[thirdparty/mdadm.git] / Manage.c

/*
 * mdadm - manage Linux "md" devices aka RAID arrays.
 *
 * Copyright (C) 2001-2006 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@cse.unsw.edu.au>
 *    Paper: Neil Brown
 *           School of Computer Science and Engineering
 *           The University of New South Wales
 *           Sydney, 2052
 *           Australia
 */

#include "mdadm.h"
#include "md_u.h"
#include "md_p.h"

#define REGISTER_DEV 		_IO (MD_MAJOR, 1)
#define START_MD     		_IO (MD_MAJOR, 2)
#define STOP_MD      		_IO (MD_MAJOR, 3)

int Manage_ro(char *devname, int fd, int readonly)
{
	/* switch to readonly or rw
	 *
	 * requires >= 0.90.0
	 * first check that array is runing
	 * use RESTART_ARRAY_RW or STOP_ARRAY_RO
	 *
	 */
	mdu_array_info_t array;

	if (md_get_version(fd) < 9000) {
		fprintf(stderr, Name ": need md driver version 0.90.0 or later\n");
		return 1;
	}
	if (ioctl(fd, GET_ARRAY_INFO, &array)) {
		fprintf(stderr, Name ": %s does not appear to be active.\n",
			devname);
		return 1;
	}

	if (readonly>0) {
		if (ioctl(fd, STOP_ARRAY_RO, NULL)) {
			fprintf(stderr, Name ": failed to set readonly for %s: %s\n",
				devname, strerror(errno));
			return 1;
		}
	} else if (readonly < 0) {
		if (ioctl(fd, RESTART_ARRAY_RW, NULL)) {
			fprintf(stderr, Name ": failed to set writable for %s: %s\n",
				devname, strerror(errno));
			return 1;
		}
	}
	return 0;
}

#ifndef MDASSEMBLE

int Manage_runstop(char *devname, int fd, int runstop, int quiet)
{
	/* Run or stop the array. array must already be configured
	 * required >= 0.90.0
	 * Only print failure messages if quiet == 0;
	 * quiet > 0 means really be quiet
	 * quiet < 0 means we will try again if it fails.
	 */
	mdu_param_t param; /* unused */

	if (runstop == -1 && md_get_version(fd) < 9000) {
		if (ioctl(fd, STOP_MD, 0)) {
			if (quiet == 0) fprintf(stderr,
						Name ": stopping device %s "
						"failed: %s\n",
						devname, strerror(errno));
			return 1;
		}
	}

	if (md_get_version(fd) < 9000) {
		fprintf(stderr, Name ": need md driver version 0.90.0 or later\n");
		return 1;
	}
	/*
	if (ioctl(fd, GET_ARRAY_INFO, &array)) {
		fprintf(stderr, Name ": %s does not appear to be active.\n",
			devname);
		return 1;
	}
	*/
	if (runstop>0) {
		if (ioctl(fd, RUN_ARRAY, &param)) {
			fprintf(stderr, Name ": failed to run array %s: %s\n",
				devname, strerror(errno));
			return 1;
		}
		if (quiet <= 0)
			fprintf(stderr, Name ": started %s\n", devname);
	} else if (runstop < 0){
		struct map_ent *map = NULL;
		struct stat stb;
		struct mdinfo *mdi;
		/* If this is an mdmon managed array, just write 'inactive'
		 * to the array state and let mdmon clear up.
		 */
		mdi = sysfs_read(fd, -1, GET_LEVEL|GET_VERSION);
		if (mdi &&
		    mdi->array.level > 0 &&
		    mdi->text_version[0] == '/') {
			char *cp;

			/* This is mdmon managed. */
			close(fd);
			if (sysfs_set_str(mdi, NULL,
					  "array_state", "inactive") < 0) {
				if (quiet == 0)
					fprintf(stderr, Name
						": failed to stop array %s: %s\n",
						devname, strerror(errno));
				return 1;
			}

			/* Give monitor a chance to act */
			cp = strchr(mdi->text_version+1, '/');
			if (*cp)
				*cp = 0;
			ping_monitor(mdi->text_version+1);

			fd = open(devname, O_RDONLY);
		} else if (mdi &&
			   mdi->array.major_version == -1 &&
			   mdi->array.minor_version == -2 &&
			   mdi->text_version[0] != '/') {
			/* container, possibly mdmon-managed.
			 * Make sure mdmon isn't opening it, which
			 * would interfere with the 'stop'
			 */
			ping_monitor(mdi->sys_name);
		}
		if (mdi)
			sysfs_free(mdi);

		if (fd >= 0 && ioctl(fd, STOP_ARRAY, NULL)) {
			if (quiet == 0) {
				fprintf(stderr, Name
					": failed to stop array %s: %s\n",
					devname, strerror(errno));
				if (errno == EBUSY)
					fprintf(stderr, "Perhaps a running "
						"process, mounted filesystem "
						"or active volume group?\n");
			}
			return 1;
		}

		if (quiet <= 0)
			fprintf(stderr, Name ": stopped %s\n", devname);
		if (fd >= 0 && fstat(fd, &stb) == 0) {
			int devnum;
			if (major(stb.st_rdev) == MD_MAJOR)
				devnum = minor(stb.st_rdev);
			else
				devnum = -1-(minor(stb.st_rdev)>>6);
			map_delete(&map, devnum);
			map_write(map);
			map_free(map);
		}
	}
	return 0;
}

int Manage_resize(char *devname, int fd, long long size, int raid_disks)
{
	mdu_array_info_t info;
	if (ioctl(fd, GET_ARRAY_INFO, &info) != 0) {
		fprintf(stderr, Name ": Cannot get array information for %s: %s\n",
			devname, strerror(errno));
		return 1;
	}
	if (size >= 0)
		info.size = size;
	if (raid_disks > 0)
		info.raid_disks = raid_disks;
	if (ioctl(fd, SET_ARRAY_INFO, &info) != 0) {
		fprintf(stderr, Name ": Cannot set device size/shape for %s: %s\n",
			devname, strerror(errno));
		return 1;
	}
	return 0;
}

int Manage_reconfig(char *devname, int fd, int layout)
{
	mdu_array_info_t info;
	if (ioctl(fd, GET_ARRAY_INFO, &info) != 0) {
		fprintf(stderr, Name ": Cannot get array information for %s: %s\n",
			devname, strerror(errno));
		return 1;
	}
	info.layout = layout;
	printf("layout set to %d\n", info.layout);
	if (ioctl(fd, SET_ARRAY_INFO, &info) != 0) {
		fprintf(stderr, Name ": Cannot set layout for %s: %s\n",
			devname, strerror(errno));
		return 1;
	}
	return 0;
}

int Manage_subdevs(char *devname, int fd,
		   mddev_dev_t devlist, int verbose)
{
	/* do something to each dev.
	 * devmode can be
	 *  'a' - add the device
	 *	   try HOT_ADD_DISK
	 *         If that fails EINVAL, try ADD_NEW_DISK
	 *  'r' - remove the device HOT_REMOVE_DISK
	 *        device can be 'faulty' or 'detached' in which case all
	 *	  matching devices are removed.
	 *  'f' - set the device faulty SET_DISK_FAULTY
	 *        device can be 'detached' in which case any device that
	 *	  is inaccessible will be marked faulty.
	 */
	mdu_array_info_t array;
	mdu_disk_info_t disc;
	unsigned long long array_size;
	mddev_dev_t dv, next = NULL;
	struct stat stb;
	int j, jnext = 0;
	int tfd;
	struct supertype *st, *tst;
	int duuid[4];
	int ouuid[4];
	int lfd = -1;

	if (ioctl(fd, GET_ARRAY_INFO, &array)) {
		fprintf(stderr, Name ": cannot get array info for %s\n",
			devname);
		return 1;
	}

	/* array.size is only 32 bit and may be truncated.
	 * So read from sysfs if possible, and record number of sectors
	 */

	array_size = get_component_size(fd);
	if (array_size <= 0)
		array_size = array.size * 2;

	tst = super_by_fd(fd);
	if (!tst) {
		fprintf(stderr, Name ": unsupport array - version %d.%d\n",
			array.major_version, array.minor_version);
		return 1;
	}

	for (dv = devlist, j=0 ; dv; dv = next, j = jnext) {
		unsigned long long ldsize;
		char dvname[20];
		char *dnprintable = dv->devname;
		int err;

		next = dv->next;
		jnext = 0;

		if (strcmp(dv->devname, "failed")==0 ||
		    strcmp(dv->devname, "faulty")==0) {
			if (dv->disposition != 'r') {
				fprintf(stderr, Name ": %s only meaningful "
					"with -r, not -%c\n",
					dv->devname, dv->disposition);
				return 1;
			}
			for (; j < array.raid_disks + array.nr_disks ; j++) {
				disc.number = j;
				if (ioctl(fd, GET_DISK_INFO, &disc))
					continue;
				if (disc.major == 0 && disc.minor == 0)
					continue;
				if ((disc.state & 1) == 0) /* faulty */
					continue;
				stb.st_rdev = makedev(disc.major, disc.minor);
				next = dv;
				jnext = j+1;
				sprintf(dvname,"%d:%d", disc.major, disc.minor);
				dnprintable = dvname;
				break;
			}
			if (jnext == 0)
				continue;
		} else if (strcmp(dv->devname, "detached") == 0) {
			if (dv->disposition != 'r' && dv->disposition != 'f') {
				fprintf(stderr, Name ": %s only meaningful "
					"with -r of -f, not -%c\n",
					dv->devname, dv->disposition);
				return 1;
			}
			for (; j < array.raid_disks + array.nr_disks; j++) {
				int sfd;
				disc.number = j;
				if (ioctl(fd, GET_DISK_INFO, &disc))
					continue;
				if (disc.major == 0 && disc.minor == 0)
					continue;
				sprintf(dvname,"%d:%d", disc.major, disc.minor);
				sfd = dev_open(dvname, O_RDONLY);
				if (sfd >= 0) {
					close(sfd);
					continue;
				}
				if (dv->disposition == 'f' &&
				    (disc.state & 1) == 1) /* already faulty */
					continue;
				if (errno != ENXIO)
					continue;
				stb.st_rdev = makedev(disc.major, disc.minor);
				next = dv;
				jnext = j+1;
				dnprintable = dvname;
				break;
			}
			if (jnext == 0)
				continue;
		} else {
			j = 0;

			if (stat(dv->devname, &stb)) {
				fprintf(stderr, Name ": cannot find %s: %s\n",
					dv->devname, strerror(errno));
				return 1;
			}
			if ((stb.st_mode & S_IFMT) != S_IFBLK) {
				fprintf(stderr, Name ": %s is not a "
					"block device.\n",
					dv->devname);
				return 1;
			}
		}
		switch(dv->disposition){
		default:
			fprintf(stderr, Name ": internal error - devmode[%s]=%d\n",
				dv->devname, dv->disposition);
			return 1;
		case 'a':
			/* add the device */
			if (tst->subarray[0]) {
				fprintf(stderr, Name ": Cannot add disks to a"
					" \'member\' array, perform this"
					" operation on the parent container\n");
				return 1;
			}
			/* Make sure it isn't in use (in 2.6 or later) */
			tfd = open(dv->devname, O_RDONLY|O_EXCL|O_DIRECT);
			if (tfd < 0) {
				fprintf(stderr, Name ": Cannot open %s: %s\n",
					dv->devname, strerror(errno));
				return 1;
			}
			remove_partitions(tfd);

			st = dup_super(tst);

			if (array.not_persistent==0)
				st->ss->load_super(st, tfd, NULL);

			if (!get_dev_size(tfd, dv->devname, &ldsize)) {
				close(tfd);
				return 1;
			}
			close(tfd);


			if (!tst->ss->external &&
			    array.major_version == 0 &&
			    md_get_version(fd)%100 < 2) {
				if (ioctl(fd, HOT_ADD_DISK,
					  (unsigned long)stb.st_rdev)==0) {
					if (verbose >= 0)
						fprintf(stderr, Name ": hot added %s\n",
							dv->devname);
					continue;
				}

				fprintf(stderr, Name ": hot add failed for %s: %s\n",
					dv->devname, strerror(errno));
				return 1;
			}

			if (array.not_persistent == 0) {

				/* Make sure device is large enough */
				if (tst->ss->avail_size(tst, ldsize/512) <
				    array_size) {
					fprintf(stderr, Name ": %s not large enough to join array\n",
						dv->devname);
					return 1;
				}

				/* need to find a sample superblock to copy, and
				 * a spare slot to use
				 */
				for (j = 0; j < tst->max_devs; j++) {
					char *dev;
					int dfd;
					disc.number = j;
					if (ioctl(fd, GET_DISK_INFO, &disc))
						continue;
					if (disc.major==0 && disc.minor==0)
						continue;
					if ((disc.state & 4)==0) continue; /* sync */
					/* Looks like a good device to try */
					dev = map_dev(disc.major, disc.minor, 1);
					if (!dev) continue;
					dfd = dev_open(dev, O_RDONLY);
					if (dfd < 0) continue;
					if (tst->ss->load_super(tst, dfd,
								NULL)) {
						close(dfd);
						continue;
					}
					close(dfd);
					break;
				}
				if (!tst->sb) {
					fprintf(stderr, Name ": cannot find valid superblock in this array - HELP\n");
					return 1;
				}
				/* Possibly this device was recently part of the array
				 * and was temporarily removed, and is now being re-added.
				 * If so, we can simply re-add it.
				 */
				tst->ss->uuid_from_super(tst, duuid);

				/* re-add doesn't work for version-1 superblocks
				 * before 2.6.18 :-(
				 */
				if (array.major_version == 1 &&
				    get_linux_version() <= 2006018)
					;
				else if (st->sb) {
					st->ss->uuid_from_super(st, ouuid);
					if (memcmp(duuid, ouuid, sizeof(ouuid))==0) {
						/* looks close enough for now.  Kernel
						 * will worry about whether a bitmap
						 * based reconstruction is possible.
						 */
						struct mdinfo mdi;
						st->ss->getinfo_super(st, &mdi);
						disc.major = major(stb.st_rdev);
						disc.minor = minor(stb.st_rdev);
						disc.number = mdi.disk.number;
						disc.raid_disk = mdi.disk.raid_disk;
						disc.state = mdi.disk.state;
						if (dv->writemostly)
							disc.state |= 1 << MD_DISK_WRITEMOSTLY;
						if (ioctl(fd, ADD_NEW_DISK, &disc) == 0) {
							if (verbose >= 0)
								fprintf(stderr, Name ": re-added %s\n", dv->devname);
							continue;
						}
						/* fall back on normal-add */
					}
				}
			} else {
				/* non-persistent. Must ensure that new drive
				 * is at least array.size big.
				 */
				if (ldsize/512 < array_size) {
					fprintf(stderr, Name ": %s not large enough to join array\n",
						dv->devname);
					return 1;
				}
			}
			/* in 2.6.17 and earlier, version-1 superblocks won't
			 * use the number we write, but will choose a free number.
			 * we must choose the same free number, which requires
			 * starting at 'raid_disks' and counting up
			 */
			for (j = array.raid_disks; j< tst->max_devs; j++) {
				disc.number = j;
				if (ioctl(fd, GET_DISK_INFO, &disc))
					break;
				if (disc.major==0 && disc.minor==0)
					break;
				if (disc.state & 8) /* removed */
					break;
			}
			disc.major = major(stb.st_rdev);
			disc.minor = minor(stb.st_rdev);
			disc.number =j;
			disc.state = 0;
			if (array.not_persistent==0) {
				int dfd;
				if (dv->writemostly)
					disc.state |= 1 << MD_DISK_WRITEMOSTLY;
				dfd = open(dv->devname, O_RDWR | O_EXCL|O_DIRECT);
				tst->ss->add_to_super(tst, &disc, dfd,
						      dv->devname);
				/* write_init_super will close 'dfd' */
				if (tst->ss->write_init_super(tst))
					return 1;
			} else if (dv->re_add) {
				/*  this had better be raid1.
				 * As we are "--re-add"ing we must find a spare slot
				 * to fill.
				 */
				char *used = malloc(array.raid_disks);
				memset(used, 0, array.raid_disks);
				for (j=0; j< tst->max_devs; j++) {
					mdu_disk_info_t disc2;
					disc2.number = j;
					if (ioctl(fd, GET_DISK_INFO, &disc2))
						continue;
					if (disc2.major==0 && disc2.minor==0)
						continue;
					if (disc2.state & 8) /* removed */
						continue;
					if (disc2.raid_disk < 0)
						continue;
					if (disc2.raid_disk > array.raid_disks)
						continue;
					used[disc2.raid_disk] = 1;
				}
				for (j=0 ; j<array.raid_disks; j++)
					if (!used[j]) {
						disc.raid_disk = j;
						disc.state |= (1<<MD_DISK_SYNC);
						break;
					}
			}
			if (dv->writemostly)
				disc.state |= (1 << MD_DISK_WRITEMOSTLY);
			if (ioctl(fd,ADD_NEW_DISK, &disc)) {
				fprintf(stderr, Name ": add new device failed for %s as %d: %s\n",
					dv->devname, j, strerror(errno));
				return 1;
			}
			if (verbose >= 0)
				fprintf(stderr, Name ": added %s\n", dv->devname);
			break;

		case 'r':
			/* hot remove */
			if (tst->subarray[0]) {
				fprintf(stderr, Name ": Cannot remove disks from a"
					" \'member\' array, perform this"
					" operation on the parent container\n");
				return 1;
			}
			if (tst->ss->external) {
				/* To remove a device from a container, we must
				 * check that it isn't in use in an array.
				 * This involves looking in the 'holders'
				 * directory - there must be just one entry,
				 * the container.
				 * To ensure that it doesn't get used as a
				 * hold spare while we are checking, we
				 * get an O_EXCL open on the container
				 */
				int dnum = fd2devnum(fd);
				lfd = open_dev_excl(dnum);
				if (lfd < 0) {
					fprintf(stderr, Name
						": Cannot get exclusive access "
						" to container - odd\n");
					return 1;
				}
				if (!sysfs_unique_holder(dnum, stb.st_rdev)) {
					fprintf(stderr, Name
						": %s is %s, cannot remove.\n",
						dnprintable,
						errno == EEXIST ? "still in use":
						"not a member");
					close(lfd);
					return 1;
				}
			}
			/* FIXME check that it is a current member */
			err = ioctl(fd, HOT_REMOVE_DISK, (unsigned long)stb.st_rdev);
			if (err && errno == ENODEV) {
				/* Old kernels rejected this if no personality
				 * registered */
				struct mdinfo *sra = sysfs_read(fd, 0, GET_DEVS);
				struct mdinfo *dv = NULL;
				if (sra)
					dv = sra->devs;
				for ( ; dv ; dv=dv->next)
					if (dv->disk.major == major(stb.st_rdev) &&
					    dv->disk.minor == minor(stb.st_rdev))
						break;
				if (dv)
					err = sysfs_set_str(sra, dv,
							    "state", "remove");
				else
					err = -1;
				if (sra)
					sysfs_free(sra);
			}
			if (err) {
				fprintf(stderr, Name ": hot remove failed "
					"for %s: %s\n",	dnprintable,
					strerror(errno));
				if (lfd >= 0)
					close(lfd);
				return 1;
			}
			close(lfd);
			if (verbose >= 0)
				fprintf(stderr, Name ": hot removed %s\n",
					dnprintable);
			break;

		case 'f': /* set faulty */
			/* FIXME check current member */
			if (ioctl(fd, SET_DISK_FAULTY, (unsigned long) stb.st_rdev)) {
				fprintf(stderr, Name ": set device faulty failed for %s:  %s\n",
					dnprintable, strerror(errno));
				return 1;
			}
			if (verbose >= 0)
				fprintf(stderr, Name ": set %s faulty in %s\n",
					dnprintable, devname);
			break;
		}
	}
	return 0;

}

int autodetect(void)
{
	/* Open any md device, and issue the RAID_AUTORUN ioctl */
	int rv = 1;
	int fd = dev_open("9:0", O_RDONLY);
	if (fd >= 0) {
		if (ioctl(fd, RAID_AUTORUN, 0) == 0)
			rv = 0;
		close(fd);
	}
	return rv;
}
#endif
Commit	Line	Data
64c4757e	1	/*
9a9dab36	2	* mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e	3	*
4f589ad0	4	* Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
64c4757e NB	5	*
	6	*
	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.
	11	*
	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.
	16	*
	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
	20	*
	21	* Author: Neil Brown
	22	* Email: <neilb@cse.unsw.edu.au>
	23	* Paper: Neil Brown
	24	* School of Computer Science and Engineering
	25	* The University of New South Wales
	26	* Sydney, 2052
	27	* Australia
	28	*/
	29
9a9dab36	30	#include "mdadm.h"
682c7051 NB	31	#include "md_u.h"
682c7051 NB	32	#include "md_p.h"
64c4757e	33
82b27616 NB	34	#define REGISTER_DEV _IO (MD_MAJOR, 1)
	35	#define START_MD _IO (MD_MAJOR, 2)
	36	#define STOP_MD _IO (MD_MAJOR, 3)
	37
64c4757e NB	38	int Manage_ro(char *devname, int fd, int readonly)
64c4757e NB	39	{
682c7051 NB	40	/* switch to readonly or rw
	41	*
	42	* requires >= 0.90.0
	43	* first check that array is runing
	44	* use RESTART_ARRAY_RW or STOP_ARRAY_RO
	45	*
	46	*/
	47	mdu_array_info_t array;
aba69144	48
682c7051 NB	49	if (md_get_version(fd) < 9000) {
	50	fprintf(stderr, Name ": need md driver version 0.90.0 or later\n");
	51	return 1;
	52	}
	53	if (ioctl(fd, GET_ARRAY_INFO, &array)) {
	54	fprintf(stderr, Name ": %s does not appear to be active.\n",
	55	devname);
	56	return 1;
	57	}
aba69144	58
682c7051 NB	59	if (readonly>0) {
	60	if (ioctl(fd, STOP_ARRAY_RO, NULL)) {
	61	fprintf(stderr, Name ": failed to set readonly for %s: %s\n",
	62	devname, strerror(errno));
	63	return 1;
	64	}
	65	} else if (readonly < 0) {
	66	if (ioctl(fd, RESTART_ARRAY_RW, NULL)) {
82b27616	67	fprintf(stderr, Name ": failed to set writable for %s: %s\n",
682c7051 NB	68	devname, strerror(errno));
	69	return 1;
	70	}
	71	}
aba69144	72	return 0;
64c4757e NB	73	}
64c4757e NB	74
435d4ebb LB	75	#ifndef MDASSEMBLE
435d4ebb LB	76
91f068bf	77	int Manage_runstop(char *devname, int fd, int runstop, int quiet)
64c4757e	78	{
682c7051 NB	79	/* Run or stop the array. array must already be configured
682c7051 NB	80	* required >= 0.90.0
4b9d39fc N	81	* Only print failure messages if quiet == 0;
	82	* quiet > 0 means really be quiet
	83	* quiet < 0 means we will try again if it fails.
682c7051	84	*/
682c7051	85	mdu_param_t param; /* unused */
82b27616 NB	86
	87	if (runstop == -1 && md_get_version(fd) < 9000) {
	88	if (ioctl(fd, STOP_MD, 0)) {
4b9d39fc N	89	if (quiet == 0) fprintf(stderr,
	90	Name ": stopping device %s "
	91	"failed: %s\n",
	92	devname, strerror(errno));
82b27616 NB	93	return 1;
	94	}
	95	}
aba69144	96
682c7051 NB	97	if (md_get_version(fd) < 9000) {
	98	fprintf(stderr, Name ": need md driver version 0.90.0 or later\n");
	99	return 1;
	100	}
82b27616	101	/*
682c7051 NB	102	if (ioctl(fd, GET_ARRAY_INFO, &array)) {
	103	fprintf(stderr, Name ": %s does not appear to be active.\n",
	104	devname);
	105	return 1;
	106	}
82b27616	107	*/
682c7051 NB	108	if (runstop>0) {
	109	if (ioctl(fd, RUN_ARRAY, &param)) {
	110	fprintf(stderr, Name ": failed to run array %s: %s\n",
	111	devname, strerror(errno));
	112	return 1;
	113	}
8382f19b NB	114	if (quiet <= 0)
8382f19b NB	115	fprintf(stderr, Name ": started %s\n", devname);
682c7051	116	} else if (runstop < 0){
8382f19b NB	117	struct map_ent *map = NULL;
8382f19b NB	118	struct stat stb;
daf7a3ce NB	119	struct mdinfo *mdi;
	120	/* If this is an mdmon managed array, just write 'inactive'
	121	* to the array state and let mdmon clear up.
	122	*/
	123	mdi = sysfs_read(fd, -1, GET_LEVEL\|GET_VERSION);
	124	if (mdi &&
	125	mdi->array.level > 0 &&
	126	mdi->text_version[0] == '/') {
	127	char *cp;
	128
	129	/* This is mdmon managed. */
	130	close(fd);
	131	if (sysfs_set_str(mdi, NULL,
	132	"array_state", "inactive") < 0) {
4b9d39fc	133	if (quiet == 0)
daf7a3ce	134	fprintf(stderr, Name
4b9d39fc	135	": failed to stop array %s: %s\n",
daf7a3ce NB	136	devname, strerror(errno));
	137	return 1;
	138	}
	139
	140	/* Give monitor a chance to act */
	141	cp = strchr(mdi->text_version+1, '/');
	142	if (*cp)
	143	*cp = 0;
	144	ping_monitor(mdi->text_version+1);
	145
	146	fd = open(devname, O_RDONLY);
ada6c239 N	147	} else if (mdi &&
	148	mdi->array.major_version == -1 &&
	149	mdi->array.minor_version == -2 &&
	150	mdi->text_version[0] != '/') {
	151	/* container, possibly mdmon-managed.
	152	* Make sure mdmon isn't opening it, which
	153	* would interfere with the 'stop'
	154	*/
	155	ping_monitor(mdi->sys_name);
daf7a3ce NB	156	}
	157	if (mdi)
	158	sysfs_free(mdi);
	159
	160	if (fd >= 0 && ioctl(fd, STOP_ARRAY, NULL)) {
37ea3936	161	if (quiet == 0) {
daf7a3ce	162	fprintf(stderr, Name
4b9d39fc	163	": failed to stop array %s: %s\n",
91f068bf	164	devname, strerror(errno));
d927f3c4 NB	165	if (errno == EBUSY)
	166	fprintf(stderr, "Perhaps a running "
	167	"process, mounted filesystem "
	168	"or active volume group?\n");
	169	}
682c7051 NB	170	return 1;
682c7051 NB	171	}
daf7a3ce	172
ab56093f NB	173	if (quiet <= 0)
ab56093f NB	174	fprintf(stderr, Name ": stopped %s\n", devname);
daf7a3ce	175	if (fd >= 0 && fstat(fd, &stb) == 0) {
8382f19b NB	176	int devnum;
	177	if (major(stb.st_rdev) == MD_MAJOR)
	178	devnum = minor(stb.st_rdev);
	179	else
	180	devnum = -1-(minor(stb.st_rdev)>>6);
	181	map_delete(&map, devnum);
	182	map_write(map);
	183	map_free(map);
	184	}
682c7051 NB	185	}
682c7051 NB	186	return 0;
64c4757e NB	187	}
64c4757e NB	188
dd0781e5 NB	189	int Manage_resize(char *devname, int fd, long long size, int raid_disks)
	190	{
	191	mdu_array_info_t info;
	192	if (ioctl(fd, GET_ARRAY_INFO, &info) != 0) {
	193	fprintf(stderr, Name ": Cannot get array information for %s: %s\n",
	194	devname, strerror(errno));
	195	return 1;
	196	}
	197	if (size >= 0)
	198	info.size = size;
	199	if (raid_disks > 0)
	200	info.raid_disks = raid_disks;
	201	if (ioctl(fd, SET_ARRAY_INFO, &info) != 0) {
	202	fprintf(stderr, Name ": Cannot set device size/shape for %s: %s\n",
	203	devname, strerror(errno));
	204	return 1;
	205	}
	206	return 0;
	207	}
	208
b5e64645 NB	209	int Manage_reconfig(char *devname, int fd, int layout)
	210	{
	211	mdu_array_info_t info;
	212	if (ioctl(fd, GET_ARRAY_INFO, &info) != 0) {
	213	fprintf(stderr, Name ": Cannot get array information for %s: %s\n",
	214	devname, strerror(errno));
	215	return 1;
	216	}
	217	info.layout = layout;
	218	printf("layout set to %d\n", info.layout);
	219	if (ioctl(fd, SET_ARRAY_INFO, &info) != 0) {
	220	fprintf(stderr, Name ": Cannot set layout for %s: %s\n",
	221	devname, strerror(errno));
	222	return 1;
	223	}
	224	return 0;
	225	}
dd0781e5	226
64c4757e	227	int Manage_subdevs(char *devname, int fd,
dab6685f	228	mddev_dev_t devlist, int verbose)
cd29a5c8	229	{
682c7051 NB	230	/* do something to each dev.
	231	* devmode can be
	232	* 'a' - add the device
	233	* try HOT_ADD_DISK
	234	* If that fails EINVAL, try ADD_NEW_DISK
	235	* 'r' - remove the device HOT_REMOVE_DISK
b80da661 NB	236	* device can be 'faulty' or 'detached' in which case all
b80da661 NB	237	* matching devices are removed.
682c7051	238	* 'f' - set the device faulty SET_DISK_FAULTY
b80da661 NB	239	* device can be 'detached' in which case any device that
b80da661 NB	240	* is inaccessible will be marked faulty.
682c7051 NB	241	*/
	242	mdu_array_info_t array;
	243	mdu_disk_info_t disc;
7a3be72f	244	unsigned long long array_size;
b80da661	245	mddev_dev_t dv, next = NULL;
682c7051	246	struct stat stb;
b80da661	247	int j, jnext = 0;
0fbf459d	248	int tfd;
3da92f27	249	struct supertype st, tst;
fe80f49b NB	250	int duuid[4];
fe80f49b NB	251	int ouuid[4];
f94d52f4	252	int lfd = -1;
682c7051 NB	253
	254	if (ioctl(fd, GET_ARRAY_INFO, &array)) {
	255	fprintf(stderr, Name ": cannot get array info for %s\n",
	256	devname);
	257	return 1;
	258	}
3da92f27	259
7a3be72f NB	260	/* array.size is only 32 bit and may be truncated.
	261	* So read from sysfs if possible, and record number of sectors
	262	*/
	263
	264	array_size = get_component_size(fd);
	265	if (array_size <= 0)
	266	array_size = array.size * 2;
	267
1686dc25	268	tst = super_by_fd(fd);
3da92f27 NB	269	if (!tst) {
	270	fprintf(stderr, Name ": unsupport array - version %d.%d\n",
	271	array.major_version, array.minor_version);
	272	return 1;
	273	}
	274
b80da661	275	for (dv = devlist, j=0 ; dv; dv = next, j = jnext) {
4a39c6f2	276	unsigned long long ldsize;
b80da661 NB	277	char dvname[20];
b80da661 NB	278	char *dnprintable = dv->devname;
f94d52f4	279	int err;
4a39c6f2	280
b80da661 NB	281	next = dv->next;
	282	jnext = 0;
	283
	284	if (strcmp(dv->devname, "failed")==0 \|\|
	285	strcmp(dv->devname, "faulty")==0) {
	286	if (dv->disposition != 'r') {
	287	fprintf(stderr, Name ": %s only meaningful "
	288	"with -r, not -%c\n",
	289	dv->devname, dv->disposition);
	290	return 1;
	291	}
	292	for (; j < array.raid_disks + array.nr_disks ; j++) {
	293	disc.number = j;
	294	if (ioctl(fd, GET_DISK_INFO, &disc))
	295	continue;
	296	if (disc.major == 0 && disc.minor == 0)
	297	continue;
	298	if ((disc.state & 1) == 0) /* faulty */
	299	continue;
	300	stb.st_rdev = makedev(disc.major, disc.minor);
	301	next = dv;
	302	jnext = j+1;
	303	sprintf(dvname,"%d:%d", disc.major, disc.minor);
	304	dnprintable = dvname;
	305	break;
	306	}
	307	if (jnext == 0)
	308	continue;
	309	} else if (strcmp(dv->devname, "detached") == 0) {
	310	if (dv->disposition != 'r' && dv->disposition != 'f') {
	311	fprintf(stderr, Name ": %s only meaningful "
	312	"with -r of -f, not -%c\n",
	313	dv->devname, dv->disposition);
	314	return 1;
	315	}
	316	for (; j < array.raid_disks + array.nr_disks; j++) {
	317	int sfd;
	318	disc.number = j;
	319	if (ioctl(fd, GET_DISK_INFO, &disc))
	320	continue;
	321	if (disc.major == 0 && disc.minor == 0)
	322	continue;
	323	sprintf(dvname,"%d:%d", disc.major, disc.minor);
	324	sfd = dev_open(dvname, O_RDONLY);
	325	if (sfd >= 0) {
	326	close(sfd);
	327	continue;
	328	}
	329	if (dv->disposition == 'f' &&
	330	(disc.state & 1) == 1) /* already faulty */
	331	continue;
	332	if (errno != ENXIO)
	333	continue;
	334	stb.st_rdev = makedev(disc.major, disc.minor);
	335	next = dv;
	336	jnext = j+1;
	337	dnprintable = dvname;
	338	break;
	339	}
	340	if (jnext == 0)
	341	continue;
	342	} else {
	343	j = 0;
	344
345	if (stat(dv->devname, &stb)) {
346	fprintf(stderr, Name ": cannot find %s: %s\n",
347	dv->devname, strerror(errno));
348	return 1;
349	}
350	if ((stb.st_mode & S_IFMT) != S_IFBLK) {
351	fprintf(stderr, Name ": %s is not a "
352	"block device.\n",
353	dv->devname);
354	return 1;
355	}
682c7051	356	}
cd29a5c8	357	switch(dv->disposition){
682c7051	358	default:
c913b90e NB	359	fprintf(stderr, Name ": internal error - devmode[%s]=%d\n",
c913b90e NB	360	dv->devname, dv->disposition);
682c7051 NB	361	return 1;
682c7051 NB	362	case 'a':
4a39c6f2	363	/* add the device */
f7e7067b	364	if (tst->subarray[0]) {
f7dd881f DW	365	fprintf(stderr, Name ": Cannot add disks to a"
	366	" \'member\' array, perform this"
	367	" operation on the parent container\n");
	368	return 1;
f94d52f4	369	}
f277ce36	370	/* Make sure it isn't in use (in 2.6 or later) */
6416d527	371	tfd = open(dv->devname, O_RDONLY\|O_EXCL\|O_DIRECT);
0fbf459d	372	if (tfd < 0) {
d7eaf49f NB	373	fprintf(stderr, Name ": Cannot open %s: %s\n",
	374	dv->devname, strerror(errno));
	375	return 1;
	376	}
25037aa8	377	remove_partitions(tfd);
3da92f27	378
1686dc25	379	st = dup_super(tst);
3da92f27	380
fe80f49b	381	if (array.not_persistent==0)
3da92f27 NB	382	st->ss->load_super(st, tfd, NULL);
3da92f27 NB	383
beae1dfe	384	if (!get_dev_size(tfd, dv->devname, &ldsize)) {
4a39c6f2 NB	385	close(tfd);
4a39c6f2 NB	386	return 1;
4a39c6f2	387	}
0fbf459d	388	close(tfd);
fe80f49b	389
f94d52f4 NB	390
	391	if (!tst->ss->external &&
	392	array.major_version == 0 &&
60d9a174	393	md_get_version(fd)%100 < 2) {
892debc8 NB	394	if (ioctl(fd, HOT_ADD_DISK,
892debc8 NB	395	(unsigned long)stb.st_rdev)==0) {
dab6685f NB	396	if (verbose >= 0)
	397	fprintf(stderr, Name ": hot added %s\n",
	398	dv->devname);
892debc8 NB	399	continue;
	400	}
	401
cd29a5c8	402	fprintf(stderr, Name ": hot add failed for %s: %s\n",
892debc8	403	dv->devname, strerror(errno));
cd29a5c8 NB	404	return 1;
cd29a5c8 NB	405	}
892debc8	406
fe80f49b NB	407	if (array.not_persistent == 0) {
fe80f49b NB	408
4a39c6f2	409	/* Make sure device is large enough */
3da92f27	410	if (tst->ss->avail_size(tst, ldsize/512) <
7a3be72f	411	array_size) {
4a39c6f2 NB	412	fprintf(stderr, Name ": %s not large enough to join array\n",
	413	dv->devname);
	414	return 1;
	415	}
	416
fe80f49b	417	/* need to find a sample superblock to copy, and
aba69144	418	* a spare slot to use
fe80f49b	419	*/
3da92f27	420	for (j = 0; j < tst->max_devs; j++) {
fe80f49b NB	421	char *dev;
	422	int dfd;
	423	disc.number = j;
	424	if (ioctl(fd, GET_DISK_INFO, &disc))
	425	continue;
	426	if (disc.major==0 && disc.minor==0)
	427	continue;
	428	if ((disc.state & 4)==0) continue; /* sync */
	429	/* Looks like a good device to try */
16c6fa80	430	dev = map_dev(disc.major, disc.minor, 1);
fe80f49b	431	if (!dev) continue;
16c6fa80	432	dfd = dev_open(dev, O_RDONLY);
fe80f49b	433	if (dfd < 0) continue;
3da92f27 NB	434	if (tst->ss->load_super(tst, dfd,
3da92f27 NB	435	NULL)) {
fe80f49b NB	436	close(dfd);
	437	continue;
	438	}
892debc8	439	close(dfd);
fe80f49b NB	440	break;
fe80f49b NB	441	}
3da92f27	442	if (!tst->sb) {
fe80f49b NB	443	fprintf(stderr, Name ": cannot find valid superblock in this array - HELP\n");
	444	return 1;
	445	}
	446	/* Possibly this device was recently part of the array
	447	* and was temporarily removed, and is now being re-added.
	448	* If so, we can simply re-add it.
	449	*/
3da92f27	450	tst->ss->uuid_from_super(tst, duuid);
7eae7080 NB	451
	452	/* re-add doesn't work for version-1 superblocks
	453	* before 2.6.18 :-(
	454	*/
	455	if (array.major_version == 1 &&
	456	get_linux_version() <= 2006018)
	457	;
3da92f27 NB	458	else if (st->sb) {
3da92f27 NB	459	st->ss->uuid_from_super(st, ouuid);
fe80f49b	460	if (memcmp(duuid, ouuid, sizeof(ouuid))==0) {
4d20d744 NB	461	/* looks close enough for now. Kernel
	462	* will worry about whether a bitmap
	463	* based reconstruction is possible.
fe80f49b NB	464	*/
fe80f49b NB	465	struct mdinfo mdi;
3da92f27	466	st->ss->getinfo_super(st, &mdi);
fe80f49b NB	467	disc.major = major(stb.st_rdev);
	468	disc.minor = minor(stb.st_rdev);
	469	disc.number = mdi.disk.number;
	470	disc.raid_disk = mdi.disk.raid_disk;
	471	disc.state = mdi.disk.state;
4d20d744 NB	472	if (dv->writemostly)
4d20d744 NB	473	disc.state \|= 1 << MD_DISK_WRITEMOSTLY;
fe80f49b NB	474	if (ioctl(fd, ADD_NEW_DISK, &disc) == 0) {
	475	if (verbose >= 0)
	476	fprintf(stderr, Name ": re-added %s\n", dv->devname);
308e1801	477	continue;
fe80f49b NB	478	}
	479	/* fall back on normal-add */
	480	}
892debc8	481	}
4a39c6f2 NB	482	} else {
	483	/* non-persistent. Must ensure that new drive
	484	* is at least array.size big.
	485	*/
7a3be72f	486	if (ldsize/512 < array_size) {
4a39c6f2 NB	487	fprintf(stderr, Name ": %s not large enough to join array\n",
	488	dv->devname);
	489	return 1;
	490	}
892debc8	491	}
7eae7080 NB	492	/* in 2.6.17 and earlier, version-1 superblocks won't
	493	* use the number we write, but will choose a free number.
	494	* we must choose the same free number, which requires
	495	* starting at 'raid_disks' and counting up
	496	*/
3da92f27	497	for (j = array.raid_disks; j< tst->max_devs; j++) {
aa88f531	498	disc.number = j;
682c7051 NB	499	if (ioctl(fd, GET_DISK_INFO, &disc))
	500	break;
	501	if (disc.major==0 && disc.minor==0)
	502	break;
	503	if (disc.state & 8) /* removed */
	504	break;
	505	}
0df46c2a NB	506	disc.major = major(stb.st_rdev);
0df46c2a NB	507	disc.minor = minor(stb.st_rdev);
892debc8 NB	508	disc.number =j;
892debc8 NB	509	disc.state = 0;
fe80f49b	510	if (array.not_persistent==0) {
111d01fc	511	int dfd;
fe80f49b NB	512	if (dv->writemostly)
fe80f49b NB	513	disc.state \|= 1 << MD_DISK_WRITEMOSTLY;
6416d527	514	dfd = open(dv->devname, O_RDWR \| O_EXCL\|O_DIRECT);
111d01fc NB	515	tst->ss->add_to_super(tst, &disc, dfd,
	516	dv->devname);
	517	/* write_init_super will close 'dfd' */
	518	if (tst->ss->write_init_super(tst))
fe80f49b NB	519	return 1;
	520	} else if (dv->re_add) {
	521	/* this had better be raid1.
	522	* As we are "--re-add"ing we must find a spare slot
	523	* to fill.
	524	*/
	525	char *used = malloc(array.raid_disks);
	526	memset(used, 0, array.raid_disks);
3da92f27	527	for (j=0; j< tst->max_devs; j++) {
fe80f49b NB	528	mdu_disk_info_t disc2;
	529	disc2.number = j;
	530	if (ioctl(fd, GET_DISK_INFO, &disc2))
	531	continue;
	532	if (disc2.major==0 && disc2.minor==0)
	533	continue;
	534	if (disc2.state & 8) /* removed */
	535	continue;
	536	if (disc2.raid_disk < 0)
	537	continue;
	538	if (disc2.raid_disk > array.raid_disks)
	539	continue;
	540	used[disc2.raid_disk] = 1;
	541	}
	542	for (j=0 ; j<array.raid_disks; j++)
	543	if (!used[j]) {
	544	disc.raid_disk = j;
	545	disc.state \|= (1<<MD_DISK_SYNC);
	546	break;
	547	}
	548	}
d8def61c NB	549	if (dv->writemostly)
d8def61c NB	550	disc.state \|= (1 << MD_DISK_WRITEMOSTLY);
682c7051	551	if (ioctl(fd,ADD_NEW_DISK, &disc)) {
f277ce36 NB	552	fprintf(stderr, Name ": add new device failed for %s as %d: %s\n",
f277ce36 NB	553	dv->devname, j, strerror(errno));
682c7051 NB	554	return 1;
682c7051 NB	555	}
dab6685f NB	556	if (verbose >= 0)
dab6685f NB	557	fprintf(stderr, Name ": added %s\n", dv->devname);
682c7051 NB	558	break;
	559
	560	case 'r':
	561	/* hot remove */
f7e7067b	562	if (tst->subarray[0]) {
f7dd881f DW	563	fprintf(stderr, Name ": Cannot remove disks from a"
	564	" \'member\' array, perform this"
	565	" operation on the parent container\n");
	566	return 1;
f94d52f4 NB	567	}
	568	if (tst->ss->external) {
	569	/* To remove a device from a container, we must
	570	* check that it isn't in use in an array.
	571	* This involves looking in the 'holders'
	572	* directory - there must be just one entry,
	573	* the container.
	574	* To ensure that it doesn't get used as a
	575	* hold spare while we are checking, we
	576	* get an O_EXCL open on the container
	577	*/
	578	int dnum = fd2devnum(fd);
	579	lfd = open_dev_excl(dnum);
	580	if (lfd < 0) {
	581	fprintf(stderr, Name
	582	": Cannot get exclusive access "
	583	" to container - odd\n");
	584	return 1;
	585	}
	586	if (!sysfs_unique_holder(dnum, stb.st_rdev)) {
	587	fprintf(stderr, Name
	588	": %s is %s, cannot remove.\n",
	589	dnprintable,
	590	errno == EEXIST ? "still in use":
	591	"not a member");
	592	close(lfd);
	593	return 1;
	594	}
	595	}
82b27616	596	/* FIXME check that it is a current member */
f94d52f4 NB	597	err = ioctl(fd, HOT_REMOVE_DISK, (unsigned long)stb.st_rdev);
	598	if (err && errno == ENODEV) {
	599	/* Old kernels rejected this if no personality
	600	* registered */
	601	struct mdinfo *sra = sysfs_read(fd, 0, GET_DEVS);
	602	struct mdinfo *dv = NULL;
	603	if (sra)
	604	dv = sra->devs;
	605	for ( ; dv ; dv=dv->next)
	606	if (dv->disk.major == major(stb.st_rdev) &&
	607	dv->disk.minor == minor(stb.st_rdev))
	608	break;
	609	if (dv)
	610	err = sysfs_set_str(sra, dv,
	611	"state", "remove");
	612	else
	613	err = -1;
	614	if (sra)
	615	sysfs_free(sra);
	616	}
	617	if (err) {
b80da661 NB	618	fprintf(stderr, Name ": hot remove failed "
	619	"for %s: %s\n", dnprintable,
	620	strerror(errno));
f94d52f4 NB	621	if (lfd >= 0)
f94d52f4 NB	622	close(lfd);
682c7051 NB	623	return 1;
682c7051 NB	624	}
f94d52f4	625	close(lfd);
dab6685f	626	if (verbose >= 0)
b80da661 NB	627	fprintf(stderr, Name ": hot removed %s\n",
b80da661 NB	628	dnprintable);
682c7051 NB	629	break;
	630
	631	case 'f': /* set faulty */
	632	/* FIXME check current member */
cd29a5c8	633	if (ioctl(fd, SET_DISK_FAULTY, (unsigned long) stb.st_rdev)) {
b83d95f3	634	fprintf(stderr, Name ": set device faulty failed for %s: %s\n",
b80da661	635	dnprintable, strerror(errno));
682c7051 NB	636	return 1;
682c7051 NB	637	}
dab6685f NB	638	if (verbose >= 0)
dab6685f NB	639	fprintf(stderr, Name ": set %s faulty in %s\n",
b80da661	640	dnprintable, devname);
682c7051 NB	641	break;
	642	}
	643	}
	644	return 0;
aba69144	645
64c4757e	646	}
1f48664b NB	647
	648	int autodetect(void)
	649	{
	650	/* Open any md device, and issue the RAID_AUTORUN ioctl */
	651	int rv = 1;
	652	int fd = dev_open("9:0", O_RDONLY);
	653	if (fd >= 0) {
	654	if (ioctl(fd, RAID_AUTORUN, 0) == 0)
	655	rv = 0;
	656	close(fd);
	657	}
	658	return rv;
	659	}
435d4ebb	660	#endif