]>
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
)
93 fmt_devname(mdi
->sys_name
, devnum
);
97 struct mdinfo
*sysfs_read(int fd
, int devnum
, unsigned long options
)
108 sra
= xcalloc(1, sizeof(*sra
));
109 sysfs_init(sra
, fd
, devnum
);
110 if (sra
->sys_name
[0] == 0) {
115 sprintf(fname
, "/sys/block/%s/md/", sra
->sys_name
);
116 base
= fname
+ strlen(fname
);
119 if (options
& GET_VERSION
) {
120 strcpy(base
, "metadata_version");
121 if (load_sys(fname
, buf
))
123 if (strncmp(buf
, "none", 4) == 0) {
124 sra
->array
.major_version
=
125 sra
->array
.minor_version
= -1;
126 strcpy(sra
->text_version
, "");
127 } else if (strncmp(buf
, "external:", 9) == 0) {
128 sra
->array
.major_version
= -1;
129 sra
->array
.minor_version
= -2;
130 strcpy(sra
->text_version
, buf
+9);
133 &sra
->array
.major_version
,
134 &sra
->array
.minor_version
);
135 strcpy(sra
->text_version
, buf
);
138 if (options
& GET_LEVEL
) {
139 strcpy(base
, "level");
140 if (load_sys(fname
, buf
))
142 sra
->array
.level
= map_name(pers
, buf
);
144 if (options
& GET_LAYOUT
) {
145 strcpy(base
, "layout");
146 if (load_sys(fname
, buf
))
148 sra
->array
.layout
= strtoul(buf
, NULL
, 0);
150 if (options
& GET_DISKS
) {
151 strcpy(base
, "raid_disks");
152 if (load_sys(fname
, buf
))
154 sra
->array
.raid_disks
= strtoul(buf
, NULL
, 0);
156 if (options
& GET_DEGRADED
) {
157 strcpy(base
, "degraded");
158 if (load_sys(fname
, buf
))
160 sra
->array
.failed_disks
= strtoul(buf
, NULL
, 0);
162 if (options
& GET_COMPONENT
) {
163 strcpy(base
, "component_size");
164 if (load_sys(fname
, buf
))
166 sra
->component_size
= strtoull(buf
, NULL
, 0);
167 /* sysfs reports "K", but we want sectors */
168 sra
->component_size
*= 2;
170 if (options
& GET_CHUNK
) {
171 strcpy(base
, "chunk_size");
172 if (load_sys(fname
, buf
))
174 sra
->array
.chunk_size
= strtoul(buf
, NULL
, 0);
176 if (options
& GET_CACHE
) {
177 strcpy(base
, "stripe_cache_size");
178 if (load_sys(fname
, buf
))
180 sra
->cache_size
= strtoul(buf
, NULL
, 0);
182 if (options
& GET_MISMATCH
) {
183 strcpy(base
, "mismatch_cnt");
184 if (load_sys(fname
, buf
))
186 sra
->mismatch_cnt
= strtoul(buf
, NULL
, 0);
188 if (options
& GET_SAFEMODE
) {
195 strcpy(base
, "safe_mode_delay");
196 if (load_sys(fname
, buf
))
199 /* remove a period, and count digits after it */
201 for (i
= 0; i
< len
; i
++) {
203 if (isdigit(buf
[i
])) {
208 } else if (buf
[i
] == '.') {
213 msec
= strtoul(buf
, NULL
, 10);
214 msec
= (msec
* 1000) / scale
;
215 sra
->safe_mode_delay
= msec
;
217 if (options
& GET_BITMAP_LOCATION
) {
218 strcpy(base
, "bitmap/location");
219 if (load_sys(fname
, buf
))
221 if (strncmp(buf
, "file", 4) == 0)
222 sra
->bitmap_offset
= 1;
223 else if (strncmp(buf
, "none", 4) == 0)
224 sra
->bitmap_offset
= 0;
225 else if (buf
[0] == '+')
226 sra
->bitmap_offset
= strtol(buf
+1, NULL
, 10);
231 if (! (options
& GET_DEVS
))
234 /* Get all the devices as well */
236 dir
= opendir(fname
);
239 sra
->array
.spare_disks
= 0;
241 while ((de
= readdir(dir
)) != NULL
) {
243 if (de
->d_ino
== 0 ||
244 strncmp(de
->d_name
, "dev-", 4) != 0)
246 strcpy(base
, de
->d_name
);
247 dbase
= base
+ strlen(base
);
250 dev
= xmalloc(sizeof(*dev
));
252 /* Always get slot, major, minor */
253 strcpy(dbase
, "slot");
254 if (load_sys(fname
, buf
)) {
255 /* hmm... unable to read 'slot' maybe the device
258 strcpy(dbase
, "block");
259 if (readlink(fname
, buf
, sizeof(buf
)) < 0 &&
260 errno
!= ENAMETOOLONG
) {
261 /* ...yup device is gone */
265 /* slot is unreadable but 'block' link
266 * still intact... something bad is happening
274 strcpy(dev
->sys_name
, de
->d_name
);
275 dev
->disk
.raid_disk
= strtoul(buf
, &ep
, 10);
276 if (*ep
) dev
->disk
.raid_disk
= -1;
278 strcpy(dbase
, "block/dev");
279 if (load_sys(fname
, buf
)) {
280 /* assume this is a stale reference to a hot
286 sscanf(buf
, "%d:%d", &dev
->disk
.major
, &dev
->disk
.minor
);
288 /* special case check for block devices that can go 'offline' */
289 strcpy(dbase
, "block/device/state");
290 if (load_sys(fname
, buf
) == 0 &&
291 strncmp(buf
, "offline", 7) == 0) {
296 /* finally add this disk to the array */
297 dev
->next
= sra
->devs
;
300 if (options
& GET_OFFSET
) {
301 strcpy(dbase
, "offset");
302 if (load_sys(fname
, buf
))
304 dev
->data_offset
= strtoull(buf
, NULL
, 0);
306 if (options
& GET_SIZE
) {
307 strcpy(dbase
, "size");
308 if (load_sys(fname
, buf
))
310 dev
->component_size
= strtoull(buf
, NULL
, 0) * 2;
312 if (options
& GET_STATE
) {
314 strcpy(dbase
, "state");
315 if (load_sys(fname
, buf
))
317 if (strstr(buf
, "in_sync"))
318 dev
->disk
.state
|= (1<<MD_DISK_SYNC
);
319 if (strstr(buf
, "faulty"))
320 dev
->disk
.state
|= (1<<MD_DISK_FAULTY
);
321 if (dev
->disk
.state
== 0)
322 sra
->array
.spare_disks
++;
324 if (options
& GET_ERROR
) {
325 strcpy(buf
, "errors");
326 if (load_sys(fname
, buf
))
328 dev
->errors
= strtoul(buf
, NULL
, 0);
341 int sysfs_attr_match(const char *attr
, const char *str
)
343 /* See if attr, read from a sysfs file, matches
344 * str. They must either be the same, or attr can
345 * have a trailing newline or comma
347 while (*attr
&& *str
&& *attr
== *str
) {
352 if (*str
|| (*attr
&& *attr
!= ',' && *attr
!= '\n'))
357 int sysfs_match_word(const char *word
, char **list
)
360 for (n
=0; list
[n
]; n
++)
361 if (sysfs_attr_match(word
, list
[n
]))
366 unsigned long long get_component_size(int fd
)
368 /* Find out the component size of the array.
369 * We cannot trust GET_ARRAY_INFO ioctl as it's
370 * size field is only 32bits.
371 * So look in /sys/block/mdXXX/md/component_size
373 * This returns in units of sectors.
378 if (fstat(fd
, &stb
)) return 0;
379 if (major(stb
.st_rdev
) != (unsigned)get_mdp_major())
380 sprintf(fname
, "/sys/block/md%d/md/component_size",
381 (int)minor(stb
.st_rdev
));
383 sprintf(fname
, "/sys/block/md_d%d/md/component_size",
384 (int)minor(stb
.st_rdev
)>>MdpMinorShift
);
385 fd
= open(fname
, O_RDONLY
);
388 n
= read(fd
, fname
, sizeof(fname
));
390 if (n
< 0 || n
== sizeof(fname
))
393 return strtoull(fname
, NULL
, 10) * 2;
396 int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
397 char *name
, char *val
)
403 sprintf(fname
, "/sys/block/%s/md/%s/%s",
404 sra
->sys_name
, dev
?dev
->sys_name
:"", name
);
405 fd
= open(fname
, O_WRONLY
);
408 n
= write(fd
, val
, strlen(val
));
410 if (n
!= strlen(val
)) {
411 dprintf(Name
": failed to write '%s' to '%s' (%s)\n",
412 val
, fname
, strerror(errno
));
418 int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
419 char *name
, unsigned long long val
)
422 sprintf(valstr
, "%llu", val
);
423 return sysfs_set_str(sra
, dev
, name
, valstr
);
426 int sysfs_set_num_signed(struct mdinfo
*sra
, struct mdinfo
*dev
,
427 char *name
, long long val
)
430 sprintf(valstr
, "%lli", val
);
431 return sysfs_set_str(sra
, dev
, name
, valstr
);
434 int sysfs_uevent(struct mdinfo
*sra
, char *event
)
440 sprintf(fname
, "/sys/block/%s/uevent",
442 fd
= open(fname
, O_WRONLY
);
445 n
= write(fd
, event
, strlen(event
));
447 if (n
!= (int)strlen(event
)) {
448 dprintf(Name
": failed to write '%s' to '%s' (%s)\n",
449 event
, fname
, strerror(errno
));
455 int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
, char *name
)
460 sprintf(fname
, "/sys/block/%s/md/%s/%s",
461 sra
->sys_name
, dev
?dev
->sys_name
:"", name
);
463 return stat(fname
, &st
) == 0;
466 int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
472 sprintf(fname
, "/sys/block/%s/md/%s/%s",
473 sra
->sys_name
, dev
?dev
->sys_name
:"", name
);
474 fd
= open(fname
, O_RDWR
);
476 fd
= open(fname
, O_RDONLY
);
480 int sysfs_fd_get_ll(int fd
, unsigned long long *val
)
487 n
= read(fd
, buf
, sizeof(buf
));
491 *val
= strtoull(buf
, &ep
, 0);
492 if (ep
== buf
|| (*ep
!= 0 && *ep
!= '\n' && *ep
!= ' '))
497 int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
498 char *name
, unsigned long long *val
)
503 fd
= sysfs_get_fd(sra
, dev
, name
);
506 n
= sysfs_fd_get_ll(fd
, val
);
511 int sysfs_fd_get_str(int fd
, char *val
, int size
)
516 n
= read(fd
, val
, size
);
523 int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
524 char *name
, char *val
, int size
)
529 fd
= sysfs_get_fd(sra
, dev
, name
);
532 n
= sysfs_fd_get_str(fd
, val
, size
);
537 int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
)
546 sprintf(delay
, "%ld.%03ld\n", sec
, msec
);
547 /* this '\n' ^ needed for kernels older than 2.6.28 */
548 return sysfs_set_str(sra
, NULL
, "safe_mode_delay", delay
);
551 int sysfs_set_array(struct mdinfo
*info
, int vers
)
555 int raid_disks
= info
->array
.raid_disks
;
558 if (info
->array
.major_version
== -1 &&
559 info
->array
.minor_version
== -2) {
562 strcat(strcpy(ver
, "external:"), info
->text_version
);
564 /* meta version might already be set if we are setting
565 * new geometry for a reshape. In that case we don't
566 * want to over-write the 'readonly' flag that is
567 * stored in the metadata version. So read the current
568 * version first, and preserve the flag
570 if (sysfs_get_str(info
, NULL
, "metadata_version",
572 if (strlen(buf
) >= 9 && buf
[9] == '-')
575 if ((vers
% 100) < 2 ||
576 sysfs_set_str(info
, NULL
, "metadata_version",
578 pr_err("This kernel does not "
579 "support external metadata.\n");
583 if (info
->array
.level
< 0)
584 return 0; /* FIXME */
585 rv
|= sysfs_set_str(info
, NULL
, "level",
586 map_num(pers
, info
->array
.level
));
587 if (info
->reshape_active
&& info
->delta_disks
!= UnSet
)
588 raid_disks
-= info
->delta_disks
;
589 rv
|= sysfs_set_num(info
, NULL
, "raid_disks", raid_disks
);
590 rv
|= sysfs_set_num(info
, NULL
, "chunk_size", info
->array
.chunk_size
);
591 rv
|= sysfs_set_num(info
, NULL
, "layout", info
->array
.layout
);
592 rv
|= sysfs_set_num(info
, NULL
, "component_size", info
->component_size
/2);
593 if (info
->custom_array_size
) {
596 rc
= sysfs_set_num(info
, NULL
, "array_size",
597 info
->custom_array_size
/2);
598 if (rc
&& errno
== ENOENT
) {
599 pr_err("This kernel does not "
600 "have the md/array_size attribute, "
601 "the array may be larger than expected\n");
607 if (info
->array
.level
> 0)
608 rv
|= sysfs_set_num(info
, NULL
, "resync_start", info
->resync_start
);
610 if (info
->reshape_active
) {
611 rv
|= sysfs_set_num(info
, NULL
, "reshape_position",
612 info
->reshape_progress
);
613 rv
|= sysfs_set_num(info
, NULL
, "chunk_size", info
->new_chunk
);
614 rv
|= sysfs_set_num(info
, NULL
, "layout", info
->new_layout
);
615 rv
|= sysfs_set_num(info
, NULL
, "raid_disks",
616 info
->array
.raid_disks
);
617 /* We don't set 'new_level' here. That can only happen
618 * once the reshape completes.
624 int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
)
631 sprintf(dv
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
632 rv
= sysfs_set_str(sra
, NULL
, "new_dev", dv
);
636 memset(nm
, 0, sizeof(nm
));
637 sprintf(dv
, "/sys/dev/block/%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
638 rv
= readlink(dv
, nm
, sizeof(nm
)-1);
642 dname
= strrchr(nm
, '/');
644 strcpy(sd
->sys_name
, "dev-");
645 strcpy(sd
->sys_name
+4, dname
);
647 /* test write to see if 'recovery_start' is available */
648 if (resume
&& sd
->recovery_start
< MaxSector
&&
649 sysfs_set_num(sra
, sd
, "recovery_start", 0)) {
650 sysfs_set_str(sra
, sd
, "state", "remove");
654 rv
= sysfs_set_num(sra
, sd
, "offset", sd
->data_offset
);
655 rv
|= sysfs_set_num(sra
, sd
, "size", (sd
->component_size
+1) / 2);
656 if (sra
->array
.level
!= LEVEL_CONTAINER
) {
657 if (sd
->recovery_start
== MaxSector
)
658 /* This can correctly fail if array isn't started,
659 * yet, so just ignore status for now.
661 sysfs_set_str(sra
, sd
, "state", "insync");
662 if (sd
->disk
.raid_disk
>= 0)
663 rv
|= sysfs_set_num(sra
, sd
, "slot", sd
->disk
.raid_disk
);
665 sysfs_set_num(sra
, sd
, "recovery_start", sd
->recovery_start
);
671 int sysfs_disk_to_sg(int fd
)
673 /* from an open block device, try find and open its corresponding
674 * scsi_generic interface
679 char sg_major_minor
[8];
683 int major
, minor
, rv
;
688 snprintf(path
, sizeof(path
), "/sys/dev/block/%d:%d/device",
689 major(st
.st_rdev
), minor(st
.st_rdev
));
697 if (strncmp("scsi_generic:", de
->d_name
,
698 strlen("scsi_generic:")) == 0)
707 snprintf(sg_path
, sizeof(sg_path
), "%s/%s/dev", path
, de
->d_name
);
708 fd
= open(sg_path
, O_RDONLY
);
712 rv
= read(fd
, sg_major_minor
, sizeof(sg_major_minor
));
717 sg_major_minor
[rv
- 1] = '\0';
719 c
= strchr(sg_major_minor
, ':');
722 major
= strtol(sg_major_minor
, NULL
, 10);
723 minor
= strtol(c
, NULL
, 10);
724 snprintf(path
, sizeof(path
), "/dev/.tmp.md.%d:%d:%d",
725 (int) getpid(), major
, minor
);
726 if (mknod(path
, S_IFCHR
|0600, makedev(major
, minor
))==0) {
727 fd
= open(path
, O_RDONLY
);
736 int sysfs_disk_to_scsi_id(int fd
, __u32
*id
)
738 /* from an open block device, try to retrieve it scsi_id */
743 int host
, bus
, target
, lun
;
748 snprintf(path
, sizeof(path
), "/sys/dev/block/%d:%d/device/scsi_device",
749 major(st
.st_rdev
), minor(st
.st_rdev
));
755 for (de
= readdir(dir
); de
; de
= readdir(dir
)) {
758 if (de
->d_type
!= DT_DIR
)
761 count
= sscanf(de
->d_name
, "%d:%d:%d:%d", &host
, &bus
, &target
, &lun
);
770 *id
= (host
<< 24) | (bus
<< 16) | (target
<< 8) | (lun
<< 0);
775 int sysfs_unique_holder(int devnum
, long rdev
)
777 /* Check that devnum is a holder of rdev,
778 * and is the only holder.
779 * we should be locked against races by
780 * an O_EXCL on devnum
787 sprintf(dirname
, "/sys/dev/block/%d:%d/holders",
788 major(rdev
), minor(rdev
));
789 dir
= opendir(dirname
);
794 while ((de
= readdir(dir
)) != NULL
) {
803 if (de
->d_name
[0] == '.')
805 strcpy(dirname
+l
, "/");
806 strcat(dirname
+l
, de
->d_name
);
807 strcat(dirname
+l
, "/dev");
808 fd
= open(dirname
, O_RDONLY
);
813 n
= read(fd
, buf
, sizeof(buf
)-1);
818 if (sscanf(buf
, "%d:%d%c", &mj
, &mn
, &c
) != 3 ||
839 int sysfs_freeze_array(struct mdinfo
*sra
)
841 /* Try to freeze resync/rebuild on this array/container.
842 * Return -1 if the array is busy,
843 * return 0 if this kernel doesn't support 'frozen'
844 * return 1 if it worked.
848 if (!sysfs_attribute_available(sra
, NULL
, "sync_action"))
849 return 1; /* no sync_action == frozen */
850 if (sysfs_get_str(sra
, NULL
, "sync_action", buf
, 20) <= 0)
852 if (strcmp(buf
, "frozen\n") == 0)
855 if (strcmp(buf
, "idle\n") != 0)
857 if (sysfs_set_str(sra
, NULL
, "sync_action", "frozen") < 0)