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
;
152 void remove_partitions(int fd
)
154 /* remove partitions from this block devices.
155 * This is used for components added to an array
157 #ifdef BLKPG_DEL_PARTITION
158 struct blkpg_ioctl_arg a
;
159 struct blkpg_partition p
;
161 a
.op
= BLKPG_DEL_PARTITION
;
163 a
.datalen
= sizeof(p
);
165 memset(a
.data
, 0, a
.datalen
);
166 for (p
.pno
=0; p
.pno
< 16; p
.pno
++)
167 ioctl(fd
, BLKPG
, &a
);
171 int enough(int level
, int raid_disks
, int layout
, int clean
,
172 char *avail
, int avail_disks
)
177 /* This is the tricky one - we need to check
178 * which actual disks are present.
180 copies
= (layout
&255)* ((layout
>>8) & 255);
183 /* there must be one of the 'copies' form 'first' */
189 first
= (first
+1) % raid_disks
;
194 } while (first
!= 0);
198 return avail_disks
>= 1;
201 return avail_disks
== raid_disks
;
203 return avail_disks
>= 1;
207 return avail_disks
>= raid_disks
-1;
209 return avail_disks
>= raid_disks
;
212 return avail_disks
>= raid_disks
-2;
214 return avail_disks
>= raid_disks
;
220 const int uuid_match_any
[4] = { ~0, ~0, ~0, ~0 };
221 int same_uuid(int a
[4], int b
[4], int swapuuid
)
223 if (memcmp(a
, uuid_match_any
, sizeof(int[4])) == 0 ||
224 memcmp(b
, uuid_match_any
, sizeof(int[4])) == 0)
228 /* parse uuids are hostendian.
229 * uuid's from some superblocks are big-ending
230 * if there is a difference, we need to swap..
232 unsigned char *ac
= (unsigned char *)a
;
233 unsigned char *bc
= (unsigned char *)b
;
235 for (i
=0; i
<16; i
+= 4) {
236 if (ac
[i
+0] != bc
[i
+3] ||
237 ac
[i
+1] != bc
[i
+2] ||
238 ac
[i
+2] != bc
[i
+1] ||
252 void copy_uuid(void *a
, int b
[4], int swapuuid
)
255 /* parse uuids are hostendian.
256 * uuid's from some superblocks are big-ending
257 * if there is a difference, we need to swap..
259 unsigned char *ac
= (unsigned char *)a
;
260 unsigned char *bc
= (unsigned char *)b
;
262 for (i
=0; i
<16; i
+= 4) {
272 char *fname_from_uuid(struct supertype
*st
, struct mdinfo
*info
, char *buf
, char sep
)
280 copy_uuid(uuid
, info
->uuid
, st
->ss
->swapuuid
);
281 for (i
= 0; i
< 4; i
++) {
285 for (j
= 3; j
>= 0; j
--) {
286 sprintf(c
,"%02x", (unsigned char) uuid
[j
+4*i
]);
294 int check_ext2(int fd
, char *name
)
297 * Check for an ext2fs file system.
298 * Superblock is always 1K at 1K offset
300 * s_magic is le16 at 56 == 0xEF53
301 * report mtime - le32 at 44
303 * logblksize - le32 at 24
305 unsigned char sb
[1024];
308 if (lseek(fd
, 1024,0)!= 1024)
310 if (read(fd
, sb
, 1024)!= 1024)
312 if (sb
[56] != 0x53 || sb
[57] != 0xef)
315 mtime
= sb
[44]|(sb
[45]|(sb
[46]|sb
[47]<<8)<<8)<<8;
316 bsize
= sb
[24]|(sb
[25]|(sb
[26]|sb
[27]<<8)<<8)<<8;
317 size
= sb
[4]|(sb
[5]|(sb
[6]|sb
[7]<<8)<<8)<<8;
318 fprintf(stderr
, Name
": %s appears to contain an ext2fs file system\n",
320 fprintf(stderr
," size=%dK mtime=%s",
321 size
*(1<<bsize
), ctime(&mtime
));
325 int check_reiser(int fd
, char *name
)
328 * superblock is at 64K
330 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
333 unsigned char sb
[1024];
335 if (lseek(fd
, 64*1024, 0) != 64*1024)
337 if (read(fd
, sb
, 1024) != 1024)
339 if (strncmp((char*)sb
+52, "ReIsErFs",8)!=0 &&
340 strncmp((char*)sb
+52, "ReIsEr2Fs",9)!=0)
342 fprintf(stderr
, Name
": %s appears to contain a reiserfs file system\n",name
);
343 size
= sb
[0]|(sb
[1]|(sb
[2]|sb
[3]<<8)<<8)<<8;
344 fprintf(stderr
, " size = %luK\n", size
*4);
349 int check_raid(int fd
, char *name
)
354 struct supertype
*st
= guess_super(fd
);
357 st
->ss
->load_super(st
, fd
, name
);
358 /* Looks like a raid array .. */
359 fprintf(stderr
, Name
": %s appears to be part of a raid array:\n",
361 st
->ss
->getinfo_super(st
, &info
);
362 st
->ss
->free_super(st
);
363 crtime
= info
.array
.ctime
;
364 level
= map_num(pers
, info
.array
.level
);
365 if (!level
) level
= "-unknown-";
366 fprintf(stderr
, " level=%s devices=%d ctime=%s",
367 level
, info
.array
.raid_disks
, ctime(&crtime
));
375 for (i
=0; i
<5; i
++) {
377 fprintf(stderr
, "%s%s", mesg
, add
);
379 if (fgets(buf
, 100, stdin
)==NULL
)
381 if (buf
[0]=='y' || buf
[0]=='Y')
383 if (buf
[0]=='n' || buf
[0]=='N')
387 fprintf(stderr
, Name
": assuming 'no'\n");
390 #endif /* MDASSEMBLE */
392 char *map_num(mapping_t
*map
, int num
)
402 int map_name(mapping_t
*map
, char *name
)
405 if (strcmp(map
->name
, name
)==0)
413 int is_standard(char *dev
, int *nump
)
415 /* tests if dev is a "standard" md dev name.
416 * i.e if the last component is "/dNN" or "/mdNN",
417 * where NN is a string of digits
418 * Returns 1 if a partitionable standard,
419 * -1 if non-partitonable,
420 * 0 if not a standard name.
422 char *d
= strrchr(dev
, '/');
427 if (strncmp(d
, "/d",2)==0)
428 d
+= 2, type
=1; /* /dev/md/dN{pM} */
429 else if (strncmp(d
, "/md_d", 5)==0)
430 d
+= 5, type
=1; /* /dev/md_dN{pM} */
431 else if (strncmp(d
, "/md", 3)==0)
432 d
+= 3, type
=-1; /* /dev/mdN */
433 else if (d
-dev
> 3 && strncmp(d
-2, "md/", 3)==0)
434 d
+= 1, type
=-1; /* /dev/md/N */
444 if (nump
) *nump
= num
;
451 * convert a major/minor pair for a block device into a name in /dev, if possible.
452 * On the first call, walk /dev collecting name.
453 * Put them in a simple linked listfor now.
460 int devlist_ready
= 0;
462 int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
)
466 if (S_ISLNK(stb
->st_mode
)) {
467 if (stat(name
, &st
) != 0)
472 if ((stb
->st_mode
&S_IFMT
)== S_IFBLK
) {
473 char *n
= strdup(name
);
474 struct devmap
*dm
= malloc(sizeof(*dm
));
475 if (strncmp(n
, "/dev/./", 7)==0)
478 dm
->major
= major(stb
->st_rdev
);
479 dm
->minor
= minor(stb
->st_rdev
);
490 int add_dev_1(const char *name
, const struct stat
*stb
, int flag
)
492 return add_dev(name
, stb
, flag
, NULL
);
494 int nftw(const char *path
, int (*han
)(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
), int nopenfd
, int flags
)
496 return ftw(path
, add_dev_1
, nopenfd
);
499 int nftw(const char *path
, int (*han
)(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
), int nopenfd
, int flags
)
503 #endif /* HAVE_FTW */
504 #endif /* HAVE_NFTW */
507 * Find a block device with the right major/minor number.
508 * If we find multiple names, choose the shortest.
509 * If we find a name in /dev/md/, we prefer that.
510 * This applies only to names for MD devices.
512 char *map_dev(int major
, int minor
, int create
)
515 char *regular
= NULL
, *preferred
=NULL
;
518 if (major
== 0 && minor
== 0)
522 if (!devlist_ready
) {
526 struct devmap
*d
= devlist
;
531 if (lstat(dev
, &stb
)==0 &&
532 S_ISLNK(stb
.st_mode
))
534 nftw(dev
, add_dev
, 10, FTW_PHYS
);
539 for (p
=devlist
; p
; p
=p
->next
)
540 if (p
->major
== major
&&
542 if (strncmp(p
->name
, "/dev/md/",8) == 0) {
543 if (preferred
== NULL
||
544 strlen(p
->name
) < strlen(preferred
))
547 if (regular
== NULL
||
548 strlen(p
->name
) < strlen(regular
))
552 if (!regular
&& !preferred
&& !did_check
) {
556 if (create
&& !regular
&& !preferred
) {
558 snprintf(buf
, sizeof(buf
), "%d:%d", major
, minor
);
562 return preferred
? preferred
: regular
;
565 unsigned long calc_csum(void *super
, int bytes
)
567 unsigned long long newcsum
= 0;
570 unsigned int *superc
= (unsigned int*) super
;
572 for(i
=0; i
<bytes
/4; i
++)
574 csum
= (newcsum
& 0xffffffff) + (newcsum
>>32);
576 /* The in-kernel checksum calculation is always 16bit on
577 * the alpha, though it is 32 bit on i386...
578 * I wonder what it is elsewhere... (it uses and API in
579 * a way that it shouldn't).
581 csum
= (csum
& 0xffff) + (csum
>> 16);
582 csum
= (csum
& 0xffff) + (csum
>> 16);
588 char *human_size(long long bytes
)
592 /* We convert bytes to either centi-M{ega,ibi}bytes or
593 * centi-G{igi,ibi}bytes, with appropriate rounding,
594 * and then print 1/100th of those as a decimal.
595 * We allow upto 2048Megabytes before converting to
596 * gigabytes, as that shows more precision and isn't
597 * too large a number.
598 * Terrabytes are not yet handled.
601 if (bytes
< 5000*1024)
603 else if (bytes
< 2*1024LL*1024LL*1024LL) {
604 long cMiB
= (bytes
/ ( (1LL<<20) / 200LL ) +1) /2;
605 long cMB
= (bytes
/ ( 1000000LL / 200LL ) +1) /2;
606 snprintf(buf
, sizeof(buf
), " (%ld.%02ld MiB %ld.%02ld MB)",
607 cMiB
/100 , cMiB
% 100,
610 long cGiB
= (bytes
/ ( (1LL<<30) / 200LL ) +1) /2;
611 long cGB
= (bytes
/ (1000000000LL/200LL ) +1) /2;
612 snprintf(buf
, sizeof(buf
), " (%ld.%02ld GiB %ld.%02ld GB)",
613 cGiB
/100 , cGiB
% 100,
619 char *human_size_brief(long long bytes
)
623 if (bytes
< 5000*1024)
624 snprintf(buf
, sizeof(buf
), "%ld.%02ldKiB",
625 (long)(bytes
>>10), (long)(((bytes
&1023)*100+512)/1024)
627 else if (bytes
< 2*1024LL*1024LL*1024LL)
628 snprintf(buf
, sizeof(buf
), "%ld.%02ldMiB",
630 (long)((bytes
&0xfffff)+0x100000/200)/(0x100000/100)
633 snprintf(buf
, sizeof(buf
), "%ld.%02ldGiB",
635 (long)(((bytes
>>10)&0xfffff)+0x100000/200)/(0x100000/100)
640 void print_r10_layout(int layout
)
642 int near
= layout
& 255;
643 int far
= (layout
>> 8) & 255;
644 int offset
= (layout
&0x10000);
648 printf("%s near=%d", sep
, near
);
652 printf("%s %s=%d", sep
, offset
?"offset":"far", far
);
654 printf("NO REDUNDANCY");
658 unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
659 int chunksize
, unsigned long long devsize
)
663 case 0: data_disks
= raid_disks
; break;
664 case 1: data_disks
= 1; break;
666 case 5: data_disks
= raid_disks
- 1; break;
667 case 6: data_disks
= raid_disks
- 2; break;
668 case 10: data_disks
= raid_disks
/ (layout
& 255) / ((layout
>>8)&255);
671 devsize
&= ~(unsigned long long)((chunksize
>>9)-1);
672 return data_disks
* devsize
;
675 int get_mdp_major(void)
677 static int mdp_major
= -1;
681 int have_devices
= 0;
686 fl
= fopen("/proc/devices", "r");
689 while ((w
= conf_word(fl
, 1))) {
690 if (have_block
&& strcmp(w
, "devices:")==0)
692 have_block
= (strcmp(w
, "Block")==0);
695 if (have_devices
&& strcmp(w
, "mdp")==0)
696 mdp_major
= last_num
;
703 #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
704 char *get_md_name(int dev
)
706 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
707 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
708 static char devname
[50];
714 int mdp
= get_mdp_major();
715 if (mdp
< 0) return NULL
;
716 rdev
= makedev(mdp
, (-1-dev
)<<6);
717 snprintf(devname
, sizeof(devname
), "/dev/md/d%d", -1-dev
);
718 if (stat(devname
, &stb
) == 0
719 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
720 && (stb
.st_rdev
== rdev
))
723 rdev
= makedev(MD_MAJOR
, dev
);
724 snprintf(devname
, sizeof(devname
), "/dev/md%d", dev
);
725 if (stat(devname
, &stb
) == 0
726 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
727 && (stb
.st_rdev
== rdev
))
730 snprintf(devname
, sizeof(devname
), "/dev/md/%d", dev
);
731 if (stat(devname
, &stb
) == 0
732 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
733 && (stb
.st_rdev
== rdev
))
736 dn
= map_dev(major(rdev
), minor(rdev
), 0);
739 snprintf(devname
, sizeof(devname
), "/dev/.tmp.md%d", dev
);
740 if (mknod(devname
, S_IFBLK
| 0600, rdev
) == -1)
744 if (stat(devname
, &stb
) == 0
745 && (S_IFMT
&stb
.st_mode
) == S_IFBLK
746 && (stb
.st_rdev
== rdev
))
752 void put_md_name(char *name
)
754 if (strncmp(name
, "/dev/.tmp.md", 12)==0)
758 int find_free_devnum(int use_partitions
)
761 for (devnum
= 127; devnum
!= 128;
762 devnum
= devnum
? devnum
-1 : (1<<20)-1) {
766 _devnum
= use_partitions
? (-1-devnum
) : devnum
;
767 if (mddev_busy(_devnum
))
769 /* make sure it is new to /dev too, at least as a
771 dn
= map_dev(dev2major(_devnum
), dev2minor(_devnum
), 0);
772 if (dn
&& ! is_standard(dn
, NULL
))
778 return use_partitions
? (-1-devnum
) : devnum
;
780 #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
782 int dev_open(char *dev
, int flags
)
784 /* like 'open', but if 'dev' matches %d:%d, create a temp
785 * block device and open that
795 major
= strtoul(dev
, &e
, 0);
796 if (e
> dev
&& *e
== ':' && e
[1] &&
797 (minor
= strtoul(e
+1, &e
, 0)) >= 0 &&
799 snprintf(devname
, sizeof(devname
), "/dev/.tmp.md.%d:%d:%d",
800 (int)getpid(), major
, minor
);
801 if (mknod(devname
, S_IFBLK
|0600, makedev(major
, minor
))==0) {
802 fd
= open(devname
, flags
|O_DIRECT
);
806 fd
= open(dev
, flags
|O_DIRECT
);
810 int open_dev(int devnum
)
814 sprintf(buf
, "%d:%d", dev2major(devnum
), dev2minor(devnum
));
815 return dev_open(buf
, O_RDWR
);
818 int open_dev_excl(int devnum
)
823 sprintf(buf
, "%d:%d", dev2major(devnum
), dev2minor(devnum
));
824 for (i
=0 ; i
<25 ; i
++) {
825 int fd
= dev_open(buf
, O_RDWR
|O_EXCL
);
835 int same_dev(char *one
, char *two
)
837 struct stat st1
, st2
;
838 if (stat(one
, &st1
) != 0)
840 if (stat(two
, &st2
) != 0)
842 if ((st1
.st_mode
& S_IFMT
) != S_IFBLK
)
844 if ((st2
.st_mode
& S_IFMT
) != S_IFBLK
)
846 return st1
.st_rdev
== st2
.st_rdev
;
849 void wait_for(char *dev
, int fd
)
852 struct stat stb_want
;
854 if (fstat(fd
, &stb_want
) != 0 ||
855 (stb_want
.st_mode
& S_IFMT
) != S_IFBLK
)
858 for (i
=0 ; i
<25 ; i
++) {
860 if (stat(dev
, &stb
) == 0 &&
861 (stb
.st_mode
& S_IFMT
) == S_IFBLK
&&
862 (stb
.st_rdev
== stb_want
.st_rdev
))
867 dprintf("%s: timeout waiting for %s\n", __func__
, dev
);
870 struct superswitch
*superlist
[] = { &super0
, &super1
, &super_ddf
, &super_imsm
, NULL
};
872 #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
874 struct supertype
*super_by_fd(int fd
)
876 mdu_array_info_t array
;
879 struct supertype
*st
= NULL
;
884 char *subarray
= NULL
;
886 sra
= sysfs_read(fd
, 0, GET_VERSION
);
889 vers
= sra
->array
.major_version
;
890 minor
= sra
->array
.minor_version
;
891 verstr
= sra
->text_version
;
893 if (ioctl(fd
, GET_ARRAY_INFO
, &array
))
894 array
.major_version
= array
.minor_version
= 0;
895 vers
= array
.major_version
;
896 minor
= array
.minor_version
;
901 sprintf(version
, "%d.%d", vers
, minor
);
904 if (minor
== -2 && is_subarray(verstr
)) {
905 char *dev
= verstr
+1;
906 subarray
= strchr(dev
, '/');
910 devnum
= devname2devnum(dev
);
911 subarray
= strdup(subarray
);
914 sra
= sysfs_read(-1, devnum
, GET_VERSION
);
915 if (sra
&& sra
->text_version
[0])
916 verstr
= sra
->text_version
;
918 verstr
= "-no-metadata-";
921 for (i
= 0; st
== NULL
&& superlist
[i
] ; i
++)
922 st
= superlist
[i
]->match_metadata_desc(verstr
);
929 strncpy(st
->subarray
, subarray
, 32);
930 st
->subarray
[31] = 0;
937 #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
940 struct supertype
*dup_super(struct supertype
*orig
)
942 struct supertype
*st
;
946 st
= malloc(sizeof(*st
));
949 memset(st
, 0, sizeof(*st
));
951 st
->max_devs
= orig
->max_devs
;
952 st
->minor_version
= orig
->minor_version
;
953 strcpy(st
->subarray
, orig
->subarray
);
959 struct supertype
*guess_super(int fd
)
961 /* try each load_super to find the best match,
962 * and return the best superswitch
964 struct superswitch
*ss
;
965 struct supertype
*st
;
966 unsigned long besttime
= 0;
970 st
= malloc(sizeof(*st
));
971 for (i
=0 ; superlist
[i
]; i
++) {
974 memset(st
, 0, sizeof(*st
));
975 rv
= ss
->load_super(st
, fd
, NULL
);
978 st
->ss
->getinfo_super(st
, &info
);
979 if (bestsuper
== -1 ||
980 besttime
< info
.array
.ctime
) {
982 besttime
= info
.array
.ctime
;
987 if (bestsuper
!= -1) {
989 memset(st
, 0, sizeof(*st
));
990 rv
= superlist
[bestsuper
]->load_super(st
, fd
, NULL
);
992 superlist
[bestsuper
]->free_super(st
);
1000 /* Return size of device in bytes */
1001 int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
)
1003 unsigned long long ldsize
;
1006 if (fstat(fd
, &st
) != -1 && S_ISREG(st
.st_mode
))
1007 ldsize
= (unsigned long long)st
.st_size
;
1010 if (ioctl(fd
, BLKGETSIZE64
, &ldsize
) != 0)
1013 unsigned long dsize
;
1014 if (ioctl(fd
, BLKGETSIZE
, &dsize
) == 0) {
1019 fprintf(stderr
, Name
": Cannot get size of %s: %s\b",
1020 dname
, strerror(errno
));
1028 void get_one_disk(int mdfd
, mdu_array_info_t
*ainf
, mdu_disk_info_t
*disk
)
1031 ioctl(mdfd
, GET_ARRAY_INFO
, ainf
);
1032 for (d
= 0 ; d
< ainf
->raid_disks
+ ainf
->nr_disks
; d
++)
1033 if (ioctl(mdfd
, GET_DISK_INFO
, disk
) == 0)
1037 int open_container(int fd
)
1039 /* 'fd' is a block device. Find out if it is in use
1040 * by a container, and return an open fd on that container.
1051 if (fstat(fd
, &st
) != 0)
1053 sprintf(path
, "/sys/dev/block/%d:%d/holders",
1054 (int)major(st
.st_rdev
), (int)minor(st
.st_rdev
));
1055 e
= path
+ strlen(path
);
1057 dir
= opendir(path
);
1060 while ((de
= readdir(dir
))) {
1063 if (de
->d_name
[0] == '.')
1065 sprintf(e
, "/%s/dev", de
->d_name
);
1066 dfd
= open(path
, O_RDONLY
);
1069 n
= read(dfd
, buf
, sizeof(buf
));
1071 if (n
<= 0 || n
>= sizeof(buf
))
1074 if (sscanf(buf
, "%d:%d", &major
, &minor
) != 2)
1076 sprintf(buf
, "%d:%d", major
, minor
);
1077 dfd
= dev_open(buf
, O_RDONLY
);
1087 int add_disk(int mdfd
, struct supertype
*st
,
1088 struct mdinfo
*sra
, struct mdinfo
*info
)
1090 /* Add a device to an array, in one of 2 ways. */
1093 if (st
->ss
->external
) {
1094 rv
= sysfs_add_disk(sra
, info
,
1095 info
->disk
.state
& (1<<MD_DISK_SYNC
));
1098 for (sd2
= sra
->devs
; sd2
; sd2
=sd2
->next
)
1102 sd2
= malloc(sizeof(*sd2
));
1104 sd2
->next
= sra
->devs
;
1110 rv
= ioctl(mdfd
, ADD_NEW_DISK
, &info
->disk
);
1114 int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
)
1116 /* Initialise kernel's knowledge of array.
1117 * This varies between externally managed arrays
1120 int vers
= md_get_version(mdfd
);
1124 if (st
->ss
->external
)
1125 rv
= sysfs_set_array(info
, vers
);
1128 if ((vers
% 100) >= 1) { /* can use different versions */
1129 mdu_array_info_t inf
;
1130 memset(&inf
, 0, sizeof(inf
));
1131 inf
.major_version
= info
->array
.major_version
;
1132 inf
.minor_version
= info
->array
.minor_version
;
1133 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, &inf
);
1135 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, NULL
);
1139 char *devnum2devname(int num
)
1143 sprintf(name
, "md%d", num
);
1145 sprintf(name
, "md_d%d", -1-num
);
1146 return strdup(name
);
1149 int devname2devnum(char *name
)
1153 if (strncmp(name
, "md_d", 4)==0)
1154 num
= -1-strtoul(name
+4, &ep
, 10);
1156 num
= strtoul(name
+2, &ep
, 10);
1160 int stat2devnum(struct stat
*st
)
1167 if ((S_IFMT
& st
->st_mode
) == S_IFBLK
) {
1168 if (major(st
->st_rdev
) == MD_MAJOR
)
1169 return minor(st
->st_rdev
);
1170 else if (major(st
->st_rdev
) == get_mdp_major())
1171 return -1- (minor(st
->st_rdev
)>>MdpMinorShift
);
1173 /* must be an extended-minor partition. Look at the
1174 * /sys/dev/block/%d:%d link which must look like
1175 * ../../block/mdXXX/mdXXXpYY
1177 sprintf(path
, "/sys/dev/block/%d:%d", major(st
->st_rdev
),
1178 minor(st
->st_rdev
));
1179 n
= readlink(path
, link
, sizeof(link
)-1);
1183 cp
= strrchr(link
, '/');
1185 cp
= strchr(link
, '/');
1186 if (cp
&& strncmp(cp
, "/md", 3) == 0)
1187 return devname2devnum(cp
+1);
1193 int fd2devnum(int fd
)
1196 if (fstat(fd
, &stb
) == 0)
1197 return stat2devnum(&stb
);
1201 int mdmon_running(int devnum
)
1207 sprintf(path
, "/var/run/mdadm/%s.pid", devnum2devname(devnum
));
1208 fd
= open(path
, O_RDONLY
, 0);
1212 n
= read(fd
, pid
, 9);
1216 if (kill(atoi(pid
), 0) == 0)
1221 int signal_mdmon(int devnum
)
1227 sprintf(path
, "/var/run/mdadm/%s.pid", devnum2devname(devnum
));
1228 fd
= open(path
, O_RDONLY
, 0);
1232 n
= read(fd
, pid
, 9);
1236 if (kill(atoi(pid
), SIGUSR1
) == 0)
1241 int start_mdmon(int devnum
)
1255 if (check_env("MDADM_NO_MDMON"))
1258 len
= readlink("/proc/self/exe", pathbuf
, sizeof(pathbuf
));
1262 sl
= strrchr(pathbuf
, '/');
1267 strcpy(sl
, "mdmon");
1273 /* FIXME yuk. CLOSE_EXEC?? */
1274 for (i
=3; i
< 100; i
++)
1276 for (i
=0; paths
[i
]; i
++)
1278 execl(paths
[i
], "mdmon",
1279 devnum2devname(devnum
),
1282 case -1: fprintf(stderr
, Name
": cannot run mdmon. "
1283 "Array remains readonly\n");
1285 default: /* parent - good */
1286 pid
= wait(&status
);
1287 if (pid
< 0 || status
!= 0)
1293 int check_env(char *name
)
1295 char *val
= getenv(name
);
1297 if (val
&& atoi(val
) == 1)
1303 __u32
random32(void)
1306 int rfd
= open("/dev/urandom", O_RDONLY
);
1307 if (rfd
< 0 || read(rfd
, &rv
, 4) != 4)
1315 int flush_metadata_updates(struct supertype
*st
)
1319 st
->update_tail
= NULL
;
1323 sfd
= connect_monitor(devnum2devname(st
->container_dev
));
1327 while (st
->updates
) {
1328 struct metadata_update
*mu
= st
->updates
;
1329 st
->updates
= mu
->next
;
1331 send_message(sfd
, mu
, 0);
1339 st
->update_tail
= NULL
;
1343 void append_metadata_update(struct supertype
*st
, void *buf
, int len
)
1346 struct metadata_update
*mu
= malloc(sizeof(*mu
));
1352 *st
->update_tail
= mu
;
1353 st
->update_tail
= &mu
->next
;
1355 #endif /* MDASSEMBLE */
1358 /* tinyc doesn't optimize this check in ioctl.h out ... */
1359 unsigned int __invalid_size_argument_for_IOC
= 0;