2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 * Email: <neilb@suse.de>
26 #define _FILE_OFFSET_BITS 64
29 extern __off64_t lseek64
__P ((int __fd
, __off64_t __offset
, int __whence
));
30 #elif !defined(lseek64)
31 # if defined(__NO_STAT64) || __WORDSIZE != 32
32 # define lseek64 lseek
36 #include <sys/types.h>
48 /* Newer glibc requires sys/sysmacros.h directly for makedev() */
49 #include <sys/sysmacros.h>
52 /* dietlibc has deprecated random and srandom!! */
59 typedef uint64_t cmap_handle_t
;
61 #include <corosync/cmap.h>
68 #define LKF_NOQUEUE 0x00000001
70 #define EUNLOCK 0x10002
72 typedef void *dlm_lshandle_t
;
82 #include <linux/kdev_t.h>
83 /*#include <linux/fs.h> */
84 #include <sys/mount.h>
85 #include <asm/types.h>
86 #include <sys/ioctl.h>
88 #define MdpMinorShift 6
91 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
94 #define DEFAULT_CHUNK 512
95 #define DEFAULT_BITMAP_CHUNK 4096
96 #define DEFAULT_BITMAP_DELAY 5
97 #define DEFAULT_MAX_WRITE_BEHIND 256
99 /* MAP_DIR should be somewhere that persists across the pivotroot
100 * from early boot to late boot.
101 * /run seems to have emerged as the best standard.
104 #define MAP_DIR "/run/mdadm"
106 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
107 * want something other than the default of "map"
110 #define MAP_FILE "map"
111 #endif /* MAP_FILE */
112 /* MDMON_DIR is where pid and socket files used for communicating
113 * with mdmon normally live. Best is /var/run/mdadm as
114 * mdmon is needed at early boot then it needs to write there prior
115 * to /var/run being mounted read/write, and it also then needs to
116 * persist beyond when /var/run is mounter read-only. So, to be
117 * safe, the default is somewhere that is read/write early in the
118 * boot process and stays up as long as possible during shutdown.
121 #define MDMON_DIR "/run/mdadm"
122 #endif /* MDMON_DIR */
124 /* FAILED_SLOTS is where to save files storing recent removal of array
125 * member in order to allow future reuse of disk inserted in the same
126 * slot for array recovery
128 #ifndef FAILED_SLOTS_DIR
129 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
130 #endif /* FAILED_SLOTS */
138 /* Redhat don't like to #include <asm/byteorder.h>, and
139 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
140 * and there is no standard conversion function so... */
141 /* And dietlibc doesn't think byteswap is ok, so.. */
142 /* #include <byteswap.h> */
143 #define __mdadm_bswap_16(x) (((x) & 0x00ffU) << 8 | \
144 ((x) & 0xff00U) >> 8)
145 #define __mdadm_bswap_32(x) (((x) & 0x000000ffU) << 24 | \
146 ((x) & 0xff000000U) >> 24 | \
147 ((x) & 0x0000ff00U) << 8 | \
148 ((x) & 0x00ff0000U) >> 8)
149 #define __mdadm_bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
150 ((x) & 0xff00000000000000ULL) >> 56 | \
151 ((x) & 0x000000000000ff00ULL) << 40 | \
152 ((x) & 0x00ff000000000000ULL) >> 40 | \
153 ((x) & 0x0000000000ff0000ULL) << 24 | \
154 ((x) & 0x0000ff0000000000ULL) >> 24 | \
155 ((x) & 0x00000000ff000000ULL) << 8 | \
156 ((x) & 0x000000ff00000000ULL) >> 8)
158 #if !defined(__KLIBC__)
159 #if BYTE_ORDER == LITTLE_ENDIAN
160 #define __cpu_to_le16(_x) (unsigned int)(_x)
161 #define __cpu_to_le32(_x) (unsigned int)(_x)
162 #define __cpu_to_le64(_x) (unsigned long long)(_x)
163 #define __le16_to_cpu(_x) (unsigned int)(_x)
164 #define __le32_to_cpu(_x) (unsigned int)(_x)
165 #define __le64_to_cpu(_x) (unsigned long long)(_x)
167 #define __cpu_to_be16(_x) __mdadm_bswap_16(_x)
168 #define __cpu_to_be32(_x) __mdadm_bswap_32(_x)
169 #define __cpu_to_be64(_x) __mdadm_bswap_64(_x)
170 #define __be16_to_cpu(_x) __mdadm_bswap_16(_x)
171 #define __be32_to_cpu(_x) __mdadm_bswap_32(_x)
172 #define __be64_to_cpu(_x) __mdadm_bswap_64(_x)
173 #elif BYTE_ORDER == BIG_ENDIAN
174 #define __cpu_to_le16(_x) __mdadm_bswap_16(_x)
175 #define __cpu_to_le32(_x) __mdadm_bswap_32(_x)
176 #define __cpu_to_le64(_x) __mdadm_bswap_64(_x)
177 #define __le16_to_cpu(_x) __mdadm_bswap_16(_x)
178 #define __le32_to_cpu(_x) __mdadm_bswap_32(_x)
179 #define __le64_to_cpu(_x) __mdadm_bswap_64(_x)
181 #define __cpu_to_be16(_x) (unsigned int)(_x)
182 #define __cpu_to_be32(_x) (unsigned int)(_x)
183 #define __cpu_to_be64(_x) (unsigned long long)(_x)
184 #define __be16_to_cpu(_x) (unsigned int)(_x)
185 #define __be32_to_cpu(_x) (unsigned int)(_x)
186 #define __be64_to_cpu(_x) (unsigned long long)(_x)
188 # error "unknown endianness."
190 #endif /* __KLIBC__ */
193 * Partially stolen from include/linux/unaligned/packed_struct.h
195 struct __una_u16
{ __u16 x
; } __attribute__ ((packed
));
196 struct __una_u32
{ __u32 x
; } __attribute__ ((packed
));
198 static inline __u16
__get_unaligned16(const void *p
)
200 const struct __una_u16
*ptr
= (const struct __una_u16
*)p
;
204 static inline __u32
__get_unaligned32(const void *p
)
206 const struct __una_u32
*ptr
= (const struct __una_u32
*)p
;
210 static inline void __put_unaligned16(__u16 val
, void *p
)
212 struct __una_u16
*ptr
= (struct __una_u16
*)p
;
216 static inline void __put_unaligned32(__u32 val
, void *p
)
218 struct __una_u32
*ptr
= (struct __una_u32
*)p
;
223 * Check at compile time that something is of a particular type.
224 * Always evaluates to 1 so you may use it easily in comparisons.
227 #define typecheck(type,x) \
229 typeof(x) __dummy2; \
230 (void)(&__dummy == &__dummy2); \
235 * These inlines deal with timer wrapping correctly.
237 * time_after(a,b) returns true if the time a is after time b.
240 #define time_after(a,b) \
241 (typecheck(unsigned int, a) && \
242 typecheck(unsigned int, b) && \
243 ((int)((b) - (a)) < 0))
245 #define time_before(a,b) time_after(b,a)
248 * min()/max()/clamp() macros that also do
249 * strict type-checking.. See the
250 * "unnecessary" pointer comparison.
252 #define min(x, y) ({ \
253 typeof(x) _min1 = (x); \
254 typeof(y) _min2 = (y); \
255 (void) (&_min1 == &_min2); \
256 _min1 < _min2 ? _min1 : _min2; })
258 #define max(x, y) ({ \
259 typeof(x) _max1 = (x); \
260 typeof(y) _max2 = (y); \
261 (void) (&_max1 == &_max2); \
262 _max1 > _max2 ? _max1 : _max2; })
264 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
266 extern const char Name
[];
269 unsigned long long sector
;
276 struct md_bb_entry
*entries
;
279 /* general information that might be extracted from a superblock */
281 mdu_array_info_t array
;
282 mdu_disk_info_t disk
;
286 unsigned long long data_offset
;
287 unsigned long long new_data_offset
;
288 unsigned long long component_size
; /* same as array.size, except in
289 * sectors and up to 64bits.
291 unsigned long long custom_array_size
; /* size for non-default sized
292 * arrays (in sectors)
295 #define VOLUME_RESHAPE 1
296 #define CONTAINER_RESHAPE 2
297 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
299 unsigned long long reshape_progress
;
300 int recovery_blocked
; /* for external metadata it
301 * indicates that there is
302 * reshape in progress in
304 * for native metadata it is
305 * reshape_active field mirror
307 int journal_device_required
;
311 CONSISTENCY_POLICY_UNKNOWN
,
312 CONSISTENCY_POLICY_NONE
,
313 CONSISTENCY_POLICY_RESYNC
,
314 CONSISTENCY_POLICY_BITMAP
,
315 CONSISTENCY_POLICY_JOURNAL
,
316 CONSISTENCY_POLICY_PPL
,
317 } consistency_policy
;
319 /* During reshape we can sometimes change the data_offset to avoid
320 * over-writing still-valid data. We need to know if there is space.
321 * So getinfo_super will fill in space_before and space_after in sectors.
322 * data_offset can be increased or decreased by this amount.
324 unsigned long long space_before
, space_after
;
326 unsigned long long resync_start
; /* per-array resync position */
327 unsigned long long recovery_start
; /* per-device rebuild position */
328 #define MaxSector (~0ULL) /* resync/recovery complete position */
330 long bitmap_offset
; /* 0 == none, 1 == a file */
331 unsigned int ppl_size
;
333 unsigned long long ppl_sector
;
334 unsigned long safe_mode_delay
; /* ms delay to mark clean */
335 int new_level
, delta_disks
, new_layout
, new_chunk
;
337 unsigned long cache_size
; /* size of raid456 stripe cache*/
339 char text_version
[50];
341 int container_member
; /* for assembling external-metatdata arrays
342 * This is to be used internally by metadata
344 int container_enough
; /* flag external handlers can set to
345 * indicate that subarrays have not enough (-1),
346 * enough to start (0), or all expected disks (1) */
351 /* Device info for mdmon: */
358 #define DS_WRITE_MOSTLY 4
360 #define DS_BLOCKED 16
361 #define DS_REMOVE 1024
362 #define DS_UNBLOCK 2048
363 int prev_state
, curr_state
, next_state
;
365 /* info read from sysfs */
390 struct supertype
*supertype
;
393 struct spare_criteria
{
394 unsigned long long min_size
;
395 unsigned int sector_size
;
411 extern char short_options
[];
412 extern char short_bitmap_options
[];
413 extern char short_bitmap_auto_options
[];
414 extern struct option long_options
[];
415 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
417 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
419 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
421 /* for option that don't have short equivilents, we assign arbitrary
422 * numbers later than any 'short' character option.
424 enum special_options
{
489 enum prefix_standard
{
501 FlagDefault
, FlagSet
, FlagClear
,
504 /* structures read from config file */
505 /* List of mddevice names and identifiers
506 * Identifiers can be:
508 * super-minor=decimal-minor-number-from-superblock
509 * devices=comma,separated,list,of,device,names,with,wildcards
511 * If multiple fields are present, the intersection of all matching
512 * devices is considered
514 #define UnSet (0xfffe)
524 char *devices
; /* comma separated list of device
525 * names with wild cards
530 struct supertype
*st
;
531 int autof
; /* 1 for normal, 2 for partitioned */
536 char *container
; /* /dev/whatever name of container, or
537 * uuid of container. You would expect
538 * this to be the 'devname' or UUID
539 * of some other entry.
541 char *member
; /* subarray within a container */
543 struct mddev_ident
*next
;
545 /* fields needed by different users of this structure */
546 int assembled
; /* set when assembly succeeds */
558 int require_homehost
;
588 unsigned long long size
;
589 int consistency_policy
;
592 /* List of device names - wildcards expanded */
595 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail,
597 * Not set for names read from .config
599 enum flag_mode writemostly
;
600 enum flag_mode failfast
;
601 int used
; /* set when used */
602 long long data_offset
;
603 struct mddev_dev
*next
;
606 typedef struct mapping
{
615 char *pattern
; /* U for up, _ for down */
616 int percent
; /* -1 if no resync */
617 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
620 char * metadata_version
;
623 struct dev_member
*next
;
625 struct mdstat_ent
*next
;
628 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
629 extern void mdstat_close(void);
630 extern void free_mdstat(struct mdstat_ent
*ms
);
631 extern void mdstat_wait(int seconds
);
632 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
633 extern int mddev_busy(char *devnm
);
634 extern struct mdstat_ent
*mdstat_by_component(char *name
);
635 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, char *container
);
638 struct map_ent
*next
;
645 extern int map_update(struct map_ent
**mpp
, char *devnm
, char *metadata
,
646 int uuid
[4], char *path
);
647 extern void map_remove(struct map_ent
**map
, char *devnm
);
648 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
649 extern struct map_ent
*map_by_devnm(struct map_ent
**map
, char *devnm
);
650 extern void map_free(struct map_ent
*map
);
651 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
652 extern void map_read(struct map_ent
**melp
);
653 extern int map_write(struct map_ent
*mel
);
654 extern void map_delete(struct map_ent
**mapp
, char *devnm
);
655 extern void map_add(struct map_ent
**melp
,
656 char *devnm
, char *metadata
, int uuid
[4], char *path
);
657 extern int map_lock(struct map_ent
**melp
);
658 extern void map_unlock(struct map_ent
**melp
);
659 extern void map_fork(void);
661 /* various details can be requested */
662 enum sysfs_read_flags
{
663 GET_LEVEL
= (1 << 0),
664 GET_LAYOUT
= (1 << 1),
665 GET_COMPONENT
= (1 << 2),
666 GET_CHUNK
= (1 << 3),
667 GET_CACHE
= (1 << 4),
668 GET_MISMATCH
= (1 << 5),
669 GET_VERSION
= (1 << 6),
670 GET_DISKS
= (1 << 7),
671 GET_SAFEMODE
= (1 << 9),
672 GET_BITMAP_LOCATION
= (1 << 10),
674 GET_DEVS
= (1 << 20), /* gets role, major, minor */
675 GET_OFFSET
= (1 << 21),
676 GET_SIZE
= (1 << 22),
677 GET_STATE
= (1 << 23),
678 GET_ERROR
= (1 << 24),
679 GET_ARRAY_STATE
= (1 << 25),
680 GET_CONSISTENCY_POLICY
= (1 << 26),
681 GET_DEVS_ALL
= (1 << 27),
684 /* If fd >= 0, get the array it is open on,
687 extern int sysfs_open(char *devnm
, char *devname
, char *attr
);
688 extern int sysfs_init(struct mdinfo
*mdi
, int fd
, char *devnm
);
689 extern void sysfs_init_dev(struct mdinfo
*mdi
, dev_t devid
);
690 extern void sysfs_free(struct mdinfo
*sra
);
691 extern struct mdinfo
*sysfs_read(int fd
, char *devnm
, unsigned long options
);
692 extern int sysfs_attr_match(const char *attr
, const char *str
);
693 extern int sysfs_match_word(const char *word
, char **list
);
694 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
695 char *name
, char *val
);
696 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
697 char *name
, unsigned long long val
);
698 extern int sysfs_set_num_signed(struct mdinfo
*sra
, struct mdinfo
*dev
,
699 char *name
, long long val
);
700 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
701 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
703 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
704 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
705 char *name
, unsigned long long *val
);
706 extern int sysfs_fd_get_two(int fd
, unsigned long long *v1
, unsigned long long *v2
);
707 extern int sysfs_get_two(struct mdinfo
*sra
, struct mdinfo
*dev
,
708 char *name
, unsigned long long *v1
, unsigned long long *v2
);
709 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
710 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
712 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
713 char *name
, char *val
, int size
);
714 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
715 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
716 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
717 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
718 extern int sysfs_unique_holder(char *devnm
, long rdev
);
719 extern int sysfs_freeze_array(struct mdinfo
*sra
);
720 extern int sysfs_wait(int fd
, int *msec
);
721 extern int load_sys(char *path
, char *buf
, int len
);
722 extern int zero_disk_range(int fd
, unsigned long long sector
, size_t count
);
723 extern int reshape_prepare_fdlist(char *devname
,
727 unsigned long blocks
,
730 unsigned long long *offsets
);
731 extern void reshape_free_fdlist(int *fdlist
,
732 unsigned long long *offsets
,
734 extern int reshape_open_backup_file(char *backup
,
739 unsigned long long *offsets
,
742 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
743 unsigned int ndata
, unsigned int odata
);
744 extern char *locate_backup(char *name
);
745 extern char *make_backup(char *name
);
747 extern int save_stripes(int *source
, unsigned long long *offsets
,
748 int raid_disks
, int chunk_size
, int level
, int layout
,
749 int nwrites
, int *dest
,
750 unsigned long long start
, unsigned long long length
,
752 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
753 int raid_disks
, int chunk_size
, int level
, int layout
,
754 int source
, unsigned long long read_offset
,
755 unsigned long long start
, unsigned long long length
,
759 #define Sendmail "/usr/lib/sendmail -t"
762 #define SYSLOG_FACILITY LOG_DAEMON
764 extern char *map_num(mapping_t
*map
, int num
);
765 extern int map_name(mapping_t
*map
, char *name
);
766 extern mapping_t r0layout
[], r5layout
[], r6layout
[],
767 pers
[], modes
[], faultylayout
[];
768 extern mapping_t consistency_policies
[], sysfs_array_states
[];
770 extern char *map_dev_preferred(int major
, int minor
, int create
,
772 static inline char *map_dev(int major
, int minor
, int create
)
774 return map_dev_preferred(major
, minor
, create
, NULL
);
778 struct metadata_update
;
780 /* 'struct reshape' records the intermediate states of
782 * The starting geometry is converted to the 'before' geometry
783 * by at most an atomic level change. They could be the same.
784 * Similarly the 'after' geometry is converted to the final
785 * geometry by at most a level change.
786 * Note that 'before' and 'after' must have the same level.
787 * 'blocks' is the minimum number of sectors for a reshape unit.
788 * This will be a multiple of the stripe size in each of the
789 * 'before' and 'after' geometries.
790 * If 'blocks' is 0, no restriping is necessary.
791 * 'min_offset_change' is the minimum change to data_offset to
792 * allow the reshape to happen. It is at least the larger of
793 * the old and new chunk sizes, and typically the same as 'blocks'
794 * divided by number of data disks.
798 int parity
; /* number of parity blocks/devices */
803 unsigned long long backup_blocks
;
804 unsigned long long min_offset_change
;
805 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
806 unsigned long long new_size
; /* New size of array in sectors */
809 /* A superswitch provides entry point to a metadata handler.
811 * The superswitch primarily operates on some "metadata" that
812 * is accessed via the 'supertype'.
813 * This metadata has one of three possible sources.
814 * 1/ It is read from a single device. In this case it may not completely
815 * describe the array or arrays as some information might be on other
817 * 2/ It is read from all devices in a container. In this case all
818 * information is present.
819 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
820 * be complete once enough ->add_to_super calls have completed.
822 * When creating an array inside a container, the metadata will be
823 * formed by a combination of 2 and 3. The metadata or the array is read,
824 * then new information is added.
826 * The metadata must sometimes have a concept of a 'current' array
827 * and a 'current' device.
828 * The 'current' array is set by init_super to be the newly created array,
829 * or is set by super_by_fd when it finds it is looking at an array inside
832 * The 'current' device is either the device that the metadata was read from
833 * in case 1, or the last device added by add_to_super in case 3.
834 * Case 2 does not identify a 'current' device.
836 extern struct superswitch
{
838 /* Used to report details of metadata read from a component
839 * device. ->load_super has been called.
841 void (*examine_super
)(struct supertype
*st
, char *homehost
);
842 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
843 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
844 void (*export_examine_super
)(struct supertype
*st
);
845 int (*examine_badblocks
)(struct supertype
*st
, int fd
, char *devname
);
846 int (*copy_metadata
)(struct supertype
*st
, int from
, int to
);
848 /* Used to report details of an active array.
849 * ->load_super was possibly given a 'component' string.
851 void (*detail_super
)(struct supertype
*st
, char *homehost
,
853 void (*brief_detail_super
)(struct supertype
*st
, char *subarray
);
854 void (*export_detail_super
)(struct supertype
*st
);
856 /* Optional: platform hardware / firmware details */
857 int (*detail_platform
)(int verbose
, int enumerate_only
, char *controller_path
);
858 int (*export_detail_platform
)(int verbose
, char *controller_path
);
861 * to get uuid to storing in bitmap metadata
862 * and 'reshape' backup-data metadata
863 * To see if a device is being re-added to an array it was part of.
865 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
867 /* Extract generic details from metadata. This could be details about
868 * the container, or about an individual array within the container.
869 * The determination is made either by:
870 * load_super being given a 'component' string.
871 * validate_geometry determining what to create.
872 * The info includes both array information and device information.
873 * The particular device should be:
874 * The last device added by add_to_super
875 * The device the metadata was loaded from by load_super
876 * If 'map' is present, then it is an array raid_disks long
877 * (raid_disk must already be set and correct) and it is filled
878 * with 1 for slots that are thought to be active and 0 for slots which
879 * appear to be failed/missing.
880 * *info is zeroed out before data is added.
882 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
883 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
884 /* Check if the given metadata is flagged as belonging to "this"
885 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
887 int (*match_home
)(struct supertype
*st
, char *homehost
);
889 /* Make one of several generic modifications to metadata
890 * prior to assembly (or other times).
891 * sparc2.2 - first bug in early 0.90 metadata
892 * super-minor - change name of 0.90 metadata
893 * summaries - 'correct' any redundant data
894 * resync - mark array as dirty to trigger a resync.
895 * uuid - set new uuid - only 0.90 or 1.x
896 * name - change the name of the array (where supported)
897 * homehost - change which host this array is tied to.
898 * devicesize - If metadata is at start of device, change recorded
899 * device size to match actual device size
900 * byteorder - swap bytes for 0.90 metadata
902 * force-one - mark that device as uptodate, not old or failed.
903 * force-array - mark array as clean if it would not otherwise
905 * assemble - not sure how this is different from force-one...
906 * linear-grow-new - add a new device to a linear array, but don't
907 * change the size: so superblock still matches
908 * linear-grow-update - now change the size of the array.
909 * writemostly - set the WriteMostly1 bit in the superblock devflags
910 * readwrite - clear the WriteMostly1 bit in the superblock devflags
911 * failfast - set the FailFast1 bit in the superblock
912 * nofailfast - clear the FailFast1 bit
913 * no-bitmap - clear any record that a bitmap is present.
914 * bbl - add a bad-block-log if possible
915 * no-bbl - remove any bad-block-log is it is empty.
916 * force-no-bbl - remove any bad-block-log even if empty.
917 * revert-reshape - If a reshape is in progress, modify metadata so
918 * it will resume going in the opposite direction.
920 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
922 char *devname
, int verbose
,
923 int uuid_set
, char *homehost
);
925 /* Create new metadata for new array as described. This could
926 * be a new container, or an array in a pre-existing container.
927 * Also used to zero metadata prior to writing it to invalidate old
930 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
931 struct shape
*s
, char *name
,
932 char *homehost
, int *uuid
,
933 unsigned long long data_offset
);
935 /* update the metadata to include new device, either at create or
936 * when hot-adding a spare.
938 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
939 int fd
, char *devname
,
940 unsigned long long data_offset
);
941 /* update the metadata to delete a device,
944 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
946 /* Write metadata to one device when fixing problems or adding
949 int (*store_super
)(struct supertype
*st
, int fd
);
951 /* Write all metadata for this array.
953 int (*write_init_super
)(struct supertype
*st
);
954 /* Check if metadata read from one device is compatible with an array,
955 * used when assembling an array, or pseudo-assembling was with
956 * "--examine --brief"
957 * If "st" has not yet been loaded the superblock from, "tst" is
958 * moved in, otherwise the superblock in 'st' is compared with
961 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
);
962 /* Load metadata from a single device. If 'devname' is not NULL
963 * print error messages as appropriate */
964 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
965 /* 'fd' is a 'container' md array - load array metadata from the
968 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
969 /* If 'arg' is a valid name of this metadata type, allocate and
970 * return a 'supertype' for the particular minor version */
971 struct supertype
* (*match_metadata_desc
)(char *arg
);
972 /* If a device has the given size, and the data_offset has been
973 * requested - work out how much space is available for data.
974 * This involves adjusting for reserved space (e.g. bitmaps)
975 * and for any rounding.
976 * 'mdadm' only calls this for existing arrays where a possible
977 * spare is being added. However some super-handlers call it
978 * internally from validate_geometry when creating an array.
980 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
,
981 unsigned long long data_offset
);
983 * Return spare criteria for array:
984 * - minimum disk size can be used in array;
985 * - sector size can be used in array.
986 * Return values: 0 - for success and -EINVAL on error.
988 int (*get_spare_criteria
)(struct supertype
*st
,
989 struct spare_criteria
*sc
);
990 /* Find somewhere to put a bitmap - possibly auto-size it - and
991 * update the metadata to record this. The array may be newly
992 * created, in which case data_size may be updated, or it might
993 * already exist. Metadata handler can know if init_super
994 * has been called, but not write_init_super.
998 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
999 int delay
, int write_behind
,
1000 unsigned long long size
, int may_change
, int major
);
1001 /* Seek 'fd' to start of write-intent-bitmap. Must be an
1002 * md-native format bitmap
1004 int (*locate_bitmap
)(struct supertype
*st
, int fd
, int node_num
);
1005 /* if add_internal_bitmap succeeded for existing array, this
1008 int (*write_bitmap
)(struct supertype
*st
, int fd
, enum bitmap_update update
);
1009 /* Free the superblock and any other allocated data */
1010 void (*free_super
)(struct supertype
*st
);
1012 /* validate_geometry is called with an st returned by
1013 * match_metadata_desc.
1014 * It should check that the geometry described is compatible with
1015 * the metadata type. It will be called repeatedly as devices
1016 * added to validate changing size and new devices. If there are
1017 * inter-device dependencies, it should record sufficient details
1018 * so these can be validated.
1019 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
1021 * 1: everything is OK
1022 * 0: not OK for some reason - if 'verbose', then error was reported.
1023 * -1: st->sb was NULL, 'subdev' is a member of a container of this
1024 * type, but array is not acceptable for some reason
1025 * message was reported even if verbose is 0.
1027 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
1029 int *chunk
, unsigned long long size
,
1030 unsigned long long data_offset
,
1031 char *subdev
, unsigned long long *freesize
,
1032 int consistency_policy
, int verbose
);
1034 /* Return a linked list of 'mdinfo' structures for all arrays
1035 * in the container. For non-containers, it is like
1036 * getinfo_super with an allocated mdinfo.*/
1037 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
1038 /* query the supertype for default geometry */
1039 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
1040 /* Permit subarray's to be deleted from inactive containers */
1041 int (*kill_subarray
)(struct supertype
*st
,
1042 char *subarray_id
); /* optional */
1043 /* Permit subarray's to be modified */
1044 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
1045 char *update
, struct mddev_ident
*ident
); /* optional */
1046 /* Check if reshape is supported for this external format.
1047 * st is obtained from super_by_fd() where st->subarray[0] is
1048 * initialized to indicate if reshape is being performed at the
1049 * container or subarray level
1051 #define APPLY_METADATA_CHANGES 1
1052 #define ROLLBACK_METADATA_CHANGES 0
1054 int (*reshape_super
)(struct supertype
*st
,
1055 unsigned long long size
, int level
,
1056 int layout
, int chunksize
, int raid_disks
,
1057 int delta_disks
, char *backup
, char *dev
,
1059 int verbose
); /* optional */
1060 int (*manage_reshape
)( /* optional */
1061 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1062 struct supertype
*st
, unsigned long blocks
,
1063 int *fds
, unsigned long long *offsets
,
1064 int dests
, int *destfd
, unsigned long long *destoffsets
);
1067 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
1070 /* Tell the metadata handler the current state of the array.
1071 * This covers whether it is known to be consistent (no pending writes)
1072 * and how far along a resync is known to have progressed
1073 * (in a->resync_start).
1074 * resync status is really irrelevant if the array is not consistent,
1075 * but some metadata (DDF!) have a place to record the distinction.
1076 * If 'consistent' is '2', then the array can mark it dirty if a
1077 * resync/recovery/whatever is required, or leave it clean if not.
1078 * Return value is 0 dirty (not consistent) and 1 if clean.
1079 * it is only really important if consistent is passed in as '2'.
1081 int (*set_array_state
)(struct active_array
*a
, int consistent
);
1083 /* When the state of a device might have changed, we call set_disk to
1084 * tell the metadata what the current state is.
1085 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1087 * set_disk might be called when the state of the particular disk has
1088 * not in fact changed.
1090 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
1091 void (*sync_metadata
)(struct supertype
*st
);
1092 void (*process_update
)(struct supertype
*st
,
1093 struct metadata_update
*update
);
1094 /* Prepare updates allocates extra memory that might be
1095 * needed. If the update cannot be understood, return 0.
1097 int (*prepare_update
)(struct supertype
*st
,
1098 struct metadata_update
*update
);
1100 /* activate_spare will check if the array is degraded and, if it
1101 * is, try to find some spare space in the container.
1102 * On success, it add appropriate updates (For process_update) to
1103 * to the 'updates' list and returns a list of 'mdinfo' identifying
1104 * the device, or devices as there might be multiple missing
1105 * devices and multiple spares available.
1107 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
1108 struct metadata_update
**updates
);
1110 * Return statically allocated string that represents metadata specific
1111 * controller domain of the disk. The domain is used in disk domain
1112 * matching functions. Disks belong to the same domain if the they have
1113 * the same domain from mdadm.conf and belong the same metadata domain.
1114 * Returning NULL or not providing this handler means that metadata
1115 * does not distinguish the differences between disks that belong to
1116 * different controllers. They are in the domain specified by
1117 * configuration file (mdadm.conf).
1118 * In case when the metadata has the notion of domains based on disk
1119 * it shall return NULL for disks that do not belong to the controller
1120 * the supported domains. Such disks will form another domain and won't
1121 * be mixed with supported ones.
1123 const char *(*get_disk_controller_domain
)(const char *path
);
1125 /* for external backup area */
1126 int (*recover_backup
)(struct supertype
*st
, struct mdinfo
*info
);
1128 /* validate container after assemble */
1129 int (*validate_container
)(struct mdinfo
*info
);
1131 /* write initial empty PPL on device */
1132 int (*write_init_ppl
)(struct supertype
*st
, struct mdinfo
*info
, int fd
);
1134 /* validate ppl before assemble */
1135 int (*validate_ppl
)(struct supertype
*st
, struct mdinfo
*info
,
1136 struct mdinfo
*disk
);
1138 /* records new bad block in metadata */
1139 int (*record_bad_block
)(struct active_array
*a
, int n
,
1140 unsigned long long sector
, int length
);
1142 /* clears bad block from metadata */
1143 int (*clear_bad_block
)(struct active_array
*a
, int n
,
1144 unsigned long long sector
, int length
);
1146 /* get list of bad blocks from metadata */
1147 struct md_bb
*(*get_bad_blocks
)(struct active_array
*a
, int n
);
1149 int swapuuid
; /* true if uuid is bigending rather than hostendian */
1151 const char *name
; /* canonical metadata name */
1154 extern struct superswitch super0
, super1
;
1155 extern struct superswitch super_imsm
, super_ddf
;
1156 extern struct superswitch mbr
, gpt
;
1158 struct metadata_update
{
1161 void *space
; /* allocated space that monitor will use */
1162 void **space_list
; /* list of allocated spaces that monitor can
1163 * use or that it returned.
1165 struct metadata_update
*next
;
1168 /* A supertype holds a particular collection of metadata.
1169 * It identifies the metadata type by the superswitch, and the particular
1170 * sub-version of that metadata type.
1171 * metadata read in or created is stored in 'sb' and 'info'.
1172 * There are also fields used by mdmon to track containers.
1174 * A supertype may refer to:
1175 * Just an array, possibly in a container
1176 * A container, not identifying any particular array
1177 * Info read from just one device, not yet fully describing the array/container.
1180 * A supertype is created by:
1186 struct superswitch
*ss
;
1189 char container_devnm
[32]; /* devnm of container */
1192 void *other
; /* Hack used to convert v0.90 to v1.0 */
1193 unsigned long long devsize
;
1194 unsigned long long data_offset
; /* used by v1.x only */
1195 int ignore_hw_compat
; /* used to inform metadata handlers that it should ignore
1196 HW/firmware related incompatability to load metadata.
1197 Used when examining metadata to display content of disk
1198 when user has no hw/firmare compatible system.
1200 struct metadata_update
*updates
;
1201 struct metadata_update
**update_tail
;
1203 /* extra stuff used by mdmon */
1204 struct active_array
*arrays
;
1205 int sock
; /* listen to external programs */
1206 char devnm
[32]; /* e.g. md0. This appears in metadata_version:
1214 struct mdinfo
*devs
;
1218 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
1219 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
1220 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
1221 static inline struct supertype
*guess_super(int fd
) {
1222 return guess_super_type(fd
, guess_any
);
1224 extern struct supertype
*dup_super(struct supertype
*st
);
1225 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
1226 extern int get_dev_sector_size(int fd
, char *dname
, unsigned int *sectsizep
);
1227 extern int must_be_container(int fd
);
1228 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
1229 extern int dev_sector_size_from_id(dev_t id
, unsigned int *size
);
1230 void wait_for(char *dev
, int fd
);
1233 * Data structures for policy management.
1234 * Each device can have a policy structure that lists
1235 * various name/value pairs each possibly with a metadata associated.
1236 * The policy list is sorted by name/value/metadata
1239 struct dev_policy
*next
;
1240 char *name
; /* None of these strings are allocated. They are
1241 * all just references to strings which are known
1242 * to exist elsewhere.
1243 * name and metadata can be compared by address equality.
1245 const char *metadata
;
1249 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
1251 /* iterate over the sublist starting at list, having the same
1252 * 'name' as 'list', and matching the given metadata (Where
1253 * NULL matches anything
1255 #define pol_for_each(item, list, _metadata) \
1257 item && item->name == list->name; \
1258 item = item->next) \
1259 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1263 * policy records read from mdadm are largely just name-value pairs.
1264 * The names are constants, not strdupped
1267 struct pol_rule
*next
;
1268 char *type
; /* rule_policy or rule_part */
1273 char *dups
; /* duplicates of 'value' with a partNN appended */
1277 extern char rule_policy
[], rule_part
[];
1278 extern char rule_path
[], rule_type
[];
1279 extern char type_part
[], type_disk
[];
1281 extern void policyline(char *line
, char *type
);
1282 extern void policy_add(char *type
, ...);
1283 extern void policy_free(void);
1285 extern struct dev_policy
*path_policy(char **paths
, char *type
);
1286 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
1287 extern struct dev_policy
*devid_policy(int devid
);
1288 extern void dev_policy_free(struct dev_policy
*p
);
1290 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1291 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
1292 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
1294 enum policy_action
{
1298 act_spare
, /* This only applies to bare devices */
1299 act_spare_same_slot
, /* this allows non-bare devices,
1300 * but only if recent removal */
1301 act_force_spare
, /* this allow non-bare devices in any case */
1305 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
1306 enum policy_action want
);
1307 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
1308 enum policy_action want
);
1311 struct domainlist
*next
;
1315 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
1316 const char *metadata
);
1317 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
1318 const char *metadata
);
1319 extern void domainlist_add_dev(struct domainlist
**dom
, int devid
,
1320 const char *metadata
);
1321 extern void domain_free(struct domainlist
*dl
);
1322 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
1323 const char *metadata
);
1324 void domain_add(struct domainlist
**domp
, char *domain
);
1326 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
1327 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
1329 extern void sysfs_rules_apply(char *devnm
, struct mdinfo
*dev
);
1330 extern void sysfsline(char *line
);
1336 #define HAVE_NFTW we assume
1340 # include <features.h>
1341 # ifndef __UCLIBC_HAS_LFS__
1342 # define lseek64 lseek
1344 # ifndef __UCLIBC_HAS_FTW__
1354 #if defined(__KLIBC__)
1370 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
1372 extern int Manage_ro(char *devname
, int fd
, int readonly
);
1373 extern int Manage_run(char *devname
, int fd
, struct context
*c
);
1374 extern int Manage_stop(char *devname
, int fd
, int quiet
,
1376 extern int Manage_subdevs(char *devname
, int fd
,
1377 struct mddev_dev
*devlist
, int verbose
, int test
,
1378 char *update
, int force
);
1379 extern int autodetect(void);
1380 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1381 extern int Grow_addbitmap(char *devname
, int fd
,
1382 struct context
*c
, struct shape
*s
);
1383 extern int Grow_reshape(char *devname
, int fd
,
1384 struct mddev_dev
*devlist
,
1385 unsigned long long data_offset
,
1386 struct context
*c
, struct shape
*s
);
1387 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1388 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1389 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1390 struct mdinfo
*info
, char *backup_file
,
1391 int forked
, int freeze_reshape
);
1392 extern int Grow_consistency_policy(char *devname
, int fd
,
1393 struct context
*c
, struct shape
*s
);
1395 extern int restore_backup(struct supertype
*st
,
1396 struct mdinfo
*content
,
1399 char **backup_filep
,
1401 extern int Grow_continue_command(char *devname
, int fd
,
1402 char *backup_file
, int verbose
);
1404 extern int Assemble(struct supertype
*st
, char *mddev
,
1405 struct mddev_ident
*ident
,
1406 struct mddev_dev
*devlist
,
1409 extern int Build(char *mddev
, struct mddev_dev
*devlist
,
1410 struct shape
*s
, struct context
*c
);
1412 extern int Create(struct supertype
*st
, char *mddev
,
1413 char *name
, int *uuid
,
1414 int subdevs
, struct mddev_dev
*devlist
,
1417 unsigned long long data_offset
);
1419 extern int Detail(char *dev
, struct context
*c
);
1420 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
, int export
, char *controller_path
);
1421 extern int Query(char *dev
);
1422 extern int ExamineBadblocks(char *devname
, int brief
, struct supertype
*forcest
);
1423 extern int Examine(struct mddev_dev
*devlist
, struct context
*c
,
1424 struct supertype
*forcest
);
1425 extern int Monitor(struct mddev_dev
*devlist
,
1426 char *mailaddr
, char *alert_cmd
,
1428 int daemonise
, int oneshot
,
1429 int dosyslog
, char *pidfile
, int increments
,
1432 extern int Kill(char *dev
, struct supertype
*st
, int force
, int verbose
, int noexcl
);
1433 extern int Kill_subarray(char *dev
, char *subarray
, int verbose
);
1434 extern int Update_subarray(char *dev
, char *subarray
, char *update
, struct mddev_ident
*ident
, int quiet
);
1435 extern int Wait(char *dev
);
1436 extern int WaitClean(char *dev
, int verbose
);
1437 extern int SetAction(char *dev
, char *action
);
1439 extern int Incremental(struct mddev_dev
*devlist
, struct context
*c
,
1440 struct supertype
*st
);
1441 extern void RebuildMap(void);
1442 extern int IncrementalScan(struct context
*c
, char *devnm
);
1443 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1444 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1445 unsigned long chunksize
, unsigned long daemon_sleep
,
1446 unsigned long write_behind
,
1447 unsigned long long array_size
,
1449 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1450 extern int Write_rules(char *rule_name
);
1451 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1453 /* calculate the size of the bitmap given the array size and bitmap chunksize */
1454 static inline unsigned long long
1455 bitmap_bits(unsigned long long array_size
, unsigned long chunksize
)
1457 return (array_size
* 512 + chunksize
- 1) / chunksize
;
1460 extern int Dump_metadata(char *dev
, char *dir
, struct context
*c
,
1461 struct supertype
*st
);
1462 extern int Restore_metadata(char *dev
, char *dir
, struct context
*c
,
1463 struct supertype
*st
, int only
);
1465 int md_array_valid(int fd
);
1466 int md_array_active(int fd
);
1467 int md_array_is_active(struct mdinfo
*info
);
1468 int md_get_array_info(int fd
, struct mdu_array_info_s
*array
);
1469 int md_set_array_info(int fd
, struct mdu_array_info_s
*array
);
1470 int md_get_disk_info(int fd
, struct mdu_disk_info_s
*disk
);
1471 extern int get_linux_version(void);
1472 extern int mdadm_version(char *version
);
1473 extern unsigned long long parse_size(char *size
);
1474 extern int parse_uuid(char *str
, int uuid
[4]);
1475 extern int is_near_layout_10(int layout
);
1476 extern int parse_layout_10(char *layout
);
1477 extern int parse_layout_faulty(char *layout
);
1478 extern long parse_num(char *num
);
1479 extern int parse_cluster_confirm_arg(char *inp
, char **devname
, int *slot
);
1480 extern int check_ext2(int fd
, char *name
);
1481 extern int check_reiser(int fd
, char *name
);
1482 extern int check_raid(int fd
, char *name
);
1483 extern int check_partitions(int fd
, char *dname
,
1484 unsigned long long freesize
,
1485 unsigned long long size
);
1486 extern int fstat_is_blkdev(int fd
, char *devname
, dev_t
*rdev
);
1487 extern int stat_is_blkdev(char *devname
, dev_t
*rdev
);
1489 extern int get_mdp_major(void);
1490 extern int get_maj_min(char *dev
, int *major
, int *minor
);
1491 extern int dev_open(char *dev
, int flags
);
1492 extern int open_dev(char *devnm
);
1493 extern void reopen_mddev(int mdfd
);
1494 extern int open_dev_flags(char *devnm
, int flags
);
1495 extern int open_dev_excl(char *devnm
);
1496 extern int is_standard(char *dev
, int *nump
);
1497 extern int same_dev(char *one
, char *two
);
1498 extern int compare_paths (char* path1
,char* path2
);
1499 extern void enable_fds(int devices
);
1501 extern int parse_auto(char *str
, char *msg
, int config
);
1502 extern struct mddev_ident
*conf_get_ident(char *dev
);
1503 extern struct mddev_dev
*conf_get_devs(void);
1504 extern int conf_test_dev(char *devname
);
1505 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1506 extern struct createinfo
*conf_get_create_info(void);
1507 extern void set_conffile(char *file
);
1508 extern char *conf_get_mailaddr(void);
1509 extern char *conf_get_mailfrom(void);
1510 extern char *conf_get_program(void);
1511 extern char *conf_get_homehost(int *require_homehostp
);
1512 extern char *conf_get_homecluster(void);
1513 extern char *conf_line(FILE *file
);
1514 extern char *conf_word(FILE *file
, int allow_key
);
1515 extern void print_quoted(char *str
);
1516 extern void print_escape(char *str
);
1517 extern int use_udev(void);
1518 extern unsigned long GCD(unsigned long a
, unsigned long b
);
1519 extern int conf_name_is_free(char *name
);
1520 extern int conf_verify_devnames(struct mddev_ident
*array_list
);
1521 extern int devname_matches(char *name
, char *match
);
1522 extern struct mddev_ident
*conf_match(struct supertype
*st
,
1523 struct mdinfo
*info
,
1525 int verbose
, int *rvp
);
1527 extern void free_line(char *line
);
1528 extern int match_oneof(char *devices
, char *devname
);
1529 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1530 extern const int uuid_zero
[4];
1531 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1532 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1533 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1534 extern char *fname_from_uuid(struct supertype
*st
,
1535 struct mdinfo
*info
, char *buf
, char sep
);
1536 extern unsigned long calc_csum(void *super
, int bytes
);
1537 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1539 extern int ask(char *mesg
);
1540 extern unsigned long long get_component_size(int fd
);
1541 extern void remove_partitions(int fd
);
1542 extern int test_partition(int fd
);
1543 extern int test_partition_from_id(dev_t id
);
1544 extern int get_data_disks(int level
, int layout
, int raid_disks
);
1545 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1546 int chunksize
, unsigned long long devsize
);
1547 extern int flush_metadata_updates(struct supertype
*st
);
1548 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1549 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1550 struct mdinfo
*content
,
1552 char *chosen_name
, int *result
);
1554 #define INCR_UNSAFE 2
1555 #define INCR_ALREADY 4
1557 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1558 struct spare_criteria
*criteria
,
1559 struct domainlist
*domlist
,
1561 const char *metadata
, int get_one
);
1562 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1563 extern int add_disk(int mdfd
, struct supertype
*st
,
1564 struct mdinfo
*sra
, struct mdinfo
*info
);
1565 extern int remove_disk(int mdfd
, struct supertype
*st
,
1566 struct mdinfo
*sra
, struct mdinfo
*info
);
1567 extern int hot_remove_disk(int mdfd
, unsigned long dev
, int force
);
1568 extern int sys_hot_remove_disk(int statefd
, int force
);
1569 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1570 unsigned long long min_recovery_start(struct mdinfo
*array
);
1572 extern char *human_size(long long bytes
);
1573 extern char *human_size_brief(long long bytes
, int prefix
);
1574 extern void print_r10_layout(int layout
);
1576 extern char *find_free_devnm(int use_partitions
);
1578 extern void put_md_name(char *name
);
1579 extern char *devid2kname(dev_t devid
);
1580 extern char *devid2devnm(dev_t devid
);
1581 extern dev_t
devnm2devid(char *devnm
);
1582 extern char *get_md_name(char *devnm
);
1584 extern char DefaultConfFile
[];
1586 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1587 char *chosen
, int block_udev
);
1588 /* values for 'trustworthy' */
1590 #define LOCAL_ANY 10
1593 extern int open_mddev(char *dev
, int report_errors
);
1594 extern int open_container(int fd
);
1595 extern int metadata_container_matches(char *metadata
, char *devnm
);
1596 extern int metadata_subdev_matches(char *metadata
, char *devnm
);
1597 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1598 extern int is_subarray_active(char *subarray
, char *devname
);
1599 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1600 extern struct superswitch
*version_to_superswitch(char *vers
);
1602 extern int mdmon_running(char *devnm
);
1603 extern int mdmon_pid(char *devnm
);
1604 extern int check_env(char *name
);
1605 extern __u32
random32(void);
1606 extern void random_uuid(__u8
*buf
);
1607 extern int start_mdmon(char *devnm
);
1609 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1610 struct supertype
*st
, unsigned long stripes
,
1611 int *fds
, unsigned long long *offsets
,
1612 int dests
, int *destfd
, unsigned long long *destoffsets
);
1613 void abort_reshape(struct mdinfo
*sra
);
1615 void *super1_make_v0(struct supertype
*st
, struct mdinfo
*info
, mdp_super_t
*sb0
);
1617 extern char *stat2kname(struct stat
*st
);
1618 extern char *fd2kname(int fd
);
1619 extern char *stat2devnm(struct stat
*st
);
1620 extern char *fd2devnm(int fd
);
1621 extern void udev_block(char *devnm
);
1622 extern void udev_unblock(void);
1624 extern int in_initrd(void);
1627 void *cmap_handle
; /* corosync lib related */
1629 int (*initialize
)(cmap_handle_t
*handle
);
1630 int (*get_string
)(cmap_handle_t handle
,
1633 int (*finalize
)(cmap_handle_t handle
);
1636 extern void set_cmap_hooks(void);
1637 extern void set_hooks(void);
1640 void *dlm_handle
; /* dlm lib related */
1642 dlm_lshandle_t (*create_lockspace
)(const char *name
,
1644 dlm_lshandle_t (*open_lockspace
)(const char *name
);
1645 int (*release_lockspace
)(const char *name
, dlm_lshandle_t ls
,
1647 int (*ls_lock
)(dlm_lshandle_t lockspace
, uint32_t mode
,
1648 struct dlm_lksb
*lksb
, uint32_t flags
,
1649 const void *name
, unsigned int namelen
,
1650 uint32_t parent
, void (*astaddr
) (void *astarg
),
1651 void *astarg
, void (*bastaddr
) (void *astarg
),
1653 int (*ls_unlock_wait
)(dlm_lshandle_t lockspace
, uint32_t lkid
,
1654 uint32_t flags
, struct dlm_lksb
*lksb
);
1655 int (*ls_get_fd
)(dlm_lshandle_t ls
);
1656 int (*dispatch
)(int fd
);
1659 extern int get_cluster_name(char **name
);
1660 extern int dlm_funs_ready(void);
1661 extern int cluster_get_dlmlock(void);
1662 extern int cluster_release_dlmlock(void);
1663 extern void set_dlm_hooks(void);
1665 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1666 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1667 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1668 (ROUND_UP((unsigned long)(ptr), base)))
1670 static inline int is_subarray(char *vers
)
1672 /* The version string for a 'subarray' (an array in a container)
1674 * /containername/componentname for normal read-write arrays
1675 * -containername/componentname for arrays which mdmon must not
1676 * reconfigure. They might be read-only
1677 * or might be undergoing reshape etc.
1678 * containername is e.g. md0, md_d1
1679 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1681 return (*vers
== '/' || *vers
== '-');
1684 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1686 return &ent
->metadata_version
[10+strlen(container
)+1];
1690 #define dprintf(fmt, arg...) \
1691 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1692 #define dprintf_cont(fmt, arg...) \
1693 fprintf(stderr, fmt, ##arg)
1695 #define dprintf(fmt, arg...) \
1696 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1697 #define dprintf_cont(fmt, arg...) \
1698 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1702 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1706 ret
= vasprintf(strp
, fmt
, ap
);
1713 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1715 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1717 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1719 void *xmalloc(size_t len
);
1720 void *xrealloc(void *ptr
, size_t len
);
1721 void *xcalloc(size_t num
, size_t size
);
1722 char *xstrdup(const char *str
);
1724 #define LEVEL_MULTIPATH (-4)
1725 #define LEVEL_LINEAR (-1)
1726 #define LEVEL_FAULTY (-5)
1728 /* kernel module doesn't know about these */
1729 #define LEVEL_CONTAINER (-100)
1730 #define LEVEL_UNSUPPORTED (-200)
1732 /* the kernel does know about this one ... */
1733 #define LEVEL_NONE (-1000000)
1737 #define WriteTransient 0
1738 #define ReadTransient 1
1739 #define WritePersistent 2
1740 #define ReadPersistent 3
1741 #define WriteAll 4 /* doesn't go to device */
1742 #define ReadFixable 5
1745 #define ClearErrors 31
1746 #define ClearFaults 30
1748 #define AllPersist 100 /* internal use only */
1749 #define NoPersist 101
1751 #define ModeMask 0x1f
1758 #define minor(x) ((x)&0xff)
1759 #define major(x) (((x)>>8)&0xff)
1760 #define makedev(M,m) (((M)<<8) | (m))
1764 RAID0_ORIG_LAYOUT
= 1,
1765 RAID0_ALT_MULTIZONE_LAYOUT
= 2,
1769 #define ALGORITHM_LEFT_ASYMMETRIC 0
1770 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1771 #define ALGORITHM_LEFT_SYMMETRIC 2
1772 #define ALGORITHM_RIGHT_SYMMETRIC 3
1774 /* Define non-rotating (raid4) algorithms. These allow
1775 * conversion of raid4 to raid5.
1777 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1778 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1780 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1781 * Firstly, the exact positioning of the parity block is slightly
1782 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1784 * Secondly, or order of datablocks over which the Q syndrome is computed
1786 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1787 * These layouts are from the DDFv1.2 spec.
1788 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1789 * leaves RLQ=3 as 'Vendor Specific'
1792 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1793 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1794 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1796 /* For every RAID5 algorithm we define a RAID6 algorithm
1797 * with exactly the same layout for data and parity, and
1798 * with the Q block always on the last device (N-1).
1799 * This allows trivial conversion from RAID5 to RAID6
1801 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1802 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1803 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1804 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1805 #define ALGORITHM_PARITY_0_6 20
1806 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1808 /* Define PATH_MAX in case we don't use glibc or standard library does
1809 * not have PATH_MAX defined. Assume max path length is 4K characters.
1812 #define PATH_MAX 4096
1815 #define RESYNC_NONE -1
1816 #define RESYNC_DELAYED -2
1817 #define RESYNC_PENDING -3
1818 #define RESYNC_UNKNOWN -4
1820 /* When using "GET_DISK_INFO" it isn't certain how high
1821 * we need to check. So we impose an absolute limit of
1822 * MAX_DISKS. This needs to be much more than the largest
1823 * number of devices any metadata can support. Currently
1824 * v1.x can support 1920
1826 #define MAX_DISKS 4096
1828 /* Sometimes the 'size' value passed needs to mean "Maximum".
1829 * In those cases with use MAX_SIZE
1833 /* We want to use unsigned numbers for sector counts, but need
1834 * a value for 'invalid'. Use '1'.
1836 #define INVALID_SECTORS 1
1837 /* And another special number needed for --data_offset=variable */
1838 #define VARIABLE_OFFSET 3