2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2013 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>
28 #include <sys/socket.h>
29 #include <sys/utsname.h>
32 #include <sys/resource.h>
34 #include <linux/magic.h>
42 * following taken from linux/blkpg.h because they aren't
43 * anywhere else and it isn't safe to #include linux/ * stuff.
46 #define BLKPG _IO(0x12,105)
48 /* The argument structure */
49 struct blkpg_ioctl_arg
{
56 /* The subfunctions (for the op field) */
57 #define BLKPG_ADD_PARTITION 1
58 #define BLKPG_DEL_PARTITION 2
60 /* Sizes of name fields. Unused at present. */
61 #define BLKPG_DEVNAMELTH 64
62 #define BLKPG_VOLNAMELTH 64
64 /* The data structure for ADD_PARTITION and DEL_PARTITION */
65 struct blkpg_partition
{
66 long long start
; /* starting offset in bytes */
67 long long length
; /* length in bytes */
68 int pno
; /* partition number */
69 char devname
[BLKPG_DEVNAMELTH
]; /* partition name, like sda5 or c0d1p2,
70 to be used in kernel messages */
71 char volname
[BLKPG_VOLNAMELTH
]; /* volume label */
76 /* Force a compilation error if condition is true */
77 #define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
79 /* Force a compilation error if condition is true, but also produce a
80 result (of value 0 and type size_t), so the expression can be used
81 e.g. in a structure initializer (or where-ever else comma expressions
83 #define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
85 static struct dlm_hooks
*dlm_hooks
= NULL
;
86 static int is_dlm_hooks_ready
= 0;
87 struct dlm_lock_resource
*dlm_lock_res
= NULL
;
88 static int ast_called
= 0;
90 struct dlm_lock_resource
{
95 int dlm_funs_ready(void)
97 return is_dlm_hooks_ready
? 1 : 0;
100 /* Using poll(2) to wait for and dispatch ASTs */
101 static int poll_for_ast(dlm_lshandle_t ls
)
105 pfd
.fd
= dlm_hooks
->ls_get_fd(ls
);
110 if (poll(&pfd
, 1, 0) < 0)
115 dlm_hooks
->dispatch(dlm_hooks
->ls_get_fd(ls
));
122 static void dlm_ast(void *arg
)
127 static char *cluster_name
= NULL
;
128 /* Create the lockspace, take bitmapXXX locks on all the bitmaps. */
129 int cluster_get_dlmlock(int *lockid
)
133 int flags
= LKF_NOQUEUE
;
135 ret
= get_cluster_name(&cluster_name
);
137 pr_err("The md can't get cluster name\n");
141 dlm_lock_res
= xmalloc(sizeof(struct dlm_lock_resource
));
142 dlm_lock_res
->ls
= dlm_hooks
->create_lockspace(cluster_name
, O_RDWR
);
143 if (!dlm_lock_res
->ls
) {
144 pr_err("%s failed to create lockspace\n", cluster_name
);
148 /* Conversions need the lockid in the LKSB */
149 if (flags
& LKF_CONVERT
)
150 dlm_lock_res
->lksb
.sb_lkid
= *lockid
;
152 snprintf(str
, 64, "bitmap%s", cluster_name
);
153 /* if flags with LKF_CONVERT causes below return ENOENT which means
154 * "No such file or directory" */
155 ret
= dlm_hooks
->ls_lock(dlm_lock_res
->ls
, LKM_PWMODE
, &dlm_lock_res
->lksb
,
156 flags
, str
, strlen(str
), 0, dlm_ast
,
157 dlm_lock_res
, NULL
, NULL
);
159 pr_err("error %d when get PW mode on lock %s\n", errno
, str
);
163 /* Wait for it to complete */
164 poll_for_ast(dlm_lock_res
->ls
);
165 *lockid
= dlm_lock_res
->lksb
.sb_lkid
;
167 errno
= dlm_lock_res
->lksb
.sb_status
;
169 pr_err("error %d happened in ast with lock %s\n", errno
, str
);
177 int cluster_release_dlmlock(int lockid
)
184 /* if flags with LKF_CONVERT causes below return EINVAL which means
185 * "Invalid argument" */
186 ret
= dlm_hooks
->ls_unlock(dlm_lock_res
->ls
, lockid
, 0,
187 &dlm_lock_res
->lksb
, dlm_lock_res
);
189 pr_err("error %d happened when unlock\n", errno
);
190 /* XXX make sure the lock is unlocked eventually */
194 /* Wait for it to complete */
195 poll_for_ast(dlm_lock_res
->ls
);
197 errno
= dlm_lock_res
->lksb
.sb_status
;
198 if (errno
!= EUNLOCK
) {
199 pr_err("error %d happened in ast when unlock lockspace\n", errno
);
200 /* XXX make sure the lockspace is unlocked eventually */
204 ret
= dlm_hooks
->release_lockspace(cluster_name
, dlm_lock_res
->ls
, 1);
206 pr_err("error %d happened when release lockspace\n", errno
);
207 /* XXX make sure the lockspace is released eventually */
217 * Parse a 128 bit uuid in 4 integers
218 * format is 32 hexx nibbles with options :.<space> separator
219 * If not exactly 32 hex digits are found, return 0
222 int parse_uuid(char *str
, int uuid
[4])
224 int hit
= 0; /* number of Hex digIT */
227 for (i
= 0; i
< 4; i
++)
230 while ((c
= *str
++) != 0) {
232 if (c
>= '0' && c
<= '9')
234 else if (c
>= 'a' && c
<= 'f')
236 else if (c
>= 'A' && c
<= 'F')
238 else if (strchr(":. -", c
))
254 * Get the md version number.
255 * We use the RAID_VERSION ioctl if it is supported
256 * If not, but we have a block device with major '9', we assume
259 * Return version number as 24 but number - assume version parts
263 int md_get_version(int fd
)
268 if (fstat(fd
, &stb
)<0)
270 if ((S_IFMT
&stb
.st_mode
) != S_IFBLK
)
273 if (ioctl(fd
, RAID_VERSION
, &vers
) == 0)
274 return (vers
.major
*10000) + (vers
.minor
*100) + vers
.patchlevel
;
277 if (major(stb
.st_rdev
) == MD_MAJOR
)
282 int get_linux_version()
286 int a
= 0, b
= 0,c
= 0;
291 a
= strtoul(cp
, &cp
, 10);
293 b
= strtoul(cp
+1, &cp
, 10);
295 c
= strtoul(cp
+1, &cp
, 10);
297 return (a
*1000000)+(b
*1000)+c
;
301 int mdadm_version(char *version
)
309 cp
= strchr(version
, '-');
310 if (!cp
|| *(cp
+1) != ' ' || *(cp
+2) != 'v')
313 a
= strtoul(cp
, &cp
, 10);
316 b
= strtoul(cp
+1, &cp
, 10);
318 c
= strtoul(cp
+1, &cp
, 10);
321 if (*cp
!= ' ' && *cp
!= '-')
323 return (a
*1000000)+(b
*1000)+c
;
326 unsigned long long parse_size(char *size
)
328 /* parse 'size' which should be a number optionally
329 * followed by 'K', 'M', or 'G'.
330 * Without a suffix, K is assumed.
331 * Number returned is in sectors (half-K)
332 * INVALID_SECTORS returned on error.
335 long long s
= strtoll(size
, &c
, 10);
349 s
*= 1024 * 1024 * 2;
351 case 's': /* sectors */
362 int parse_layout_10(char *layout
)
366 /* Parse the layout string for raid10 */
367 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
368 if ((layout
[0] != 'n' && layout
[0] != 'f' && layout
[0] != 'o') ||
369 (copies
= strtoul(layout
+1, &cp
, 10)) < 1 ||
373 if (layout
[0] == 'n')
375 else if (layout
[0] == 'o')
376 rv
= 0x10000 + (copies
<<8) + 1;
378 rv
= 1 + (copies
<<8);
382 int parse_layout_faulty(char *layout
)
384 /* Parse the layout string for 'faulty' */
385 int ln
= strcspn(layout
, "0123456789");
386 char *m
= xstrdup(layout
);
389 mode
= map_name(faultylayout
, m
);
393 return mode
| (atoi(layout
+ln
)<< ModeShift
);
396 long parse_num(char *num
)
398 /* Either return a valid number, or -1 */
400 long rv
= strtol(num
, &c
, 10);
401 if (rv
< 0 || *c
|| !num
[0])
408 int parse_cluster_confirm_arg(char *input
, char **devname
, int *slot
)
411 *slot
= strtoul(input
, &dev
, 10);
412 if (dev
== input
|| dev
[0] != ':')
418 void remove_partitions(int fd
)
420 /* remove partitions from this block devices.
421 * This is used for components added to an array
423 #ifdef BLKPG_DEL_PARTITION
424 struct blkpg_ioctl_arg a
;
425 struct blkpg_partition p
;
427 a
.op
= BLKPG_DEL_PARTITION
;
429 a
.datalen
= sizeof(p
);
431 memset(a
.data
, 0, a
.datalen
);
432 for (p
.pno
= 0; p
.pno
< 16; p
.pno
++)
433 ioctl(fd
, BLKPG
, &a
);
437 int test_partition(int fd
)
439 /* Check if fd is a whole-disk or a partition.
440 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
441 * will return ENXIO on an invalid partition number.
443 struct blkpg_ioctl_arg a
;
444 struct blkpg_partition p
;
445 a
.op
= BLKPG_DEL_PARTITION
;
447 a
.datalen
= sizeof(p
);
449 memset(a
.data
, 0, a
.datalen
);
451 if (ioctl(fd
, BLKPG
, &a
) == 0)
452 /* Very unlikely, but not a partition */
454 if (errno
== ENXIO
|| errno
== ENOTTY
)
455 /* not a partition */
461 int test_partition_from_id(dev_t id
)
466 sprintf(buf
, "%d:%d", major(id
), minor(id
));
467 fd
= dev_open(buf
, O_RDONLY
);
470 rv
= test_partition(fd
);
475 int enough(int level
, int raid_disks
, int layout
, int clean
, char *avail
)
481 for (i
= 0; i
< raid_disks
; i
++)
482 avail_disks
+= !!avail
[i
];
486 /* This is the tricky one - we need to check
487 * which actual disks are present.
489 copies
= (layout
&255)* ((layout
>>8) & 255);
492 /* there must be one of the 'copies' form 'first' */
499 this = (this+1) % raid_disks
;
503 first
= (first
+(layout
&255)) % raid_disks
;
504 } while (first
!= 0);
507 case LEVEL_MULTIPATH
:
508 return avail_disks
>= 1;
511 return avail_disks
== raid_disks
;
513 return avail_disks
>= 1;
515 if (avail_disks
== raid_disks
- 1 &&
516 !avail
[raid_disks
- 1])
517 /* If just the parity device is missing, then we
518 * have enough, even if not clean
524 return avail_disks
>= raid_disks
-1;
526 return avail_disks
>= raid_disks
;
529 return avail_disks
>= raid_disks
-2;
531 return avail_disks
>= raid_disks
;
537 int enough_fd(int fd
)
539 struct mdu_array_info_s array
;
540 struct mdu_disk_info_s disk
;
544 if (ioctl(fd
, GET_ARRAY_INFO
, &array
) != 0 ||
545 array
.raid_disks
<= 0)
547 avail
= xcalloc(array
.raid_disks
, 1);
548 for (i
= 0; i
< MAX_DISKS
&& array
.nr_disks
> 0; i
++) {
550 if (ioctl(fd
, GET_DISK_INFO
, &disk
) != 0)
552 if (disk
.major
== 0 && disk
.minor
== 0)
556 if (! (disk
.state
& (1<<MD_DISK_SYNC
)))
558 if (disk
.raid_disk
< 0 || disk
.raid_disk
>= array
.raid_disks
)
560 avail
[disk
.raid_disk
] = 1;
562 /* This is used on an active array, so assume it is clean */
563 rv
= enough(array
.level
, array
.raid_disks
, array
.layout
,
569 const int uuid_zero
[4] = { 0, 0, 0, 0 };
571 int same_uuid(int a
[4], int b
[4], int swapuuid
)
574 /* parse uuids are hostendian.
575 * uuid's from some superblocks are big-ending
576 * if there is a difference, we need to swap..
578 unsigned char *ac
= (unsigned char *)a
;
579 unsigned char *bc
= (unsigned char *)b
;
581 for (i
= 0; i
< 16; i
+= 4) {
582 if (ac
[i
+0] != bc
[i
+3] ||
583 ac
[i
+1] != bc
[i
+2] ||
584 ac
[i
+2] != bc
[i
+1] ||
599 void copy_uuid(void *a
, int b
[4], int swapuuid
)
602 /* parse uuids are hostendian.
603 * uuid's from some superblocks are big-ending
604 * if there is a difference, we need to swap..
606 unsigned char *ac
= (unsigned char *)a
;
607 unsigned char *bc
= (unsigned char *)b
;
609 for (i
= 0; i
< 16; i
+= 4) {
619 char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
)
626 copy_uuid(uuid
, id
, swap
);
627 for (i
= 0; i
< 4; i
++) {
630 for (j
= 3; j
>= 0; j
--) {
631 sprintf(c
,"%02x", (unsigned char) uuid
[j
+4*i
]);
639 char *fname_from_uuid(struct supertype
*st
, struct mdinfo
*info
, char *buf
, char sep
)
641 // dirty hack to work around an issue with super1 superblocks...
642 // super1 superblocks need swapuuid set in order for assembly to
643 // work, but can't have it set if we want this printout to match
644 // all the other uuid printouts in super1.c, so we force swapuuid
645 // to 1 to make our printout match the rest of super1
646 return __fname_from_uuid(info
->uuid
, (st
->ss
== &super1
) ? 1 : st
->ss
->swapuuid
, buf
, sep
);
650 int check_ext2(int fd
, char *name
)
653 * Check for an ext2fs file system.
654 * Superblock is always 1K at 1K offset
656 * s_magic is le16 at 56 == 0xEF53
657 * report mtime - le32 at 44
659 * logblksize - le32 at 24
661 unsigned char sb
[1024];
663 unsigned long long size
;
665 if (lseek(fd
, 1024,0)!= 1024)
667 if (read(fd
, sb
, 1024)!= 1024)
669 if (sb
[56] != 0x53 || sb
[57] != 0xef)
672 mtime
= sb
[44]|(sb
[45]|(sb
[46]|sb
[47]<<8)<<8)<<8;
673 bsize
= sb
[24]|(sb
[25]|(sb
[26]|sb
[27]<<8)<<8)<<8;
674 size
= sb
[4]|(sb
[5]|(sb
[6]|sb
[7]<<8)<<8)<<8;
676 pr_err("%s appears to contain an ext2fs file system\n",
678 cont_err("size=%lluK mtime=%s", size
, ctime(&mtime
));
682 int check_reiser(int fd
, char *name
)
685 * superblock is at 64K
687 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
690 unsigned char sb
[1024];
691 unsigned long long size
;
692 if (lseek(fd
, 64*1024, 0) != 64*1024)
694 if (read(fd
, sb
, 1024) != 1024)
696 if (strncmp((char*)sb
+52, "ReIsErFs",8) != 0 &&
697 strncmp((char*)sb
+52, "ReIsEr2Fs",9) != 0)
699 pr_err("%s appears to contain a reiserfs file system\n",name
);
700 size
= sb
[0]|(sb
[1]|(sb
[2]|sb
[3]<<8)<<8)<<8;
701 cont_err("size = %lluK\n", size
*4);
706 int check_raid(int fd
, char *name
)
711 struct supertype
*st
= guess_super(fd
);
715 st
->ss
->load_super(st
, fd
, name
);
716 /* Looks like a raid array .. */
717 pr_err("%s appears to be part of a raid array:\n",
719 st
->ss
->getinfo_super(st
, &info
, NULL
);
720 st
->ss
->free_super(st
);
721 crtime
= info
.array
.ctime
;
722 level
= map_num(pers
, info
.array
.level
);
723 if (!level
) level
= "-unknown-";
724 cont_err("level=%s devices=%d ctime=%s",
725 level
, info
.array
.raid_disks
, ctime(&crtime
));
733 for (i
= 0; i
< 5; i
++) {
735 fprintf(stderr
, "%s%s", mesg
, add
);
737 if (fgets(buf
, 100, stdin
)==NULL
)
739 if (buf
[0]=='y' || buf
[0]=='Y')
741 if (buf
[0]=='n' || buf
[0]=='N')
745 pr_err("assuming 'no'\n");
748 #endif /* MDASSEMBLE */
750 int is_standard(char *dev
, int *nump
)
752 /* tests if dev is a "standard" md dev name.
753 * i.e if the last component is "/dNN" or "/mdNN",
754 * where NN is a string of digits
755 * Returns 1 if a partitionable standard,
756 * -1 if non-partitonable,
757 * 0 if not a standard name.
759 char *d
= strrchr(dev
, '/');
764 if (strncmp(d
, "/d",2) == 0)
765 d
+= 2, type
= 1; /* /dev/md/dN{pM} */
766 else if (strncmp(d
, "/md_d", 5) == 0)
767 d
+= 5, type
= 1; /* /dev/md_dN{pM} */
768 else if (strncmp(d
, "/md", 3) == 0)
769 d
+= 3, type
= -1; /* /dev/mdN */
770 else if (d
-dev
> 3 && strncmp(d
-2, "md/", 3) == 0)
771 d
+= 1, type
= -1; /* /dev/md/N */
781 if (nump
) *nump
= num
;
786 unsigned long calc_csum(void *super
, int bytes
)
788 unsigned long long newcsum
= 0;
791 unsigned int *superc
= (unsigned int*) super
;
793 for(i
= 0; i
< bytes
/4; i
++)
794 newcsum
+= superc
[i
];
795 csum
= (newcsum
& 0xffffffff) + (newcsum
>>32);
797 /* The in-kernel checksum calculation is always 16bit on
798 * the alpha, though it is 32 bit on i386...
799 * I wonder what it is elsewhere... (it uses an API in
800 * a way that it shouldn't).
802 csum
= (csum
& 0xffff) + (csum
>> 16);
803 csum
= (csum
& 0xffff) + (csum
>> 16);
809 char *human_size(long long bytes
)
813 /* We convert bytes to either centi-M{ega,ibi}bytes or
814 * centi-G{igi,ibi}bytes, with appropriate rounding,
815 * and then print 1/100th of those as a decimal.
816 * We allow upto 2048Megabytes before converting to
817 * gigabytes, as that shows more precision and isn't
818 * too large a number.
819 * Terabytes are not yet handled.
822 if (bytes
< 5000*1024)
824 else if (bytes
< 2*1024LL*1024LL*1024LL) {
825 long cMiB
= (bytes
* 200LL / (1LL<<20) + 1) / 2;
826 long cMB
= (bytes
/ ( 1000000LL / 200LL ) +1) /2;
827 snprintf(buf
, sizeof(buf
), " (%ld.%02ld MiB %ld.%02ld MB)",
828 cMiB
/100 , cMiB
% 100,
831 long cGiB
= (bytes
* 200LL / (1LL<<30) +1) / 2;
832 long cGB
= (bytes
/ (1000000000LL/200LL ) +1) /2;
833 snprintf(buf
, sizeof(buf
), " (%ld.%02ld GiB %ld.%02ld GB)",
834 cGiB
/100 , cGiB
% 100,
840 char *human_size_brief(long long bytes
, int prefix
)
844 /* We convert bytes to either centi-M{ega,ibi}bytes or
845 * centi-G{igi,ibi}bytes, with appropriate rounding,
846 * and then print 1/100th of those as a decimal.
847 * We allow upto 2048Megabytes before converting to
848 * gigabytes, as that shows more precision and isn't
849 * too large a number.
850 * Terabytes are not yet handled.
852 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
853 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
856 if (bytes
< 5000*1024)
858 else if (prefix
== IEC
) {
859 if (bytes
< 2*1024LL*1024LL*1024LL) {
860 long cMiB
= (bytes
* 200LL / (1LL<<20) +1) /2;
861 snprintf(buf
, sizeof(buf
), "%ld.%02ldMiB",
862 cMiB
/100 , cMiB
% 100);
864 long cGiB
= (bytes
* 200LL / (1LL<<30) +1) /2;
865 snprintf(buf
, sizeof(buf
), "%ld.%02ldGiB",
866 cGiB
/100 , cGiB
% 100);
869 else if (prefix
== JEDEC
) {
870 if (bytes
< 2*1024LL*1024LL*1024LL) {
871 long cMB
= (bytes
/ ( 1000000LL / 200LL ) +1) /2;
872 snprintf(buf
, sizeof(buf
), "%ld.%02ldMB",
875 long cGB
= (bytes
/ (1000000000LL/200LL ) +1) /2;
876 snprintf(buf
, sizeof(buf
), "%ld.%02ldGB",
877 cGB
/100 , cGB
% 100);
886 void print_r10_layout(int layout
)
888 int near
= layout
& 255;
889 int far
= (layout
>> 8) & 255;
890 int offset
= (layout
&0x10000);
894 printf("%s near=%d", sep
, near
);
898 printf("%s %s=%d", sep
, offset
?"offset":"far", far
);
900 printf("NO REDUNDANCY");
904 unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
905 int chunksize
, unsigned long long devsize
)
909 devsize
&= ~(unsigned long long)((chunksize
>>9)-1);
910 return get_data_disks(level
, layout
, raid_disks
) * devsize
;
913 int get_data_disks(int level
, int layout
, int raid_disks
)
917 case 0: data_disks
= raid_disks
;
919 case 1: data_disks
= 1;
922 case 5: data_disks
= raid_disks
- 1;
924 case 6: data_disks
= raid_disks
- 2;
926 case 10: data_disks
= raid_disks
/ (layout
& 255) / ((layout
>>8)&255);
933 int devnm2devid(char *devnm
)
935 /* First look in /sys/block/$DEVNM/dev for %d:%d
936 * If that fails, try parsing out a number
943 sprintf(path
, "/sys/block/%s/dev", devnm
);
944 fd
= open(path
, O_RDONLY
);
947 int n
= read(fd
, buf
, sizeof(buf
));
951 if (n
> 0 && sscanf(buf
, "%d:%d\n", &mjr
, &mnr
) == 2)
952 return makedev(mjr
, mnr
);
954 if (strncmp(devnm
, "md_d", 4) == 0 &&
956 (mnr
= strtoul(devnm
+4, &ep
, 10)) >= 0 &&
957 ep
> devnm
&& *ep
== 0)
958 return makedev(get_mdp_major(), mnr
<< MdpMinorShift
);
960 if (strncmp(devnm
, "md", 2) == 0 &&
962 (mnr
= strtoul(devnm
+2, &ep
, 10)) >= 0 &&
963 ep
> devnm
&& *ep
== 0)
964 return makedev(MD_MAJOR
, mnr
);
969 #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
970 char *get_md_name(char *devnm
)
972 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
973 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
975 static char devname
[50];
977 dev_t rdev
= devnm2devid(devnm
);
982 if (strncmp(devnm
, "md_", 3) == 0) {
983 snprintf(devname
, sizeof(devname
), "/dev/md/%s",
985 if (stat(devname
, &stb
) == 0
986 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
987 && (stb
.st_rdev
== rdev
))
990 snprintf(devname
, sizeof(devname
), "/dev/%s", devnm
);
991 if (stat(devname
, &stb
) == 0
992 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
993 && (stb
.st_rdev
== rdev
))
996 snprintf(devname
, sizeof(devname
), "/dev/md/%s", devnm
+2);
997 if (stat(devname
, &stb
) == 0
998 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
999 && (stb
.st_rdev
== rdev
))
1002 dn
= map_dev(major(rdev
), minor(rdev
), 0);
1005 snprintf(devname
, sizeof(devname
), "/dev/.tmp.%s", devnm
);
1006 if (mknod(devname
, S_IFBLK
| 0600, rdev
) == -1)
1007 if (errno
!= EEXIST
)
1010 if (stat(devname
, &stb
) == 0
1011 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
1012 && (stb
.st_rdev
== rdev
))
1018 void put_md_name(char *name
)
1020 if (strncmp(name
, "/dev/.tmp.md", 12) == 0)
1023 #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1025 int get_maj_min(char *dev
, int *major
, int *minor
)
1028 *major
= strtoul(dev
, &e
, 0);
1029 return (e
> dev
&& *e
== ':' && e
[1] &&
1030 (*minor
= strtoul(e
+1, &e
, 0)) >= 0 &&
1034 int dev_open(char *dev
, int flags
)
1036 /* like 'open', but if 'dev' matches %d:%d, create a temp
1037 * block device and open that
1044 if (!dev
) return -1;
1047 if (get_maj_min(dev
, &major
, &minor
)) {
1048 snprintf(devname
, sizeof(devname
), "/dev/.tmp.md.%d:%d:%d",
1049 (int)getpid(), major
, minor
);
1050 if (mknod(devname
, S_IFBLK
|0600, makedev(major
, minor
)) == 0) {
1051 fd
= open(devname
, flags
);
1055 /* Try /tmp as /dev appear to be read-only */
1056 snprintf(devname
, sizeof(devname
), "/tmp/.tmp.md.%d:%d:%d",
1057 (int)getpid(), major
, minor
);
1058 if (mknod(devname
, S_IFBLK
|0600, makedev(major
, minor
)) == 0) {
1059 fd
= open(devname
, flags
);
1064 fd
= open(dev
, flags
);
1068 int open_dev_flags(char *devnm
, int flags
)
1073 devid
= devnm2devid(devnm
);
1074 sprintf(buf
, "%d:%d", major(devid
), minor(devid
));
1075 return dev_open(buf
, flags
);
1078 int open_dev(char *devnm
)
1080 return open_dev_flags(devnm
, O_RDONLY
);
1083 int open_dev_excl(char *devnm
)
1088 int devid
= devnm2devid(devnm
);
1091 sprintf(buf
, "%d:%d", major(devid
), minor(devid
));
1092 for (i
= 0 ; i
< 25 ; i
++) {
1093 int fd
= dev_open(buf
, flags
|O_EXCL
);
1096 if (errno
== EACCES
&& flags
== O_RDWR
) {
1109 int same_dev(char *one
, char *two
)
1111 struct stat st1
, st2
;
1112 if (stat(one
, &st1
) != 0)
1114 if (stat(two
, &st2
) != 0)
1116 if ((st1
.st_mode
& S_IFMT
) != S_IFBLK
)
1118 if ((st2
.st_mode
& S_IFMT
) != S_IFBLK
)
1120 return st1
.st_rdev
== st2
.st_rdev
;
1123 void wait_for(char *dev
, int fd
)
1126 struct stat stb_want
;
1129 if (fstat(fd
, &stb_want
) != 0 ||
1130 (stb_want
.st_mode
& S_IFMT
) != S_IFBLK
)
1133 for (i
= 0 ; i
< 25 ; i
++) {
1135 if (stat(dev
, &stb
) == 0 &&
1136 (stb
.st_mode
& S_IFMT
) == S_IFBLK
&&
1137 (stb
.st_rdev
== stb_want
.st_rdev
))
1144 dprintf("timeout waiting for %s\n", dev
);
1147 struct superswitch
*superlist
[] =
1150 &super_ddf
, &super_imsm
,
1154 #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
1156 struct supertype
*super_by_fd(int fd
, char **subarrayp
)
1158 mdu_array_info_t array
;
1161 struct supertype
*st
= NULL
;
1166 char *subarray
= NULL
;
1167 char container
[32] = "";
1169 sra
= sysfs_read(fd
, NULL
, GET_VERSION
);
1172 vers
= sra
->array
.major_version
;
1173 minor
= sra
->array
.minor_version
;
1174 verstr
= sra
->text_version
;
1176 if (ioctl(fd
, GET_ARRAY_INFO
, &array
))
1177 array
.major_version
= array
.minor_version
= 0;
1178 vers
= array
.major_version
;
1179 minor
= array
.minor_version
;
1184 sprintf(version
, "%d.%d", vers
, minor
);
1187 if (minor
== -2 && is_subarray(verstr
)) {
1188 char *dev
= verstr
+1;
1190 subarray
= strchr(dev
, '/');
1193 subarray
= xstrdup(subarray
);
1195 strcpy(container
, dev
);
1198 sra
= sysfs_read(-1, container
, GET_VERSION
);
1199 if (sra
&& sra
->text_version
[0])
1200 verstr
= sra
->text_version
;
1202 verstr
= "-no-metadata-";
1205 for (i
= 0; st
== NULL
&& superlist
[i
] ; i
++)
1206 st
= superlist
[i
]->match_metadata_desc(verstr
);
1213 *subarrayp
= subarray
;
1214 strcpy(st
->container_devnm
, container
);
1215 strcpy(st
->devnm
, fd2devnm(fd
));
1221 #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1223 int dev_size_from_id(dev_t id
, unsigned long long *size
)
1228 sprintf(buf
, "%d:%d", major(id
), minor(id
));
1229 fd
= dev_open(buf
, O_RDONLY
);
1232 if (get_dev_size(fd
, NULL
, size
)) {
1240 struct supertype
*dup_super(struct supertype
*orig
)
1242 struct supertype
*st
;
1246 st
= xcalloc(1, sizeof(*st
));
1248 st
->max_devs
= orig
->max_devs
;
1249 st
->minor_version
= orig
->minor_version
;
1250 st
->ignore_hw_compat
= orig
->ignore_hw_compat
;
1251 st
->data_offset
= orig
->data_offset
;
1257 struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
)
1259 /* try each load_super to find the best match,
1260 * and return the best superswitch
1262 struct superswitch
*ss
;
1263 struct supertype
*st
;
1264 time_t besttime
= 0;
1268 st
= xcalloc(1, sizeof(*st
));
1269 st
->container_devnm
[0] = 0;
1271 for (i
= 0 ; superlist
[i
]; i
++) {
1274 if (guess_type
== guess_array
&& ss
->add_to_super
== NULL
)
1276 if (guess_type
== guess_partitions
&& ss
->add_to_super
!= NULL
)
1278 memset(st
, 0, sizeof(*st
));
1279 st
->ignore_hw_compat
= 1;
1280 rv
= ss
->load_super(st
, fd
, NULL
);
1283 st
->ss
->getinfo_super(st
, &info
, NULL
);
1284 if (bestsuper
== -1 ||
1285 besttime
< info
.array
.ctime
) {
1287 besttime
= info
.array
.ctime
;
1292 if (bestsuper
!= -1) {
1294 memset(st
, 0, sizeof(*st
));
1295 st
->ignore_hw_compat
= 1;
1296 rv
= superlist
[bestsuper
]->load_super(st
, fd
, NULL
);
1298 superlist
[bestsuper
]->free_super(st
);
1306 /* Return size of device in bytes */
1307 int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
)
1309 unsigned long long ldsize
;
1312 if (fstat(fd
, &st
) != -1 && S_ISREG(st
.st_mode
))
1313 ldsize
= (unsigned long long)st
.st_size
;
1316 if (ioctl(fd
, BLKGETSIZE64
, &ldsize
) != 0)
1319 unsigned long dsize
;
1320 if (ioctl(fd
, BLKGETSIZE
, &dsize
) == 0) {
1325 pr_err("Cannot get size of %s: %s\b",
1326 dname
, strerror(errno
));
1334 /* Return true if this can only be a container, not a member device.
1335 * i.e. is and md device and size is zero
1337 int must_be_container(int fd
)
1339 unsigned long long size
;
1340 if (md_get_version(fd
) < 0)
1342 if (get_dev_size(fd
, NULL
, &size
) == 0)
1349 /* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1350 * Returns: 1 if successful
1351 * -1 for unknown partition type
1352 * 0 for other errors
1354 static int get_gpt_last_partition_end(int fd
, unsigned long long *endofpart
)
1357 unsigned char empty_gpt_entry
[16]= {0};
1358 struct GPT_part_entry
*part
;
1360 unsigned long long curr_part_end
;
1361 unsigned all_partitions
, entry_size
;
1366 BUILD_BUG_ON(sizeof(gpt
) != 512);
1367 /* skip protective MBR */
1368 lseek(fd
, 512, SEEK_SET
);
1369 /* read GPT header */
1370 if (read(fd
, &gpt
, 512) != 512)
1373 /* get the number of partition entries and the entry size */
1374 all_partitions
= __le32_to_cpu(gpt
.part_cnt
);
1375 entry_size
= __le32_to_cpu(gpt
.part_size
);
1377 /* Check GPT signature*/
1378 if (gpt
.magic
!= GPT_SIGNATURE_MAGIC
)
1382 if (all_partitions
> 1024 ||
1383 entry_size
> sizeof(buf
))
1386 part
= (struct GPT_part_entry
*)buf
;
1388 for (part_nr
= 0; part_nr
< all_partitions
; part_nr
++) {
1389 /* read partition entry */
1390 if (read(fd
, buf
, entry_size
) != (ssize_t
)entry_size
)
1393 /* is this valid partition? */
1394 if (memcmp(part
->type_guid
, empty_gpt_entry
, 16) != 0) {
1395 /* check the last lba for the current partition */
1396 curr_part_end
= __le64_to_cpu(part
->ending_lba
);
1397 if (curr_part_end
> *endofpart
)
1398 *endofpart
= curr_part_end
;
1405 /* Sets endofpart parameter to the last block used by the last partition on the device.
1406 * Returns: 1 if successful
1407 * -1 for unknown partition type
1408 * 0 for other errors
1410 static int get_last_partition_end(int fd
, unsigned long long *endofpart
)
1412 struct MBR boot_sect
;
1413 struct MBR_part_record
*part
;
1414 unsigned long long curr_part_end
;
1420 BUILD_BUG_ON(sizeof(boot_sect
) != 512);
1423 if (read(fd
, &boot_sect
, 512) != 512)
1426 /* check MBP signature */
1427 if (boot_sect
.magic
== MBR_SIGNATURE_MAGIC
) {
1429 /* found the correct signature */
1430 part
= boot_sect
.parts
;
1432 for (part_nr
= 0; part_nr
< MBR_PARTITIONS
; part_nr
++) {
1433 /* check for GPT type */
1434 if (part
->part_type
== MBR_GPT_PARTITION_TYPE
) {
1435 retval
= get_gpt_last_partition_end(fd
, endofpart
);
1438 /* check the last used lba for the current partition */
1439 curr_part_end
= __le32_to_cpu(part
->first_sect_lba
) +
1440 __le32_to_cpu(part
->blocks_num
);
1441 if (curr_part_end
> *endofpart
)
1442 *endofpart
= curr_part_end
;
1447 /* Unknown partition table */
1454 int check_partitions(int fd
, char *dname
, unsigned long long freesize
,
1455 unsigned long long size
)
1458 * Check where the last partition ends
1460 unsigned long long endofpart
;
1463 if ((ret
= get_last_partition_end(fd
, &endofpart
)) > 0) {
1464 /* There appears to be a partition table here */
1465 if (freesize
== 0) {
1466 /* partitions will not be visible in new device */
1467 pr_err("partition table exists on %s but will be lost or\n"
1468 " meaningless after creating array\n",
1471 } else if (endofpart
> freesize
) {
1472 /* last partition overlaps metadata */
1473 pr_err("metadata will over-write last partition on %s.\n",
1476 } else if (size
&& endofpart
> size
) {
1477 /* partitions will be truncated in new device */
1478 pr_err("array size is too small to cover all partitions on %s.\n",
1486 int open_container(int fd
)
1488 /* 'fd' is a block device. Find out if it is in use
1489 * by a container, and return an open fd on that container.
1500 if (fstat(fd
, &st
) != 0)
1502 sprintf(path
, "/sys/dev/block/%d:%d/holders",
1503 (int)major(st
.st_rdev
), (int)minor(st
.st_rdev
));
1504 e
= path
+ strlen(path
);
1506 dir
= opendir(path
);
1509 while ((de
= readdir(dir
))) {
1512 if (de
->d_name
[0] == '.')
1514 /* Need to make sure it is a container and not a volume */
1515 sprintf(e
, "/%s/md/metadata_version", de
->d_name
);
1516 dfd
= open(path
, O_RDONLY
);
1519 n
= read(dfd
, buf
, sizeof(buf
));
1521 if (n
<= 0 || (unsigned)n
>= sizeof(buf
))
1524 if (strncmp(buf
, "external", 8) != 0 ||
1528 sprintf(e
, "/%s/dev", de
->d_name
);
1529 dfd
= open(path
, O_RDONLY
);
1532 n
= read(dfd
, buf
, sizeof(buf
));
1534 if (n
<= 0 || (unsigned)n
>= sizeof(buf
))
1537 if (sscanf(buf
, "%d:%d", &major
, &minor
) != 2)
1539 sprintf(buf
, "%d:%d", major
, minor
);
1540 dfd
= dev_open(buf
, O_RDONLY
);
1550 struct superswitch
*version_to_superswitch(char *vers
)
1554 for (i
= 0; superlist
[i
]; i
++) {
1555 struct superswitch
*ss
= superlist
[i
];
1557 if (strcmp(vers
, ss
->name
) == 0)
1564 int metadata_container_matches(char *metadata
, char *devnm
)
1566 /* Check if 'devnm' is the container named in 'metadata'
1568 * /containername/componentname or
1569 * -containername/componentname
1572 if (*metadata
!= '/' && *metadata
!= '-')
1575 if (strncmp(metadata
+1, devnm
, l
) != 0)
1577 if (metadata
[l
+1] != '/')
1582 int metadata_subdev_matches(char *metadata
, char *devnm
)
1584 /* Check if 'devnm' is the subdev named in 'metadata'
1586 * /containername/subdev or
1587 * -containername/subdev
1590 if (*metadata
!= '/' && *metadata
!= '-')
1592 sl
= strchr(metadata
+1, '/');
1595 if (strcmp(sl
+1, devnm
) == 0)
1600 int is_container_member(struct mdstat_ent
*mdstat
, char *container
)
1602 if (mdstat
->metadata_version
== NULL
||
1603 strncmp(mdstat
->metadata_version
, "external:", 9) != 0 ||
1604 !metadata_container_matches(mdstat
->metadata_version
+9, container
))
1610 int is_subarray_active(char *subarray
, char *container
)
1612 struct mdstat_ent
*mdstat
= mdstat_read(0, 0);
1613 struct mdstat_ent
*ent
;
1615 for (ent
= mdstat
; ent
; ent
= ent
->next
)
1616 if (is_container_member(ent
, container
))
1617 if (strcmp(to_subarray(ent
, container
), subarray
) == 0)
1620 free_mdstat(mdstat
);
1625 /* open_subarray - opens a subarray in a container
1626 * @dev: container device name
1627 * @st: empty supertype
1628 * @quiet: block reporting errors flag
1630 * On success returns an fd to a container and fills in *st
1632 int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
)
1635 struct mdinfo
*info
;
1639 fd
= open(dev
, O_RDWR
|O_EXCL
);
1642 pr_err("Couldn't open %s, aborting\n",
1647 _devnm
= fd2devnm(fd
);
1648 if (_devnm
== NULL
) {
1650 pr_err("Failed to determine device number for %s\n",
1654 strcpy(st
->devnm
, _devnm
);
1656 mdi
= sysfs_read(fd
, st
->devnm
, GET_VERSION
|GET_LEVEL
);
1659 pr_err("Failed to read sysfs for %s\n",
1664 if (mdi
->array
.level
!= UnSet
) {
1666 pr_err("%s is not a container\n", dev
);
1670 st
->ss
= version_to_superswitch(mdi
->text_version
);
1673 pr_err("Operation not supported for %s metadata\n",
1678 if (st
->devnm
[0] == 0) {
1680 pr_err("Failed to allocate device name\n");
1684 if (!st
->ss
->load_container
) {
1686 pr_err("%s is not a container\n", dev
);
1690 if (st
->ss
->load_container(st
, fd
, NULL
)) {
1692 pr_err("Failed to load metadata for %s\n",
1697 info
= st
->ss
->container_content(st
, subarray
);
1700 pr_err("Failed to find subarray-%s in %s\n",
1710 st
->ss
->free_super(st
);
1723 int add_disk(int mdfd
, struct supertype
*st
,
1724 struct mdinfo
*sra
, struct mdinfo
*info
)
1726 /* Add a device to an array, in one of 2 ways. */
1729 if (st
->ss
->external
) {
1730 if (info
->disk
.state
& (1<<MD_DISK_SYNC
))
1731 info
->recovery_start
= MaxSector
;
1733 info
->recovery_start
= 0;
1734 rv
= sysfs_add_disk(sra
, info
, 0);
1737 for (sd2
= sra
->devs
; sd2
; sd2
=sd2
->next
)
1741 sd2
= xmalloc(sizeof(*sd2
));
1743 sd2
->next
= sra
->devs
;
1749 rv
= ioctl(mdfd
, ADD_NEW_DISK
, &info
->disk
);
1753 int remove_disk(int mdfd
, struct supertype
*st
,
1754 struct mdinfo
*sra
, struct mdinfo
*info
)
1757 /* Remove the disk given by 'info' from the array */
1759 if (st
->ss
->external
)
1760 rv
= sysfs_set_str(sra
, info
, "slot", "none");
1763 rv
= ioctl(mdfd
, HOT_REMOVE_DISK
, makedev(info
->disk
.major
,
1768 int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
)
1770 /* Initialise kernel's knowledge of array.
1771 * This varies between externally managed arrays
1774 int vers
= md_get_version(mdfd
);
1778 if (st
->ss
->external
)
1779 rv
= sysfs_set_array(info
, vers
);
1782 if ((vers
% 100) >= 1) { /* can use different versions */
1783 mdu_array_info_t inf
;
1784 memset(&inf
, 0, sizeof(inf
));
1785 inf
.major_version
= info
->array
.major_version
;
1786 inf
.minor_version
= info
->array
.minor_version
;
1787 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, &inf
);
1789 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, NULL
);
1793 unsigned long long min_recovery_start(struct mdinfo
*array
)
1795 /* find the minimum recovery_start in an array for metadata
1796 * formats that only record per-array recovery progress instead
1799 unsigned long long recovery_start
= MaxSector
;
1802 for (d
= array
->devs
; d
; d
= d
->next
)
1803 recovery_start
= min(recovery_start
, d
->recovery_start
);
1805 return recovery_start
;
1808 int mdmon_pid(char *devnm
)
1815 sprintf(path
, "%s/%s.pid", MDMON_DIR
, devnm
);
1817 fd
= open(path
, O_RDONLY
| O_NOATIME
, 0);
1821 n
= read(fd
, pid
, 9);
1828 int mdmon_running(char *devnm
)
1830 int pid
= mdmon_pid(devnm
);
1833 if (kill(pid
, 0) == 0)
1838 int start_mdmon(char *devnm
)
1852 if (check_env("MDADM_NO_MDMON"))
1855 len
= readlink("/proc/self/exe", pathbuf
, sizeof(pathbuf
)-1);
1859 sl
= strrchr(pathbuf
, '/');
1864 strcpy(sl
, "mdmon");
1868 /* First try to run systemctl */
1869 if (!check_env("MDADM_NO_SYSTEMCTL"))
1872 /* FIXME yuk. CLOSE_EXEC?? */
1874 for (i
= 3; skipped
< 20; i
++)
1880 /* Don't want to see error messages from
1881 * systemctl. If the service doesn't exist,
1882 * we start mdmon ourselves.
1885 open("/dev/null", O_WRONLY
);
1886 snprintf(pathbuf
, sizeof(pathbuf
), "mdmon@%s.service",
1888 status
= execl("/usr/bin/systemctl", "systemctl",
1891 status
= execl("/bin/systemctl", "systemctl", "start",
1894 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
1896 default: /* parent - good */
1897 pid
= wait(&status
);
1898 if (pid
>= 0 && status
== 0)
1902 /* That failed, try running mdmon directly */
1905 /* FIXME yuk. CLOSE_EXEC?? */
1907 for (i
= 3; skipped
< 20; i
++)
1913 for (i
= 0; paths
[i
]; i
++)
1915 execl(paths
[i
], paths
[i
],
1919 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
1921 default: /* parent - good */
1922 pid
= wait(&status
);
1923 if (pid
< 0 || status
!= 0) {
1924 pr_err("failed to launch mdmon. Array remains readonly\n");
1931 __u32
random32(void)
1934 int rfd
= open("/dev/urandom", O_RDONLY
);
1935 if (rfd
< 0 || read(rfd
, &rv
, 4) != 4)
1943 int flush_metadata_updates(struct supertype
*st
)
1947 st
->update_tail
= NULL
;
1951 sfd
= connect_monitor(st
->container_devnm
);
1955 while (st
->updates
) {
1956 struct metadata_update
*mu
= st
->updates
;
1957 st
->updates
= mu
->next
;
1959 send_message(sfd
, mu
, 0);
1967 st
->update_tail
= NULL
;
1971 void append_metadata_update(struct supertype
*st
, void *buf
, int len
)
1974 struct metadata_update
*mu
= xmalloc(sizeof(*mu
));
1979 mu
->space_list
= NULL
;
1981 *st
->update_tail
= mu
;
1982 st
->update_tail
= &mu
->next
;
1984 #endif /* MDASSEMBLE */
1987 /* tinyc doesn't optimize this check in ioctl.h out ... */
1988 unsigned int __invalid_size_argument_for_IOC
= 0;
1991 int experimental(void)
1993 if (check_env("MDADM_EXPERIMENTAL"))
1996 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
2001 /* Pick all spares matching given criteria from a container
2002 * if min_size == 0 do not check size
2003 * if domlist == NULL do not check domains
2004 * if spare_group given add it to domains of each spare
2005 * metadata allows to test domains using metadata of destination array */
2006 struct mdinfo
*container_choose_spares(struct supertype
*st
,
2007 unsigned long long min_size
,
2008 struct domainlist
*domlist
,
2010 const char *metadata
, int get_one
)
2012 struct mdinfo
*d
, **dp
, *disks
= NULL
;
2014 /* get list of all disks in container */
2015 if (st
->ss
->getinfo_super_disks
)
2016 disks
= st
->ss
->getinfo_super_disks(st
);
2020 /* find spare devices on the list */
2022 disks
->array
.spare_disks
= 0;
2026 if (d
->disk
.state
== 0) {
2027 /* check if size is acceptable */
2028 unsigned long long dev_size
;
2029 dev_t dev
= makedev(d
->disk
.major
,d
->disk
.minor
);
2032 (dev_size_from_id(dev
, &dev_size
) &&
2033 dev_size
>= min_size
))
2035 /* check if domain matches */
2036 if (found
&& domlist
) {
2037 struct dev_policy
*pol
= devid_policy(dev
);
2039 pol_add(&pol
, pol_domain
,
2041 if (domain_test(domlist
, pol
, metadata
) != 1)
2043 dev_policy_free(pol
);
2048 disks
->array
.spare_disks
++;
2062 /* Checks if paths point to the same device
2063 * Returns 0 if they do.
2064 * Returns 1 if they don't.
2065 * Returns -1 if something went wrong,
2066 * e.g. paths are empty or the files
2067 * they point to don't exist */
2068 int compare_paths (char* path1
, char* path2
)
2070 struct stat st1
,st2
;
2072 if (path1
== NULL
|| path2
== NULL
)
2074 if (stat(path1
,&st1
) != 0)
2076 if (stat(path2
,&st2
) != 0)
2078 if ((st1
.st_ino
== st2
.st_ino
) && (st1
.st_dev
== st2
.st_dev
))
2083 /* Make sure we can open as many devices as needed */
2084 void enable_fds(int devices
)
2086 unsigned int fds
= 20 + devices
;
2088 if (getrlimit(RLIMIT_NOFILE
, &lim
) != 0
2089 || lim
.rlim_cur
>= fds
)
2091 if (lim
.rlim_max
< fds
)
2094 setrlimit(RLIMIT_NOFILE
, &lim
);
2099 /* This is based on similar function in systemd. */
2101 /* statfs.f_type is signed long on s390x and MIPS, causing all
2102 sorts of sign extension problems with RAMFS_MAGIC being
2103 defined as 0x858458f6 */
2104 return statfs("/", &s
) >= 0 &&
2105 ((unsigned long)s
.f_type
== TMPFS_MAGIC
||
2106 ((unsigned long)s
.f_type
& 0xFFFFFFFFUL
) ==
2107 ((unsigned long)RAMFS_MAGIC
& 0xFFFFFFFFUL
));
2110 void reopen_mddev(int mdfd
)
2112 /* Re-open without any O_EXCL, but keep
2117 devnm
= fd2devnm(mdfd
);
2119 fd
= open_dev(devnm
);
2120 if (fd
>= 0 && fd
!= mdfd
)
2124 static struct cmap_hooks
*cmap_hooks
= NULL
;
2125 static int is_cmap_hooks_ready
= 0;
2128 void set_cmap_hooks(void)
2130 cmap_hooks
= xmalloc(sizeof(struct cmap_hooks
));
2131 cmap_hooks
->cmap_handle
= dlopen("libcmap.so.4", RTLD_NOW
| RTLD_LOCAL
);
2132 if (!cmap_hooks
->cmap_handle
)
2135 cmap_hooks
->initialize
= dlsym(cmap_hooks
->cmap_handle
, "cmap_initialize");
2136 cmap_hooks
->get_string
= dlsym(cmap_hooks
->cmap_handle
, "cmap_get_string");
2137 cmap_hooks
->finalize
= dlsym(cmap_hooks
->cmap_handle
, "cmap_finalize");
2139 if (!cmap_hooks
->initialize
|| !cmap_hooks
->get_string
||
2140 !cmap_hooks
->finalize
)
2141 dlclose(cmap_hooks
->cmap_handle
);
2143 is_cmap_hooks_ready
= 1;
2146 int get_cluster_name(char **cluster_name
)
2149 cmap_handle_t handle
;
2151 if (!is_cmap_hooks_ready
)
2154 rv
= cmap_hooks
->initialize(&handle
);
2158 rv
= cmap_hooks
->get_string(handle
, "totem.cluster_name", cluster_name
);
2160 free(*cluster_name
);
2167 cmap_hooks
->finalize(handle
);
2172 void set_dlm_hooks(void)
2174 dlm_hooks
= xmalloc(sizeof(struct dlm_hooks
));
2175 dlm_hooks
->dlm_handle
= dlopen("libdlm_lt.so.3", RTLD_NOW
| RTLD_LOCAL
);
2176 if (!dlm_hooks
->dlm_handle
)
2179 dlm_hooks
->create_lockspace
= dlsym(dlm_hooks
->dlm_handle
, "dlm_create_lockspace");
2180 dlm_hooks
->release_lockspace
= dlsym(dlm_hooks
->dlm_handle
, "dlm_release_lockspace");
2181 dlm_hooks
->ls_lock
= dlsym(dlm_hooks
->dlm_handle
, "dlm_ls_lock");
2182 dlm_hooks
->ls_unlock
= dlsym(dlm_hooks
->dlm_handle
, "dlm_ls_unlock");
2183 dlm_hooks
->ls_get_fd
= dlsym(dlm_hooks
->dlm_handle
, "dlm_ls_get_fd");
2184 dlm_hooks
->dispatch
= dlsym(dlm_hooks
->dlm_handle
, "dlm_dispatch");
2186 if (!dlm_hooks
->create_lockspace
|| !dlm_hooks
->ls_lock
||
2187 !dlm_hooks
->ls_unlock
|| !dlm_hooks
->release_lockspace
||
2188 !dlm_hooks
->ls_get_fd
|| !dlm_hooks
->dispatch
)
2189 dlclose(dlm_hooks
->dlm_handle
);
2191 is_dlm_hooks_ready
= 1;
2194 void set_hooks(void)