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>
27 #include <sys/socket.h>
28 #include <sys/utsname.h>
36 * following taken from linux/blkpg.h because they aren't
37 * anywhere else and it isn't safe to #include linux/ * stuff.
40 #define BLKPG _IO(0x12,105)
42 /* The argument structure */
43 struct blkpg_ioctl_arg
{
50 /* The subfunctions (for the op field) */
51 #define BLKPG_ADD_PARTITION 1
52 #define BLKPG_DEL_PARTITION 2
54 /* Sizes of name fields. Unused at present. */
55 #define BLKPG_DEVNAMELTH 64
56 #define BLKPG_VOLNAMELTH 64
58 /* The data structure for ADD_PARTITION and DEL_PARTITION */
59 struct blkpg_partition
{
60 long long start
; /* starting offset in bytes */
61 long long length
; /* length in bytes */
62 int pno
; /* partition number */
63 char devname
[BLKPG_DEVNAMELTH
]; /* partition name, like sda5 or c0d1p2,
64 to be used in kernel messages */
65 char volname
[BLKPG_VOLNAMELTH
]; /* volume label */
69 * Parse a 128 bit uuid in 4 integers
70 * format is 32 hexx nibbles with options :.<space> separator
71 * If not exactly 32 hex digits are found, return 0
74 int parse_uuid(char *str
, int uuid
[4])
76 int hit
= 0; /* number of Hex digIT */
79 for (i
=0; i
<4; i
++) uuid
[i
]=0;
85 else if (c
>='a' && c
<= 'f')
87 else if (c
>='A' && c
<= 'F')
89 else if (strchr(":. -", c
))
106 * Get the md version number.
107 * We use the RAID_VERSION ioctl if it is supported
108 * If not, but we have a block device with major '9', we assume
111 * Return version number as 24 but number - assume version parts
115 int md_get_version(int fd
)
120 if (fstat(fd
, &stb
)<0)
122 if ((S_IFMT
&stb
.st_mode
) != S_IFBLK
)
125 if (ioctl(fd
, RAID_VERSION
, &vers
) == 0)
126 return (vers
.major
*10000) + (vers
.minor
*100) + vers
.patchlevel
;
129 if (major(stb
.st_rdev
) == MD_MAJOR
)
134 int get_linux_version()
143 a
= strtoul(cp
, &cp
, 10);
144 if (*cp
!= '.') return -1;
145 b
= strtoul(cp
+1, &cp
, 10);
146 if (*cp
!= '.') return -1;
147 c
= strtoul(cp
+1, NULL
, 10);
149 return (a
*1000000)+(b
*1000)+c
;
153 long long parse_size(char *size
)
155 /* parse 'size' which should be a number optionally
156 * followed by 'K', 'M', or 'G'.
157 * Without a suffix, K is assumed.
158 * Number returned is in sectors (half-K)
161 long long s
= strtoll(size
, &c
, 10);
175 s
*= 1024 * 1024 * 2;
184 int parse_layout_10(char *layout
)
188 /* Parse the layout string for raid10 */
189 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
190 if ((layout
[0] != 'n' && layout
[0] != 'f' && layout
[0] != 'o') ||
191 (copies
= strtoul(layout
+1, &cp
, 10)) < 1 ||
195 if (layout
[0] == 'n')
197 else if (layout
[0] == 'o')
198 rv
= 0x10000 + (copies
<<8) + 1;
200 rv
= 1 + (copies
<<8);
204 int parse_layout_faulty(char *layout
)
206 /* Parse the layout string for 'faulty' */
207 int ln
= strcspn(layout
, "0123456789");
208 char *m
= strdup(layout
);
211 mode
= map_name(faultylayout
, m
);
215 return mode
| (atoi(layout
+ln
)<< ModeShift
);
219 void remove_partitions(int fd
)
221 /* remove partitions from this block devices.
222 * This is used for components added to an array
224 #ifdef BLKPG_DEL_PARTITION
225 struct blkpg_ioctl_arg a
;
226 struct blkpg_partition p
;
228 a
.op
= BLKPG_DEL_PARTITION
;
230 a
.datalen
= sizeof(p
);
232 memset(a
.data
, 0, a
.datalen
);
233 for (p
.pno
=0; p
.pno
< 16; p
.pno
++)
234 ioctl(fd
, BLKPG
, &a
);
238 int enough(int level
, int raid_disks
, int layout
, int clean
,
239 char *avail
, int avail_disks
)
244 /* This is the tricky one - we need to check
245 * which actual disks are present.
247 copies
= (layout
&255)* ((layout
>>8) & 255);
250 /* there must be one of the 'copies' form 'first' */
256 first
= (first
+1) % raid_disks
;
261 } while (first
!= 0);
264 case LEVEL_MULTIPATH
:
265 return avail_disks
>= 1;
268 return avail_disks
== raid_disks
;
270 return avail_disks
>= 1;
274 return avail_disks
>= raid_disks
-1;
276 return avail_disks
>= raid_disks
;
279 return avail_disks
>= raid_disks
-2;
281 return avail_disks
>= raid_disks
;
287 const int uuid_match_any
[4] = { ~0, ~0, ~0, ~0 };
288 int same_uuid(int a
[4], int b
[4], int swapuuid
)
290 if (memcmp(a
, uuid_match_any
, sizeof(int[4])) == 0 ||
291 memcmp(b
, uuid_match_any
, sizeof(int[4])) == 0)
295 /* parse uuids are hostendian.
296 * uuid's from some superblocks are big-ending
297 * if there is a difference, we need to swap..
299 unsigned char *ac
= (unsigned char *)a
;
300 unsigned char *bc
= (unsigned char *)b
;
302 for (i
=0; i
<16; i
+= 4) {
303 if (ac
[i
+0] != bc
[i
+3] ||
304 ac
[i
+1] != bc
[i
+2] ||
305 ac
[i
+2] != bc
[i
+1] ||
319 void copy_uuid(void *a
, int b
[4], int swapuuid
)
322 /* parse uuids are hostendian.
323 * uuid's from some superblocks are big-ending
324 * if there is a difference, we need to swap..
326 unsigned char *ac
= (unsigned char *)a
;
327 unsigned char *bc
= (unsigned char *)b
;
329 for (i
=0; i
<16; i
+= 4) {
339 char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
)
346 copy_uuid(uuid
, id
, swap
);
347 for (i
= 0; i
< 4; i
++) {
350 for (j
= 3; j
>= 0; j
--) {
351 sprintf(c
,"%02x", (unsigned char) uuid
[j
+4*i
]);
359 char *fname_from_uuid(struct supertype
*st
, struct mdinfo
*info
, char *buf
, char sep
)
361 return __fname_from_uuid(info
->uuid
, st
->ss
->swapuuid
, buf
, sep
);
365 int check_ext2(int fd
, char *name
)
368 * Check for an ext2fs file system.
369 * Superblock is always 1K at 1K offset
371 * s_magic is le16 at 56 == 0xEF53
372 * report mtime - le32 at 44
374 * logblksize - le32 at 24
376 unsigned char sb
[1024];
379 if (lseek(fd
, 1024,0)!= 1024)
381 if (read(fd
, sb
, 1024)!= 1024)
383 if (sb
[56] != 0x53 || sb
[57] != 0xef)
386 mtime
= sb
[44]|(sb
[45]|(sb
[46]|sb
[47]<<8)<<8)<<8;
387 bsize
= sb
[24]|(sb
[25]|(sb
[26]|sb
[27]<<8)<<8)<<8;
388 size
= sb
[4]|(sb
[5]|(sb
[6]|sb
[7]<<8)<<8)<<8;
389 fprintf(stderr
, Name
": %s appears to contain an ext2fs file system\n",
391 fprintf(stderr
," size=%dK mtime=%s",
392 size
*(1<<bsize
), ctime(&mtime
));
396 int check_reiser(int fd
, char *name
)
399 * superblock is at 64K
401 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
404 unsigned char sb
[1024];
406 if (lseek(fd
, 64*1024, 0) != 64*1024)
408 if (read(fd
, sb
, 1024) != 1024)
410 if (strncmp((char*)sb
+52, "ReIsErFs",8)!=0 &&
411 strncmp((char*)sb
+52, "ReIsEr2Fs",9)!=0)
413 fprintf(stderr
, Name
": %s appears to contain a reiserfs file system\n",name
);
414 size
= sb
[0]|(sb
[1]|(sb
[2]|sb
[3]<<8)<<8)<<8;
415 fprintf(stderr
, " size = %luK\n", size
*4);
420 int check_raid(int fd
, char *name
)
425 struct supertype
*st
= guess_super(fd
);
428 st
->ss
->load_super(st
, fd
, name
);
429 /* Looks like a raid array .. */
430 fprintf(stderr
, Name
": %s appears to be part of a raid array:\n",
432 st
->ss
->getinfo_super(st
, &info
);
433 st
->ss
->free_super(st
);
434 crtime
= info
.array
.ctime
;
435 level
= map_num(pers
, info
.array
.level
);
436 if (!level
) level
= "-unknown-";
437 fprintf(stderr
, " level=%s devices=%d ctime=%s",
438 level
, info
.array
.raid_disks
, ctime(&crtime
));
446 for (i
=0; i
<5; i
++) {
448 fprintf(stderr
, "%s%s", mesg
, add
);
450 if (fgets(buf
, 100, stdin
)==NULL
)
452 if (buf
[0]=='y' || buf
[0]=='Y')
454 if (buf
[0]=='n' || buf
[0]=='N')
458 fprintf(stderr
, Name
": assuming 'no'\n");
461 #endif /* MDASSEMBLE */
463 char *map_num(mapping_t
*map
, int num
)
473 int map_name(mapping_t
*map
, char *name
)
476 if (strcmp(map
->name
, name
)==0)
484 int is_standard(char *dev
, int *nump
)
486 /* tests if dev is a "standard" md dev name.
487 * i.e if the last component is "/dNN" or "/mdNN",
488 * where NN is a string of digits
489 * Returns 1 if a partitionable standard,
490 * -1 if non-partitonable,
491 * 0 if not a standard name.
493 char *d
= strrchr(dev
, '/');
498 if (strncmp(d
, "/d",2)==0)
499 d
+= 2, type
=1; /* /dev/md/dN{pM} */
500 else if (strncmp(d
, "/md_d", 5)==0)
501 d
+= 5, type
=1; /* /dev/md_dN{pM} */
502 else if (strncmp(d
, "/md", 3)==0)
503 d
+= 3, type
=-1; /* /dev/mdN */
504 else if (d
-dev
> 3 && strncmp(d
-2, "md/", 3)==0)
505 d
+= 1, type
=-1; /* /dev/md/N */
515 if (nump
) *nump
= num
;
522 * convert a major/minor pair for a block device into a name in /dev, if possible.
523 * On the first call, walk /dev collecting name.
524 * Put them in a simple linked listfor now.
531 int devlist_ready
= 0;
533 int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
)
537 if (S_ISLNK(stb
->st_mode
)) {
538 if (stat(name
, &st
) != 0)
543 if ((stb
->st_mode
&S_IFMT
)== S_IFBLK
) {
544 char *n
= strdup(name
);
545 struct devmap
*dm
= malloc(sizeof(*dm
));
546 if (strncmp(n
, "/dev/./", 7)==0)
549 dm
->major
= major(stb
->st_rdev
);
550 dm
->minor
= minor(stb
->st_rdev
);
561 int add_dev_1(const char *name
, const struct stat
*stb
, int flag
)
563 return add_dev(name
, stb
, flag
, NULL
);
565 int nftw(const char *path
, int (*han
)(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
), int nopenfd
, int flags
)
567 return ftw(path
, add_dev_1
, nopenfd
);
570 int nftw(const char *path
, int (*han
)(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
), int nopenfd
, int flags
)
574 #endif /* HAVE_FTW */
575 #endif /* HAVE_NFTW */
578 * Find a block device with the right major/minor number.
579 * If we find multiple names, choose the shortest.
580 * If we find a name in /dev/md/, we prefer that.
581 * This applies only to names for MD devices.
583 char *map_dev(int major
, int minor
, int create
)
586 char *regular
= NULL
, *preferred
=NULL
;
589 if (major
== 0 && minor
== 0)
593 if (!devlist_ready
) {
597 struct devmap
*d
= devlist
;
602 if (lstat(dev
, &stb
)==0 &&
603 S_ISLNK(stb
.st_mode
))
605 nftw(dev
, add_dev
, 10, FTW_PHYS
);
610 for (p
=devlist
; p
; p
=p
->next
)
611 if (p
->major
== major
&&
613 if (strncmp(p
->name
, "/dev/md/",8) == 0) {
614 if (preferred
== NULL
||
615 strlen(p
->name
) < strlen(preferred
))
618 if (regular
== NULL
||
619 strlen(p
->name
) < strlen(regular
))
623 if (!regular
&& !preferred
&& !did_check
) {
627 if (create
&& !regular
&& !preferred
) {
629 snprintf(buf
, sizeof(buf
), "%d:%d", major
, minor
);
633 return preferred
? preferred
: regular
;
636 unsigned long calc_csum(void *super
, int bytes
)
638 unsigned long long newcsum
= 0;
641 unsigned int *superc
= (unsigned int*) super
;
643 for(i
=0; i
<bytes
/4; i
++)
645 csum
= (newcsum
& 0xffffffff) + (newcsum
>>32);
647 /* The in-kernel checksum calculation is always 16bit on
648 * the alpha, though it is 32 bit on i386...
649 * I wonder what it is elsewhere... (it uses and API in
650 * a way that it shouldn't).
652 csum
= (csum
& 0xffff) + (csum
>> 16);
653 csum
= (csum
& 0xffff) + (csum
>> 16);
659 char *human_size(long long bytes
)
663 /* We convert bytes to either centi-M{ega,ibi}bytes or
664 * centi-G{igi,ibi}bytes, with appropriate rounding,
665 * and then print 1/100th of those as a decimal.
666 * We allow upto 2048Megabytes before converting to
667 * gigabytes, as that shows more precision and isn't
668 * too large a number.
669 * Terrabytes are not yet handled.
672 if (bytes
< 5000*1024)
674 else if (bytes
< 2*1024LL*1024LL*1024LL) {
675 long cMiB
= (bytes
/ ( (1LL<<20) / 200LL ) +1) /2;
676 long cMB
= (bytes
/ ( 1000000LL / 200LL ) +1) /2;
677 snprintf(buf
, sizeof(buf
), " (%ld.%02ld MiB %ld.%02ld MB)",
678 cMiB
/100 , cMiB
% 100,
681 long cGiB
= (bytes
/ ( (1LL<<30) / 200LL ) +1) /2;
682 long cGB
= (bytes
/ (1000000000LL/200LL ) +1) /2;
683 snprintf(buf
, sizeof(buf
), " (%ld.%02ld GiB %ld.%02ld GB)",
684 cGiB
/100 , cGiB
% 100,
690 char *human_size_brief(long long bytes
)
694 if (bytes
< 5000*1024)
695 snprintf(buf
, sizeof(buf
), "%ld.%02ldKiB",
696 (long)(bytes
>>10), (long)(((bytes
&1023)*100+512)/1024)
698 else if (bytes
< 2*1024LL*1024LL*1024LL)
699 snprintf(buf
, sizeof(buf
), "%ld.%02ldMiB",
701 (long)((bytes
&0xfffff)+0x100000/200)/(0x100000/100)
704 snprintf(buf
, sizeof(buf
), "%ld.%02ldGiB",
706 (long)(((bytes
>>10)&0xfffff)+0x100000/200)/(0x100000/100)
711 void print_r10_layout(int layout
)
713 int near
= layout
& 255;
714 int far
= (layout
>> 8) & 255;
715 int offset
= (layout
&0x10000);
719 printf("%s near=%d", sep
, near
);
723 printf("%s %s=%d", sep
, offset
?"offset":"far", far
);
725 printf("NO REDUNDANCY");
729 unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
730 int chunksize
, unsigned long long devsize
)
734 case 0: data_disks
= raid_disks
; break;
735 case 1: data_disks
= 1; break;
737 case 5: data_disks
= raid_disks
- 1; break;
738 case 6: data_disks
= raid_disks
- 2; break;
739 case 10: data_disks
= raid_disks
/ (layout
& 255) / ((layout
>>8)&255);
742 devsize
&= ~(unsigned long long)((chunksize
>>9)-1);
743 return data_disks
* devsize
;
746 int get_mdp_major(void)
748 static int mdp_major
= -1;
752 int have_devices
= 0;
757 fl
= fopen("/proc/devices", "r");
760 while ((w
= conf_word(fl
, 1))) {
761 if (have_block
&& strcmp(w
, "devices:")==0)
763 have_block
= (strcmp(w
, "Block")==0);
766 if (have_devices
&& strcmp(w
, "mdp")==0)
767 mdp_major
= last_num
;
774 #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
775 char *get_md_name(int dev
)
777 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
778 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
779 static char devname
[50];
785 int mdp
= get_mdp_major();
786 if (mdp
< 0) return NULL
;
787 rdev
= makedev(mdp
, (-1-dev
)<<6);
788 snprintf(devname
, sizeof(devname
), "/dev/md/d%d", -1-dev
);
789 if (stat(devname
, &stb
) == 0
790 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
791 && (stb
.st_rdev
== rdev
))
794 rdev
= makedev(MD_MAJOR
, dev
);
795 snprintf(devname
, sizeof(devname
), "/dev/md%d", dev
);
796 if (stat(devname
, &stb
) == 0
797 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
798 && (stb
.st_rdev
== rdev
))
801 snprintf(devname
, sizeof(devname
), "/dev/md/%d", dev
);
802 if (stat(devname
, &stb
) == 0
803 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
804 && (stb
.st_rdev
== rdev
))
807 dn
= map_dev(major(rdev
), minor(rdev
), 0);
810 snprintf(devname
, sizeof(devname
), "/dev/.tmp.md%d", dev
);
811 if (mknod(devname
, S_IFBLK
| 0600, rdev
) == -1)
815 if (stat(devname
, &stb
) == 0
816 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
817 && (stb
.st_rdev
== rdev
))
823 void put_md_name(char *name
)
825 if (strncmp(name
, "/dev/.tmp.md", 12)==0)
829 int find_free_devnum(int use_partitions
)
832 for (devnum
= 127; devnum
!= 128;
833 devnum
= devnum
? devnum
-1 : (1<<20)-1) {
837 _devnum
= use_partitions
? (-1-devnum
) : devnum
;
838 if (mddev_busy(_devnum
))
840 /* make sure it is new to /dev too, at least as a
842 dn
= map_dev(dev2major(_devnum
), dev2minor(_devnum
), 0);
843 if (dn
&& ! is_standard(dn
, NULL
))
849 return use_partitions
? (-1-devnum
) : devnum
;
851 #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
853 int dev_open(char *dev
, int flags
)
855 /* like 'open', but if 'dev' matches %d:%d, create a temp
856 * block device and open that
866 major
= strtoul(dev
, &e
, 0);
867 if (e
> dev
&& *e
== ':' && e
[1] &&
868 (minor
= strtoul(e
+1, &e
, 0)) >= 0 &&
870 snprintf(devname
, sizeof(devname
), "/dev/.tmp.md.%d:%d:%d",
871 (int)getpid(), major
, minor
);
872 if (mknod(devname
, S_IFBLK
|0600, makedev(major
, minor
))==0) {
873 fd
= open(devname
, flags
|O_DIRECT
);
877 fd
= open(dev
, flags
|O_DIRECT
);
881 int open_dev(int devnum
)
885 sprintf(buf
, "%d:%d", dev2major(devnum
), dev2minor(devnum
));
886 return dev_open(buf
, O_RDWR
);
889 int open_dev_excl(int devnum
)
894 sprintf(buf
, "%d:%d", dev2major(devnum
), dev2minor(devnum
));
895 for (i
=0 ; i
<25 ; i
++) {
896 int fd
= dev_open(buf
, O_RDWR
|O_EXCL
);
906 int same_dev(char *one
, char *two
)
908 struct stat st1
, st2
;
909 if (stat(one
, &st1
) != 0)
911 if (stat(two
, &st2
) != 0)
913 if ((st1
.st_mode
& S_IFMT
) != S_IFBLK
)
915 if ((st2
.st_mode
& S_IFMT
) != S_IFBLK
)
917 return st1
.st_rdev
== st2
.st_rdev
;
920 void wait_for(char *dev
, int fd
)
923 struct stat stb_want
;
925 if (fstat(fd
, &stb_want
) != 0 ||
926 (stb_want
.st_mode
& S_IFMT
) != S_IFBLK
)
929 for (i
=0 ; i
<25 ; i
++) {
931 if (stat(dev
, &stb
) == 0 &&
932 (stb
.st_mode
& S_IFMT
) == S_IFBLK
&&
933 (stb
.st_rdev
== stb_want
.st_rdev
))
938 dprintf("%s: timeout waiting for %s\n", __func__
, dev
);
941 struct superswitch
*superlist
[] = { &super0
, &super1
, &super_ddf
, &super_imsm
, NULL
};
943 #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
945 struct supertype
*super_by_fd(int fd
)
947 mdu_array_info_t array
;
950 struct supertype
*st
= NULL
;
955 char *subarray
= NULL
;
957 sra
= sysfs_read(fd
, 0, GET_VERSION
);
960 vers
= sra
->array
.major_version
;
961 minor
= sra
->array
.minor_version
;
962 verstr
= sra
->text_version
;
964 if (ioctl(fd
, GET_ARRAY_INFO
, &array
))
965 array
.major_version
= array
.minor_version
= 0;
966 vers
= array
.major_version
;
967 minor
= array
.minor_version
;
972 sprintf(version
, "%d.%d", vers
, minor
);
975 if (minor
== -2 && is_subarray(verstr
)) {
976 char *dev
= verstr
+1;
977 subarray
= strchr(dev
, '/');
981 devnum
= devname2devnum(dev
);
982 subarray
= strdup(subarray
);
985 sra
= sysfs_read(-1, devnum
, GET_VERSION
);
986 if (sra
&& sra
->text_version
[0])
987 verstr
= sra
->text_version
;
989 verstr
= "-no-metadata-";
992 for (i
= 0; st
== NULL
&& superlist
[i
] ; i
++)
993 st
= superlist
[i
]->match_metadata_desc(verstr
);
1000 strncpy(st
->subarray
, subarray
, 32);
1001 st
->subarray
[31] = 0;
1004 st
->subarray
[0] = 0;
1008 #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1011 struct supertype
*dup_super(struct supertype
*orig
)
1013 struct supertype
*st
;
1017 st
= malloc(sizeof(*st
));
1020 memset(st
, 0, sizeof(*st
));
1022 st
->max_devs
= orig
->max_devs
;
1023 st
->minor_version
= orig
->minor_version
;
1024 strcpy(st
->subarray
, orig
->subarray
);
1030 struct supertype
*guess_super(int fd
)
1032 /* try each load_super to find the best match,
1033 * and return the best superswitch
1035 struct superswitch
*ss
;
1036 struct supertype
*st
;
1037 unsigned long besttime
= 0;
1041 st
= malloc(sizeof(*st
));
1042 for (i
=0 ; superlist
[i
]; i
++) {
1045 memset(st
, 0, sizeof(*st
));
1046 rv
= ss
->load_super(st
, fd
, NULL
);
1049 st
->ss
->getinfo_super(st
, &info
);
1050 if (bestsuper
== -1 ||
1051 besttime
< info
.array
.ctime
) {
1053 besttime
= info
.array
.ctime
;
1058 if (bestsuper
!= -1) {
1060 memset(st
, 0, sizeof(*st
));
1061 rv
= superlist
[bestsuper
]->load_super(st
, fd
, NULL
);
1063 superlist
[bestsuper
]->free_super(st
);
1071 /* Return size of device in bytes */
1072 int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
)
1074 unsigned long long ldsize
;
1077 if (fstat(fd
, &st
) != -1 && S_ISREG(st
.st_mode
))
1078 ldsize
= (unsigned long long)st
.st_size
;
1081 if (ioctl(fd
, BLKGETSIZE64
, &ldsize
) != 0)
1084 unsigned long dsize
;
1085 if (ioctl(fd
, BLKGETSIZE
, &dsize
) == 0) {
1090 fprintf(stderr
, Name
": Cannot get size of %s: %s\b",
1091 dname
, strerror(errno
));
1099 void get_one_disk(int mdfd
, mdu_array_info_t
*ainf
, mdu_disk_info_t
*disk
)
1102 ioctl(mdfd
, GET_ARRAY_INFO
, ainf
);
1103 for (d
= 0 ; d
< ainf
->raid_disks
+ ainf
->nr_disks
; d
++)
1104 if (ioctl(mdfd
, GET_DISK_INFO
, disk
) == 0)
1108 int open_container(int fd
)
1110 /* 'fd' is a block device. Find out if it is in use
1111 * by a container, and return an open fd on that container.
1122 if (fstat(fd
, &st
) != 0)
1124 sprintf(path
, "/sys/dev/block/%d:%d/holders",
1125 (int)major(st
.st_rdev
), (int)minor(st
.st_rdev
));
1126 e
= path
+ strlen(path
);
1128 dir
= opendir(path
);
1131 while ((de
= readdir(dir
))) {
1134 if (de
->d_name
[0] == '.')
1136 sprintf(e
, "/%s/dev", de
->d_name
);
1137 dfd
= open(path
, O_RDONLY
);
1140 n
= read(dfd
, buf
, sizeof(buf
));
1142 if (n
<= 0 || n
>= sizeof(buf
))
1145 if (sscanf(buf
, "%d:%d", &major
, &minor
) != 2)
1147 sprintf(buf
, "%d:%d", major
, minor
);
1148 dfd
= dev_open(buf
, O_RDONLY
);
1158 int add_disk(int mdfd
, struct supertype
*st
,
1159 struct mdinfo
*sra
, struct mdinfo
*info
)
1161 /* Add a device to an array, in one of 2 ways. */
1164 if (st
->ss
->external
) {
1165 rv
= sysfs_add_disk(sra
, info
,
1166 info
->disk
.state
& (1<<MD_DISK_SYNC
));
1169 for (sd2
= sra
->devs
; sd2
; sd2
=sd2
->next
)
1173 sd2
= malloc(sizeof(*sd2
));
1175 sd2
->next
= sra
->devs
;
1181 rv
= ioctl(mdfd
, ADD_NEW_DISK
, &info
->disk
);
1185 int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
)
1187 /* Initialise kernel's knowledge of array.
1188 * This varies between externally managed arrays
1191 int vers
= md_get_version(mdfd
);
1195 if (st
->ss
->external
)
1196 rv
= sysfs_set_array(info
, vers
);
1199 if ((vers
% 100) >= 1) { /* can use different versions */
1200 mdu_array_info_t inf
;
1201 memset(&inf
, 0, sizeof(inf
));
1202 inf
.major_version
= info
->array
.major_version
;
1203 inf
.minor_version
= info
->array
.minor_version
;
1204 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, &inf
);
1206 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, NULL
);
1210 char *devnum2devname(int num
)
1214 sprintf(name
, "md%d", num
);
1216 sprintf(name
, "md_d%d", -1-num
);
1217 return strdup(name
);
1220 int devname2devnum(char *name
)
1224 if (strncmp(name
, "md_d", 4)==0)
1225 num
= -1-strtoul(name
+4, &ep
, 10);
1227 num
= strtoul(name
+2, &ep
, 10);
1231 int stat2devnum(struct stat
*st
)
1238 if ((S_IFMT
& st
->st_mode
) == S_IFBLK
) {
1239 if (major(st
->st_rdev
) == MD_MAJOR
)
1240 return minor(st
->st_rdev
);
1241 else if (major(st
->st_rdev
) == get_mdp_major())
1242 return -1- (minor(st
->st_rdev
)>>MdpMinorShift
);
1244 /* must be an extended-minor partition. Look at the
1245 * /sys/dev/block/%d:%d link which must look like
1246 * ../../block/mdXXX/mdXXXpYY
1248 sprintf(path
, "/sys/dev/block/%d:%d", major(st
->st_rdev
),
1249 minor(st
->st_rdev
));
1250 n
= readlink(path
, link
, sizeof(link
)-1);
1254 cp
= strrchr(link
, '/');
1256 cp
= strchr(link
, '/');
1257 if (cp
&& strncmp(cp
, "/md", 3) == 0)
1258 return devname2devnum(cp
+1);
1264 int fd2devnum(int fd
)
1267 if (fstat(fd
, &stb
) == 0)
1268 return stat2devnum(&stb
);
1272 int mdmon_running(int devnum
)
1278 sprintf(path
, "/var/run/mdadm/%s.pid", devnum2devname(devnum
));
1279 fd
= open(path
, O_RDONLY
, 0);
1283 n
= read(fd
, pid
, 9);
1287 if (kill(atoi(pid
), 0) == 0)
1292 int signal_mdmon(int devnum
)
1298 sprintf(path
, "/var/run/mdadm/%s.pid", devnum2devname(devnum
));
1299 fd
= open(path
, O_RDONLY
, 0);
1303 n
= read(fd
, pid
, 9);
1307 if (kill(atoi(pid
), SIGUSR1
) == 0)
1312 int start_mdmon(int devnum
)
1326 if (check_env("MDADM_NO_MDMON"))
1329 len
= readlink("/proc/self/exe", pathbuf
, sizeof(pathbuf
));
1333 sl
= strrchr(pathbuf
, '/');
1338 strcpy(sl
, "mdmon");
1344 /* FIXME yuk. CLOSE_EXEC?? */
1345 for (i
=3; i
< 100; i
++)
1347 for (i
=0; paths
[i
]; i
++)
1349 execl(paths
[i
], "mdmon",
1350 devnum2devname(devnum
),
1353 case -1: fprintf(stderr
, Name
": cannot run mdmon. "
1354 "Array remains readonly\n");
1356 default: /* parent - good */
1357 pid
= wait(&status
);
1358 if (pid
< 0 || status
!= 0)
1364 int check_env(char *name
)
1366 char *val
= getenv(name
);
1368 if (val
&& atoi(val
) == 1)
1374 __u32
random32(void)
1377 int rfd
= open("/dev/urandom", O_RDONLY
);
1378 if (rfd
< 0 || read(rfd
, &rv
, 4) != 4)
1386 int flush_metadata_updates(struct supertype
*st
)
1390 st
->update_tail
= NULL
;
1394 sfd
= connect_monitor(devnum2devname(st
->container_dev
));
1398 while (st
->updates
) {
1399 struct metadata_update
*mu
= st
->updates
;
1400 st
->updates
= mu
->next
;
1402 send_message(sfd
, mu
, 0);
1410 st
->update_tail
= NULL
;
1414 void append_metadata_update(struct supertype
*st
, void *buf
, int len
)
1417 struct metadata_update
*mu
= malloc(sizeof(*mu
));
1423 *st
->update_tail
= mu
;
1424 st
->update_tail
= &mu
->next
;
1426 #endif /* MDASSEMBLE */
1429 /* tinyc doesn't optimize this check in ioctl.h out ... */
1430 unsigned int __invalid_size_argument_for_IOC
= 0;