[thirdparty/mdadm.git] / Create.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"
#include	<ctype.h>

int Create(struct supertype *st, char *mddev, int mdfd,
	   int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
	   char *name, char *homehost, int *uuid,
	   int subdevs, mddev_dev_t devlist,
	   int runstop, int verbose, int force, int assume_clean,
	   char *bitmap_file, int bitmap_chunk, int write_behind, int delay)
{
	/*
	 * Create a new raid array.
	 *
	 * First check that necessary details are available
	 * (i.e. level, raid-disks)
	 *
	 * Then check each disk to see what might be on it
	 * and report anything interesting.
	 *
	 * If anything looks odd, and runstop not set,
	 * abort.
	 *
	 * SET_ARRAY_INFO and ADD_NEW_DISK, and
	 * if runstop==run, or raiddisks disks were used,
	 * RUN_ARRAY
	 */
	unsigned long long minsize=0, maxsize=0;
	char *mindisc = NULL;
	char *maxdisc = NULL;
	int dnum;
	mddev_dev_t dv;
	int fail=0, warn=0;
	struct stat stb;
	int first_missing = subdevs * 2;
	int second_missing = subdevs * 2;
	int missing_disks = 0;
	int insert_point = subdevs * 2; /* where to insert a missing drive */
	int pass;
	int vers;
	int rv;
	int bitmap_fd;
	unsigned long long bitmapsize;
	struct mdinfo *sra;
	struct mdinfo info;

	int major_num = BITMAP_MAJOR_HI;

	memset(&info, 0, sizeof(info));

	vers = md_get_version(mdfd);
	if (vers < 9000) {
		fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
		return 1;
	} else {
		mdu_array_info_t inf;
		memset(&inf, 0, sizeof(inf));
		ioctl(mdfd, GET_ARRAY_INFO, &inf);
		if (inf.working_disks != 0) {
			fprintf(stderr, Name ": another array by this name"
				" is already running.\n");
			return 1;
		}
	}
	if (level == UnSet) {
		fprintf(stderr,
			Name ": a RAID level is needed to create an array.\n");
		return 1;
	}
	if (raiddisks < 4 && level == 6) {
		fprintf(stderr,
			Name ": at least 4 raid-devices needed for level 6\n");
		return 1;
	}
	if (raiddisks > 256 && level == 6) {
		fprintf(stderr,
			Name ": no more than 256 raid-devices supported for level 6\n");
		return 1;
	}
	if (raiddisks < 2 && level >= 4) {
		fprintf(stderr,
			Name ": at least 2 raid-devices needed for level 4 or 5\n");
		return 1;
	}
	if (level <= 0 && sparedisks) {
		fprintf(stderr,
			Name ": This level does not support spare devices\n");
		return 1;
	}
	if (subdevs > raiddisks+sparedisks) {
		fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks);
		return 1;
	}
	if (subdevs < raiddisks+sparedisks) {
		fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n");
		return 1;
	}
	if (bitmap_file && level <= 0) {
		fprintf(stderr, Name ": bitmaps not meaningful with level %s\n",
			map_num(pers, level)?:"given");
		return 1;
	}

	/* now set some defaults */
	if (layout == UnSet)
		switch(level) {
		default: /* no layout */
			layout = 0;
			break;
		case 10:
			layout = 0x102; /* near=2, far=1 */
			if (verbose > 0)
				fprintf(stderr,
					Name ": layout defaults to n1\n");
			break;
		case 5:
		case 6:
			layout = map_name(r5layout, "default");
			if (verbose > 0)
				fprintf(stderr,
					Name ": layout defaults to %s\n", map_num(r5layout, layout));
			break;
		case LEVEL_FAULTY:
			layout = map_name(faultylayout, "default");

			if (verbose > 0)
				fprintf(stderr,
					Name ": layout defaults to %s\n", map_num(faultylayout, layout));
			break;
		}

	if (level == 10)
		/* check layout fits in array*/
		if ((layout&255) * ((layout>>8)&255) > raiddisks) {
			fprintf(stderr, Name ": that layout requires at least %d devices\n",
				(layout&255) * ((layout>>8)&255));
			return 1;
		}

	switch(level) {
	case 4:
	case 5:
	case 10:
	case 6:
	case 0:
	case LEVEL_LINEAR: /* linear */
		if (chunk == 0) {
			chunk = 64;
			if (verbose > 0)
				fprintf(stderr, Name ": chunk size defaults to 64K\n");
		}
		break;
	case 1:
	case LEVEL_FAULTY:
	case LEVEL_MULTIPATH:
	case LEVEL_CONTAINER:
		if (chunk) {
			chunk = 0;
			if (verbose > 0)
				fprintf(stderr, Name ": chunk size ignored for this level\n");
		}
		break;
	default:
		fprintf(stderr, Name ": unknown level %d\n", level);
		return 1;
	}

	if (st && ! st->ss->validate_geometry(st, level, layout, raiddisks,
					      chunk, size, NULL, NULL))
		return 1;

	/* now look at the subdevs */
	info.array.active_disks = 0;
	info.array.working_disks = 0;
	dnum = 0;
	for (dv=devlist; dv; dv=dv->next, dnum++) {
		char *dname = dv->devname;
		unsigned long long freesize;
		if (strcasecmp(dname, "missing")==0) {
			if (first_missing > dnum)
				first_missing = dnum;
			if (second_missing > dnum && dnum > first_missing)
				second_missing = dnum;
			missing_disks ++;
			continue;
		}
		info.array.working_disks++;
		if (dnum < raiddisks)
			info.array.active_disks++;
		if (st == NULL) {
			struct createinfo *ci = conf_get_create_info();
			if (ci)
				st = ci->supertype;
		}
		if (st == NULL) {
			/* Need to choose a default metadata, which is different
			 * depending on geometry of array.
			 */
			int i;
			char *name = "default";
			for(i=0; !st && superlist[i]; i++) {
				st = superlist[i]->match_metadata_desc(name);
				if (st && !st->ss->validate_geometry
					    	(st, level, layout, raiddisks,
						 chunk, size, dname, &freesize))
					st = NULL;
			}

			if (!st) {
				fprintf(stderr, Name ": device %s not suitable "
					"for any style of array\n",
					dname);
				exit(2);
			}
			if (st->ss->major != 0 ||
			    st->minor_version != 90)
				fprintf(stderr, Name ": Defaulting to version"
					" %d.%d metadata\n",
					st->ss->major,
					st->minor_version);
		} else {
			if (!st->ss->validate_geometry(st, level, layout,
						       raiddisks,
						       chunk, size, dname,
						       &freesize)) {

				fprintf(stderr,
					Name ": %s is not suitable for "
					"this array.\n",
					dname);
				fail = 1;
				continue;
			}
		}

		freesize /= 2; /* convert to K */
		if (chunk) {
			/* round to chunk size */
			freesize = freesize & ~(chunk-1);
		}

		if (size && freesize < size) {
			fprintf(stderr, Name ": %s is smaller that given size."
				" %lluK < %lluK + metadata\n",
				dname, freesize, size);
			fail = 1;
			continue;
		}
		if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
			maxdisc = dname;
			maxsize = freesize;
		}
		if (mindisc ==NULL || (mindisc && freesize < minsize)) {
			mindisc = dname;
			minsize = freesize;
		}
		if (runstop != 1 || verbose >= 0) {
			int fd = open(dname, O_RDONLY, 0);
			if (fd <0 ) {
				fprintf(stderr, Name ": Cannot open %s: %s\n",
					dname, strerror(errno));
				fail=1;
				continue;
			}
			warn |= check_ext2(fd, dname);
			warn |= check_reiser(fd, dname);
			warn |= check_raid(fd, dname);
			close(fd);
		}
	}
	if (fail) {
		fprintf(stderr, Name ": create aborted\n");
		return 1;
	}
	if (size == 0) {
		if (mindisc == NULL) {
			fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
			return 1;
		}
		if (level > 0 || level == LEVEL_MULTIPATH || level == LEVEL_FAULTY) {
			/* size is meaningful */
			if (minsize > 0x100000000ULL && st->ss->major == 0) {
				fprintf(stderr, Name ": devices too large for RAID level %d\n", level);
				return 1;
			}
			size = minsize;
			if (verbose > 0)
				fprintf(stderr, Name ": size set to %lluK\n", size);
		}
	}
	if (level > 0 && ((maxsize-size)*100 > maxsize)) {
		if (runstop != 1 || verbose >= 0)
			fprintf(stderr, Name ": largest drive (%s) exceed size (%lluK) by more than 1%%\n",
				maxdisc, size);
		warn = 1;
	}

	if (warn) {
		if (runstop!= 1) {
			if (!ask("Continue creating array? ")) {
				fprintf(stderr, Name ": create aborted.\n");
				return 1;
			}
		} else {
			if (verbose > 0)
				fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
		}
	}

	/* If this is raid4/5, we want to configure the last active slot
	 * as missing, so that a reconstruct happens (faster than re-parity)
	 * FIX: Can we do this for raid6 as well?
	 */
	if (assume_clean==0 && force == 0 && first_missing >= raiddisks) {
		switch ( level ) {
		case 4:
		case 5:
			insert_point = raiddisks-1;
			sparedisks++;
			info.array.active_disks--;
			missing_disks++;
			break;
		default:
			break;
		}
	}
	/* For raid6, if creating with 1 missing drive, make a good drive
	 * into a spare, else the create will fail
	 */
	if (assume_clean == 0 && force == 0 && first_missing < raiddisks &&
	    second_missing >= raiddisks && level == 6) {
		insert_point = raiddisks - 1;
		if (insert_point == first_missing)
			insert_point--;
		sparedisks ++;
		info.array.active_disks--;
		missing_disks++;
	}

	if (level <= 0 && first_missing != subdevs * 2) {
		fprintf(stderr,
			Name ": This level does not support missing devices\n");
		return 1;
	}

	/* Ok, lets try some ioctls */

	info.array.level = level;
	info.array.size = size;
	info.array.raid_disks = raiddisks;
	/* The kernel should *know* what md_minor we are dealing
	 * with, but it chooses to trust me instead. Sigh
	 */
	info.array.md_minor = 0;
	if (fstat(mdfd, &stb)==0)
		info.array.md_minor = minor(stb.st_rdev);
	info.array.not_persistent = 0;

	if ( ( (level == 4 || level == 5) &&
	       (insert_point < raiddisks || first_missing < raiddisks) )
	     ||
	     ( level == 6 && (insert_point < raiddisks
			      || second_missing < raiddisks))
	     ||
	     assume_clean
		)
		info.array.state = 1; /* clean, but one+ drive will be missing*/
	else
		info.array.state = 0; /* not clean, but no errors */

	if (level == 10) {
		/* for raid10, the bitmap size is the capacity of the array,
		 * which is array.size * raid_disks / ncopies;
		 * .. but convert to sectors.
		 */
		int ncopies = ((layout>>8) & 255) * (layout & 255);
		bitmapsize = (unsigned long long)size * raiddisks / ncopies * 2;
/*		printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/
	} else
		bitmapsize = (unsigned long long)size * 2;

	/* There is lots of redundancy in these disk counts,
	 * raid_disks is the most meaningful value
	 *          it describes the geometry of the array
	 *          it is constant
	 * nr_disks is total number of used slots.
	 *          it should be raid_disks+spare_disks
	 * spare_disks is the number of extra disks present
	 *          see above
	 * active_disks is the number of working disks in
	 *          active slots. (With raid_disks)
	 * working_disks is the total number of working disks,
	 *          including spares
	 * failed_disks is the number of disks marked failed
	 *
         * Ideally, the kernel would keep these (except raid_disks)
	 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
	 * So for now, we assume that all raid and spare
	 * devices will be given.
	 */
	info.array.spare_disks=sparedisks;
	info.array.failed_disks=missing_disks;
	info.array.nr_disks = info.array.working_disks
		+ info.array.failed_disks;
	info.array.layout = layout;
	info.array.chunk_size = chunk*1024;
	info.array.major_version = st->ss->major;

	if (name == NULL || *name == 0) {
		/* base name on mddev */
		/*  /dev/md0 -> 0
		 *  /dev/md_d0 -> d0
		 *  /dev/md/1 -> 1
		 *  /dev/md/d1 -> d1
		 *  /dev/md/home -> home
		 *  /dev/mdhome -> home
		 */
		name = strrchr(mddev, '/');
		if (name) {
			name++;
			if (strncmp(name, "md_d", 4)==0 &&
			    strlen(name) > 4 &&
			    isdigit(name[4]) &&
			    (name-mddev) == 5 /* /dev/ */)
				name += 3;
			else if (strncmp(name, "md", 2)==0 &&
				 strlen(name) > 2 &&
				 isdigit(name[2]) &&
				 (name-mddev) == 5 /* /dev/ */)
				name += 2;
		}
	}
	if (!st->ss->init_super(st, &info.array, size, name, homehost, uuid))
		return 1;

	if (bitmap_file && vers < 9003) {
		major_num = BITMAP_MAJOR_HOSTENDIAN;
#ifdef __BIG_ENDIAN
		fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n"
			"  between different architectured.  Consider upgrading the Linux kernel.\n");
#endif
	}

	if (bitmap_file && strcmp(bitmap_file, "internal")==0) {
		if ((vers%100) < 2) {
			fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n");
			return 1;
		}
		if (!st->ss->add_internal_bitmap(st, &bitmap_chunk,
						 delay, write_behind,
						 bitmapsize, 1, major_num)) {
			fprintf(stderr, Name ": Given bitmap chunk size not supported.\n");
			return 1;
		}
		bitmap_file = NULL;
	}


	sra = sysfs_read(mdfd, 0, 0);

	if (st->ss->external) {
		char ver[100];
		strcat(strcpy(ver, "external:"), st->ss->text_version);
		if ((vers % 100) < 2 ||
		    sra == NULL ||
		    sysfs_set_str(sra, NULL, "metadata_version",
				  ver) < 0) {
			fprintf(stderr, Name ": This kernel does not "
				"support external metadata.\n");
			return 1;
		}
		rv = 0;
	} else	if ((vers % 100) >= 1) { /* can use different versions */
		mdu_array_info_t inf;
		memset(&inf, 0, sizeof(inf));
		inf.major_version = st->ss->major;
		inf.minor_version = st->minor_version;
		rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
	} else
		rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
	if (rv) {
		fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
			mddev, strerror(errno));
		return 1;
	}

	if (bitmap_file) {
		int uuid[4];

		st->ss->uuid_from_super(st, uuid);
		if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk,
				 delay, write_behind,
				 bitmapsize,
				 major_num)) {
			return 1;
		}
		bitmap_fd = open(bitmap_file, O_RDWR);
		if (bitmap_fd < 0) {
			fprintf(stderr, Name ": weird: %s cannot be openned\n",
				bitmap_file);
			return 1;
		}
		if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) {
			fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n",
				mddev, strerror(errno));
			return 1;
		}
	}


	for (pass=1; pass <=2 ; pass++) {
		mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */

		for (dnum=0, dv = devlist ; dv ;
		     dv=(dv->next)?(dv->next):moved_disk, dnum++) {
			int fd;
			struct stat stb;

			info.disk.number = dnum;
			if (dnum == insert_point) {
				moved_disk = dv;
			}
			info.disk.raid_disk = info.disk.number;
			if (info.disk.raid_disk < raiddisks)
				info.disk.state = (1<<MD_DISK_ACTIVE) |
						(1<<MD_DISK_SYNC);
			else
				info.disk.state = 0;
			if (dv->writemostly)
				info.disk.state |= (1<<MD_DISK_WRITEMOSTLY);

			if (dnum == insert_point ||
			    strcasecmp(dv->devname, "missing")==0)
				continue;

			fd = open(dv->devname, O_RDWR|O_EXCL, 0);
			if (fd < 0) {
				fprintf(stderr, Name ": failed to open %s "
					"after earlier success - aborting\n",
					dv->devname);
				return 1;
			}
			fstat(fd, &stb);
			info.disk.major = major(stb.st_rdev);
			info.disk.minor = minor(stb.st_rdev);

			switch(pass){
			case 1:
				remove_partitions(fd);
				st->ss->add_to_super(st, &info.disk,
						     fd, dv->devname);
				break;
			case 2:
				close(fd);

				if (st->ss->external) {
					char dv[100];
					sprintf(dv, "%d:%d\n",
						info.disk.major,
						info.disk.minor);
					sysfs_set_str(sra, NULL, "new_dev", dv);
					/* FIXME check error */
					/*FIXME find that device and set it up*/
				} else if (ioctl(mdfd, ADD_NEW_DISK,
						 &info.disk)) {
					fprintf(stderr,
						Name ": ADD_NEW_DISK for %s "
						"failed: %s\n",
						dv->devname, strerror(errno));
					st->ss->free_super(st);
					return 1;
				}
				break;
			}
			if (dv == moved_disk && dnum != insert_point) break;
		}
		if (pass == 1)
			st->ss->write_init_super(st);
	}
	st->ss->free_super(st);

	/* param is not actually used */
	if (level == LEVEL_CONTAINER)
		/* No need to start */
		;
	else if (runstop == 1 || subdevs >= raiddisks) {
		mdu_param_t param;
		if (ioctl(mdfd, RUN_ARRAY, &param)) {
			fprintf(stderr, Name ": RUN_ARRAY failed: %s\n",
				strerror(errno));
			Manage_runstop(mddev, mdfd, -1, 0);
			return 1;
		}
		if (verbose >= 0)
			fprintf(stderr, Name ": array %s started.\n", mddev);
	} else {
		fprintf(stderr, Name ": not starting array - not enough devices.\n");
	}
	return 0;
}
Commit	Line	Data
	1	/*
	2	* mdadm - manage Linux "md" devices aka RAID arrays.
	3	*
	4	* Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
	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
	30	#include "mdadm.h"
	31	#include "md_u.h"
	32	#include "md_p.h"
	33	#include <ctype.h>
	34
	35	int Create(struct supertype st, char mddev, int mdfd,
	36	int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
	37	char name, char homehost, int *uuid,
	38	int subdevs, mddev_dev_t devlist,
	39	int runstop, int verbose, int force, int assume_clean,
	40	char *bitmap_file, int bitmap_chunk, int write_behind, int delay)
	41	{
	42	/*
	43	* Create a new raid array.
	44	*
	45	* First check that necessary details are available
	46	* (i.e. level, raid-disks)
	47	*
	48	* Then check each disk to see what might be on it
	49	* and report anything interesting.
	50	*
	51	* If anything looks odd, and runstop not set,
	52	* abort.
	53	*
	54	* SET_ARRAY_INFO and ADD_NEW_DISK, and
	55	* if runstop==run, or raiddisks disks were used,
	56	* RUN_ARRAY
	57	*/
	58	unsigned long long minsize=0, maxsize=0;
	59	char *mindisc = NULL;
	60	char *maxdisc = NULL;
	61	int dnum;
	62	mddev_dev_t dv;
	63	int fail=0, warn=0;
	64	struct stat stb;
	65	int first_missing = subdevs * 2;
	66	int second_missing = subdevs * 2;
	67	int missing_disks = 0;
	68	int insert_point = subdevs * 2; /* where to insert a missing drive */
	69	int pass;
	70	int vers;
	71	int rv;
	72	int bitmap_fd;
	73	unsigned long long bitmapsize;
	74	struct mdinfo *sra;
	75	struct mdinfo info;
	76
	77	int major_num = BITMAP_MAJOR_HI;
	78
	79	memset(&info, 0, sizeof(info));
	80
	81	vers = md_get_version(mdfd);
	82	if (vers < 9000) {
	83	fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
	84	return 1;
	85	} else {
	86	mdu_array_info_t inf;
	87	memset(&inf, 0, sizeof(inf));
	88	ioctl(mdfd, GET_ARRAY_INFO, &inf);
	89	if (inf.working_disks != 0) {
	90	fprintf(stderr, Name ": another array by this name"
	91	" is already running.\n");
	92	return 1;
	93	}
	94	}
	95	if (level == UnSet) {
	96	fprintf(stderr,
	97	Name ": a RAID level is needed to create an array.\n");
	98	return 1;
	99	}
	100	if (raiddisks < 4 && level == 6) {
	101	fprintf(stderr,
	102	Name ": at least 4 raid-devices needed for level 6\n");
	103	return 1;
	104	}
	105	if (raiddisks > 256 && level == 6) {
	106	fprintf(stderr,
	107	Name ": no more than 256 raid-devices supported for level 6\n");
	108	return 1;
	109	}
	110	if (raiddisks < 2 && level >= 4) {
	111	fprintf(stderr,
	112	Name ": at least 2 raid-devices needed for level 4 or 5\n");
	113	return 1;
	114	}
	115	if (level <= 0 && sparedisks) {
	116	fprintf(stderr,
	117	Name ": This level does not support spare devices\n");
	118	return 1;
	119	}
	120	if (subdevs > raiddisks+sparedisks) {
	121	fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks);
	122	return 1;
	123	}
	124	if (subdevs < raiddisks+sparedisks) {
	125	fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n");
	126	return 1;
	127	}
	128	if (bitmap_file && level <= 0) {
	129	fprintf(stderr, Name ": bitmaps not meaningful with level %s\n",
	130	map_num(pers, level)?:"given");
	131	return 1;
	132	}
	133
	134	/* now set some defaults */
	135	if (layout == UnSet)
	136	switch(level) {
	137	default: /* no layout */
	138	layout = 0;
	139	break;
	140	case 10:
	141	layout = 0x102; /* near=2, far=1 */
	142	if (verbose > 0)
	143	fprintf(stderr,
	144	Name ": layout defaults to n1\n");
	145	break;
	146	case 5:
	147	case 6:
	148	layout = map_name(r5layout, "default");
	149	if (verbose > 0)
	150	fprintf(stderr,
	151	Name ": layout defaults to %s\n", map_num(r5layout, layout));
	152	break;
	153	case LEVEL_FAULTY:
	154	layout = map_name(faultylayout, "default");
	155
	156	if (verbose > 0)
	157	fprintf(stderr,
	158	Name ": layout defaults to %s\n", map_num(faultylayout, layout));
	159	break;
	160	}
	161
	162	if (level == 10)
	163	/* check layout fits in array*/
	164	if ((layout&255) * ((layout>>8)&255) > raiddisks) {
	165	fprintf(stderr, Name ": that layout requires at least %d devices\n",
	166	(layout&255) * ((layout>>8)&255));
	167	return 1;
	168	}
	169
	170	switch(level) {
	171	case 4:
	172	case 5:
	173	case 10:
	174	case 6:
	175	case 0:
	176	case LEVEL_LINEAR: /* linear */
	177	if (chunk == 0) {
	178	chunk = 64;
	179	if (verbose > 0)
	180	fprintf(stderr, Name ": chunk size defaults to 64K\n");
	181	}
	182	break;
	183	case 1:
	184	case LEVEL_FAULTY:
	185	case LEVEL_MULTIPATH:
	186	case LEVEL_CONTAINER:
	187	if (chunk) {
	188	chunk = 0;
	189	if (verbose > 0)
	190	fprintf(stderr, Name ": chunk size ignored for this level\n");
	191	}
	192	break;
	193	default:
	194	fprintf(stderr, Name ": unknown level %d\n", level);
	195	return 1;
	196	}
	197
	198	if (st && ! st->ss->validate_geometry(st, level, layout, raiddisks,
	199	chunk, size, NULL, NULL))
	200	return 1;
	201
	202	/* now look at the subdevs */
	203	info.array.active_disks = 0;
	204	info.array.working_disks = 0;
	205	dnum = 0;
	206	for (dv=devlist; dv; dv=dv->next, dnum++) {
	207	char *dname = dv->devname;
	208	unsigned long long freesize;
	209	if (strcasecmp(dname, "missing")==0) {
	210	if (first_missing > dnum)
	211	first_missing = dnum;
	212	if (second_missing > dnum && dnum > first_missing)
	213	second_missing = dnum;
	214	missing_disks ++;
	215	continue;
	216	}
	217	info.array.working_disks++;
	218	if (dnum < raiddisks)
	219	info.array.active_disks++;
	220	if (st == NULL) {
	221	struct createinfo *ci = conf_get_create_info();
	222	if (ci)
	223	st = ci->supertype;
	224	}
	225	if (st == NULL) {
	226	/* Need to choose a default metadata, which is different
	227	* depending on geometry of array.
	228	*/
	229	int i;
	230	char *name = "default";
	231	for(i=0; !st && superlist[i]; i++) {
	232	st = superlist[i]->match_metadata_desc(name);
	233	if (st && !st->ss->validate_geometry
	234	(st, level, layout, raiddisks,
	235	chunk, size, dname, &freesize))
	236	st = NULL;
	237	}
	238
	239	if (!st) {
	240	fprintf(stderr, Name ": device %s not suitable "
	241	"for any style of array\n",
	242	dname);
	243	exit(2);
	244	}
	245	if (st->ss->major != 0 \|\|
	246	st->minor_version != 90)
	247	fprintf(stderr, Name ": Defaulting to version"
	248	" %d.%d metadata\n",
	249	st->ss->major,
	250	st->minor_version);
	251	} else {
	252	if (!st->ss->validate_geometry(st, level, layout,
	253	raiddisks,
	254	chunk, size, dname,
	255	&freesize)) {
	256
	257	fprintf(stderr,
	258	Name ": %s is not suitable for "
	259	"this array.\n",
	260	dname);
	261	fail = 1;
	262	continue;
	263	}
	264	}
	265
	266	freesize /= 2; /* convert to K */
	267	if (chunk) {
	268	/* round to chunk size */
	269	freesize = freesize & ~(chunk-1);
	270	}
	271
	272	if (size && freesize < size) {
	273	fprintf(stderr, Name ": %s is smaller that given size."
	274	" %lluK < %lluK + metadata\n",
	275	dname, freesize, size);
	276	fail = 1;
	277	continue;
	278	}
	279	if (maxdisc == NULL \|\| (maxdisc && freesize > maxsize)) {
	280	maxdisc = dname;
	281	maxsize = freesize;
	282	}
	283	if (mindisc ==NULL \|\| (mindisc && freesize < minsize)) {
	284	mindisc = dname;
	285	minsize = freesize;
	286	}
	287	if (runstop != 1 \|\| verbose >= 0) {
	288	int fd = open(dname, O_RDONLY, 0);
	289	if (fd <0 ) {
	290	fprintf(stderr, Name ": Cannot open %s: %s\n",
	291	dname, strerror(errno));
	292	fail=1;
	293	continue;
	294	}
	295	warn \|= check_ext2(fd, dname);
	296	warn \|= check_reiser(fd, dname);
	297	warn \|= check_raid(fd, dname);
	298	close(fd);
	299	}
	300	}
	301	if (fail) {
	302	fprintf(stderr, Name ": create aborted\n");
	303	return 1;
	304	}
	305	if (size == 0) {
	306	if (mindisc == NULL) {
	307	fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
	308	return 1;
	309	}
	310	if (level > 0 \|\| level == LEVEL_MULTIPATH \|\| level == LEVEL_FAULTY) {
	311	/* size is meaningful */
	312	if (minsize > 0x100000000ULL && st->ss->major == 0) {
	313	fprintf(stderr, Name ": devices too large for RAID level %d\n", level);
	314	return 1;
	315	}
	316	size = minsize;
	317	if (verbose > 0)
	318	fprintf(stderr, Name ": size set to %lluK\n", size);
	319	}
	320	}
	321	if (level > 0 && ((maxsize-size)*100 > maxsize)) {
	322	if (runstop != 1 \|\| verbose >= 0)
	323	fprintf(stderr, Name ": largest drive (%s) exceed size (%lluK) by more than 1%%\n",
	324	maxdisc, size);
	325	warn = 1;
	326	}
	327
	328	if (warn) {
	329	if (runstop!= 1) {
	330	if (!ask("Continue creating array? ")) {
	331	fprintf(stderr, Name ": create aborted.\n");
	332	return 1;
	333	}
	334	} else {
	335	if (verbose > 0)
	336	fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
	337	}
	338	}
	339
	340	/* If this is raid4/5, we want to configure the last active slot
	341	* as missing, so that a reconstruct happens (faster than re-parity)
	342	* FIX: Can we do this for raid6 as well?
	343	*/
	344	if (assume_clean==0 && force == 0 && first_missing >= raiddisks) {
	345	switch ( level ) {
	346	case 4:
	347	case 5:
	348	insert_point = raiddisks-1;
	349	sparedisks++;
	350	info.array.active_disks--;
	351	missing_disks++;
	352	break;
	353	default:
	354	break;
	355	}
	356	}
	357	/* For raid6, if creating with 1 missing drive, make a good drive
	358	* into a spare, else the create will fail
	359	*/
	360	if (assume_clean == 0 && force == 0 && first_missing < raiddisks &&
	361	second_missing >= raiddisks && level == 6) {
	362	insert_point = raiddisks - 1;
	363	if (insert_point == first_missing)
	364	insert_point--;
	365	sparedisks ++;
	366	info.array.active_disks--;
	367	missing_disks++;
	368	}
	369
	370	if (level <= 0 && first_missing != subdevs * 2) {
	371	fprintf(stderr,
	372	Name ": This level does not support missing devices\n");
	373	return 1;
	374	}
	375
	376	/* Ok, lets try some ioctls */
	377
	378	info.array.level = level;
	379	info.array.size = size;
	380	info.array.raid_disks = raiddisks;
	381	/* The kernel should know what md_minor we are dealing
	382	* with, but it chooses to trust me instead. Sigh
	383	*/
	384	info.array.md_minor = 0;
	385	if (fstat(mdfd, &stb)==0)
	386	info.array.md_minor = minor(stb.st_rdev);
	387	info.array.not_persistent = 0;
	388
	389	if ( ( (level == 4 \|\| level == 5) &&
	390	(insert_point < raiddisks \|\| first_missing < raiddisks) )
	391	\|\|
	392	( level == 6 && (insert_point < raiddisks
	393	\|\| second_missing < raiddisks))
	394	\|\|
	395	assume_clean
	396	)
	397	info.array.state = 1; /* clean, but one+ drive will be missing*/
	398	else
	399	info.array.state = 0; /* not clean, but no errors */
	400
	401	if (level == 10) {
	402	/* for raid10, the bitmap size is the capacity of the array,
	403	* which is array.size * raid_disks / ncopies;
	404	* .. but convert to sectors.
	405	*/
	406	int ncopies = ((layout>>8) & 255) * (layout & 255);
	407	bitmapsize = (unsigned long long)size * raiddisks / ncopies * 2;
	408	/* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/
	409	} else
	410	bitmapsize = (unsigned long long)size * 2;
	411
	412	/* There is lots of redundancy in these disk counts,
	413	* raid_disks is the most meaningful value
	414	* it describes the geometry of the array
	415	* it is constant
	416	* nr_disks is total number of used slots.
	417	* it should be raid_disks+spare_disks
	418	* spare_disks is the number of extra disks present
	419	* see above
	420	* active_disks is the number of working disks in
	421	* active slots. (With raid_disks)
	422	* working_disks is the total number of working disks,
	423	* including spares
	424	* failed_disks is the number of disks marked failed
	425	*
	426	* Ideally, the kernel would keep these (except raid_disks)
	427	* up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
	428	* So for now, we assume that all raid and spare
	429	* devices will be given.
	430	*/
	431	info.array.spare_disks=sparedisks;
	432	info.array.failed_disks=missing_disks;
	433	info.array.nr_disks = info.array.working_disks
	434	+ info.array.failed_disks;
	435	info.array.layout = layout;
	436	info.array.chunk_size = chunk*1024;
	437	info.array.major_version = st->ss->major;
	438
	439	if (name == NULL \|\| *name == 0) {
	440	/* base name on mddev */
	441	/* /dev/md0 -> 0
	442	* /dev/md_d0 -> d0
	443	* /dev/md/1 -> 1
	444	* /dev/md/d1 -> d1
	445	* /dev/md/home -> home
	446	* /dev/mdhome -> home
	447	*/
	448	name = strrchr(mddev, '/');
	449	if (name) {
	450	name++;
	451	if (strncmp(name, "md_d", 4)==0 &&
	452	strlen(name) > 4 &&
	453	isdigit(name[4]) &&
	454	(name-mddev) == 5 /* /dev/ */)
	455	name += 3;
	456	else if (strncmp(name, "md", 2)==0 &&
	457	strlen(name) > 2 &&
	458	isdigit(name[2]) &&
	459	(name-mddev) == 5 /* /dev/ */)
	460	name += 2;
	461	}
	462	}
	463	if (!st->ss->init_super(st, &info.array, size, name, homehost, uuid))
	464	return 1;
	465
	466	if (bitmap_file && vers < 9003) {
	467	major_num = BITMAP_MAJOR_HOSTENDIAN;
	468	#ifdef __BIG_ENDIAN
	469	fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n"
	470	" between different architectured. Consider upgrading the Linux kernel.\n");
	471	#endif
	472	}
	473
	474	if (bitmap_file && strcmp(bitmap_file, "internal")==0) {
	475	if ((vers%100) < 2) {
	476	fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n");
	477	return 1;
	478	}
	479	if (!st->ss->add_internal_bitmap(st, &bitmap_chunk,
	480	delay, write_behind,
	481	bitmapsize, 1, major_num)) {
	482	fprintf(stderr, Name ": Given bitmap chunk size not supported.\n");
	483	return 1;
	484	}
	485	bitmap_file = NULL;
	486	}
	487
	488
	489	sra = sysfs_read(mdfd, 0, 0);
	490
	491	if (st->ss->external) {
	492	char ver[100];
	493	strcat(strcpy(ver, "external:"), st->ss->text_version);
	494	if ((vers % 100) < 2 \|\|
	495	sra == NULL \|\|
	496	sysfs_set_str(sra, NULL, "metadata_version",
	497	ver) < 0) {
	498	fprintf(stderr, Name ": This kernel does not "
	499	"support external metadata.\n");
	500	return 1;
	501	}
	502	rv = 0;
	503	} else if ((vers % 100) >= 1) { /* can use different versions */
	504	mdu_array_info_t inf;
	505	memset(&inf, 0, sizeof(inf));
	506	inf.major_version = st->ss->major;
	507	inf.minor_version = st->minor_version;
	508	rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
	509	} else
	510	rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
	511	if (rv) {
	512	fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
	513	mddev, strerror(errno));
	514	return 1;
	515	}
	516
	517	if (bitmap_file) {
	518	int uuid[4];
	519
	520	st->ss->uuid_from_super(st, uuid);
	521	if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk,
	522	delay, write_behind,
	523	bitmapsize,
	524	major_num)) {
	525	return 1;
	526	}
	527	bitmap_fd = open(bitmap_file, O_RDWR);
	528	if (bitmap_fd < 0) {
	529	fprintf(stderr, Name ": weird: %s cannot be openned\n",
	530	bitmap_file);
	531	return 1;
	532	}
	533	if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) {
	534	fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n",
	535	mddev, strerror(errno));
	536	return 1;
	537	}
	538	}
	539
	540
	541	for (pass=1; pass <=2 ; pass++) {
	542	mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */
	543
	544	for (dnum=0, dv = devlist ; dv ;
	545	dv=(dv->next)?(dv->next):moved_disk, dnum++) {
	546	int fd;
	547	struct stat stb;
	548
	549	info.disk.number = dnum;
	550	if (dnum == insert_point) {
	551	moved_disk = dv;
	552	}
	553	info.disk.raid_disk = info.disk.number;
	554	if (info.disk.raid_disk < raiddisks)
	555	info.disk.state = (1<<MD_DISK_ACTIVE) \|
	556	(1<<MD_DISK_SYNC);
	557	else
	558	info.disk.state = 0;
	559	if (dv->writemostly)
	560	info.disk.state \|= (1<<MD_DISK_WRITEMOSTLY);
	561
	562	if (dnum == insert_point \|\|
	563	strcasecmp(dv->devname, "missing")==0)
	564	continue;
	565
	566	fd = open(dv->devname, O_RDWR\|O_EXCL, 0);
	567	if (fd < 0) {
	568	fprintf(stderr, Name ": failed to open %s "
	569	"after earlier success - aborting\n",
	570	dv->devname);
	571	return 1;
	572	}
	573	fstat(fd, &stb);
	574	info.disk.major = major(stb.st_rdev);
	575	info.disk.minor = minor(stb.st_rdev);
	576
	577	switch(pass){
	578	case 1:
	579	remove_partitions(fd);
	580	st->ss->add_to_super(st, &info.disk,
	581	fd, dv->devname);
	582	break;
	583	case 2:
	584	close(fd);
	585
	586	if (st->ss->external) {
	587	char dv[100];
	588	sprintf(dv, "%d:%d\n",
	589	info.disk.major,
	590	info.disk.minor);
	591	sysfs_set_str(sra, NULL, "new_dev", dv);
	592	/* FIXME check error */
	593	/FIXME find that device and set it up/
	594	} else if (ioctl(mdfd, ADD_NEW_DISK,
	595	&info.disk)) {
	596	fprintf(stderr,
	597	Name ": ADD_NEW_DISK for %s "
	598	"failed: %s\n",
	599	dv->devname, strerror(errno));
	600	st->ss->free_super(st);
	601	return 1;
	602	}
	603	break;
	604	}
	605	if (dv == moved_disk && dnum != insert_point) break;
	606	}
	607	if (pass == 1)
	608	st->ss->write_init_super(st);
	609	}
	610	st->ss->free_super(st);
	611
	612	/* param is not actually used */
	613	if (level == LEVEL_CONTAINER)
	614	/* No need to start */
	615	;
	616	else if (runstop == 1 \|\| subdevs >= raiddisks) {
	617	mdu_param_t param;
	618	if (ioctl(mdfd, RUN_ARRAY, &param)) {
	619	fprintf(stderr, Name ": RUN_ARRAY failed: %s\n",
	620	strerror(errno));
	621	Manage_runstop(mddev, mdfd, -1, 0);
	622	return 1;
	623	}
	624	if (verbose >= 0)
	625	fprintf(stderr, Name ": array %s started.\n", mddev);
	626	} else {
	627	fprintf(stderr, Name ": not starting array - not enough devices.\n");
	628	}
	629	return 0;
	630	}