struct map_ent
**melp
);
744 extern void map_unlock(struct map_ent
**melp
);
745 extern void map_fork(void);
747 /* various details can be requested */
748 enum sysfs_read_flags
{
749 GET_LEVEL
= (1 << 0),
750 GET_LAYOUT
= (1 << 1),
751 GET_COMPONENT
= (1 << 2),
752 GET_CHUNK
= (1 << 3),
753 GET_CACHE
= (1 << 4),
754 GET_MISMATCH
= (1 << 5),
755 GET_VERSION
= (1 << 6),
756 GET_DISKS
= (1 << 7),
757 GET_SAFEMODE
= (1 << 9),
758 GET_BITMAP_LOCATION
= (1 << 10),
760 GET_DEVS
= (1 << 20), /* gets role, major, minor */
761 GET_OFFSET
= (1 << 21),
762 GET_SIZE
= (1 << 22),
763 GET_STATE
= (1 << 23),
764 GET_ERROR
= (1 << 24),
765 GET_ARRAY_STATE
= (1 << 25),
766 GET_CONSISTENCY_POLICY
= (1 << 26),
767 GET_DEVS_ALL
= (1 << 27),
770 /* If fd >= 0, get the array it is open on,
773 extern int sysfs_open(char *devnm
, char *devname
, char *attr
);
774 extern int sysfs_init(struct mdinfo
*mdi
, int fd
, char *devnm
);
775 extern void sysfs_init_dev(struct mdinfo
*mdi
, dev_t devid
);
776 extern void sysfs_free(struct mdinfo
*sra
);
777 extern struct mdinfo
*sysfs_read(int fd
, char *devnm
, unsigned long options
);
778 extern int sysfs_attr_match(const char *attr
, const char *str
);
779 extern int sysfs_match_word(const char *word
, char **list
);
780 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
781 char *name
, char *val
);
782 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
783 char *name
, unsigned long long val
);
784 extern int sysfs_set_num_signed(struct mdinfo
*sra
, struct mdinfo
*dev
,
785 char *name
, long long val
);
786 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
787 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
789 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
790 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
791 char *name
, unsigned long long *val
);
792 extern int sysfs_fd_get_two(int fd
, unsigned long long *v1
, unsigned long long *v2
);
793 extern int sysfs_get_two(struct mdinfo
*sra
, struct mdinfo
*dev
,
794 char *name
, unsigned long long *v1
, unsigned long long *v2
);
795 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
796 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
798 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
799 char *name
, char *val
, int size
);
800 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
801 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
802 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
803 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
804 extern int sysfs_unique_holder(char *devnm
, long rdev
);
805 extern int sysfs_freeze_array(struct mdinfo
*sra
);
806 extern int sysfs_wait(int fd
, int *msec
);
807 extern int load_sys(char *path
, char *buf
, int len
);
808 extern int zero_disk_range(int fd
, unsigned long long sector
, size_t count
);
809 extern int reshape_prepare_fdlist(char *devname
,
813 unsigned long blocks
,
816 unsigned long long *offsets
);
817 extern void reshape_free_fdlist(int *fdlist
,
818 unsigned long long *offsets
,
820 extern int reshape_open_backup_file(char *backup
,
825 unsigned long long *offsets
,
828 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
829 unsigned int ndata
, unsigned int odata
);
830 extern char *locate_backup(char *name
);
831 extern char *make_backup(char *name
);
833 extern int save_stripes(int *source
, unsigned long long *offsets
,
834 int raid_disks
, int chunk_size
, int level
, int layout
,
835 int nwrites
, int *dest
,
836 unsigned long long start
, unsigned long long length
,
838 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
839 int raid_disks
, int chunk_size
, int level
, int layout
,
840 int source
, unsigned long long read_offset
,
841 unsigned long long start
, unsigned long long length
,
845 #define Sendmail "/usr/lib/sendmail -t"
848 #define SYSLOG_FACILITY LOG_DAEMON
849 extern char *map_num_s(mapping_t
*map
, int num
);
850 extern char *map_num(mapping_t
*map
, int num
);
851 extern int map_name(mapping_t
*map
, char *name
);
852 extern mapping_t r0layout
[], r5layout
[], r6layout
[],
853 pers
[], modes
[], faultylayout
[];
854 extern mapping_t consistency_policies
[], sysfs_array_states
[], update_options
[];
856 extern char *map_dev_preferred(int major
, int minor
, int create
,
858 static inline char *map_dev(int major
, int minor
, int create
)
860 return map_dev_preferred(major
, minor
, create
, NULL
);
864 * is_fd_valid() - check file descriptor.
865 * @fd: file descriptor.
867 * The function checks if @fd is nonnegative integer and shall be used only
868 * to verify open() result.
870 static inline int is_fd_valid(int fd
)
876 * is_level456() - check whether given level is between inclusive 4 and 6.
877 * @level: level to check.
879 * Return: true if condition is met, false otherwise
881 static inline bool is_level456(int level
)
883 return (level
>= 4 && level
<= 6);
887 * close_fd() - verify, close and unset file descriptor.
888 * @fd: pointer to file descriptor.
890 * The function closes and invalidates file descriptor if appropriative. It
891 * ignores incorrect file descriptor quitely to simplify error handling.
893 static inline void close_fd(int *fd
)
895 if (is_fd_valid(*fd
) && close(*fd
) == 0)
900 struct metadata_update
;
902 /* 'struct reshape' records the intermediate states of
904 * The starting geometry is converted to the 'before' geometry
905 * by at most an atomic level change. They could be the same.
906 * Similarly the 'after' geometry is converted to the final
907 * geometry by at most a level change.
908 * Note that 'before' and 'after' must have the same level.
909 * 'blocks' is the minimum number of sectors for a reshape unit.
910 * This will be a multiple of the stripe size in each of the
911 * 'before' and 'after' geometries.
912 * If 'blocks' is 0, no restriping is necessary.
913 * 'min_offset_change' is the minimum change to data_offset to
914 * allow the reshape to happen. It is at least the larger of
915 * the old and new chunk sizes, and typically the same as 'blocks'
916 * divided by number of data disks.
920 int parity
; /* number of parity blocks/devices */
925 unsigned long long backup_blocks
;
926 unsigned long long min_offset_change
;
927 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
928 unsigned long long new_size
; /* New size of array in sectors */
931 /* A superswitch provides entry point to a metadata handler.
933 * The superswitch primarily operates on some "metadata" that
934 * is accessed via the 'supertype'.
935 * This metadata has one of three possible sources.
936 * 1/ It is read from a single device. In this case it may not completely
937 * describe the array or arrays as some information might be on other
939 * 2/ It is read from all devices in a container. In this case all
940 * information is present.
941 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
942 * be complete once enough ->add_to_super calls have completed.
944 * When creating an array inside a container, the metadata will be
945 * formed by a combination of 2 and 3. The metadata or the array is read,
946 * then new information is added.
948 * The metadata must sometimes have a concept of a 'current' array
949 * and a 'current' device.
950 * The 'current' array is set by init_super to be the newly created array,
951 * or is set by super_by_fd when it finds it is looking at an array inside
954 * The 'current' device is either the device that the metadata was read from
955 * in case 1, or the last device added by add_to_super in case 3.
956 * Case 2 does not identify a 'current' device.
958 extern struct superswitch
{
960 /* Used to report details of metadata read from a component
961 * device. ->load_super has been called.
963 void (*examine_super
)(struct supertype
*st
, char *homehost
);
964 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
965 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
966 void (*export_examine_super
)(struct supertype
*st
);
967 int (*examine_badblocks
)(struct supertype
*st
, int fd
, char *devname
);
968 int (*copy_metadata
)(struct supertype
*st
, int from
, int to
);
970 /* Used to report details of an active array.
971 * ->load_super was possibly given a 'component' string.
973 void (*detail_super
)(struct supertype
*st
, char *homehost
,
975 void (*brief_detail_super
)(struct supertype
*st
, char *subarray
);
976 void (*export_detail_super
)(struct supertype
*st
);
978 /* Optional: platform hardware / firmware details */
979 int (*detail_platform
)(int verbose
, int enumerate_only
, char *controller_path
);
980 int (*export_detail_platform
)(int verbose
, char *controller_path
);
983 * to get uuid to storing in bitmap metadata
984 * and 'reshape' backup-data metadata
985 * To see if a device is being re-added to an array it was part of.
987 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
989 /* Extract generic details from metadata. This could be details about
990 * the container, or about an individual array within the container.
991 * The determination is made either by:
992 * load_super being given a 'component' string.
993 * validate_geometry determining what to create.
994 * The info includes both array information and device information.
995 * The particular device should be:
996 * The last device added by add_to_super
997 * The device the metadata was loaded from by load_super
998 * If 'map' is present, then it is an array raid_disks long
999 * (raid_disk must already be set and correct) and it is filled
1000 * with 1 for slots that are thought to be active and 0 for slots which
1001 * appear to be failed/missing.
1002 * *info is zeroed out before data is added.
1004 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
1005 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
1006 /* Check if the given metadata is flagged as belonging to "this"
1007 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
1009 int (*match_home
)(struct supertype
*st
, char *homehost
);
1011 /* Make one of several generic modifications to metadata
1012 * prior to assembly (or other times).
1013 * sparc2.2 - first bug in early 0.90 metadata
1014 * super-minor - change name of 0.90 metadata
1015 * summaries - 'correct' any redundant data
1016 * resync - mark array as dirty to trigger a resync.
1017 * uuid - set new uuid - only 0.90 or 1.x
1018 * name - change the name of the array (where supported)
1019 * homehost - change which host this array is tied to.
1020 * devicesize - If metadata is at start of device, change recorded
1021 * device size to match actual device size
1022 * byteorder - swap bytes for 0.90 metadata
1024 * force-one - mark that device as uptodate, not old or failed.
1025 * force-array - mark array as clean if it would not otherwise
1027 * assemble - not sure how this is different from force-one...
1028 * linear-grow-new - add a new device to a linear array, but don't
1029 * change the size: so superblock still matches
1030 * linear-grow-update - now change the size of the array.
1031 * writemostly - set the WriteMostly1 bit in the superblock devflags
1032 * readwrite - clear the WriteMostly1 bit in the superblock devflags
1033 * failfast - set the FailFast1 bit in the superblock
1034 * nofailfast - clear the FailFast1 bit
1035 * no-bitmap - clear any record that a bitmap is present.
1036 * bbl - add a bad-block-log if possible
1037 * no-bbl - remove any bad-block-log is it is empty.
1038 * force-no-bbl - remove any bad-block-log even if empty.
1039 * revert-reshape - If a reshape is in progress, modify metadata so
1040 * it will resume going in the opposite direction.
1042 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
1043 enum update_opt update
,
1044 char *devname
, int verbose
,
1045 int uuid_set
, char *homehost
);
1047 /* Create new metadata for new array as described. This could
1048 * be a new container, or an array in a pre-existing container.
1049 * Also used to zero metadata prior to writing it to invalidate old
1052 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
1053 struct shape
*s
, char *name
,
1054 char *homehost
, int *uuid
,
1055 unsigned long long data_offset
);
1057 /* update the metadata to include new device, either at create or
1058 * when hot-adding a spare.
1060 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
1061 int fd
, char *devname
,
1062 unsigned long long data_offset
);
1063 /* update the metadata to delete a device,
1064 * when hot-removing.
1066 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
1068 /* Write metadata to one device when fixing problems or adding
1071 int (*store_super
)(struct supertype
*st
, int fd
);
1073 /* Write all metadata for this array.
1075 int (*write_init_super
)(struct supertype
*st
);
1076 /* Check if metadata read from one device is compatible with an array,
1077 * used when assembling an array, or pseudo-assembling was with
1078 * "--examine --brief"
1079 * If "st" has not yet been loaded the superblock from, "tst" is
1080 * moved in, otherwise the superblock in 'st' is compared with
1083 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
,
1085 /* Load metadata from a single device. If 'devname' is not NULL
1086 * print error messages as appropriate */
1087 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
1088 /* 'fd' is a 'container' md array - load array metadata from the
1091 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
1092 /* If 'arg' is a valid name of this metadata type, allocate and
1093 * return a 'supertype' for the particular minor version */
1094 struct supertype
* (*match_metadata_desc
)(char *arg
);
1095 /* If a device has the given size, and the data_offset has been
1096 * requested - work out how much space is available for data.
1097 * This involves adjusting for reserved space (e.g. bitmaps)
1098 * and for any rounding.
1099 * 'mdadm' only calls this for existing arrays where a possible
1100 * spare is being added. However some super-handlers call it
1101 * internally from validate_geometry when creating an array.
1103 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
,
1104 unsigned long long data_offset
);
1106 * Return spare criteria for array:
1107 * - minimum disk size can be used in array;
1108 * - sector size can be used in array.
1109 * Return values: 0 - for success and -EINVAL on error.
1111 int (*get_spare_criteria
)(struct supertype
*st
,
1112 struct spare_criteria
*sc
);
1113 /* Find somewhere to put a bitmap - possibly auto-size it - and
1114 * update the metadata to record this. The array may be newly
1115 * created, in which case data_size may be updated, or it might
1116 * already exist. Metadata handler can know if init_super
1117 * has been called, but not write_init_super.
1121 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
1122 int delay
, int write_behind
,
1123 unsigned long long size
, int may_change
, int major
);
1124 /* Perform additional setup required to activate a bitmap.
1126 int (*set_bitmap
)(struct supertype
*st
, struct mdinfo
*info
);
1127 /* Seek 'fd' to start of write-intent-bitmap. Must be an
1128 * md-native format bitmap
1130 int (*locate_bitmap
)(struct supertype
*st
, int fd
, int node_num
);
1131 /* if add_internal_bitmap succeeded for existing array, this
1134 int (*write_bitmap
)(struct supertype
*st
, int fd
, enum bitmap_update update
);
1135 /* Free the superblock and any other allocated data */
1136 void (*free_super
)(struct supertype
*st
);
1138 /* validate_geometry is called with an st returned by
1139 * match_metadata_desc.
1140 * It should check that the geometry described is compatible with
1141 * the metadata type. It will be called repeatedly as devices
1142 * added to validate changing size and new devices. If there are
1143 * inter-device dependencies, it should record sufficient details
1144 * so these can be validated.
1145 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
1147 * 1: everything is OK
1148 * 0: not OK for some reason - if 'verbose', then error was reported.
1149 * -1: st->sb was NULL, 'subdev' is a member of a container of this
1150 * type, but array is not acceptable for some reason
1151 * message was reported even if verbose is 0.
1153 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
1155 int *chunk
, unsigned long long size
,
1156 unsigned long long data_offset
,
1157 char *subdev
, unsigned long long *freesize
,
1158 int consistency_policy
, int verbose
);
1160 /* Return a linked list of 'mdinfo' structures for all arrays
1161 * in the container. For non-containers, it is like
1162 * getinfo_super with an allocated mdinfo.*/
1163 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
1164 /* query the supertype for default geometry */
1165 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
1166 /* Permit subarray's to be deleted from inactive containers */
1167 int (*kill_subarray
)(struct supertype
*st
,
1168 char *subarray_id
); /* optional */
1170 * update_subarray() - Permit subarray to be modified.
1172 * @subarray: Subarray name.
1173 * @update: Update option.
1174 * @ident: Optional identifiers.
1176 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
1177 enum update_opt update
, struct mddev_ident
*ident
);
1178 /* Check if reshape is supported for this external format.
1179 * st is obtained from super_by_fd() where st->subarray[0] is
1180 * initialized to indicate if reshape is being performed at the
1181 * container or subarray level
1183 #define APPLY_METADATA_CHANGES 1
1184 #define ROLLBACK_METADATA_CHANGES 0
1186 int (*reshape_super
)(struct supertype
*st
,
1187 unsigned long long size
, int level
,
1188 int layout
, int chunksize
, int raid_disks
,
1189 int delta_disks
, char *backup
, char *dev
,
1191 int verbose
); /* optional */
1192 int (*manage_reshape
)( /* optional */
1193 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1194 struct supertype
*st
, unsigned long blocks
,
1195 int *fds
, unsigned long long *offsets
,
1196 int dests
, int *destfd
, unsigned long long *destoffsets
);
1199 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
1202 /* Tell the metadata handler the current state of the array.
1203 * This covers whether it is known to be consistent (no pending writes)
1204 * and how far along a resync is known to have progressed
1205 * (in a->resync_start).
1206 * resync status is really irrelevant if the array is not consistent,
1207 * but some metadata (DDF!) have a place to record the distinction.
1208 * If 'consistent' is '2', then the array can mark it dirty if a
1209 * resync/recovery/whatever is required, or leave it clean if not.
1210 * Return value is 0 dirty (not consistent) and 1 if clean.
1211 * it is only really important if consistent is passed in as '2'.
1213 int (*set_array_state
)(struct active_array
*a
, int consistent
);
1215 /* When the state of a device might have changed, we call set_disk to
1216 * tell the metadata what the current state is.
1217 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1219 * set_disk might be called when the state of the particular disk has
1220 * not in fact changed.
1222 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
1223 void (*sync_metadata
)(struct supertype
*st
);
1224 void (*process_update
)(struct supertype
*st
,
1225 struct metadata_update
*update
);
1226 /* Prepare updates allocates extra memory that might be
1227 * needed. If the update cannot be understood, return 0.
1229 int (*prepare_update
)(struct supertype
*st
,
1230 struct metadata_update
*update
);
1232 /* activate_spare will check if the array is degraded and, if it
1233 * is, try to find some spare space in the container.
1234 * On success, it add appropriate updates (For process_update) to
1235 * to the 'updates' list and returns a list of 'mdinfo' identifying
1236 * the device, or devices as there might be multiple missing
1237 * devices and multiple spares available.
1239 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
1240 struct metadata_update
**updates
);
1242 * Return statically allocated string that represents metadata specific
1243 * controller domain of the disk. The domain is used in disk domain
1244 * matching functions. Disks belong to the same domain if the they have
1245 * the same domain from mdadm.conf and belong the same metadata domain.
1246 * Returning NULL or not providing this handler means that metadata
1247 * does not distinguish the differences between disks that belong to
1248 * different controllers. They are in the domain specified by
1249 * configuration file (mdadm.conf).
1250 * In case when the metadata has the notion of domains based on disk
1251 * it shall return NULL for disks that do not belong to the controller
1252 * the supported domains. Such disks will form another domain and won't
1253 * be mixed with supported ones.
1255 const char *(*get_disk_controller_domain
)(const char *path
);
1257 /* for external backup area */
1258 int (*recover_backup
)(struct supertype
*st
, struct mdinfo
*info
);
1260 /* validate container after assemble */
1261 int (*validate_container
)(struct mdinfo
*info
);
1263 /* write initial empty PPL on device */
1264 int (*write_init_ppl
)(struct supertype
*st
, struct mdinfo
*info
, int fd
);
1266 /* validate ppl before assemble */
1267 int (*validate_ppl
)(struct supertype
*st
, struct mdinfo
*info
,
1268 struct mdinfo
*disk
);
1270 /* records new bad block in metadata */
1271 int (*record_bad_block
)(struct active_array
*a
, int n
,
1272 unsigned long long sector
, int length
);
1274 /* clears bad block from metadata */
1275 int (*clear_bad_block
)(struct active_array
*a
, int n
,
1276 unsigned long long sector
, int length
);
1278 /* get list of bad blocks from metadata */
1279 struct md_bb
*(*get_bad_blocks
)(struct active_array
*a
, int n
);
1281 int swapuuid
; /* true if uuid is bigending rather than hostendian */
1283 const char *name
; /* canonical metadata name */
1286 extern struct superswitch super0
, super1
;
1287 extern struct superswitch super_imsm
, super_ddf
;
1288 extern struct superswitch mbr
, gpt
;
1290 void imsm_set_no_platform(int v
);
1292 struct metadata_update
{
1295 void *space
; /* allocated space that monitor will use */
1296 void **space_list
; /* list of allocated spaces that monitor can
1297 * use or that it returned.
1299 struct metadata_update
*next
;
1302 /* A supertype holds a particular collection of metadata.
1303 * It identifies the metadata type by the superswitch, and the particular
1304 * sub-version of that metadata type.
1305 * metadata read in or created is stored in 'sb' and 'info'.
1306 * There are also fields used by mdmon to track containers.
1308 * A supertype may refer to:
1309 * Just an array, possibly in a container
1310 * A container, not identifying any particular array
1311 * Info read from just one device, not yet fully describing the array/container.
1314 * A supertype is created by:
1320 struct superswitch
*ss
;
1323 char container_devnm
[32]; /* devnm of container */
1326 void *other
; /* Hack used to convert v0.90 to v1.0 */
1327 unsigned long long devsize
;
1328 unsigned long long data_offset
; /* used by v1.x only */
1329 int ignore_hw_compat
; /* used to inform metadata handlers that it should ignore
1330 HW/firmware related incompatability to load metadata.
1331 Used when examining metadata to display content of disk
1332 when user has no hw/firmare compatible system.
1334 struct metadata_update
*updates
;
1335 struct metadata_update
**update_tail
;
1337 /* extra stuff used by mdmon */
1338 struct active_array
*arrays
;
1339 int sock
; /* listen to external programs */
1340 char devnm
[32]; /* e.g. md0. This appears in metadata_version:
1348 struct mdinfo
*devs
;
1352 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
1353 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
1354 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
1355 static inline struct supertype
*guess_super(int fd
) {
1356 return guess_super_type(fd
, guess_any
);
1358 extern struct supertype
*dup_super(struct supertype
*st
);
1359 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
1360 extern int get_dev_sector_size(int fd
, char *dname
, unsigned int *sectsizep
);
1361 extern int must_be_container(int fd
);
1362 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
1363 extern int dev_sector_size_from_id(dev_t id
, unsigned int *size
);
1364 void wait_for(char *dev
, int fd
);
1367 * Data structures for policy management.
1368 * Each device can have a policy structure that lists
1369 * various name/value pairs each possibly with a metadata associated.
1370 * The policy list is sorted by name/value/metadata
1373 struct dev_policy
*next
;
1374 char *name
; /* None of these strings are allocated. They are
1375 * all just references to strings which are known
1376 * to exist elsewhere.
1377 * name and metadata can be compared by address equality.
1379 const char *metadata
;
1383 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
1385 /* iterate over the sublist starting at list, having the same
1386 * 'name' as 'list', and matching the given metadata (Where
1387 * NULL matches anything
1389 #define pol_for_each(item, list, _metadata) \
1391 item && item->name == list->name; \
1392 item = item->next) \
1393 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1397 * policy records read from mdadm are largely just name-value pairs.
1398 * The names are constants, not strdupped
1401 struct pol_rule
*next
;
1402 char *type
; /* rule_policy or rule_part */
1407 char *dups
; /* duplicates of 'value' with a partNN appended */
1411 extern char rule_policy
[], rule_part
[];
1412 extern char rule_path
[], rule_type
[];
1413 extern char type_part
[], type_disk
[];
1415 extern void policyline(char *line
, char *type
);
1416 extern void policy_add(char *type
, ...);
1417 extern void policy_free(void);
1419 extern struct dev_policy
*path_policy(char **paths
, char *type
);
1420 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
1421 extern struct dev_policy
*devid_policy(int devid
);
1422 extern void dev_policy_free(struct dev_policy
*p
);
1424 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1425 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
1426 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
1428 enum policy_action
{
1432 act_spare
, /* This only applies to bare devices */
1433 act_spare_same_slot
, /* this allows non-bare devices,
1434 * but only if recent removal */
1435 act_force_spare
, /* this allow non-bare devices in any case */
1439 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
1440 enum policy_action want
);
1441 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
1442 enum policy_action want
);
1445 struct domainlist
*next
;
1449 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
1450 const char *metadata
);
1451 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
1452 const char *metadata
);
1453 extern void domainlist_add_dev(struct domainlist
**dom
, int devid
,
1454 const char *metadata
);
1455 extern void domain_free(struct domainlist
*dl
);
1456 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
1457 const char *metadata
);
1458 void domain_add(struct domainlist
**domp
, char *domain
);
1460 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
1461 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
1463 extern void sysfs_rules_apply(char *devnm
, struct mdinfo
*dev
);
1464 extern void sysfsline(char *line
);
1470 #define HAVE_NFTW we assume
1474 # include <features.h>
1475 # ifndef __UCLIBC_HAS_LFS__
1476 # define lseek64 lseek
1478 # ifndef __UCLIBC_HAS_FTW__
1488 #if defined(__KLIBC__)
1504 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
1506 extern int Manage_ro(char *devname
, int fd
, int readonly
);
1507 extern int Manage_run(char *devname
, int fd
, struct context
*c
);
1508 extern int Manage_stop(char *devname
, int fd
, int quiet
,
1510 extern int Manage_subdevs(char *devname
, int fd
,
1511 struct mddev_dev
*devlist
, int verbose
, int test
,
1512 enum update_opt update
, int force
);
1513 extern int autodetect(void);
1514 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1515 extern int Grow_addbitmap(char *devname
, int fd
,
1516 struct context
*c
, struct shape
*s
);
1517 extern int Grow_reshape(char *devname
, int fd
,
1518 struct mddev_dev
*devlist
,
1519 struct context
*c
, struct shape
*s
);
1520 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1521 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1522 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1523 struct mdinfo
*info
, char *backup_file
,
1524 int forked
, int freeze_reshape
);
1525 extern int Grow_consistency_policy(char *devname
, int fd
,
1526 struct context
*c
, struct shape
*s
);
1528 extern int restore_backup(struct supertype
*st
,
1529 struct mdinfo
*content
,
1532 char **backup_filep
,
1534 extern int Grow_continue_command(char *devname
, int fd
,
1535 char *backup_file
, int verbose
);
1537 extern int Assemble(struct supertype
*st
, char *mddev
,
1538 struct mddev_ident
*ident
,
1539 struct mddev_dev
*devlist
,
1542 extern int Build(struct mddev_ident
*ident
, struct mddev_dev
*devlist
, struct shape
*s
,
1545 extern int Create(struct supertype
*st
, struct mddev_ident
*ident
, int subdevs
,
1546 struct mddev_dev
*devlist
, struct shape
*s
, struct context
*c
);
1548 extern int Detail(char *dev
, struct context
*c
);
1549 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
, int export
, char *controller_path
);
1550 extern int Query(char *dev
);
1551 extern int ExamineBadblocks(char *devname
, int brief
, struct supertype
*forcest
);
1552 extern int Examine(struct mddev_dev
*devlist
, struct context
*c
,
1553 struct supertype
*forcest
);
1554 extern int Monitor(struct mddev_dev
*devlist
,
1555 char *mailaddr
, char *alert_cmd
,
1557 int daemonise
, int oneshot
,
1558 int dosyslog
, char *pidfile
, int increments
,
1561 extern int Kill(char *dev
, struct supertype
*st
, int force
, int verbose
, int noexcl
);
1562 extern int Kill_subarray(char *dev
, char *subarray
, int verbose
);
1563 extern int Update_subarray(char *dev
, char *subarray
, enum update_opt update
, struct mddev_ident
*ident
, int quiet
);
1564 extern int Wait(char *dev
);
1565 extern int WaitClean(char *dev
, int verbose
);
1566 extern int SetAction(char *dev
, char *action
);
1568 extern int Incremental(struct mddev_dev
*devlist
, struct context
*c
,
1569 struct supertype
*st
);
1570 extern void RebuildMap(void);
1571 extern int IncrementalScan(struct context
*c
, char *devnm
);
1572 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1573 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1574 unsigned long chunksize
, unsigned long daemon_sleep
,
1575 unsigned long write_behind
,
1576 unsigned long long array_size
,
1578 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1579 extern int IsBitmapDirty(char *filename
);
1580 extern int Write_rules(char *rule_name
);
1581 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1583 /* calculate the size of the bitmap given the array size and bitmap chunksize */
1584 static inline unsigned long long
1585 bitmap_bits(unsigned long long array_size
, unsigned long chunksize
)
1587 return (array_size
* 512 + chunksize
- 1) / chunksize
;
1590 extern int Dump_metadata(char *dev
, char *dir
, struct context
*c
,
1591 struct supertype
*st
);
1592 extern int Restore_metadata(char *dev
, char *dir
, struct context
*c
,
1593 struct supertype
*st
, int only
);
1595 int md_array_valid(int fd
);
1596 int md_array_active(int fd
);
1597 int md_array_is_active(struct mdinfo
*info
);
1598 int md_get_array_info(int fd
, struct mdu_array_info_s
*array
);
1599 int md_set_array_info(int fd
, struct mdu_array_info_s
*array
);
1600 int md_get_disk_info(int fd
, struct mdu_disk_info_s
*disk
);
1601 extern int get_linux_version(void);
1602 extern int mdadm_version(char *version
);
1603 extern unsigned long long parse_size(char *size
);
1604 extern int parse_uuid(char *str
, int uuid
[4]);
1605 int default_layout(struct supertype
*st
, int level
, int verbose
);
1606 extern int is_near_layout_10(int layout
);
1607 extern int parse_layout_10(char *layout
);
1608 extern int parse_layout_faulty(char *layout
);
1609 extern int parse_num(int *dest
, const char *num
);
1610 extern int parse_cluster_confirm_arg(char *inp
, char **devname
, int *slot
);
1611 extern int check_ext2(int fd
, char *name
);
1612 extern int check_reiser(int fd
, char *name
);
1613 extern int check_raid(int fd
, char *name
);
1614 extern int check_partitions(int fd
, char *dname
,
1615 unsigned long long freesize
,
1616 unsigned long long size
);
1617 extern bool is_name_posix_compatible(const char *path
);
1618 extern int fstat_is_blkdev(int fd
, char *devname
, dev_t
*rdev
);
1619 extern int stat_is_blkdev(char *devname
, dev_t
*rdev
);
1621 extern bool is_string_lq(const char * const str
, size_t max_len
);
1622 extern bool is_dev_alive(char *path
);
1623 extern int get_mdp_major(void);
1624 extern int get_maj_min(char *dev
, int *major
, int *minor
);
1625 extern bool is_bit_set(int *val
, unsigned char index
);
1626 extern int dev_open(char *dev
, int flags
);
1627 extern int open_dev(char *devnm
);
1628 extern void reopen_mddev(int mdfd
);
1629 extern int open_dev_flags(char *devnm
, int flags
);
1630 extern int open_dev_excl(char *devnm
);
1631 extern int is_standard(char *dev
, int *nump
);
1632 extern int same_dev(char *one
, char *two
);
1633 extern int compare_paths (char* path1
,char* path2
);
1634 extern void enable_fds(int devices
);
1635 extern void manage_fork_fds(int close_all
);
1636 extern int continue_via_systemd(char *devnm
, char *service_name
, char *prefix
);
1638 extern void ident_init(struct mddev_ident
*ident
);
1639 extern mdadm_status_t
ident_set_devname(struct mddev_ident
*ident
, const char *devname
);
1640 extern mdadm_status_t
ident_set_name(struct mddev_ident
*ident
, const char *name
);
1642 extern int parse_auto(char *str
, char *msg
, int config
);
1643 extern struct mddev_ident
*conf_get_ident(char *dev
);
1644 extern struct mddev_dev
*conf_get_devs(void);
1645 extern int conf_test_dev(char *devname
);
1646 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1647 extern struct createinfo
*conf_get_create_info(void);
1648 extern void set_conffile(char *file
);
1649 extern char *conf_get_mailaddr(void);
1650 extern char *conf_get_mailfrom(void);
1651 extern char *conf_get_program(void);
1652 extern char *conf_get_homehost(int *require_homehostp
);
1653 extern char *conf_get_homecluster(void);
1654 extern int conf_get_monitor_delay(void);
1655 extern char *conf_line(FILE *file
);
1656 extern char *conf_word(FILE *file
, int allow_key
);
1657 extern void print_quoted(char *str
);
1658 extern int use_udev(void);
1659 extern void print_escape(char *str
);
1660 extern unsigned long GCD(unsigned long a
, unsigned long b
);
1661 extern int conf_name_is_free(char *name
);
1662 extern bool is_devname_ignore(const char *devname
);
1663 extern bool is_devname_md_numbered(const char *devname
);
1664 extern bool is_devname_md_d_numbered(const char *devname
);
1665 extern int conf_verify_devnames(struct mddev_ident
*array_list
);
1666 extern int devname_matches(char *name
, char *match
);
1667 extern struct mddev_ident
*conf_match(struct supertype
*st
,
1668 struct mdinfo
*info
,
1670 int verbose
, int *rvp
);
1672 extern void free_line(char *line
);
1673 extern int match_oneof(char *devices
, char *devname
);
1674 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1675 extern const int uuid_zero
[4];
1676 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1677 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1678 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1679 extern char *fname_from_uuid(struct supertype
*st
,
1680 struct mdinfo
*info
, char *buf
, char sep
);
1681 extern unsigned long calc_csum(void *super
, int bytes
);
1682 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1684 extern int ask(char *mesg
);
1685 extern unsigned long long get_component_size(int fd
);
1686 extern void remove_partitions(int fd
);
1687 extern int test_partition(int fd
);
1688 extern int test_partition_from_id(dev_t id
);
1689 extern int get_data_disks(int level
, int layout
, int raid_disks
);
1690 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1691 int chunksize
, unsigned long long devsize
);
1692 extern int flush_metadata_updates(struct supertype
*st
);
1693 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1694 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1695 struct mdinfo
*content
,
1697 char *chosen_name
, int *result
);
1699 #define INCR_UNSAFE 2
1700 #define INCR_ALREADY 4
1702 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1703 struct spare_criteria
*criteria
,
1704 struct domainlist
*domlist
,
1706 const char *metadata
, int get_one
);
1707 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1708 extern int add_disk(int mdfd
, struct supertype
*st
,
1709 struct mdinfo
*sra
, struct mdinfo
*info
);
1710 extern int remove_disk(int mdfd
, struct supertype
*st
,
1711 struct mdinfo
*sra
, struct mdinfo
*info
);
1712 extern int hot_remove_disk(int mdfd
, unsigned long dev
, int force
);
1713 extern int sys_hot_remove_disk(int statefd
, int force
);
1714 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1715 unsigned long long min_recovery_start(struct mdinfo
*array
);
1717 extern char *human_size(long long bytes
);
1718 extern char *human_size_brief(long long bytes
, int prefix
);
1719 extern void print_r10_layout(int layout
);
1721 extern char *find_free_devnm(int use_partitions
);
1723 extern void put_md_name(char *name
);
1724 extern char *devid2kname(dev_t devid
);
1725 extern char *devid2devnm(dev_t devid
);
1726 extern dev_t
devnm2devid(char *devnm
);
1727 extern char *get_md_name(char *devnm
);
1729 extern char DefaultConfFile
[];
1731 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1732 char *chosen
, int block_udev
);
1733 /* values for 'trustworthy' */
1735 #define LOCAL_ANY 10
1738 extern int open_mddev(char *dev
, int report_errors
);
1739 extern int is_mddev(char *dev
);
1740 extern int open_container(int fd
);
1741 extern int metadata_container_matches(char *metadata
, char *devnm
);
1742 extern int metadata_subdev_matches(char *metadata
, char *devnm
);
1743 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1744 extern int is_subarray_active(char *subarray
, char *devname
);
1745 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1746 extern struct superswitch
*version_to_superswitch(char *vers
);
1748 extern int mdmon_running(char *devnm
);
1749 extern int mdmon_pid(char *devnm
);
1750 extern int check_env(char *name
);
1751 extern __u32
random32(void);
1752 extern void random_uuid(__u8
*buf
);
1753 extern int start_mdmon(char *devnm
);
1755 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1756 struct supertype
*st
, unsigned long stripes
,
1757 int *fds
, unsigned long long *offsets
,
1758 int dests
, int *destfd
, unsigned long long *destoffsets
);
1759 void abort_reshape(struct mdinfo
*sra
);
1761 void *super1_make_v0(struct supertype
*st
, struct mdinfo
*info
, mdp_super_t
*sb0
);
1763 extern char *stat2kname(struct stat
*st
);
1764 extern char *fd2kname(int fd
);
1765 extern char *stat2devnm(struct stat
*st
);
1766 bool stat_is_md_dev(struct stat
*st
);
1767 extern char *fd2devnm(int fd
);
1769 extern int in_initrd(void);
1772 void *cmap_handle
; /* corosync lib related */
1774 int (*initialize
)(cmap_handle_t
*handle
);
1775 int (*get_string
)(cmap_handle_t handle
,
1778 int (*finalize
)(cmap_handle_t handle
);
1781 extern void set_cmap_hooks(void);
1782 extern void set_hooks(void);
1785 void *dlm_handle
; /* dlm lib related */
1787 dlm_lshandle_t (*create_lockspace
)(const char *name
,
1789 dlm_lshandle_t (*open_lockspace
)(const char *name
);
1790 int (*release_lockspace
)(const char *name
, dlm_lshandle_t ls
,
1792 int (*ls_lock
)(dlm_lshandle_t lockspace
, uint32_t mode
,
1793 struct dlm_lksb
*lksb
, uint32_t flags
,
1794 const void *name
, unsigned int namelen
,
1795 uint32_t parent
, void (*astaddr
) (void *astarg
),
1796 void *astarg
, void (*bastaddr
) (void *astarg
),
1798 int (*ls_unlock_wait
)(dlm_lshandle_t lockspace
, uint32_t lkid
,
1799 uint32_t flags
, struct dlm_lksb
*lksb
);
1800 int (*ls_get_fd
)(dlm_lshandle_t ls
);
1801 int (*dispatch
)(int fd
);
1804 extern int get_cluster_name(char **name
);
1805 extern int dlm_funs_ready(void);
1806 extern int cluster_get_dlmlock(void);
1807 extern int cluster_release_dlmlock(void);
1808 extern void set_dlm_hooks(void);
1810 #define MSEC_TO_NSEC(msec) ((msec) * 1000000)
1811 #define USEC_TO_NSEC(usec) ((usec) * 1000)
1812 extern void sleep_for(unsigned int sec
, long nsec
, bool wake_after_interrupt
);
1813 extern bool is_directory(const char *path
);
1814 extern bool is_file(const char *path
);
1815 extern int s_gethostname(char *buf
, int buf_len
);
1817 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1818 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1819 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1820 (ROUND_UP((unsigned long)(ptr), base)))
1822 static inline int is_subarray(char *vers
)
1824 /* The version string for a 'subarray' (an array in a container)
1826 * /containername/componentname for normal read-write arrays
1827 * -containername/componentname for arrays which mdmon must not
1828 * reconfigure. They might be read-only
1829 * or might be undergoing reshape etc.
1830 * containername is e.g. md0, md_d1
1831 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1833 return (*vers
== '/' || *vers
== '-');
1836 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1838 return &ent
->metadata_version
[10+strlen(container
)+1];
1842 * signal_s() - Wrapper for sigaction() with signal()-like interface.
1843 * @sig: The signal to set the signal handler to.
1844 * @handler: The signal handler.
1846 * Return: previous handler or SIG_ERR on failure.
1848 static inline sighandler_t
signal_s(int sig
, sighandler_t handler
)
1850 struct sigaction new_act
;
1851 struct sigaction old_act
;
1853 new_act
.sa_handler
= handler
;
1854 new_act
.sa_flags
= 0;
1856 if (sigaction(sig
, &new_act
, &old_act
) == 0)
1857 return old_act
.sa_handler
;
1863 #define dprintf(fmt, arg...) \
1864 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1865 #define dprintf_cont(fmt, arg...) \
1866 fprintf(stderr, fmt, ##arg)
1868 #define dprintf(fmt, arg...) \
1869 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1870 #define dprintf_cont(fmt, arg...) \
1871 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1874 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1878 ret
= vasprintf(strp
, fmt
, ap
);
1885 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1887 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1889 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1891 #define pr_info(fmt, args...) printf("%s: "fmt, Name, ##args)
1893 void *xmalloc(size_t len
);
1894 void *xrealloc(void *ptr
, size_t len
);
1895 void *xcalloc(size_t num
, size_t size
);
1896 char *xstrdup(const char *str
);
1898 #define LEVEL_MULTIPATH (-4)
1899 #define LEVEL_LINEAR (-1)
1900 #define LEVEL_FAULTY (-5)
1902 /* kernel module doesn't know about these */
1903 #define LEVEL_CONTAINER (-100)
1904 #define LEVEL_UNSUPPORTED (-200)
1906 /* the kernel does know about this one ... */
1907 #define LEVEL_NONE (-1000000)
1911 #define WriteTransient 0
1912 #define ReadTransient 1
1913 #define WritePersistent 2
1914 #define ReadPersistent 3
1915 #define WriteAll 4 /* doesn't go to device */
1916 #define ReadFixable 5
1919 #define ClearErrors 31
1920 #define ClearFaults 30
1922 #define AllPersist 100 /* internal use only */
1923 #define NoPersist 101
1925 #define ModeMask 0x1f
1932 #define minor(x) ((x)&0xff)
1933 #define major(x) (((x)>>8)&0xff)
1934 #define makedev(M,m) (((M)<<8) | (m))
1938 RAID0_ORIG_LAYOUT
= 1,
1939 RAID0_ALT_MULTIZONE_LAYOUT
= 2,
1943 #define ALGORITHM_LEFT_ASYMMETRIC 0
1944 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1945 #define ALGORITHM_LEFT_SYMMETRIC 2
1946 #define ALGORITHM_RIGHT_SYMMETRIC 3
1948 /* Define non-rotating (raid4) algorithms. These allow
1949 * conversion of raid4 to raid5.
1951 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1952 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1954 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1955 * Firstly, the exact positioning of the parity block is slightly
1956 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1958 * Secondly, or order of datablocks over which the Q syndrome is computed
1960 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1961 * These layouts are from the DDFv1.2 spec.
1962 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1963 * leaves RLQ=3 as 'Vendor Specific'
1966 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1967 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1968 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1970 /* For every RAID5 algorithm we define a RAID6 algorithm
1971 * with exactly the same layout for data and parity, and
1972 * with the Q block always on the last device (N-1).
1973 * This allows trivial conversion from RAID5 to RAID6
1975 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1976 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1977 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1978 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1979 #define ALGORITHM_PARITY_0_6 20
1980 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1982 /* Define PATH_MAX in case we don't use glibc or standard library does
1983 * not have PATH_MAX defined. Assume max path length is 4K characters.
1986 #define PATH_MAX 4096
1989 #define RESYNC_NONE -1
1990 #define RESYNC_DELAYED -2
1991 #define RESYNC_PENDING -3
1992 #define RESYNC_REMOTE -4
1993 #define RESYNC_UNKNOWN -5
1995 /* When using "GET_DISK_INFO" it isn't certain how high
1996 * we need to check. So we impose an absolute limit of
1997 * MAX_DISKS. This needs to be much more than the largest
1998 * number of devices any metadata can support. Currently
1999 * v1.x can support 1920
2001 #define MAX_DISKS 4096
2003 /* Sometimes the 'size' value passed needs to mean "Maximum".
2004 * In those cases with use MAX_SIZE
2008 /* We want to use unsigned numbers for sector counts, but need
2009 * a value for 'invalid'. Use '1'.
2011 #define INVALID_SECTORS 1
2012 /* And another special number needed for --data_offset=variable */
2013 #define VARIABLE_OFFSET 3
2016 * is_container() - check if @level is &LEVEL_CONTAINER
2017 * @level: level value
2020 * 1 if level is equal to &LEVEL_CONTAINER, 0 otherwise.
2022 static inline int is_container(const int level
)
2024 if (level
== LEVEL_CONTAINER
)