[thirdparty/mdadm.git] / sysfs.c

/*
 * sysfs - extract md related information from sysfs.  Part of:
 * mdadm - manage Linux "md" devices aka RAID arrays.
 *
 * Copyright (C) 2006 Neil Brown <neilb@suse.de>
 *
 *
 *    This program is free software; you can redistribute it and/or modify
 *    it under the terms of the GNU General Public License as published by
 *    the Free Software Foundation; either version 2 of the License, or
 *    (at your option) any later version.
 *
 *    This program is distributed in the hope that it will be useful,
 *    but WITHOUT ANY WARRANTY; without even the implied warranty of
 *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *    GNU General Public License for more details.
 *
 *    You should have received a copy of the GNU General Public License
 *    along with this program; if not, write to the Free Software
 *    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 *    Author: Neil Brown
 *    Email: <neilb@suse.de>
 */

#include	"mdadm.h"
#include	<dirent.h>
#include	<ctype.h>

int load_sys(char *path, char *buf)
{
	int fd = open(path, O_RDONLY);
	int n;
	if (fd < 0)
		return -1;
	n = read(fd, buf, 1024);
	close(fd);
	if (n <0 || n >= 1024)
		return -1;
	buf[n] = 0;
	if (n && buf[n-1] == '\n')
		buf[n-1] = 0;
	return 0;
}

void sysfs_free(struct mdinfo *sra)
{
	while (sra) {
		struct mdinfo *sra2 = sra->next;
		while (sra->devs) {
			struct mdinfo *d = sra->devs;
			sra->devs = d->next;
			free(d);
		}
		free(sra);
		sra = sra2;
	}
}

int sysfs_open(int devnum, char *devname, char *attr)
{
	char fname[50];
	int fd;
	char *mdname = devnum2devname(devnum);

	if (!mdname)
		return -1;

	sprintf(fname, "/sys/block/%s/md/", mdname);
	if (devname) {
		strcat(fname, devname);
		strcat(fname, "/");
	}
	strcat(fname, attr);
	fd = open(fname, O_RDWR);
	if (fd < 0 && errno == EACCES)
		fd = open(fname, O_RDONLY);
	free(mdname);
	return fd;
}

void sysfs_init(struct mdinfo *mdi, int fd, int devnum)
{
	if (fd >= 0) {
		struct stat stb;
		mdu_version_t vers;
 		if (fstat(fd, &stb))
			return;
		if (ioctl(fd, RAID_VERSION, &vers) != 0)
			return;
		if (major(stb.st_rdev) == MD_MAJOR)
			sprintf(mdi->sys_name, "md%d", (int)minor(stb.st_rdev));
		else if (major(stb.st_rdev) == get_mdp_major())
			sprintf(mdi->sys_name, "md_d%d",
				(int)minor(stb.st_rdev)>>MdpMinorShift);
		else {
			/* must be an extended-minor partition. Look at the
			 * /sys/dev/block/%d:%d link which must look like
			 * ../../block/mdXXX/mdXXXpYY
			 */
			char path[30];
			char link[200];
			char *cp;
			int n;
			sprintf(path, "/sys/dev/block/%d:%d", major(stb.st_rdev),
				minor(stb.st_rdev));
			n = readlink(path, link, sizeof(link)-1);
			if (n <= 0)
				return;
			link[n] = 0;
			cp = strrchr(link, '/');
			if (cp) *cp = 0;
			cp = strchr(link, '/');
			if (cp && strncmp(cp, "/md", 3) == 0)
				strcpy(mdi->sys_name, cp+1);
			else
				return;
		}
	} else {
		if (devnum >= 0)
			sprintf(mdi->sys_name, "md%d", devnum);
		else
			sprintf(mdi->sys_name, "md_d%d",
				-1-devnum);
	}
}

struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options)
{
	/* Longest possible name in sysfs, mounted at /sys, is
	 *  /sys/block/md_dXXX/md/dev-XXXXX/block/dev
	 *  /sys/block/md_dXXX/md/metadata_version
	 * which is about 41 characters.  50 should do for now
	 */
	char fname[50];
	char buf[1024];
	char *base;
	char *dbase;
	struct mdinfo *sra;
	struct mdinfo *dev;
	DIR *dir = NULL;
	struct dirent *de;

	sra = malloc(sizeof(*sra));
	if (sra == NULL)
		return sra;
	memset(sra, 0, sizeof(*sra));
	sysfs_init(sra, fd, devnum);

	sprintf(fname, "/sys/block/%s/md/", sra->sys_name);
	base = fname + strlen(fname);

	sra->devs = NULL;
	if (options & GET_VERSION) {
		strcpy(base, "metadata_version");
		if (load_sys(fname, buf))
			goto abort;
		if (strncmp(buf, "none", 4) == 0) {
			sra->array.major_version =
				sra->array.minor_version = -1;
			strcpy(sra->text_version, "");
		} else if (strncmp(buf, "external:", 9) == 0) {
			sra->array.major_version = -1;
			sra->array.minor_version = -2;
			strcpy(sra->text_version, buf+9);
		} else {
			sscanf(buf, "%d.%d",
			       &sra->array.major_version,
			       &sra->array.minor_version);
			strcpy(sra->text_version, buf);
		}
	}
	if (options & GET_LEVEL) {
		strcpy(base, "level");
		if (load_sys(fname, buf))
			goto abort;
		sra->array.level = map_name(pers, buf);
	}
	if (options & GET_LAYOUT) {
		strcpy(base, "layout");
		if (load_sys(fname, buf))
			goto abort;
		sra->array.layout = strtoul(buf, NULL, 0);
	}
	if (options & GET_DISKS) {
		strcpy(base, "raid_disks");
		if (load_sys(fname, buf))
			goto abort;
		sra->array.raid_disks = strtoul(buf, NULL, 0);
	}
	if (options & GET_DEGRADED) {
		strcpy(base, "degraded");
		if (load_sys(fname, buf))
			goto abort;
		sra->array.failed_disks = strtoul(buf, NULL, 0);
	}
	if (options & GET_COMPONENT) {
		strcpy(base, "component_size");
		if (load_sys(fname, buf))
			goto abort;
		sra->component_size = strtoull(buf, NULL, 0);
		/* sysfs reports "K", but we want sectors */
		sra->component_size *= 2;
	}
	if (options & GET_CHUNK) {
		strcpy(base, "chunk_size");
		if (load_sys(fname, buf))
			goto abort;
		sra->array.chunk_size = strtoul(buf, NULL, 0);
	}
	if (options & GET_CACHE) {
		strcpy(base, "stripe_cache_size");
		if (load_sys(fname, buf))
			goto abort;
		sra->cache_size = strtoul(buf, NULL, 0);
	}
	if (options & GET_MISMATCH) {
		strcpy(base, "mismatch_cnt");
		if (load_sys(fname, buf))
			goto abort;
		sra->mismatch_cnt = strtoul(buf, NULL, 0);
	}
	if (options & GET_SAFEMODE) {
		int scale = 1;
		int dot = 0;
		int i;
		unsigned long msec;
		size_t len;

		strcpy(base, "safe_mode_delay");
		if (load_sys(fname, buf))
			goto abort;

		/* remove a period, and count digits after it */
		len = strlen(buf);
		for (i = 0; i < len; i++) {
			if (dot) {
				if (isdigit(buf[i])) {
					buf[i-1] = buf[i];
					scale *= 10;
				}
				buf[i] = 0;
			} else if (buf[i] == '.') {
				dot=1;
				buf[i] = 0;
			}
		}
		msec = strtoul(buf, NULL, 10);
		msec = (msec * 1000) / scale;
		sra->safe_mode_delay = msec;
	}

	if (! (options & GET_DEVS))
		return sra;

	/* Get all the devices as well */
	*base = 0;
	dir = opendir(fname);
	if (!dir)
		goto abort;
	sra->array.spare_disks = 0;

	while ((de = readdir(dir)) != NULL) {
		char *ep;
		if (de->d_ino == 0 ||
		    strncmp(de->d_name, "dev-", 4) != 0)
			continue;
		strcpy(base, de->d_name);
		dbase = base + strlen(base);
		*dbase++ = '/';

		dev = malloc(sizeof(*dev));
		if (!dev)
			goto abort;

		/* Always get slot, major, minor */
		strcpy(dbase, "slot");
		if (load_sys(fname, buf)) {
			/* hmm... unable to read 'slot' maybe the device
			 * is going away?
			 */
			strcpy(dbase, "block");
			if (readlink(fname, buf, sizeof(buf)) < 0 &&
			    errno != ENAMETOOLONG) {
				/* ...yup device is gone */
				free(dev);
				continue;
			} else {
				/* slot is unreadable but 'block' link
				 * still intact... something bad is happening
				 * so abort
				 */
				free(dev);
				goto abort;
			}
			
		}
		dev->next = sra->devs;
		sra->devs = dev;

		strcpy(dev->sys_name, de->d_name);
		dev->disk.raid_disk = strtoul(buf, &ep, 10);
		if (*ep) dev->disk.raid_disk = -1;

		strcpy(dbase, "block/dev");
		if (load_sys(fname, buf))
			goto abort;
		sscanf(buf, "%d:%d", &dev->disk.major, &dev->disk.minor);

		if (options & GET_OFFSET) {
			strcpy(dbase, "offset");
			if (load_sys(fname, buf))
				goto abort;
			dev->data_offset = strtoull(buf, NULL, 0);
		}
		if (options & GET_SIZE) {
			strcpy(dbase, "size");
			if (load_sys(fname, buf))
				goto abort;
			dev->component_size = strtoull(buf, NULL, 0) * 2;
		}
		if (options & GET_STATE) {
			dev->disk.state = 0;
			strcpy(dbase, "state");
			if (load_sys(fname, buf))
				goto abort;
			if (strstr(buf, "in_sync"))
				dev->disk.state |= (1<<MD_DISK_SYNC);
			if (strstr(buf, "faulty"))
				dev->disk.state |= (1<<MD_DISK_FAULTY);
			if (dev->disk.state == 0)
				sra->array.spare_disks++;
		}
		if (options & GET_ERROR) {
			strcpy(buf, "errors");
			if (load_sys(fname, buf))
				goto abort;
			dev->errors = strtoul(buf, NULL, 0);
		}
	}
	closedir(dir);
	return sra;

 abort:
	if (dir)
		closedir(dir);
	sysfs_free(sra);
	return NULL;
}

int sysfs_attr_match(const char *attr, const char *str)
{
	/* See if attr, read from a sysfs file, matches
	 * str.  They must either be the same, or attr can
	 * have a trailing newline or comma
	 */
	while (*attr && *str && *attr == *str) {
		attr++;
		str++;
	}

	if (*str || (*attr && *attr != ',' && *attr != '\n'))
		return 0;
	return 1;
}

int sysfs_match_word(const char *word, char **list)
{
	int n;
	for (n=0; list[n]; n++)
		if (sysfs_attr_match(word, list[n]))
			break;
	return n;
}

unsigned long long get_component_size(int fd)
{
	/* Find out the component size of the array.
	 * We cannot trust GET_ARRAY_INFO ioctl as it's
	 * size field is only 32bits.
	 * So look in /sys/block/mdXXX/md/component_size
	 *
	 * This returns in units of sectors.
	 */
	struct stat stb;
	char fname[50];
	int n;
	if (fstat(fd, &stb)) return 0;
	if (major(stb.st_rdev) != get_mdp_major())
		sprintf(fname, "/sys/block/md%d/md/component_size",
			(int)minor(stb.st_rdev));
	else
		sprintf(fname, "/sys/block/md_d%d/md/component_size",
			(int)minor(stb.st_rdev)>>MdpMinorShift);
	fd = open(fname, O_RDONLY);
	if (fd < 0)
		return 0;
	n = read(fd, fname, sizeof(fname));
	close(fd);
	if (n == sizeof(fname))
		return 0;
	fname[n] = 0;
	return strtoull(fname, NULL, 10) * 2;
}

int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
		  char *name, char *val)
{
	char fname[50];
	int n;
	int fd;

	sprintf(fname, "/sys/block/%s/md/%s/%s",
		sra->sys_name, dev?dev->sys_name:"", name);
	fd = open(fname, O_WRONLY);
	if (fd < 0)
		return -1;
	n = write(fd, val, strlen(val));
	close(fd);
	if (n != strlen(val)) {
		dprintf(Name ": failed to write '%s' to '%s' (%s)\n",
			val, fname, strerror(errno));
		return -1;
	}
	return 0;
}

int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
		  char *name, unsigned long long val)
{
	char valstr[50];
	sprintf(valstr, "%llu", val);
	return sysfs_set_str(sra, dev, name, valstr);
}

int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
		       char *name, unsigned long long *val)
{
	char fname[50];
	char buf[50];
	int n;
	int fd;
	char *ep;
	sprintf(fname, "/sys/block/%s/md/%s/%s",
		sra->sys_name, dev?dev->sys_name:"", name);
	fd = open(fname, O_RDONLY);
	if (fd < 0)
		return -1;
	n = read(fd, buf, sizeof(buf));
	close(fd);
	if (n <= 0)
		return -1;
	buf[n] = 0;
	*val = strtoull(buf, &ep, 0);
	if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
		return -1;
	return 0;
}

int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms)
{
	unsigned long sec;
	unsigned long msec;
	char delay[30];

	sec = ms / 1000;
	msec = ms % 1000;

	sprintf(delay, "%ld.%03ld\n", sec, msec);
	/*             this '\n' ^ needed for kernels older than 2.6.28 */
	return sysfs_set_str(sra, NULL, "safe_mode_delay", delay);
}

int sysfs_set_array(struct mdinfo *info, int vers)
{
	int rv = 0;
	char ver[100];

	ver[0] = 0;
	if (info->array.major_version == -1 &&
	    info->array.minor_version == -2) {
		strcat(strcpy(ver, "external:"), info->text_version);

		if ((vers % 100) < 2 ||
		    sysfs_set_str(info, NULL, "metadata_version",
				  ver) < 0) {
			fprintf(stderr, Name ": This kernel does not "
				"support external metadata.\n");
			return 1;
		}
	}
	if (info->array.level < 0)
		return 0; /* FIXME */
	rv |= sysfs_set_str(info, NULL, "level",
			    map_num(pers, info->array.level));
	rv |= sysfs_set_num(info, NULL, "raid_disks", info->array.raid_disks);
	rv |= sysfs_set_num(info, NULL, "chunk_size", info->array.chunk_size);
	rv |= sysfs_set_num(info, NULL, "layout", info->array.layout);
	rv |= sysfs_set_num(info, NULL, "component_size", info->component_size/2);
	if (info->array.level > 0)
		rv |= sysfs_set_num(info, NULL, "resync_start", info->resync_start);
	return rv;
}

int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd)
{
	char dv[100];
	char nm[100];
	char *dname;
	int rv;

	sprintf(dv, "%d:%d", sd->disk.major, sd->disk.minor);
	rv = sysfs_set_str(sra, NULL, "new_dev", dv);
	if (rv)
		return rv;

	memset(nm, 0, sizeof(nm));
	sprintf(dv, "/sys/dev/block/%d:%d", sd->disk.major, sd->disk.minor);
	rv = readlink(dv, nm, sizeof(nm));
	if (rv <= 0)
		return -1;
	nm[rv] = '\0';
	dname = strrchr(nm, '/');
	if (dname) dname++;
	strcpy(sd->sys_name, "dev-");
	strcpy(sd->sys_name+4, dname);

	rv = sysfs_set_num(sra, sd, "offset", sd->data_offset);
	rv |= sysfs_set_num(sra, sd, "size", (sd->component_size+1) / 2);
	if (sra->array.level != LEVEL_CONTAINER) {
		rv |= sysfs_set_num(sra, sd, "slot", sd->disk.raid_disk);
//		rv |= sysfs_set_str(sra, sd, "state", "in_sync");
	}
	return rv;
}

#if 0
int sysfs_disk_to_sg(int fd)
{
	/* from an open block device, try find and open its corresponding
	 * scsi_generic interface
	 */
	struct stat st;
	char path[256];
	char sg_path[256];
	char sg_major_minor[8];
	char *c;
	DIR *dir;
	struct dirent *de;
	int major, minor, rv;

	if (fstat(fd, &st))
		return -1;

	snprintf(path, sizeof(path), "/sys/dev/block/%d:%d/device",
		 major(st.st_rdev), minor(st.st_rdev));

	dir = opendir(path);
	if (!dir)
		return -1;

	de = readdir(dir);
	while (de) {
		if (strncmp("scsi_generic:", de->d_name,
			    strlen("scsi_generic:")) == 0)
			break;
		de = readdir(dir);
	}
	closedir(dir);

	if (!de)
		return -1;

	snprintf(sg_path, sizeof(sg_path), "%s/%s/dev", path, de->d_name);
	fd = open(sg_path, O_RDONLY);
	if (fd < 0)
		return fd;

	rv = read(fd, sg_major_minor, sizeof(sg_major_minor));
	close(fd);
	if (rv < 0)
		return -1;
	else
		sg_major_minor[rv - 1] = '\0';

	c = strchr(sg_major_minor, ':');
	*c = '\0';
	c++;
	major = strtol(sg_major_minor, NULL, 10);
	minor = strtol(c, NULL, 10);
	snprintf(path, sizeof(path), "/dev/.tmp.md.%d:%d:%d",
		 (int) getpid(), major, minor);
	if (mknod(path, S_IFCHR|0600, makedev(major, minor))==0) {
			fd = open(path, O_RDONLY);
			unlink(path);
			return fd;
	}

	return -1;
}
#endif

int sysfs_disk_to_scsi_id(int fd, __u32 *id)
{
	/* from an open block device, try to retrieve it scsi_id */
	struct stat st;
	char path[256];
	char *c1, *c2;
	DIR *dir;
	struct dirent *de;

	if (fstat(fd, &st))
		return 1;

	snprintf(path, sizeof(path), "/sys/dev/block/%d:%d/device",
		 major(st.st_rdev), minor(st.st_rdev));

	dir = opendir(path);
	if (!dir)
		return 1;

	de = readdir(dir);
	while (de) {
		if (strncmp("scsi_disk:", de->d_name,
			    strlen("scsi_disk:")) == 0)
			break;
		de = readdir(dir);
	}
	closedir(dir);

	if (!de)
		return 1;

	c1 = strchr(de->d_name, ':');
	c1++;
	c2 = strchr(c1, ':');
	*c2 = '\0';
	*id = strtol(c1, NULL, 10) << 24; /* host */
	c1 = c2 + 1;
	c2 = strchr(c1, ':');
	*c2 = '\0';
	*id |= strtol(c1, NULL, 10) << 16; /* channel */
	c1 = c2 + 1;
	c2 = strchr(c1, ':');
	*c2 = '\0';
	*id |= strtol(c1, NULL, 10) << 8; /* lun */
	c1 = c2 + 1;
	*id |= strtol(c1, NULL, 10); /* id */

	return 0;
}


int sysfs_unique_holder(int devnum, long rdev)
{
	/* Check that devnum is a holder of rdev,
	 * and is the only holder.
	 * we should be locked against races by
	 * an O_EXCL on devnum
	 */
	DIR *dir;
	struct dirent *de;
	char dirname[100];
	char l;
	int found = 0;
	sprintf(dirname, "/sys/dev/block/%d:%d/holders",
		major(rdev), minor(rdev));
	dir = opendir(dirname);
	errno = ENOENT;
	if (!dir)
		return 0;
	l = strlen(dirname);
	while ((de = readdir(dir)) != NULL) {
		char buf[10];
		int n;
		int mj, mn;
		char c;
		int fd;

		if (de->d_ino == 0)
			continue;
		if (de->d_name[0] == '.')
			continue;
		strcpy(dirname+l, "/");
		strcat(dirname+l, de->d_name);
		strcat(dirname+l, "/dev");
		fd = open(dirname, O_RDONLY);
		if (fd < 0) {
			errno = ENOENT;
			break;
		}
		n = read(fd, buf, sizeof(buf)-1);
		close(fd);
		buf[n] = 0;
		if (sscanf(buf, "%d:%d%c", &mj, &mn, &c) != 3 ||
		    c != '\n') {
			errno = ENOENT;
			break;
		}
		if (mj != MD_MAJOR)
			mn = -1-(mn>>6);

		if (devnum != mn) {
			errno = EEXIST;
			break;
		}
		found = 1;
	}
	closedir(dir);
	if (de)
		return 0;
	else
		return found;
}
Commit	Line	Data
e86c9dd6 NB	1	/*
	2	* sysfs - extract md related information from sysfs. Part of:
	3	* mdadm - manage Linux "md" devices aka RAID arrays.
	4	*
	5	* Copyright (C) 2006 Neil Brown <neilb@suse.de>
	6	*
	7	*
	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.
	12	*
	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.
	17	*
	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
	21	*
	22	* Author: Neil Brown
	23	* Email: <neilb@suse.de>
	24	*/
	25
	26	#include "mdadm.h"
	27	#include <dirent.h>
1770662b	28	#include <ctype.h>
e86c9dd6 NB	29
	30	int load_sys(char path, char buf)
	31	{
	32	int fd = open(path, O_RDONLY);
	33	int n;
	34	if (fd < 0)
	35	return -1;
	36	n = read(fd, buf, 1024);
	37	close(fd);
2f6079dc	38	if (n <0 \|\| n >= 1024)
e86c9dd6 NB	39	return -1;
e86c9dd6 NB	40	buf[n] = 0;
8dfb8619	41	if (n && buf[n-1] == '\n')
e86c9dd6 NB	42	buf[n-1] = 0;
	43	return 0;
	44	}
	45
7e0f6979	46	void sysfs_free(struct mdinfo *sra)
8382f19b	47	{
7e0f6979 NB	48	while (sra) {
	49	struct mdinfo *sra2 = sra->next;
	50	while (sra->devs) {
	51	struct mdinfo *d = sra->devs;
	52	sra->devs = d->next;
	53	free(d);
	54	}
	55	free(sra);
	56	sra = sra2;
8382f19b	57	}
8382f19b NB	58	}
8382f19b NB	59
549e9569 NB	60	int sysfs_open(int devnum, char devname, char attr)
	61	{
	62	char fname[50];
549e9569	63	int fd;
1e4bc070	64	char *mdname = devnum2devname(devnum);
549e9569	65
1e4bc070 DW	66	if (!mdname)
	67	return -1;
	68
	69	sprintf(fname, "/sys/block/%s/md/", mdname);
549e9569 NB	70	if (devname) {
	71	strcat(fname, devname);
	72	strcat(fname, "/");
	73	}
	74	strcat(fname, attr);
	75	fd = open(fname, O_RDWR);
ea6d09b0	76	if (fd < 0 && errno == EACCES)
549e9569	77	fd = open(fname, O_RDONLY);
1e4bc070	78	free(mdname);
549e9569 NB	79	return fd;
	80	}
	81
f35f2525 N	82	void sysfs_init(struct mdinfo *mdi, int fd, int devnum)
	83	{
	84	if (fd >= 0) {
	85	struct stat stb;
	86	mdu_version_t vers;
	87	if (fstat(fd, &stb))
	88	return;
	89	if (ioctl(fd, RAID_VERSION, &vers) != 0)
	90	return;
4ebd3237	91	if (major(stb.st_rdev) == MD_MAJOR)
f35f2525	92	sprintf(mdi->sys_name, "md%d", (int)minor(stb.st_rdev));
4ebd3237	93	else if (major(stb.st_rdev) == get_mdp_major())
f35f2525 N	94	sprintf(mdi->sys_name, "md_d%d",
f35f2525 N	95	(int)minor(stb.st_rdev)>>MdpMinorShift);
4ebd3237 N	96	else {
	97	/* must be an extended-minor partition. Look at the
	98	* /sys/dev/block/%d:%d link which must look like
	99	* ../../block/mdXXX/mdXXXpYY
	100	*/
	101	char path[30];
	102	char link[200];
	103	char *cp;
	104	int n;
	105	sprintf(path, "/sys/dev/block/%d:%d", major(stb.st_rdev),
	106	minor(stb.st_rdev));
	107	n = readlink(path, link, sizeof(link)-1);
	108	if (n <= 0)
9c322006	109	return;
4ebd3237 N	110	link[n] = 0;
	111	cp = strrchr(link, '/');
	112	if (cp) *cp = 0;
	113	cp = strchr(link, '/');
	114	if (cp && strncmp(cp, "/md", 3) == 0)
9c322006	115	strcpy(mdi->sys_name, cp+1);
4ebd3237	116	else
9c322006	117	return;
4ebd3237	118	}
f35f2525 N	119	} else {
	120	if (devnum >= 0)
	121	sprintf(mdi->sys_name, "md%d", devnum);
	122	else
	123	sprintf(mdi->sys_name, "md_d%d",
	124	-1-devnum);
	125	}
	126	}
	127
7e0f6979	128	struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options)
e86c9dd6 NB	129	{
	130	/* Longest possible name in sysfs, mounted at /sys, is
	131	* /sys/block/md_dXXX/md/dev-XXXXX/block/dev
	132	* /sys/block/md_dXXX/md/metadata_version
	133	* which is about 41 characters. 50 should do for now
	134	*/
	135	char fname[50];
	136	char buf[1024];
	137	char *base;
	138	char *dbase;
7e0f6979	139	struct mdinfo *sra;
06c7f68e	140	struct mdinfo *dev;
355726fa	141	DIR *dir = NULL;
e86c9dd6 NB	142	struct dirent *de;
	143
	144	sra = malloc(sizeof(*sra));
	145	if (sra == NULL)
	146	return sra;
f35f2525 N	147	memset(sra, 0, sizeof(*sra));
f35f2525 N	148	sysfs_init(sra, fd, devnum);
e86c9dd6	149
7e0f6979	150	sprintf(fname, "/sys/block/%s/md/", sra->sys_name);
e86c9dd6 NB	151	base = fname + strlen(fname);
	152
	153	sra->devs = NULL;
8382f19b NB	154	if (options & GET_VERSION) {
	155	strcpy(base, "metadata_version");
	156	if (load_sys(fname, buf))
	157	goto abort;
294d6f45	158	if (strncmp(buf, "none", 4) == 0) {
7e0f6979 NB	159	sra->array.major_version =
7e0f6979 NB	160	sra->array.minor_version = -1;
294d6f45 NB	161	strcpy(sra->text_version, "");
294d6f45 NB	162	} else if (strncmp(buf, "external:", 9) == 0) {
142cb9e1 NB	163	sra->array.major_version = -1;
	164	sra->array.minor_version = -2;
	165	strcpy(sra->text_version, buf+9);
b8ac1967	166	} else {
8382f19b	167	sscanf(buf, "%d.%d",
7e0f6979 NB	168	&sra->array.major_version,
7e0f6979 NB	169	&sra->array.minor_version);
b8ac1967 NB	170	strcpy(sra->text_version, buf);
b8ac1967 NB	171	}
8382f19b	172	}
e86c9dd6 NB	173	if (options & GET_LEVEL) {
	174	strcpy(base, "level");
	175	if (load_sys(fname, buf))
	176	goto abort;
7e0f6979	177	sra->array.level = map_name(pers, buf);
e86c9dd6 NB	178	}
	179	if (options & GET_LAYOUT) {
	180	strcpy(base, "layout");
	181	if (load_sys(fname, buf))
	182	goto abort;
7e0f6979	183	sra->array.layout = strtoul(buf, NULL, 0);
e86c9dd6	184	}
549e9569 NB	185	if (options & GET_DISKS) {
	186	strcpy(base, "raid_disks");
	187	if (load_sys(fname, buf))
	188	goto abort;
	189	sra->array.raid_disks = strtoul(buf, NULL, 0);
f1d26766 DW	190	}
	191	if (options & GET_DEGRADED) {
	192	strcpy(base, "degraded");
	193	if (load_sys(fname, buf))
	194	goto abort;
	195	sra->array.failed_disks = strtoul(buf, NULL, 0);
549e9569	196	}
e86c9dd6 NB	197	if (options & GET_COMPONENT) {
	198	strcpy(base, "component_size");
	199	if (load_sys(fname, buf))
	200	goto abort;
	201	sra->component_size = strtoull(buf, NULL, 0);
353632d9 NB	202	/* sysfs reports "K", but we want sectors */
353632d9 NB	203	sra->component_size *= 2;
e86c9dd6 NB	204	}
	205	if (options & GET_CHUNK) {
	206	strcpy(base, "chunk_size");
	207	if (load_sys(fname, buf))
	208	goto abort;
7e0f6979	209	sra->array.chunk_size = strtoul(buf, NULL, 0);
e86c9dd6	210	}
758d3a8e NB	211	if (options & GET_CACHE) {
	212	strcpy(base, "stripe_cache_size");
	213	if (load_sys(fname, buf))
	214	goto abort;
	215	sra->cache_size = strtoul(buf, NULL, 0);
	216	}
37dfc3d6 NB	217	if (options & GET_MISMATCH) {
	218	strcpy(base, "mismatch_cnt");
	219	if (load_sys(fname, buf))
	220	goto abort;
	221	sra->mismatch_cnt = strtoul(buf, NULL, 0);
	222	}
1770662b DW	223	if (options & GET_SAFEMODE) {
	224	int scale = 1;
	225	int dot = 0;
	226	int i;
	227	unsigned long msec;
	228	size_t len;
	229
	230	strcpy(base, "safe_mode_delay");
	231	if (load_sys(fname, buf))
	232	goto abort;
	233
	234	/* remove a period, and count digits after it */
	235	len = strlen(buf);
	236	for (i = 0; i < len; i++) {
	237	if (dot) {
	238	if (isdigit(buf[i])) {
	239	buf[i-1] = buf[i];
	240	scale *= 10;
	241	}
	242	buf[i] = 0;
	243	} else if (buf[i] == '.') {
	244	dot=1;
	245	buf[i] = 0;
	246	}
	247	}
	248	msec = strtoul(buf, NULL, 10);
	249	msec = (msec * 1000) / scale;
	250	sra->safe_mode_delay = msec;
	251	}
e86c9dd6 NB	252
	253	if (! (options & GET_DEVS))
	254	return sra;
	255
	256	/* Get all the devices as well */
	257	*base = 0;
	258	dir = opendir(fname);
	259	if (!dir)
	260	goto abort;
7e0f6979	261	sra->array.spare_disks = 0;
e86c9dd6 NB	262
	263	while ((de = readdir(dir)) != NULL) {
	264	char *ep;
	265	if (de->d_ino == 0 \|\|
	266	strncmp(de->d_name, "dev-", 4) != 0)
	267	continue;
	268	strcpy(base, de->d_name);
	269	dbase = base + strlen(base);
	270	*dbase++ = '/';
	271
	272	dev = malloc(sizeof(*dev));
	273	if (!dev)
	274	goto abort;
e86c9dd6 NB	275
	276	/* Always get slot, major, minor */
	277	strcpy(dbase, "slot");
4795982e DW	278	if (load_sys(fname, buf)) {
	279	/* hmm... unable to read 'slot' maybe the device
	280	* is going away?
	281	*/
	282	strcpy(dbase, "block");
	283	if (readlink(fname, buf, sizeof(buf)) < 0 &&
	284	errno != ENAMETOOLONG) {
	285	/* ...yup device is gone */
	286	free(dev);
	287	continue;
	288	} else {
	289	/* slot is unreadable but 'block' link
	290	* still intact... something bad is happening
	291	* so abort
	292	*/
	293	free(dev);
	294	goto abort;
	295	}
	296
	297	}
	298	dev->next = sra->devs;
	299	sra->devs = dev;
	300
	301	strcpy(dev->sys_name, de->d_name);
06c7f68e NB	302	dev->disk.raid_disk = strtoul(buf, &ep, 10);
06c7f68e NB	303	if (*ep) dev->disk.raid_disk = -1;
e86c9dd6 NB	304
	305	strcpy(dbase, "block/dev");
	306	if (load_sys(fname, buf))
	307	goto abort;
06c7f68e	308	sscanf(buf, "%d:%d", &dev->disk.major, &dev->disk.minor);
e86c9dd6 NB	309
	310	if (options & GET_OFFSET) {
	311	strcpy(dbase, "offset");
	312	if (load_sys(fname, buf))
	313	goto abort;
06c7f68e	314	dev->data_offset = strtoull(buf, NULL, 0);
e86c9dd6 NB	315	}
	316	if (options & GET_SIZE) {
	317	strcpy(dbase, "size");
	318	if (load_sys(fname, buf))
	319	goto abort;
047d2e49	320	dev->component_size = strtoull(buf, NULL, 0) * 2;
e86c9dd6 NB	321	}
e86c9dd6 NB	322	if (options & GET_STATE) {
06c7f68e	323	dev->disk.state = 0;
e86c9dd6 NB	324	strcpy(dbase, "state");
	325	if (load_sys(fname, buf))
	326	goto abort;
	327	if (strstr(buf, "in_sync"))
06c7f68e	328	dev->disk.state \|= (1<<MD_DISK_SYNC);
e86c9dd6	329	if (strstr(buf, "faulty"))
06c7f68e NB	330	dev->disk.state \|= (1<<MD_DISK_FAULTY);
06c7f68e NB	331	if (dev->disk.state == 0)
7e0f6979	332	sra->array.spare_disks++;
e86c9dd6 NB	333	}
	334	if (options & GET_ERROR) {
	335	strcpy(buf, "errors");
	336	if (load_sys(fname, buf))
	337	goto abort;
	338	dev->errors = strtoul(buf, NULL, 0);
	339	}
	340	}
355726fa	341	closedir(dir);
e86c9dd6 NB	342	return sra;
	343
	344	abort:
355726fa NB	345	if (dir)
355726fa NB	346	closedir(dir);
8382f19b	347	sysfs_free(sra);
e86c9dd6 NB	348	return NULL;
	349	}
	350
1770662b DW	351	int sysfs_attr_match(const char attr, const char str)
	352	{
	353	/* See if attr, read from a sysfs file, matches
	354	* str. They must either be the same, or attr can
	355	* have a trailing newline or comma
	356	*/
	357	while (attr && str && attr == str) {
	358	attr++;
	359	str++;
	360	}
	361
	362	if (str \|\| (attr && attr != ',' && attr != '\n'))
	363	return 0;
	364	return 1;
	365	}
	366
	367	int sysfs_match_word(const char word, char *list)
	368	{
	369	int n;
	370	for (n=0; list[n]; n++)
	371	if (sysfs_attr_match(word, list[n]))
	372	break;
	373	return n;
	374	}
	375
e86c9dd6 NB	376	unsigned long long get_component_size(int fd)
	377	{
	378	/* Find out the component size of the array.
	379	* We cannot trust GET_ARRAY_INFO ioctl as it's
	380	* size field is only 32bits.
	381	* So look in /sys/block/mdXXX/md/component_size
353632d9	382	*
8686f3ed	383	* This returns in units of sectors.
e86c9dd6 NB	384	*/
	385	struct stat stb;
	386	char fname[50];
	387	int n;
	388	if (fstat(fd, &stb)) return 0;
4ebd3237	389	if (major(stb.st_rdev) != get_mdp_major())
e86c9dd6	390	sprintf(fname, "/sys/block/md%d/md/component_size",
ea24acd0	391	(int)minor(stb.st_rdev));
e86c9dd6 NB	392	else
e86c9dd6 NB	393	sprintf(fname, "/sys/block/md_d%d/md/component_size",
ea24acd0	394	(int)minor(stb.st_rdev)>>MdpMinorShift);
e86c9dd6 NB	395	fd = open(fname, O_RDONLY);
	396	if (fd < 0)
	397	return 0;
	398	n = read(fd, fname, sizeof(fname));
	399	close(fd);
	400	if (n == sizeof(fname))
	401	return 0;
	402	fname[n] = 0;
8686f3ed	403	return strtoull(fname, NULL, 10) * 2;
e86c9dd6 NB	404	}
e86c9dd6 NB	405
7e0f6979	406	int sysfs_set_str(struct mdinfo sra, struct mdinfo dev,
e86c9dd6 NB	407	char name, char val)
	408	{
	409	char fname[50];
	410	int n;
	411	int fd;
7e1432fb	412
e86c9dd6	413	sprintf(fname, "/sys/block/%s/md/%s/%s",
7e0f6979	414	sra->sys_name, dev?dev->sys_name:"", name);
e86c9dd6 NB	415	fd = open(fname, O_WRONLY);
	416	if (fd < 0)
	417	return -1;
	418	n = write(fd, val, strlen(val));
	419	close(fd);
2a24d7b6 DW	420	if (n != strlen(val)) {
	421	dprintf(Name ": failed to write '%s' to '%s' (%s)\n",
	422	val, fname, strerror(errno));
e86c9dd6	423	return -1;
2a24d7b6	424	}
e86c9dd6 NB	425	return 0;
	426	}
	427
7e0f6979	428	int sysfs_set_num(struct mdinfo sra, struct mdinfo dev,
e86c9dd6 NB	429	char *name, unsigned long long val)
	430	{
	431	char valstr[50];
	432	sprintf(valstr, "%llu", val);
	433	return sysfs_set_str(sra, dev, name, valstr);
	434	}
	435
7e0f6979	436	int sysfs_get_ll(struct mdinfo sra, struct mdinfo dev,
e86c9dd6 NB	437	char name, unsigned long long val)
	438	{
	439	char fname[50];
	440	char buf[50];
	441	int n;
	442	int fd;
	443	char *ep;
	444	sprintf(fname, "/sys/block/%s/md/%s/%s",
7e0f6979	445	sra->sys_name, dev?dev->sys_name:"", name);
e86c9dd6 NB	446	fd = open(fname, O_RDONLY);
	447	if (fd < 0)
	448	return -1;
	449	n = read(fd, buf, sizeof(buf));
	450	close(fd);
	451	if (n <= 0)
	452	return -1;
	453	buf[n] = 0;
	454	*val = strtoull(buf, &ep, 0);
	455	if (ep == buf \|\| (ep != 0 && ep != '\n' && *ep != ' '))
	456	return -1;
	457	return 0;
	458	}
2503d23b	459
8ed3e5e1 DW	460	int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms)
	461	{
	462	unsigned long sec;
	463	unsigned long msec;
	464	char delay[30];
	465
	466	sec = ms / 1000;
0dd3ba30	467	msec = ms % 1000;
8ed3e5e1	468
0dd3ba30 DW	469	sprintf(delay, "%ld.%03ld\n", sec, msec);
0dd3ba30 DW	470	/* this '\n' ^ needed for kernels older than 2.6.28 */
8ed3e5e1 DW	471	return sysfs_set_str(sra, NULL, "safe_mode_delay", delay);
	472	}
	473
f35f2525	474	int sysfs_set_array(struct mdinfo *info, int vers)
2503d23b NB	475	{
2503d23b NB	476	int rv = 0;
f35f2525 N	477	char ver[100];
	478
	479	ver[0] = 0;
	480	if (info->array.major_version == -1 &&
	481	info->array.minor_version == -2) {
	482	strcat(strcpy(ver, "external:"), info->text_version);
	483
	484	if ((vers % 100) < 2 \|\|
	485	sysfs_set_str(info, NULL, "metadata_version",
	486	ver) < 0) {
	487	fprintf(stderr, Name ": This kernel does not "
	488	"support external metadata.\n");
	489	return 1;
	490	}
	491	}
2503d23b NB	492	if (info->array.level < 0)
2503d23b NB	493	return 0; /* FIXME */
f35f2525	494	rv \|= sysfs_set_str(info, NULL, "level",
2503d23b	495	map_num(pers, info->array.level));
f35f2525 N	496	rv \|= sysfs_set_num(info, NULL, "raid_disks", info->array.raid_disks);
	497	rv \|= sysfs_set_num(info, NULL, "chunk_size", info->array.chunk_size);
	498	rv \|= sysfs_set_num(info, NULL, "layout", info->array.layout);
	499	rv \|= sysfs_set_num(info, NULL, "component_size", info->component_size/2);
	500	if (info->array.level > 0)
	501	rv \|= sysfs_set_num(info, NULL, "resync_start", info->resync_start);
2503d23b NB	502	return rv;
	503	}
	504
2318b9f0	505	int sysfs_add_disk(struct mdinfo sra, struct mdinfo sd)
2503d23b NB	506	{
	507	char dv[100];
	508	char nm[100];
2503d23b NB	509	char *dname;
	510	int rv;
	511
	512	sprintf(dv, "%d:%d", sd->disk.major, sd->disk.minor);
	513	rv = sysfs_set_str(sra, NULL, "new_dev", dv);
	514	if (rv)
	515	return rv;
	516
	517	memset(nm, 0, sizeof(nm));
	518	sprintf(dv, "/sys/dev/block/%d:%d", sd->disk.major, sd->disk.minor);
73649188 N	519	rv = readlink(dv, nm, sizeof(nm));
73649188 N	520	if (rv <= 0)
2503d23b	521	return -1;
73649188	522	nm[rv] = '\0';
2503d23b NB	523	dname = strrchr(nm, '/');
	524	if (dname) dname++;
	525	strcpy(sd->sys_name, "dev-");
	526	strcpy(sd->sys_name+4, dname);
	527
73649188	528	rv = sysfs_set_num(sra, sd, "offset", sd->data_offset);
2503d23b NB	529	rv \|= sysfs_set_num(sra, sd, "size", (sd->component_size+1) / 2);
	530	if (sra->array.level != LEVEL_CONTAINER) {
	531	rv \|= sysfs_set_num(sra, sd, "slot", sd->disk.raid_disk);
	532	// rv \|= sysfs_set_str(sra, sd, "state", "in_sync");
	533	}
2503d23b NB	534	return rv;
2503d23b NB	535	}
90c8b707	536
755c99fa	537	#if 0
90c8b707 DW	538	int sysfs_disk_to_sg(int fd)
	539	{
	540	/* from an open block device, try find and open its corresponding
	541	* scsi_generic interface
	542	*/
	543	struct stat st;
	544	char path[256];
	545	char sg_path[256];
	546	char sg_major_minor[8];
	547	char *c;
	548	DIR *dir;
	549	struct dirent *de;
	550	int major, minor, rv;
	551
	552	if (fstat(fd, &st))
	553	return -1;
	554
	555	snprintf(path, sizeof(path), "/sys/dev/block/%d:%d/device",
	556	major(st.st_rdev), minor(st.st_rdev));
	557
	558	dir = opendir(path);
	559	if (!dir)
	560	return -1;
	561
	562	de = readdir(dir);
	563	while (de) {
	564	if (strncmp("scsi_generic:", de->d_name,
	565	strlen("scsi_generic:")) == 0)
	566	break;
	567	de = readdir(dir);
	568	}
	569	closedir(dir);
	570
	571	if (!de)
	572	return -1;
	573
	574	snprintf(sg_path, sizeof(sg_path), "%s/%s/dev", path, de->d_name);
	575	fd = open(sg_path, O_RDONLY);
	576	if (fd < 0)
	577	return fd;
	578
	579	rv = read(fd, sg_major_minor, sizeof(sg_major_minor));
	580	close(fd);
	581	if (rv < 0)
	582	return -1;
	583	else
	584	sg_major_minor[rv - 1] = '\0';
	585
	586	c = strchr(sg_major_minor, ':');
	587	*c = '\0';
	588	c++;
	589	major = strtol(sg_major_minor, NULL, 10);
	590	minor = strtol(c, NULL, 10);
	591	snprintf(path, sizeof(path), "/dev/.tmp.md.%d:%d:%d",
	592	(int) getpid(), major, minor);
	593	if (mknod(path, S_IFCHR\|0600, makedev(major, minor))==0) {
	594	fd = open(path, O_RDONLY);
	595	unlink(path);
	596	return fd;
	597	}
	598
	599	return -1;
	600	}
755c99fa	601	#endif
90c8b707	602
f1665f72 DW	603	int sysfs_disk_to_scsi_id(int fd, __u32 *id)
	604	{
	605	/* from an open block device, try to retrieve it scsi_id */
	606	struct stat st;
	607	char path[256];
	608	char c1, c2;
	609	DIR *dir;
	610	struct dirent *de;
	611
	612	if (fstat(fd, &st))
	613	return 1;
	614
	615	snprintf(path, sizeof(path), "/sys/dev/block/%d:%d/device",
	616	major(st.st_rdev), minor(st.st_rdev));
	617
	618	dir = opendir(path);
	619	if (!dir)
	620	return 1;
	621
	622	de = readdir(dir);
	623	while (de) {
	624	if (strncmp("scsi_disk:", de->d_name,
	625	strlen("scsi_disk:")) == 0)
	626	break;
	627	de = readdir(dir);
	628	}
	629	closedir(dir);
	630
	631	if (!de)
	632	return 1;
	633
	634	c1 = strchr(de->d_name, ':');
	635	c1++;
	636	c2 = strchr(c1, ':');
	637	*c2 = '\0';
	638	id = strtol(c1, NULL, 10) << 24; / host */
	639	c1 = c2 + 1;
	640	c2 = strchr(c1, ':');
	641	*c2 = '\0';
	642	id \|= strtol(c1, NULL, 10) << 16; / channel */
	643	c1 = c2 + 1;
	644	c2 = strchr(c1, ':');
	645	*c2 = '\0';
	646	id \|= strtol(c1, NULL, 10) << 8; / lun */
	647	c1 = c2 + 1;
	648	id \|= strtol(c1, NULL, 10); / id */
	649
	650	return 0;
	651	}
f94d52f4 NB	652
	653
	654	int sysfs_unique_holder(int devnum, long rdev)
	655	{
	656	/* Check that devnum is a holder of rdev,
	657	* and is the only holder.
	658	* we should be locked against races by
	659	* an O_EXCL on devnum
	660	*/
	661	DIR *dir;
	662	struct dirent *de;
	663	char dirname[100];
	664	char l;
	665	int found = 0;
	666	sprintf(dirname, "/sys/dev/block/%d:%d/holders",
	667	major(rdev), minor(rdev));
	668	dir = opendir(dirname);
	669	errno = ENOENT;
	670	if (!dir)
	671	return 0;
	672	l = strlen(dirname);
	673	while ((de = readdir(dir)) != NULL) {
	674	char buf[10];
	675	int n;
	676	int mj, mn;
	677	char c;
	678	int fd;
	679
	680	if (de->d_ino == 0)
	681	continue;
	682	if (de->d_name[0] == '.')
	683	continue;
	684	strcpy(dirname+l, "/");
	685	strcat(dirname+l, de->d_name);
	686	strcat(dirname+l, "/dev");
	687	fd = open(dirname, O_RDONLY);
	688	if (fd < 0) {
	689	errno = ENOENT;
	690	break;
	691	}
	692	n = read(fd, buf, sizeof(buf)-1);
	693	close(fd);
	694	buf[n] = 0;
	695	if (sscanf(buf, "%d:%d%c", &mj, &mn, &c) != 3 \|\|
	696	c != '\n') {
	697	errno = ENOENT;
	698	break;
	699	}
	700	if (mj != MD_MAJOR)
	701	mn = -1-(mn>>6);
	702
	703	if (devnum != mn) {
	704	errno = EEXIST;
	705	break;
	706	}
	707	found = 1;
	708	}
	709	closedir(dir);
	710	if (de)
	711	return 0;
	712	else
	713	return found;
	714	}