]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - sysfs.c
2 * sysfs - extract md related information from sysfs. Part of:
3 * mdadm - manage Linux "md" devices aka RAID arrays.
5 * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de>
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with this program; if not, write to the Free Software
20 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 * Email: <neilb@suse.de>
30 int load_sys(char *path
, char *buf
)
32 int fd
= open(path
, O_RDONLY
);
36 n
= read(fd
, buf
, 1024);
38 if (n
<0 || n
>= 1024)
41 if (n
&& buf
[n
-1] == '\n')
46 void sysfs_free(struct mdinfo
*sra
)
49 struct mdinfo
*sra2
= sra
->next
;
51 struct mdinfo
*d
= sra
->devs
;
60 int sysfs_open(int devnum
, char *devname
, char *attr
)
64 char *mdname
= devnum2devname(devnum
);
69 sprintf(fname
, "/sys/block/%s/md/", mdname
);
71 strcat(fname
, devname
);
75 fd
= open(fname
, O_RDWR
);
76 if (fd
< 0 && errno
== EACCES
)
77 fd
= open(fname
, O_RDONLY
);
82 void sysfs_init(struct mdinfo
*mdi
, int fd
, int devnum
)
87 if (ioctl(fd
, RAID_VERSION
, &vers
) != 0)
89 devnum
= fd2devnum(fd
);
91 if (devnum
== NoMdDev
)
94 sprintf(mdi
->sys_name
, "md%d", devnum
);
96 sprintf(mdi
->sys_name
, "md_d%d",
101 struct mdinfo
*sysfs_read(int fd
, int devnum
, unsigned long options
)
103 char fname
[PATH_MAX
];
112 sra
= malloc(sizeof(*sra
));
115 memset(sra
, 0, sizeof(*sra
));
116 sysfs_init(sra
, fd
, devnum
);
117 if (sra
->sys_name
[0] == 0) {
122 sprintf(fname
, "/sys/block/%s/md/", sra
->sys_name
);
123 base
= fname
+ strlen(fname
);
126 if (options
& GET_VERSION
) {
127 strcpy(base
, "metadata_version");
128 if (load_sys(fname
, buf
))
130 if (strncmp(buf
, "none", 4) == 0) {
131 sra
->array
.major_version
=
132 sra
->array
.minor_version
= -1;
133 strcpy(sra
->text_version
, "");
134 } else if (strncmp(buf
, "external:", 9) == 0) {
135 sra
->array
.major_version
= -1;
136 sra
->array
.minor_version
= -2;
137 strcpy(sra
->text_version
, buf
+9);
140 &sra
->array
.major_version
,
141 &sra
->array
.minor_version
);
142 strcpy(sra
->text_version
, buf
);
145 if (options
& GET_LEVEL
) {
146 strcpy(base
, "level");
147 if (load_sys(fname
, buf
))
149 sra
->array
.level
= map_name(pers
, buf
);
151 if (options
& GET_LAYOUT
) {
152 strcpy(base
, "layout");
153 if (load_sys(fname
, buf
))
155 sra
->array
.layout
= strtoul(buf
, NULL
, 0);
157 if (options
& GET_DISKS
) {
158 strcpy(base
, "raid_disks");
159 if (load_sys(fname
, buf
))
161 sra
->array
.raid_disks
= strtoul(buf
, NULL
, 0);
163 if (options
& GET_DEGRADED
) {
164 strcpy(base
, "degraded");
165 if (load_sys(fname
, buf
))
167 sra
->array
.failed_disks
= strtoul(buf
, NULL
, 0);
169 if (options
& GET_COMPONENT
) {
170 strcpy(base
, "component_size");
171 if (load_sys(fname
, buf
))
173 sra
->component_size
= strtoull(buf
, NULL
, 0);
174 /* sysfs reports "K", but we want sectors */
175 sra
->component_size
*= 2;
177 if (options
& GET_CHUNK
) {
178 strcpy(base
, "chunk_size");
179 if (load_sys(fname
, buf
))
181 sra
->array
.chunk_size
= strtoul(buf
, NULL
, 0);
183 if (options
& GET_CACHE
) {
184 strcpy(base
, "stripe_cache_size");
185 if (load_sys(fname
, buf
))
187 sra
->cache_size
= strtoul(buf
, NULL
, 0);
189 if (options
& GET_MISMATCH
) {
190 strcpy(base
, "mismatch_cnt");
191 if (load_sys(fname
, buf
))
193 sra
->mismatch_cnt
= strtoul(buf
, NULL
, 0);
195 if (options
& GET_SAFEMODE
) {
202 strcpy(base
, "safe_mode_delay");
203 if (load_sys(fname
, buf
))
206 /* remove a period, and count digits after it */
208 for (i
= 0; i
< len
; i
++) {
210 if (isdigit(buf
[i
])) {
215 } else if (buf
[i
] == '.') {
220 msec
= strtoul(buf
, NULL
, 10);
221 msec
= (msec
* 1000) / scale
;
222 sra
->safe_mode_delay
= msec
;
225 if (! (options
& GET_DEVS
))
228 /* Get all the devices as well */
230 dir
= opendir(fname
);
233 sra
->array
.spare_disks
= 0;
235 while ((de
= readdir(dir
)) != NULL
) {
237 if (de
->d_ino
== 0 ||
238 strncmp(de
->d_name
, "dev-", 4) != 0)
240 strcpy(base
, de
->d_name
);
241 dbase
= base
+ strlen(base
);
244 dev
= malloc(sizeof(*dev
));
248 /* Always get slot, major, minor */
249 strcpy(dbase
, "slot");
250 if (load_sys(fname
, buf
)) {
251 /* hmm... unable to read 'slot' maybe the device
254 strcpy(dbase
, "block");
255 if (readlink(fname
, buf
, sizeof(buf
)) < 0 &&
256 errno
!= ENAMETOOLONG
) {
257 /* ...yup device is gone */
261 /* slot is unreadable but 'block' link
262 * still intact... something bad is happening
270 strcpy(dev
->sys_name
, de
->d_name
);
271 dev
->disk
.raid_disk
= strtoul(buf
, &ep
, 10);
272 if (*ep
) dev
->disk
.raid_disk
= -1;
274 strcpy(dbase
, "block/dev");
275 if (load_sys(fname
, buf
)) {
276 /* assume this is a stale reference to a hot
282 sscanf(buf
, "%d:%d", &dev
->disk
.major
, &dev
->disk
.minor
);
284 /* special case check for block devices that can go 'offline' */
285 strcpy(dbase
, "block/device/state");
286 if (load_sys(fname
, buf
) == 0 &&
287 strncmp(buf
, "offline", 7) == 0) {
292 /* finally add this disk to the array */
293 dev
->next
= sra
->devs
;
296 if (options
& GET_OFFSET
) {
297 strcpy(dbase
, "offset");
298 if (load_sys(fname
, buf
))
300 dev
->data_offset
= strtoull(buf
, NULL
, 0);
302 if (options
& GET_SIZE
) {
303 strcpy(dbase
, "size");
304 if (load_sys(fname
, buf
))
306 dev
->component_size
= strtoull(buf
, NULL
, 0) * 2;
308 if (options
& GET_STATE
) {
310 strcpy(dbase
, "state");
311 if (load_sys(fname
, buf
))
313 if (strstr(buf
, "in_sync"))
314 dev
->disk
.state
|= (1<<MD_DISK_SYNC
);
315 if (strstr(buf
, "faulty"))
316 dev
->disk
.state
|= (1<<MD_DISK_FAULTY
);
317 if (dev
->disk
.state
== 0)
318 sra
->array
.spare_disks
++;
320 if (options
& GET_ERROR
) {
321 strcpy(buf
, "errors");
322 if (load_sys(fname
, buf
))
324 dev
->errors
= strtoul(buf
, NULL
, 0);
337 int sysfs_attr_match(const char *attr
, const char *str
)
339 /* See if attr, read from a sysfs file, matches
340 * str. They must either be the same, or attr can
341 * have a trailing newline or comma
343 while (*attr
&& *str
&& *attr
== *str
) {
348 if (*str
|| (*attr
&& *attr
!= ',' && *attr
!= '\n'))
353 int sysfs_match_word(const char *word
, char **list
)
356 for (n
=0; list
[n
]; n
++)
357 if (sysfs_attr_match(word
, list
[n
]))
362 unsigned long long get_component_size(int fd
)
364 /* Find out the component size of the array.
365 * We cannot trust GET_ARRAY_INFO ioctl as it's
366 * size field is only 32bits.
367 * So look in /sys/block/mdXXX/md/component_size
369 * This returns in units of sectors.
374 if (fstat(fd
, &stb
)) return 0;
375 if (major(stb
.st_rdev
) != (unsigned)get_mdp_major())
376 sprintf(fname
, "/sys/block/md%d/md/component_size",
377 (int)minor(stb
.st_rdev
));
379 sprintf(fname
, "/sys/block/md_d%d/md/component_size",
380 (int)minor(stb
.st_rdev
)>>MdpMinorShift
);
381 fd
= open(fname
, O_RDONLY
);
384 n
= read(fd
, fname
, sizeof(fname
));
386 if (n
== sizeof(fname
))
389 return strtoull(fname
, NULL
, 10) * 2;
392 int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
393 char *name
, char *val
)
399 sprintf(fname
, "/sys/block/%s/md/%s/%s",
400 sra
->sys_name
, dev
?dev
->sys_name
:"", name
);
401 fd
= open(fname
, O_WRONLY
);
404 n
= write(fd
, val
, strlen(val
));
406 if (n
!= strlen(val
)) {
407 dprintf(Name
": failed to write '%s' to '%s' (%s)\n",
408 val
, fname
, strerror(errno
));
414 int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
415 char *name
, unsigned long long val
)
418 sprintf(valstr
, "%llu", val
);
419 return sysfs_set_str(sra
, dev
, name
, valstr
);
422 int sysfs_uevent(struct mdinfo
*sra
, char *event
)
428 sprintf(fname
, "/sys/block/%s/uevent",
430 fd
= open(fname
, O_WRONLY
);
433 n
= write(fd
, event
, strlen(event
));
438 int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
444 sprintf(fname
, "/sys/block/%s/md/%s/%s",
445 sra
->sys_name
, dev
?dev
->sys_name
:"", name
);
446 fd
= open(fname
, O_RDWR
);
448 fd
= open(fname
, O_RDONLY
);
452 int sysfs_fd_get_ll(int fd
, unsigned long long *val
)
459 n
= read(fd
, buf
, sizeof(buf
));
463 *val
= strtoull(buf
, &ep
, 0);
464 if (ep
== buf
|| (*ep
!= 0 && *ep
!= '\n' && *ep
!= ' '))
469 int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
470 char *name
, unsigned long long *val
)
475 fd
= sysfs_get_fd(sra
, dev
, name
);
478 n
= sysfs_fd_get_ll(fd
, val
);
483 int sysfs_fd_get_str(int fd
, char *val
, int size
)
488 n
= read(fd
, val
, size
);
495 int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
496 char *name
, char *val
, int size
)
501 fd
= sysfs_get_fd(sra
, dev
, name
);
504 n
= sysfs_fd_get_str(fd
, val
, size
);
509 int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
)
518 sprintf(delay
, "%ld.%03ld\n", sec
, msec
);
519 /* this '\n' ^ needed for kernels older than 2.6.28 */
520 return sysfs_set_str(sra
, NULL
, "safe_mode_delay", delay
);
523 int sysfs_set_array(struct mdinfo
*info
, int vers
)
529 if (info
->array
.major_version
== -1 &&
530 info
->array
.minor_version
== -2) {
531 strcat(strcpy(ver
, "external:"), info
->text_version
);
533 if ((vers
% 100) < 2 ||
534 sysfs_set_str(info
, NULL
, "metadata_version",
536 fprintf(stderr
, Name
": This kernel does not "
537 "support external metadata.\n");
541 if (info
->array
.level
< 0)
542 return 0; /* FIXME */
543 rv
|= sysfs_set_str(info
, NULL
, "level",
544 map_num(pers
, info
->array
.level
));
545 rv
|= sysfs_set_num(info
, NULL
, "raid_disks", info
->array
.raid_disks
);
546 rv
|= sysfs_set_num(info
, NULL
, "chunk_size", info
->array
.chunk_size
);
547 rv
|= sysfs_set_num(info
, NULL
, "layout", info
->array
.layout
);
548 rv
|= sysfs_set_num(info
, NULL
, "component_size", info
->component_size
/2);
549 if (info
->custom_array_size
) {
552 rc
= sysfs_set_num(info
, NULL
, "array_size",
553 info
->custom_array_size
/2);
554 if (rc
&& errno
== ENOENT
) {
555 fprintf(stderr
, Name
": This kernel does not "
556 "have the md/array_size attribute, "
557 "the array may be larger than expected\n");
563 if (info
->array
.level
> 0)
564 rv
|= sysfs_set_num(info
, NULL
, "resync_start", info
->resync_start
);
568 int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
)
575 sprintf(dv
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
576 rv
= sysfs_set_str(sra
, NULL
, "new_dev", dv
);
580 memset(nm
, 0, sizeof(nm
));
581 sprintf(dv
, "/sys/dev/block/%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
582 rv
= readlink(dv
, nm
, sizeof(nm
));
586 dname
= strrchr(nm
, '/');
588 strcpy(sd
->sys_name
, "dev-");
589 strcpy(sd
->sys_name
+4, dname
);
591 /* test write to see if 'recovery_start' is available */
592 if (resume
&& sd
->recovery_start
< MaxSector
&&
593 sysfs_set_num(sra
, sd
, "recovery_start", 0)) {
594 sysfs_set_str(sra
, sd
, "state", "remove");
598 rv
= sysfs_set_num(sra
, sd
, "offset", sd
->data_offset
);
599 rv
|= sysfs_set_num(sra
, sd
, "size", (sd
->component_size
+1) / 2);
600 if (sra
->array
.level
!= LEVEL_CONTAINER
) {
601 if (sd
->recovery_start
== MaxSector
)
602 /* This can correctly fail if array isn't started,
603 * yet, so just ignore status for now.
605 sysfs_set_str(sra
, sd
, "state", "insync");
606 rv
|= sysfs_set_num(sra
, sd
, "slot", sd
->disk
.raid_disk
);
608 sysfs_set_num(sra
, sd
, "recovery_start", sd
->recovery_start
);
614 int sysfs_disk_to_sg(int fd
)
616 /* from an open block device, try find and open its corresponding
617 * scsi_generic interface
622 char sg_major_minor
[8];
626 int major
, minor
, rv
;
631 snprintf(path
, sizeof(path
), "/sys/dev/block/%d:%d/device",
632 major(st
.st_rdev
), minor(st
.st_rdev
));
640 if (strncmp("scsi_generic:", de
->d_name
,
641 strlen("scsi_generic:")) == 0)
650 snprintf(sg_path
, sizeof(sg_path
), "%s/%s/dev", path
, de
->d_name
);
651 fd
= open(sg_path
, O_RDONLY
);
655 rv
= read(fd
, sg_major_minor
, sizeof(sg_major_minor
));
660 sg_major_minor
[rv
- 1] = '\0';
662 c
= strchr(sg_major_minor
, ':');
665 major
= strtol(sg_major_minor
, NULL
, 10);
666 minor
= strtol(c
, NULL
, 10);
667 snprintf(path
, sizeof(path
), "/dev/.tmp.md.%d:%d:%d",
668 (int) getpid(), major
, minor
);
669 if (mknod(path
, S_IFCHR
|0600, makedev(major
, minor
))==0) {
670 fd
= open(path
, O_RDONLY
);
679 int sysfs_disk_to_scsi_id(int fd
, __u32
*id
)
681 /* from an open block device, try to retrieve it scsi_id */
691 snprintf(path
, sizeof(path
), "/sys/dev/block/%d:%d/device",
692 major(st
.st_rdev
), minor(st
.st_rdev
));
700 if (strncmp("scsi_disk:", de
->d_name
,
701 strlen("scsi_disk:")) == 0)
710 c1
= strchr(de
->d_name
, ':');
712 c2
= strchr(c1
, ':');
714 *id
= strtol(c1
, NULL
, 10) << 24; /* host */
716 c2
= strchr(c1
, ':');
718 *id
|= strtol(c1
, NULL
, 10) << 16; /* channel */
720 c2
= strchr(c1
, ':');
722 *id
|= strtol(c1
, NULL
, 10) << 8; /* lun */
724 *id
|= strtol(c1
, NULL
, 10); /* id */
730 int sysfs_unique_holder(int devnum
, long rdev
)
732 /* Check that devnum is a holder of rdev,
733 * and is the only holder.
734 * we should be locked against races by
735 * an O_EXCL on devnum
742 sprintf(dirname
, "/sys/dev/block/%d:%d/holders",
743 major(rdev
), minor(rdev
));
744 dir
= opendir(dirname
);
749 while ((de
= readdir(dir
)) != NULL
) {
758 if (de
->d_name
[0] == '.')
760 strcpy(dirname
+l
, "/");
761 strcat(dirname
+l
, de
->d_name
);
762 strcat(dirname
+l
, "/dev");
763 fd
= open(dirname
, O_RDONLY
);
768 n
= read(fd
, buf
, sizeof(buf
)-1);
771 if (sscanf(buf
, "%d:%d%c", &mj
, &mn
, &c
) != 3 ||
794 static char *clean_states
[] = {
795 "clear", "inactive", "readonly", "read-auto", "clean", NULL
};
797 int WaitClean(char *dev
, int sock
, int verbose
)
804 fd
= open(dev
, O_RDONLY
);
807 fprintf(stderr
, Name
": Couldn't open %s: %s\n", dev
, strerror(errno
));
811 devnum
= fd2devnum(fd
);
812 mdi
= sysfs_read(fd
, devnum
, GET_VERSION
|GET_LEVEL
|GET_SAFEMODE
);
815 fprintf(stderr
, Name
": Failed to read sysfs attributes for "
821 switch(mdi
->array
.level
) {
823 case LEVEL_MULTIPATH
:
825 /* safemode delay is irrelevant for these levels */
830 /* for internal metadata the kernel handles the final clean
831 * transition, containers can never be dirty
833 if (!is_subarray(mdi
->text_version
))
836 /* safemode disabled ? */
837 if (mdi
->safe_mode_delay
== 0)
841 int state_fd
= sysfs_open(fd2devnum(fd
), NULL
, "array_state");
846 /* minimize the safe_mode_delay and prepare to wait up to 5s
847 * for writes to quiesce
849 sysfs_set_safemode(mdi
, 1);
855 /* wait for array_state to be clean */
857 rv
= read(state_fd
, buf
, sizeof(buf
));
860 if (sysfs_match_word(buf
, clean_states
) <= 4)
862 FD_SET(state_fd
, &fds
);
863 rv
= select(state_fd
+ 1, NULL
, NULL
, &fds
, &tm
);
864 if (rv
< 0 && errno
!= EINTR
)
866 lseek(state_fd
, 0, SEEK_SET
);
870 else if (fping_monitor(sock
) == 0 ||
871 ping_monitor(mdi
->text_version
) == 0) {
872 /* we need to ping to close the window between array
873 * state transitioning to clean and the metadata being
880 fprintf(stderr
, Name
": Error waiting for %s to be clean\n",
883 /* restore the original safe_mode_delay */
884 sysfs_set_safemode(mdi
, mdi
->safe_mode_delay
);
893 #endif /* MDASSEMBLE */