2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 * Email: <neilb@suse.de>
26 #define _FILE_OFFSET_BITS 64
29 extern __off64_t lseek64
__P ((int __fd
, __off64_t __offset
, int __whence
));
30 #elif !defined(lseek64)
31 # if defined(__NO_STAT64) || __WORDSIZE != 32
32 # define lseek64 lseek
36 #include <sys/types.h>
49 /* Newer glibc requires sys/sysmacros.h directly for makedev() */
50 #include <sys/sysmacros.h>
54 /* dietlibc has deprecated random and srandom!! */
61 typedef uint64_t cmap_handle_t
;
63 #include <corosync/cmap.h>
70 #define LKF_NOQUEUE 0x00000001
72 #define EUNLOCK 0x10002
74 typedef void *dlm_lshandle_t
;
84 #include <linux/kdev_t.h>
85 /*#include <linux/fs.h> */
86 #include <sys/mount.h>
87 #include <asm/types.h>
88 #include <sys/ioctl.h>
90 #define MdpMinorShift 6
93 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
96 #define DEFAULT_CHUNK 512
97 #define DEFAULT_BITMAP_CHUNK 4096
98 #define DEFAULT_BITMAP_DELAY 5
99 #define DEFAULT_MAX_WRITE_BEHIND 256
101 /* MAP_DIR should be somewhere that persists across the pivotroot
102 * from early boot to late boot.
103 * /run seems to have emerged as the best standard.
106 #define MAP_DIR "/run/mdadm"
108 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
109 * want something other than the default of "map"
112 #define MAP_FILE "map"
113 #endif /* MAP_FILE */
114 /* MDMON_DIR is where pid and socket files used for communicating
115 * with mdmon normally live. Best is /var/run/mdadm as
116 * mdmon is needed at early boot then it needs to write there prior
117 * to /var/run being mounted read/write, and it also then needs to
118 * persist beyond when /var/run is mounter read-only. So, to be
119 * safe, the default is somewhere that is read/write early in the
120 * boot process and stays up as long as possible during shutdown.
123 #define MDMON_DIR "/run/mdadm"
124 #endif /* MDMON_DIR */
126 /* FAILED_SLOTS is where to save files storing recent removal of array
127 * member in order to allow future reuse of disk inserted in the same
128 * slot for array recovery
130 #ifndef FAILED_SLOTS_DIR
131 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
132 #endif /* FAILED_SLOTS */
140 /* Redhat don't like to #include <asm/byteorder.h>, and
141 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
142 * and there is no standard conversion function so... */
143 /* And dietlibc doesn't think byteswap is ok, so.. */
144 /* #include <byteswap.h> */
145 #define __mdadm_bswap_16(x) (((x) & 0x00ffU) << 8 | \
146 ((x) & 0xff00U) >> 8)
147 #define __mdadm_bswap_32(x) (((x) & 0x000000ffU) << 24 | \
148 ((x) & 0xff000000U) >> 24 | \
149 ((x) & 0x0000ff00U) << 8 | \
150 ((x) & 0x00ff0000U) >> 8)
151 #define __mdadm_bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
152 ((x) & 0xff00000000000000ULL) >> 56 | \
153 ((x) & 0x000000000000ff00ULL) << 40 | \
154 ((x) & 0x00ff000000000000ULL) >> 40 | \
155 ((x) & 0x0000000000ff0000ULL) << 24 | \
156 ((x) & 0x0000ff0000000000ULL) >> 24 | \
157 ((x) & 0x00000000ff000000ULL) << 8 | \
158 ((x) & 0x000000ff00000000ULL) >> 8)
160 #if !defined(__KLIBC__)
161 #if BYTE_ORDER == LITTLE_ENDIAN
162 #define __cpu_to_le16(_x) (unsigned int)(_x)
163 #define __cpu_to_le32(_x) (unsigned int)(_x)
164 #define __cpu_to_le64(_x) (unsigned long long)(_x)
165 #define __le16_to_cpu(_x) (unsigned int)(_x)
166 #define __le32_to_cpu(_x) (unsigned int)(_x)
167 #define __le64_to_cpu(_x) (unsigned long long)(_x)
169 #define __cpu_to_be16(_x) __mdadm_bswap_16(_x)
170 #define __cpu_to_be32(_x) __mdadm_bswap_32(_x)
171 #define __cpu_to_be64(_x) __mdadm_bswap_64(_x)
172 #define __be16_to_cpu(_x) __mdadm_bswap_16(_x)
173 #define __be32_to_cpu(_x) __mdadm_bswap_32(_x)
174 #define __be64_to_cpu(_x) __mdadm_bswap_64(_x)
175 #elif BYTE_ORDER == BIG_ENDIAN
176 #define __cpu_to_le16(_x) __mdadm_bswap_16(_x)
177 #define __cpu_to_le32(_x) __mdadm_bswap_32(_x)
178 #define __cpu_to_le64(_x) __mdadm_bswap_64(_x)
179 #define __le16_to_cpu(_x) __mdadm_bswap_16(_x)
180 #define __le32_to_cpu(_x) __mdadm_bswap_32(_x)
181 #define __le64_to_cpu(_x) __mdadm_bswap_64(_x)
183 #define __cpu_to_be16(_x) (unsigned int)(_x)
184 #define __cpu_to_be32(_x) (unsigned int)(_x)
185 #define __cpu_to_be64(_x) (unsigned long long)(_x)
186 #define __be16_to_cpu(_x) (unsigned int)(_x)
187 #define __be32_to_cpu(_x) (unsigned int)(_x)
188 #define __be64_to_cpu(_x) (unsigned long long)(_x)
190 # error "unknown endianness."
192 #endif /* __KLIBC__ */
195 * Check at compile time that something is of a particular type.
196 * Always evaluates to 1 so you may use it easily in comparisons.
199 #define typecheck(type,x) \
201 typeof(x) __dummy2; \
202 (void)(&__dummy == &__dummy2); \
207 * These inlines deal with timer wrapping correctly.
209 * time_after(a,b) returns true if the time a is after time b.
212 #define time_after(a,b) \
213 (typecheck(unsigned int, a) && \
214 typecheck(unsigned int, b) && \
215 ((int)((b) - (a)) < 0))
217 #define time_before(a,b) time_after(b,a)
220 * min()/max()/clamp() macros that also do
221 * strict type-checking.. See the
222 * "unnecessary" pointer comparison.
224 #define min(x, y) ({ \
225 typeof(x) _min1 = (x); \
226 typeof(y) _min2 = (y); \
227 (void) (&_min1 == &_min2); \
228 _min1 < _min2 ? _min1 : _min2; })
230 #define max(x, y) ({ \
231 typeof(x) _max1 = (x); \
232 typeof(y) _max2 = (y); \
233 (void) (&_max1 == &_max2); \
234 _max1 > _max2 ? _max1 : _max2; })
236 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
238 extern const char Name
[];
241 unsigned long long sector
;
248 struct md_bb_entry
*entries
;
251 /* general information that might be extracted from a superblock */
253 mdu_array_info_t array
;
254 mdu_disk_info_t disk
;
258 unsigned long long data_offset
;
259 unsigned long long new_data_offset
;
260 unsigned long long component_size
; /* same as array.size, except in
261 * sectors and up to 64bits.
263 unsigned long long custom_array_size
; /* size for non-default sized
264 * arrays (in sectors)
267 #define VOLUME_RESHAPE 1
268 #define CONTAINER_RESHAPE 2
269 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
271 unsigned long long reshape_progress
;
272 int recovery_blocked
; /* for external metadata it
273 * indicates that there is
274 * reshape in progress in
276 * for native metadata it is
277 * reshape_active field mirror
279 int journal_device_required
;
283 CONSISTENCY_POLICY_UNKNOWN
,
284 CONSISTENCY_POLICY_NONE
,
285 CONSISTENCY_POLICY_RESYNC
,
286 CONSISTENCY_POLICY_BITMAP
,
287 CONSISTENCY_POLICY_JOURNAL
,
288 CONSISTENCY_POLICY_PPL
,
289 } consistency_policy
;
291 /* During reshape we can sometimes change the data_offset to avoid
292 * over-writing still-valid data. We need to know if there is space.
293 * So getinfo_super will fill in space_before and space_after in sectors.
294 * data_offset can be increased or decreased by this amount.
296 unsigned long long space_before
, space_after
;
298 unsigned long long resync_start
; /* per-array resync position */
299 unsigned long long recovery_start
; /* per-device rebuild position */
300 #define MaxSector (~0ULL) /* resync/recovery complete position */
302 long bitmap_offset
; /* 0 == none, 1 == a file */
303 unsigned int ppl_size
;
305 unsigned long long ppl_sector
;
306 unsigned long safe_mode_delay
; /* ms delay to mark clean */
307 int new_level
, delta_disks
, new_layout
, new_chunk
;
309 unsigned long cache_size
; /* size of raid456 stripe cache*/
311 char text_version
[50];
313 int container_member
; /* for assembling external-metatdata arrays
314 * This is to be used internally by metadata
316 int container_enough
; /* flag external handlers can set to
317 * indicate that subarrays have not enough (-1),
318 * enough to start (0), or all expected disks (1) */
323 /* Device info for mdmon: */
330 #define DS_WRITE_MOSTLY 4
332 #define DS_BLOCKED 16
333 #define DS_REMOVE 1024
334 #define DS_UNBLOCK 2048
335 int prev_state
, curr_state
, next_state
;
337 /* info read from sysfs */
338 char sysfs_array_state
[20];
351 struct supertype
*supertype
;
367 extern char short_options
[];
368 extern char short_bitmap_options
[];
369 extern char short_bitmap_auto_options
[];
370 extern struct option long_options
[];
371 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
373 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
375 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
377 /* for option that don't have short equivilents, we assign arbitrary
378 * numbers later than any 'short' character option.
380 enum special_options
{
444 enum prefix_standard
{
456 FlagDefault
, FlagSet
, FlagClear
,
459 /* structures read from config file */
460 /* List of mddevice names and identifiers
461 * Identifiers can be:
463 * super-minor=decimal-minor-number-from-superblock
464 * devices=comma,separated,list,of,device,names,with,wildcards
466 * If multiple fields are present, the intersection of all matching
467 * devices is considered
469 #define UnSet (0xfffe)
479 char *devices
; /* comma separated list of device
480 * names with wild cards
485 struct supertype
*st
;
486 int autof
; /* 1 for normal, 2 for partitioned */
491 char *container
; /* /dev/whatever name of container, or
492 * uuid of container. You would expect
493 * this to be the 'devname' or UUID
494 * of some other entry.
496 char *member
; /* subarray within a container */
498 struct mddev_ident
*next
;
500 /* fields needed by different users of this structure */
501 int assembled
; /* set when assembly succeeds */
512 int require_homehost
;
542 unsigned long long size
;
543 int consistency_policy
;
546 /* List of device names - wildcards expanded */
549 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail,
551 * Not set for names read from .config
553 enum flag_mode writemostly
;
554 enum flag_mode failfast
;
555 int used
; /* set when used */
556 long long data_offset
;
557 struct mddev_dev
*next
;
560 typedef struct mapping
{
569 char *pattern
; /* U for up, _ for down */
570 int percent
; /* -1 if no resync */
571 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
574 char * metadata_version
;
577 struct dev_member
*next
;
579 struct mdstat_ent
*next
;
582 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
583 extern void mdstat_close(void);
584 extern void free_mdstat(struct mdstat_ent
*ms
);
585 extern void mdstat_wait(int seconds
);
586 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
587 extern int mddev_busy(char *devnm
);
588 extern struct mdstat_ent
*mdstat_by_component(char *name
);
589 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, char *container
);
592 struct map_ent
*next
;
599 extern int map_update(struct map_ent
**mpp
, char *devnm
, char *metadata
,
600 int uuid
[4], char *path
);
601 extern void map_remove(struct map_ent
**map
, char *devnm
);
602 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
603 extern struct map_ent
*map_by_devnm(struct map_ent
**map
, char *devnm
);
604 extern void map_free(struct map_ent
*map
);
605 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
606 extern void map_read(struct map_ent
**melp
);
607 extern int map_write(struct map_ent
*mel
);
608 extern void map_delete(struct map_ent
**mapp
, char *devnm
);
609 extern void map_add(struct map_ent
**melp
,
610 char *devnm
, char *metadata
, int uuid
[4], char *path
);
611 extern int map_lock(struct map_ent
**melp
);
612 extern void map_unlock(struct map_ent
**melp
);
613 extern void map_fork(void);
615 /* various details can be requested */
616 enum sysfs_read_flags
{
617 GET_LEVEL
= (1 << 0),
618 GET_LAYOUT
= (1 << 1),
619 GET_COMPONENT
= (1 << 2),
620 GET_CHUNK
= (1 << 3),
621 GET_CACHE
= (1 << 4),
622 GET_MISMATCH
= (1 << 5),
623 GET_VERSION
= (1 << 6),
624 GET_DISKS
= (1 << 7),
625 GET_DEGRADED
= (1 << 8),
626 GET_SAFEMODE
= (1 << 9),
627 GET_BITMAP_LOCATION
= (1 << 10),
629 GET_DEVS
= (1 << 20), /* gets role, major, minor */
630 GET_OFFSET
= (1 << 21),
631 GET_SIZE
= (1 << 22),
632 GET_STATE
= (1 << 23),
633 GET_ERROR
= (1 << 24),
634 GET_ARRAY_STATE
= (1 << 25),
635 GET_CONSISTENCY_POLICY
= (1 << 26),
638 /* If fd >= 0, get the array it is open on,
641 extern int sysfs_open(char *devnm
, char *devname
, char *attr
);
642 extern int sysfs_init(struct mdinfo
*mdi
, int fd
, char *devnm
);
643 extern void sysfs_init_dev(struct mdinfo
*mdi
, unsigned long devid
);
644 extern void sysfs_free(struct mdinfo
*sra
);
645 extern struct mdinfo
*sysfs_read(int fd
, char *devnm
, unsigned long options
);
646 extern int sysfs_attr_match(const char *attr
, const char *str
);
647 extern int sysfs_match_word(const char *word
, char **list
);
648 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
649 char *name
, char *val
);
650 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
651 char *name
, unsigned long long val
);
652 extern int sysfs_set_num_signed(struct mdinfo
*sra
, struct mdinfo
*dev
,
653 char *name
, long long val
);
654 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
655 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
657 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
658 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
659 char *name
, unsigned long long *val
);
660 extern int sysfs_fd_get_two(int fd
, unsigned long long *v1
, unsigned long long *v2
);
661 extern int sysfs_get_two(struct mdinfo
*sra
, struct mdinfo
*dev
,
662 char *name
, unsigned long long *v1
, unsigned long long *v2
);
663 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
664 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
666 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
667 char *name
, char *val
, int size
);
668 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
669 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
670 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
671 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
672 extern int sysfs_unique_holder(char *devnm
, long rdev
);
673 extern int sysfs_freeze_array(struct mdinfo
*sra
);
674 extern int sysfs_wait(int fd
, int *msec
);
675 extern int load_sys(char *path
, char *buf
, int len
);
676 extern int reshape_prepare_fdlist(char *devname
,
680 unsigned long blocks
,
683 unsigned long long *offsets
);
684 extern void reshape_free_fdlist(int *fdlist
,
685 unsigned long long *offsets
,
687 extern int reshape_open_backup_file(char *backup
,
692 unsigned long long *offsets
,
695 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
696 unsigned int ndata
, unsigned int odata
);
697 extern char *locate_backup(char *name
);
698 extern char *make_backup(char *name
);
700 extern int save_stripes(int *source
, unsigned long long *offsets
,
701 int raid_disks
, int chunk_size
, int level
, int layout
,
702 int nwrites
, int *dest
,
703 unsigned long long start
, unsigned long long length
,
705 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
706 int raid_disks
, int chunk_size
, int level
, int layout
,
707 int source
, unsigned long long read_offset
,
708 unsigned long long start
, unsigned long long length
,
712 #define Sendmail "/usr/lib/sendmail -t"
715 #define SYSLOG_FACILITY LOG_DAEMON
717 extern char *map_num(mapping_t
*map
, int num
);
718 extern int map_name(mapping_t
*map
, char *name
);
719 extern mapping_t r5layout
[], r6layout
[], pers
[], modes
[], faultylayout
[], consistency_policies
[];
721 extern char *map_dev_preferred(int major
, int minor
, int create
,
723 static inline char *map_dev(int major
, int minor
, int create
)
725 return map_dev_preferred(major
, minor
, create
, NULL
);
729 struct metadata_update
;
731 /* 'struct reshape' records the intermediate states of
733 * The starting geometry is converted to the 'before' geometry
734 * by at most an atomic level change. They could be the same.
735 * Similarly the 'after' geometry is converted to the final
736 * geometry by at most a level change.
737 * Note that 'before' and 'after' must have the same level.
738 * 'blocks' is the minimum number of sectors for a reshape unit.
739 * This will be a multiple of the stripe size in each of the
740 * 'before' and 'after' geometries.
741 * If 'blocks' is 0, no restriping is necessary.
742 * 'min_offset_change' is the minimum change to data_offset to
743 * allow the reshape to happen. It is at least the larger of
744 * the old and new chunk sizes, and typically the same as 'blocks'
745 * divided by number of data disks.
749 int parity
; /* number of parity blocks/devices */
754 unsigned long long backup_blocks
;
755 unsigned long long min_offset_change
;
756 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
757 unsigned long long new_size
; /* New size of array in sectors */
760 /* A superswitch provides entry point to a metadata handler.
762 * The superswitch primarily operates on some "metadata" that
763 * is accessed via the 'supertype'.
764 * This metadata has one of three possible sources.
765 * 1/ It is read from a single device. In this case it may not completely
766 * describe the array or arrays as some information might be on other
768 * 2/ It is read from all devices in a container. In this case all
769 * information is present.
770 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
771 * be complete once enough ->add_to_super calls have completed.
773 * When creating an array inside a container, the metadata will be
774 * formed by a combination of 2 and 3. The metadata or the array is read,
775 * then new information is added.
777 * The metadata must sometimes have a concept of a 'current' array
778 * and a 'current' device.
779 * The 'current' array is set by init_super to be the newly created array,
780 * or is set by super_by_fd when it finds it is looking at an array inside
783 * The 'current' device is either the device that the metadata was read from
784 * in case 1, or the last device added by add_to_super in case 3.
785 * Case 2 does not identify a 'current' device.
787 extern struct superswitch
{
789 /* Used to report details of metadata read from a component
790 * device. ->load_super has been called.
792 void (*examine_super
)(struct supertype
*st
, char *homehost
);
793 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
794 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
795 void (*export_examine_super
)(struct supertype
*st
);
796 int (*examine_badblocks
)(struct supertype
*st
, int fd
, char *devname
);
797 int (*copy_metadata
)(struct supertype
*st
, int from
, int to
);
799 /* Used to report details of an active array.
800 * ->load_super was possibly given a 'component' string.
802 void (*detail_super
)(struct supertype
*st
, char *homehost
);
803 void (*brief_detail_super
)(struct supertype
*st
);
804 void (*export_detail_super
)(struct supertype
*st
);
806 /* Optional: platform hardware / firmware details */
807 int (*detail_platform
)(int verbose
, int enumerate_only
, char *controller_path
);
808 int (*export_detail_platform
)(int verbose
, char *controller_path
);
811 * to get uuid to storing in bitmap metadata
812 * and 'reshape' backup-data metadata
813 * To see if a device is being re-added to an array it was part of.
815 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
817 /* Extract generic details from metadata. This could be details about
818 * the container, or about an individual array within the container.
819 * The determination is made either by:
820 * load_super being given a 'component' string.
821 * validate_geometry determining what to create.
822 * The info includes both array information and device information.
823 * The particular device should be:
824 * The last device added by add_to_super
825 * The device the metadata was loaded from by load_super
826 * If 'map' is present, then it is an array raid_disks long
827 * (raid_disk must already be set and correct) and it is filled
828 * with 1 for slots that are thought to be active and 0 for slots which
829 * appear to be failed/missing.
830 * *info is zeroed out before data is added.
832 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
833 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
834 /* Check if the given metadata is flagged as belonging to "this"
835 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
837 int (*match_home
)(struct supertype
*st
, char *homehost
);
839 /* Make one of several generic modifications to metadata
840 * prior to assembly (or other times).
841 * sparc2.2 - first bug in early 0.90 metadata
842 * super-minor - change name of 0.90 metadata
843 * summaries - 'correct' any redundant data
844 * resync - mark array as dirty to trigger a resync.
845 * uuid - set new uuid - only 0.90 or 1.x
846 * name - change the name of the array (where supported)
847 * homehost - change which host this array is tied to.
848 * devicesize - If metadata is at start of device, change recorded
849 * device size to match actual device size
850 * byteorder - swap bytes for 0.90 metadata
852 * force-one - mark that device as uptodate, not old or failed.
853 * force-array - mark array as clean if it would not otherwise
855 * assemble - not sure how this is different from force-one...
856 * linear-grow-new - add a new device to a linear array, but don't
857 * change the size: so superblock still matches
858 * linear-grow-update - now change the size of the array.
859 * writemostly - set the WriteMostly1 bit in the superblock devflags
860 * readwrite - clear the WriteMostly1 bit in the superblock devflags
861 * failfast - set the FailFast1 bit in the superblock
862 * nofailfast - clear the FailFast1 bit
863 * no-bitmap - clear any record that a bitmap is present.
864 * bbl - add a bad-block-log if possible
865 * no-bbl - remove any bad-block-log is it is empty.
866 * force-no-bbl - remove any bad-block-log even if empty.
867 * revert-reshape - If a reshape is in progress, modify metadata so
868 * it will resume going in the opposite direction.
870 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
872 char *devname
, int verbose
,
873 int uuid_set
, char *homehost
);
875 /* Create new metadata for new array as described. This could
876 * be a new container, or an array in a pre-existing container.
877 * Also used to zero metadata prior to writing it to invalidate old
880 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
881 struct shape
*s
, char *name
,
882 char *homehost
, int *uuid
,
883 unsigned long long data_offset
);
885 /* update the metadata to include new device, either at create or
886 * when hot-adding a spare.
888 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
889 int fd
, char *devname
,
890 unsigned long long data_offset
);
891 /* update the metadata to delete a device,
894 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
896 /* Write metadata to one device when fixing problems or adding
899 int (*store_super
)(struct supertype
*st
, int fd
);
901 /* Write all metadata for this array.
903 int (*write_init_super
)(struct supertype
*st
);
904 /* Check if metadata read from one device is compatible with an array,
905 * used when assembling an array, or pseudo-assembling was with
906 * "--examine --brief"
907 * If "st" has not yet been loaded the superblock from, "tst" is
908 * moved in, otherwise the superblock in 'st' is compared with
911 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
);
912 /* Load metadata from a single device. If 'devname' is not NULL
913 * print error messages as appropriate */
914 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
915 /* 'fd' is a 'container' md array - load array metadata from the
918 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
919 /* If 'arg' is a valid name of this metadata type, allocate and
920 * return a 'supertype' for the particular minor version */
921 struct supertype
* (*match_metadata_desc
)(char *arg
);
922 /* If a device has the given size, and the data_offset has been
923 * requested - work out how much space is available for data.
924 * This involves adjusting for reserved space (e.g. bitmaps)
925 * and for any rounding.
926 * 'mdadm' only calls this for existing arrays where a possible
927 * spare is being added. However some super-handlers call it
928 * internally from validate_geometry when creating an array.
930 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
,
931 unsigned long long data_offset
);
932 /* This is similar to 'avail_size' in purpose, but is used for
933 * containers for which there is no 'component size' to compare.
934 * This reports that whole-device size which is a minimum
936 unsigned long long (*min_acceptable_spare_size
)(struct supertype
*st
);
937 /* Find somewhere to put a bitmap - possibly auto-size it - and
938 * update the metadata to record this. The array may be newly
939 * created, in which case data_size may be updated, or it might
940 * already exist. Metadata handler can know if init_super
941 * has been called, but not write_init_super.
945 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
946 int delay
, int write_behind
,
947 unsigned long long size
, int may_change
, int major
);
948 /* Seek 'fd' to start of write-intent-bitmap. Must be an
949 * md-native format bitmap
951 int (*locate_bitmap
)(struct supertype
*st
, int fd
, int node_num
);
952 /* if add_internal_bitmap succeeded for existing array, this
955 int (*write_bitmap
)(struct supertype
*st
, int fd
, enum bitmap_update update
);
956 /* Free the superblock and any other allocated data */
957 void (*free_super
)(struct supertype
*st
);
959 /* validate_geometry is called with an st returned by
960 * match_metadata_desc.
961 * It should check that the geometry described is compatible with
962 * the metadata type. It will be called repeatedly as devices
963 * added to validate changing size and new devices. If there are
964 * inter-device dependencies, it should record sufficient details
965 * so these can be validated.
966 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
968 * 1: everything is OK
969 * 0: not OK for some reason - if 'verbose', then error was reported.
970 * -1: st->sb was NULL, 'subdev' is a member of a container of this
971 * type, but array is not acceptable for some reason
972 * message was reported even if verbose is 0.
974 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
976 int *chunk
, unsigned long long size
,
977 unsigned long long data_offset
,
978 char *subdev
, unsigned long long *freesize
,
979 int consistency_policy
, int verbose
);
981 /* Return a linked list of 'mdinfo' structures for all arrays
982 * in the container. For non-containers, it is like
983 * getinfo_super with an allocated mdinfo.*/
984 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
985 /* query the supertype for default geometry */
986 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
987 /* Permit subarray's to be deleted from inactive containers */
988 int (*kill_subarray
)(struct supertype
*st
); /* optional */
989 /* Permit subarray's to be modified */
990 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
991 char *update
, struct mddev_ident
*ident
); /* optional */
992 /* Check if reshape is supported for this external format.
993 * st is obtained from super_by_fd() where st->subarray[0] is
994 * initialized to indicate if reshape is being performed at the
995 * container or subarray level
997 #define APPLY_METADATA_CHANGES 1
998 #define ROLLBACK_METADATA_CHANGES 0
1000 int (*reshape_super
)(struct supertype
*st
,
1001 unsigned long long size
, int level
,
1002 int layout
, int chunksize
, int raid_disks
,
1003 int delta_disks
, char *backup
, char *dev
,
1005 int verbose
); /* optional */
1006 int (*manage_reshape
)( /* optional */
1007 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1008 struct supertype
*st
, unsigned long blocks
,
1009 int *fds
, unsigned long long *offsets
,
1010 int dests
, int *destfd
, unsigned long long *destoffsets
);
1013 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
1016 /* Tell the metadata handler the current state of the array.
1017 * This covers whether it is known to be consistent (no pending writes)
1018 * and how far along a resync is known to have progressed
1019 * (in a->resync_start).
1020 * resync status is really irrelevant if the array is not consistent,
1021 * but some metadata (DDF!) have a place to record the distinction.
1022 * If 'consistent' is '2', then the array can mark it dirty if a
1023 * resync/recovery/whatever is required, or leave it clean if not.
1024 * Return value is 0 dirty (not consistent) and 1 if clean.
1025 * it is only really important if consistent is passed in as '2'.
1027 int (*set_array_state
)(struct active_array
*a
, int consistent
);
1029 /* When the state of a device might have changed, we call set_disk to
1030 * tell the metadata what the current state is.
1031 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1033 * set_disk might be called when the state of the particular disk has
1034 * not in fact changed.
1036 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
1037 void (*sync_metadata
)(struct supertype
*st
);
1038 void (*process_update
)(struct supertype
*st
,
1039 struct metadata_update
*update
);
1040 /* Prepare updates allocates extra memory that might be
1041 * needed. If the update cannot be understood, return 0.
1043 int (*prepare_update
)(struct supertype
*st
,
1044 struct metadata_update
*update
);
1046 /* activate_spare will check if the array is degraded and, if it
1047 * is, try to find some spare space in the container.
1048 * On success, it add appropriate updates (For process_update) to
1049 * to the 'updates' list and returns a list of 'mdinfo' identifying
1050 * the device, or devices as there might be multiple missing
1051 * devices and multiple spares available.
1053 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
1054 struct metadata_update
**updates
);
1056 * Return statically allocated string that represents metadata specific
1057 * controller domain of the disk. The domain is used in disk domain
1058 * matching functions. Disks belong to the same domain if the they have
1059 * the same domain from mdadm.conf and belong the same metadata domain.
1060 * Returning NULL or not providing this handler means that metadata
1061 * does not distinguish the differences between disks that belong to
1062 * different controllers. They are in the domain specified by
1063 * configuration file (mdadm.conf).
1064 * In case when the metadata has the notion of domains based on disk
1065 * it shall return NULL for disks that do not belong to the controller
1066 * the supported domains. Such disks will form another domain and won't
1067 * be mixed with supported ones.
1069 const char *(*get_disk_controller_domain
)(const char *path
);
1071 /* for external backup area */
1072 int (*recover_backup
)(struct supertype
*st
, struct mdinfo
*info
);
1074 /* validate container after assemble */
1075 int (*validate_container
)(struct mdinfo
*info
);
1077 /* write initial empty PPL on device */
1078 int (*write_init_ppl
)(struct supertype
*st
, struct mdinfo
*info
, int fd
);
1080 /* validate ppl before assemble */
1081 int (*validate_ppl
)(struct supertype
*st
, struct mdinfo
*info
,
1082 struct mdinfo
*disk
);
1084 /* records new bad block in metadata */
1085 int (*record_bad_block
)(struct active_array
*a
, int n
,
1086 unsigned long long sector
, int length
);
1088 /* clears bad block from metadata */
1089 int (*clear_bad_block
)(struct active_array
*a
, int n
,
1090 unsigned long long sector
, int length
);
1092 /* get list of bad blocks from metadata */
1093 struct md_bb
*(*get_bad_blocks
)(struct active_array
*a
, int n
);
1095 int swapuuid
; /* true if uuid is bigending rather than hostendian */
1097 const char *name
; /* canonical metadata name */
1100 extern struct superswitch super0
, super1
;
1101 extern struct superswitch super_imsm
, super_ddf
;
1102 extern struct superswitch mbr
, gpt
;
1104 struct metadata_update
{
1107 void *space
; /* allocated space that monitor will use */
1108 void **space_list
; /* list of allocated spaces that monitor can
1109 * use or that it returned.
1111 struct metadata_update
*next
;
1114 /* A supertype holds a particular collection of metadata.
1115 * It identifies the metadata type by the superswitch, and the particular
1116 * sub-version of that metadata type.
1117 * metadata read in or created is stored in 'sb' and 'info'.
1118 * There are also fields used by mdmon to track containers.
1120 * A supertype may refer to:
1121 * Just an array, possibly in a container
1122 * A container, not identifying any particular array
1123 * Info read from just one device, not yet fully describing the array/container.
1126 * A supertype is created by:
1132 struct superswitch
*ss
;
1135 char container_devnm
[32]; /* devnm of container */
1138 void *other
; /* Hack used to convert v0.90 to v1.0 */
1139 unsigned long long devsize
;
1140 unsigned long long data_offset
; /* used by v1.x only */
1141 int ignore_hw_compat
; /* used to inform metadata handlers that it should ignore
1142 HW/firmware related incompatability to load metadata.
1143 Used when examining metadata to display content of disk
1144 when user has no hw/firmare compatible system.
1146 struct metadata_update
*updates
;
1147 struct metadata_update
**update_tail
;
1149 /* extra stuff used by mdmon */
1150 struct active_array
*arrays
;
1151 int sock
; /* listen to external programs */
1152 char devnm
[32]; /* e.g. md0. This appears in metadata_version:
1160 struct mdinfo
*devs
;
1164 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
1165 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
1166 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
1167 static inline struct supertype
*guess_super(int fd
) {
1168 return guess_super_type(fd
, guess_any
);
1170 extern struct supertype
*dup_super(struct supertype
*st
);
1171 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
1172 extern int get_dev_sector_size(int fd
, char *dname
, unsigned int *sectsizep
);
1173 extern int must_be_container(int fd
);
1174 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
1175 void wait_for(char *dev
, int fd
);
1178 * Data structures for policy management.
1179 * Each device can have a policy structure that lists
1180 * various name/value pairs each possibly with a metadata associated.
1181 * The policy list is sorted by name/value/metadata
1184 struct dev_policy
*next
;
1185 char *name
; /* None of these strings are allocated. They are
1186 * all just references to strings which are known
1187 * to exist elsewhere.
1188 * name and metadata can be compared by address equality.
1190 const char *metadata
;
1194 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
1196 /* iterate over the sublist starting at list, having the same
1197 * 'name' as 'list', and matching the given metadata (Where
1198 * NULL matches anything
1200 #define pol_for_each(item, list, _metadata) \
1202 item && item->name == list->name; \
1203 item = item->next) \
1204 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1208 * policy records read from mdadm are largely just name-value pairs.
1209 * The names are constants, not strdupped
1212 struct pol_rule
*next
;
1213 char *type
; /* rule_policy or rule_part */
1218 char *dups
; /* duplicates of 'value' with a partNN appended */
1222 extern char rule_policy
[], rule_part
[];
1223 extern char rule_path
[], rule_type
[];
1224 extern char type_part
[], type_disk
[];
1226 extern void policyline(char *line
, char *type
);
1227 extern void policy_add(char *type
, ...);
1228 extern void policy_free(void);
1230 extern struct dev_policy
*path_policy(char *path
, char *type
);
1231 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
1232 extern struct dev_policy
*devid_policy(int devid
);
1233 extern void dev_policy_free(struct dev_policy
*p
);
1235 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1236 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
1237 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
1239 enum policy_action
{
1243 act_spare
, /* This only applies to bare devices */
1244 act_spare_same_slot
, /* this allows non-bare devices,
1245 * but only if recent removal */
1246 act_force_spare
, /* this allow non-bare devices in any case */
1250 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
1251 enum policy_action want
);
1252 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
1253 enum policy_action want
);
1256 struct domainlist
*next
;
1260 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
1261 const char *metadata
);
1262 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
1263 const char *metadata
);
1264 extern void domainlist_add_dev(struct domainlist
**dom
, int devid
,
1265 const char *metadata
);
1266 extern void domain_free(struct domainlist
*dl
);
1267 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
1268 const char *metadata
);
1269 void domain_add(struct domainlist
**domp
, char *domain
);
1271 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
1272 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
1278 #define HAVE_NFTW we assume
1282 # include <features.h>
1283 # ifndef __UCLIBC_HAS_LFS__
1284 # define lseek64 lseek
1286 # ifndef __UCLIBC_HAS_FTW__
1296 #if defined(__KLIBC__)
1312 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
1314 extern int Manage_ro(char *devname
, int fd
, int readonly
);
1315 extern int Manage_run(char *devname
, int fd
, struct context
*c
);
1316 extern int Manage_stop(char *devname
, int fd
, int quiet
,
1318 extern int Manage_subdevs(char *devname
, int fd
,
1319 struct mddev_dev
*devlist
, int verbose
, int test
,
1320 char *update
, int force
);
1321 extern int autodetect(void);
1322 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1323 extern int Grow_addbitmap(char *devname
, int fd
,
1324 struct context
*c
, struct shape
*s
);
1325 extern int Grow_reshape(char *devname
, int fd
,
1326 struct mddev_dev
*devlist
,
1327 unsigned long long data_offset
,
1328 struct context
*c
, struct shape
*s
);
1329 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1330 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1331 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1332 struct mdinfo
*info
, char *backup_file
,
1333 int forked
, int freeze_reshape
);
1334 extern int Grow_consistency_policy(char *devname
, int fd
,
1335 struct context
*c
, struct shape
*s
);
1337 extern int restore_backup(struct supertype
*st
,
1338 struct mdinfo
*content
,
1341 char **backup_filep
,
1343 extern int Grow_continue_command(char *devname
, int fd
,
1344 char *backup_file
, int verbose
);
1346 extern int Assemble(struct supertype
*st
, char *mddev
,
1347 struct mddev_ident
*ident
,
1348 struct mddev_dev
*devlist
,
1351 extern int Build(char *mddev
, struct mddev_dev
*devlist
,
1352 struct shape
*s
, struct context
*c
);
1354 extern int Create(struct supertype
*st
, char *mddev
,
1355 char *name
, int *uuid
,
1356 int subdevs
, struct mddev_dev
*devlist
,
1359 unsigned long long data_offset
);
1361 extern int Detail(char *dev
, struct context
*c
);
1362 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
, int export
, char *controller_path
);
1363 extern int Query(char *dev
);
1364 extern int ExamineBadblocks(char *devname
, int brief
, struct supertype
*forcest
);
1365 extern int Examine(struct mddev_dev
*devlist
, struct context
*c
,
1366 struct supertype
*forcest
);
1367 extern int Monitor(struct mddev_dev
*devlist
,
1368 char *mailaddr
, char *alert_cmd
,
1370 int daemonise
, int oneshot
,
1371 int dosyslog
, char *pidfile
, int increments
,
1374 extern int Kill(char *dev
, struct supertype
*st
, int force
, int verbose
, int noexcl
);
1375 extern int Kill_subarray(char *dev
, char *subarray
, int verbose
);
1376 extern int Update_subarray(char *dev
, char *subarray
, char *update
, struct mddev_ident
*ident
, int quiet
);
1377 extern int Wait(char *dev
);
1378 extern int WaitClean(char *dev
, int sock
, int verbose
);
1379 extern int SetAction(char *dev
, char *action
);
1381 extern int Incremental(struct mddev_dev
*devlist
, struct context
*c
,
1382 struct supertype
*st
);
1383 extern void RebuildMap(void);
1384 extern int IncrementalScan(struct context
*c
, char *devnm
);
1385 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1386 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1387 unsigned long chunksize
, unsigned long daemon_sleep
,
1388 unsigned long write_behind
,
1389 unsigned long long array_size
,
1391 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1392 extern int Write_rules(char *rule_name
);
1393 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1395 /* calculate the size of the bitmap given the array size and bitmap chunksize */
1396 static inline unsigned long long
1397 bitmap_bits(unsigned long long array_size
, unsigned long chunksize
)
1399 return (array_size
* 512 + chunksize
- 1) / chunksize
;
1402 extern int Dump_metadata(char *dev
, char *dir
, struct context
*c
,
1403 struct supertype
*st
);
1404 extern int Restore_metadata(char *dev
, char *dir
, struct context
*c
,
1405 struct supertype
*st
, int only
);
1407 extern int md_get_version(int fd
);
1408 int md_get_array_info(int fd
, struct mdu_array_info_s
*array
);
1409 int md_set_array_info(int fd
, struct mdu_array_info_s
*array
);
1410 int md_get_disk_info(int fd
, struct mdu_disk_info_s
*disk
);
1411 extern int get_linux_version(void);
1412 extern int mdadm_version(char *version
);
1413 extern unsigned long long parse_size(char *size
);
1414 extern int parse_uuid(char *str
, int uuid
[4]);
1415 extern int parse_layout_10(char *layout
);
1416 extern int parse_layout_faulty(char *layout
);
1417 extern long parse_num(char *num
);
1418 extern int parse_cluster_confirm_arg(char *inp
, char **devname
, int *slot
);
1419 extern int check_ext2(int fd
, char *name
);
1420 extern int check_reiser(int fd
, char *name
);
1421 extern int check_raid(int fd
, char *name
);
1422 extern int check_partitions(int fd
, char *dname
,
1423 unsigned long long freesize
,
1424 unsigned long long size
);
1426 extern int get_mdp_major(void);
1427 extern int get_maj_min(char *dev
, int *major
, int *minor
);
1428 extern int dev_open(char *dev
, int flags
);
1429 extern int open_dev(char *devnm
);
1430 extern void reopen_mddev(int mdfd
);
1431 extern int open_dev_flags(char *devnm
, int flags
);
1432 extern int open_dev_excl(char *devnm
);
1433 extern int is_standard(char *dev
, int *nump
);
1434 extern int same_dev(char *one
, char *two
);
1435 extern int compare_paths (char* path1
,char* path2
);
1436 extern void enable_fds(int devices
);
1438 extern int parse_auto(char *str
, char *msg
, int config
);
1439 extern struct mddev_ident
*conf_get_ident(char *dev
);
1440 extern struct mddev_dev
*conf_get_devs(void);
1441 extern int conf_test_dev(char *devname
);
1442 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1443 extern struct createinfo
*conf_get_create_info(void);
1444 extern void set_conffile(char *file
);
1445 extern char *conf_get_mailaddr(void);
1446 extern char *conf_get_mailfrom(void);
1447 extern char *conf_get_program(void);
1448 extern char *conf_get_homehost(int *require_homehostp
);
1449 extern char *conf_get_homecluster(void);
1450 extern char *conf_line(FILE *file
);
1451 extern char *conf_word(FILE *file
, int allow_key
);
1452 extern void print_quoted(char *str
);
1453 extern void print_escape(char *str
);
1454 extern int use_udev(void);
1455 extern unsigned long GCD(unsigned long a
, unsigned long b
);
1456 extern int conf_name_is_free(char *name
);
1457 extern int conf_verify_devnames(struct mddev_ident
*array_list
);
1458 extern int devname_matches(char *name
, char *match
);
1459 extern struct mddev_ident
*conf_match(struct supertype
*st
,
1460 struct mdinfo
*info
,
1462 int verbose
, int *rvp
);
1463 extern int experimental(void);
1465 extern void free_line(char *line
);
1466 extern int match_oneof(char *devices
, char *devname
);
1467 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1468 extern const int uuid_zero
[4];
1469 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1470 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1471 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1472 extern char *fname_from_uuid(struct supertype
*st
,
1473 struct mdinfo
*info
, char *buf
, char sep
);
1474 extern unsigned long calc_csum(void *super
, int bytes
);
1475 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1477 extern int enough_fd(int fd
);
1478 extern int ask(char *mesg
);
1479 extern unsigned long long get_component_size(int fd
);
1480 extern void remove_partitions(int fd
);
1481 extern int test_partition(int fd
);
1482 extern int test_partition_from_id(dev_t id
);
1483 extern int get_data_disks(int level
, int layout
, int raid_disks
);
1484 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1485 int chunksize
, unsigned long long devsize
);
1486 extern int flush_metadata_updates(struct supertype
*st
);
1487 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1488 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1489 struct mdinfo
*content
,
1491 char *chosen_name
, int *result
);
1493 #define INCR_UNSAFE 2
1494 #define INCR_ALREADY 4
1496 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1497 unsigned long long min_size
,
1498 struct domainlist
*domlist
,
1500 const char *metadata
, int get_one
);
1501 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1502 extern int add_disk(int mdfd
, struct supertype
*st
,
1503 struct mdinfo
*sra
, struct mdinfo
*info
);
1504 extern int remove_disk(int mdfd
, struct supertype
*st
,
1505 struct mdinfo
*sra
, struct mdinfo
*info
);
1506 extern int hot_remove_disk(int mdfd
, unsigned long dev
, int force
);
1507 extern int sys_hot_remove_disk(int statefd
, int force
);
1508 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1509 unsigned long long min_recovery_start(struct mdinfo
*array
);
1511 extern char *human_size(long long bytes
);
1512 extern char *human_size_brief(long long bytes
, int prefix
);
1513 extern void print_r10_layout(int layout
);
1515 extern char *find_free_devnm(int use_partitions
);
1517 extern void put_md_name(char *name
);
1518 extern char *devid2kname(int devid
);
1519 extern char *devid2devnm(dev_t devid
);
1520 extern dev_t
devnm2devid(char *devnm
);
1521 extern char *get_md_name(char *devnm
);
1523 extern char DefaultConfFile
[];
1525 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1527 /* values for 'trustworthy' */
1529 #define LOCAL_ANY 10
1532 extern int open_mddev(char *dev
, int report_errors
);
1533 extern int open_container(int fd
);
1534 extern int metadata_container_matches(char *metadata
, char *devnm
);
1535 extern int metadata_subdev_matches(char *metadata
, char *devnm
);
1536 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1537 extern int is_subarray_active(char *subarray
, char *devname
);
1538 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1539 extern struct superswitch
*version_to_superswitch(char *vers
);
1541 extern int mdmon_running(char *devnm
);
1542 extern int mdmon_pid(char *devnm
);
1543 extern int check_env(char *name
);
1544 extern __u32
random32(void);
1545 extern void random_uuid(__u8
*buf
);
1546 extern int start_mdmon(char *devnm
);
1548 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1549 struct supertype
*st
, unsigned long stripes
,
1550 int *fds
, unsigned long long *offsets
,
1551 int dests
, int *destfd
, unsigned long long *destoffsets
);
1552 void abort_reshape(struct mdinfo
*sra
);
1554 void *super1_make_v0(struct supertype
*st
, struct mdinfo
*info
, mdp_super_t
*sb0
);
1556 extern char *stat2kname(struct stat
*st
);
1557 extern char *fd2kname(int fd
);
1558 extern char *stat2devnm(struct stat
*st
);
1559 extern char *fd2devnm(int fd
);
1561 extern int in_initrd(void);
1564 void *cmap_handle
; /* corosync lib related */
1566 int (*initialize
)(cmap_handle_t
*handle
);
1567 int (*get_string
)(cmap_handle_t handle
,
1570 int (*finalize
)(cmap_handle_t handle
);
1573 extern void set_cmap_hooks(void);
1574 extern void set_hooks(void);
1577 void *dlm_handle
; /* dlm lib related */
1579 dlm_lshandle_t (*create_lockspace
)(const char *name
,
1581 int (*release_lockspace
)(const char *name
, dlm_lshandle_t ls
,
1583 int (*ls_lock
)(dlm_lshandle_t lockspace
, uint32_t mode
,
1584 struct dlm_lksb
*lksb
, uint32_t flags
,
1585 const void *name
, unsigned int namelen
,
1586 uint32_t parent
, void (*astaddr
) (void *astarg
),
1587 void *astarg
, void (*bastaddr
) (void *astarg
),
1589 int (*ls_unlock
)(dlm_lshandle_t lockspace
, uint32_t lkid
,
1590 uint32_t flags
, struct dlm_lksb
*lksb
,
1592 int (*ls_get_fd
)(dlm_lshandle_t ls
);
1593 int (*dispatch
)(int fd
);
1596 extern int get_cluster_name(char **name
);
1597 extern int dlm_funs_ready(void);
1598 extern int cluster_get_dlmlock(int *lockid
);
1599 extern int cluster_release_dlmlock(int lockid
);
1600 extern void set_dlm_hooks(void);
1602 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1603 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1604 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1605 (ROUND_UP((unsigned long)(ptr), base)))
1607 static inline int is_subarray(char *vers
)
1609 /* The version string for a 'subarray' (an array in a container)
1611 * /containername/componentname for normal read-write arrays
1612 * -containername/componentname for arrays which mdmon must not
1613 * reconfigure. They might be read-only
1614 * or might be undergoing reshape etc.
1615 * containername is e.g. md0, md_d1
1616 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1618 return (*vers
== '/' || *vers
== '-');
1621 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1623 return &ent
->metadata_version
[10+strlen(container
)+1];
1627 #define dprintf(fmt, arg...) \
1628 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1629 #define dprintf_cont(fmt, arg...) \
1630 fprintf(stderr, fmt, ##arg)
1632 #define dprintf(fmt, arg...) \
1633 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1634 #define dprintf_cont(fmt, arg...) \
1635 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1639 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1643 ret
= vasprintf(strp
, fmt
, ap
);
1650 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1652 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1654 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1656 void *xmalloc(size_t len
);
1657 void *xrealloc(void *ptr
, size_t len
);
1658 void *xcalloc(size_t num
, size_t size
);
1659 char *xstrdup(const char *str
);
1661 #define LEVEL_MULTIPATH (-4)
1662 #define LEVEL_LINEAR (-1)
1663 #define LEVEL_FAULTY (-5)
1665 /* kernel module doesn't know about these */
1666 #define LEVEL_CONTAINER (-100)
1667 #define LEVEL_UNSUPPORTED (-200)
1669 /* the kernel does know about this one ... */
1670 #define LEVEL_NONE (-1000000)
1674 #define WriteTransient 0
1675 #define ReadTransient 1
1676 #define WritePersistent 2
1677 #define ReadPersistent 3
1678 #define WriteAll 4 /* doesn't go to device */
1679 #define ReadFixable 5
1682 #define ClearErrors 31
1683 #define ClearFaults 30
1685 #define AllPersist 100 /* internal use only */
1686 #define NoPersist 101
1688 #define ModeMask 0x1f
1695 #define minor(x) ((x)&0xff)
1696 #define major(x) (((x)>>8)&0xff)
1697 #define makedev(M,m) (((M)<<8) | (m))
1701 #define ALGORITHM_LEFT_ASYMMETRIC 0
1702 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1703 #define ALGORITHM_LEFT_SYMMETRIC 2
1704 #define ALGORITHM_RIGHT_SYMMETRIC 3
1706 /* Define non-rotating (raid4) algorithms. These allow
1707 * conversion of raid4 to raid5.
1709 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1710 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1712 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1713 * Firstly, the exact positioning of the parity block is slightly
1714 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1716 * Secondly, or order of datablocks over which the Q syndrome is computed
1718 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1719 * These layouts are from the DDFv1.2 spec.
1720 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1721 * leaves RLQ=3 as 'Vendor Specific'
1724 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1725 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1726 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1728 /* For every RAID5 algorithm we define a RAID6 algorithm
1729 * with exactly the same layout for data and parity, and
1730 * with the Q block always on the last device (N-1).
1731 * This allows trivial conversion from RAID5 to RAID6
1733 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1734 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1735 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1736 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1737 #define ALGORITHM_PARITY_0_6 20
1738 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1740 /* Define PATH_MAX in case we don't use glibc or standard library does
1741 * not have PATH_MAX defined. Assume max path length is 4K characters.
1744 #define PATH_MAX 4096
1747 #define RESYNC_NONE -1
1748 #define RESYNC_DELAYED -2
1749 #define RESYNC_PENDING -3
1750 #define RESYNC_UNKNOWN -4
1752 /* When using "GET_DISK_INFO" it isn't certain how high
1753 * we need to check. So we impose an absolute limit of
1754 * MAX_DISKS. This needs to be much more than the largest
1755 * number of devices any metadata can support. Currently
1756 * v1.x can support 1920
1758 #define MAX_DISKS 4096
1760 /* Sometimes the 'size' value passed needs to mean "Maximum".
1761 * In those cases with use MAX_SIZE
1765 /* We want to use unsigned numbers for sector counts, but need
1766 * a value for 'invalid'. Use '1'.
1768 #define INVALID_SECTORS 1
1769 /* And another special number needed for --data_offset=variable */
1770 #define VARIABLE_OFFSET 3