[thirdparty/mdadm.git] / Create.c

/*
 * mdadm - manage Linux "md" devices aka RAID arrays.
 *
 * Copyright (C) 2001-2002 Neil Brown <neilb@cse.unsw.edu.au>
 *
 *
 *    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"

int Create(struct supertype *st, char *mddev, int mdfd,
	   int chunk, int level, int layout, unsigned long size, int raiddisks, int sparedisks,
	   char *name,
	   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 missing_disks = 0;
	int insert_point = subdevs * 2; /* where to insert a missing drive */
	void *super;
	int pass;
	int vers;
	int rv;
	int bitmap_fd;
	unsigned long long bitmapsize;

	mdu_array_info_t array;
	int major = BITMAP_MAJOR_HI;

	vers = md_get_version(mdfd);
	if (vers < 9000) {
		fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
		return 1;
	}
	if (level == UnSet) {
		fprintf(stderr,
			Name ": a RAID level is needed to create an array.\n");
		return 1;
	}
	if (raiddisks < 1) {
		fprintf(stderr,
			Name ": a number of --raid-devices must be given 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 (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;
	}

	/* 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 -1: /* linear */
		if (chunk == 0) {
			chunk = 64;
			if (verbose > 0)
				fprintf(stderr, Name ": chunk size defaults to 64K\n");
		}
		break;
	default: /* raid1, multipath */
		if (chunk) {
			chunk = 0;
			if (verbose > 0)
				fprintf(stderr, Name ": chunk size ignored for this level\n");
		}
		break;
	}

	/* now look at the subdevs */
	array.active_disks = 0;
	array.working_disks = 0;
	dnum = 0;
	for (dv=devlist; dv; dv=dv->next, dnum++) {
		char *dname = dv->devname;
		unsigned long dsize;
		unsigned long long ldsize, freesize;
		int fd;
		if (strcasecmp(dname, "missing")==0) {
			if (first_missing > dnum)
				first_missing = dnum;
			missing_disks ++;
			continue;
		}
		array.working_disks++;
		if (dnum < raiddisks)
			array.active_disks++;
		fd = open(dname, O_RDONLY|O_EXCL, 0);
		if (fd <0 ) {
			fprintf(stderr, Name ": Cannot open %s: %s\n",
				dname, strerror(errno));
			fail=1;
			continue;
		}
#ifdef BLKGETSIZE64
		if (ioctl(fd, BLKGETSIZE64, &ldsize)==0)
			;
		else
#endif
		if (ioctl(fd, BLKGETSIZE, &dsize)) {
			fprintf(stderr, Name ": Cannot get size of %s: %s\n",
				dname, strerror(errno));
			fail = 1;
			close(fd);
			continue;
		}
		else {
			ldsize = dsize;
			ldsize <<= 9;
		}
		freesize = st->ss->avail_size(st, ldsize >> 9);
		if (freesize == 0) {
			fprintf(stderr, Name ": %s is too small: %luK\n",
				dname, (unsigned long)(ldsize>>10));
			fail = 1;
			close(fd);
			continue;
		}

		freesize /= 2; /* convert to K */

		if (size && freesize < size) {
			fprintf(stderr, Name ": %s is smaller that given size."
				" %lluK < %luK + superblock\n", dname, freesize, size);
			fail = 1;
			close(fd);
			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) {
			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) {
				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 %luK\n", size);
		}
	}
	if (level > 0 && ((maxsize-size)*100 > maxsize)) {
		if (runstop != 1 || verbose >= 0)
			fprintf(stderr, Name ": largest drive (%s) exceed size (%luK) 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  raid5, 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 5:
			insert_point = raiddisks-1;
			sparedisks++;
			array.active_disks--;
			missing_disks++;
			break;
		default:
			break;
		}
	}
	
	/* Ok, lets try some ioctls */

	array.level = level;
	array.size = size;
	array.raid_disks = raiddisks;
	/* The kernel should *know* what md_minor we are dealing
	 * with, but it chooses to trust me instead. Sigh
	 */
	array.md_minor = 0;
	if (fstat(mdfd, &stb)==0)
		array.md_minor = minor(stb.st_rdev);
	array.not_persistent = 0;
	/*** FIX: Need to do something about RAID-6 here ***/
	if ( ( (level == 5) &&
	       (insert_point < raiddisks || first_missing < raiddisks) )
	     ||
	     ( level == 6 && missing_disks == 2)
	     ||
	     assume_clean
		)
		array.state = 1; /* clean, but one+ drive will be missing */
	else
		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) * (layout & 255);
		bitmapsize = (unsigned long long)array.size * raiddisks / ncopies * 2;
		printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, array.size, raiddisks, ncopies);
	} else
		bitmapsize = (unsigned long long)array.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.
	 */
	array.spare_disks=sparedisks;
	array.failed_disks=missing_disks;
	array.nr_disks = array.working_disks + array.failed_disks;
	array.layout = layout;
	array.chunk_size = chunk*1024;


	if (!st->ss->init_super(st, &super, &array, name))
		return 1;

	if (bitmap_file && vers < 9003) {
		major = 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, super, bitmap_chunk, delay, write_behind,
						 bitmapsize, 1, major)) {
			fprintf(stderr, Name ": Given bitmap chunk size not supported.\n");
			return 1;
		}
		bitmap_file = NULL;
	}


	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];

		if (bitmap_chunk == UnSet)
			bitmap_chunk = DEFAULT_BITMAP_CHUNK;

		st->ss->uuid_from_super(uuid, super);
		if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk,
				 delay, write_behind,
				 bitmapsize,
				 major)) {
			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;
			mdu_disk_info_t disk;

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

			if (dnum == insert_point ||
			    strcasecmp(dv->devname, "missing")==0) {
				disk.major = 0;
				disk.minor = 0;
				disk.state = (1<<MD_DISK_FAULTY);
			} else {
				fd = open(dv->devname, O_RDONLY|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);
				disk.major = major(stb.st_rdev);
				disk.minor = minor(stb.st_rdev);
				close(fd);
			}
			switch(pass){
			case 1:
				st->ss->add_to_super(super, &disk);
				break;
			case 2:
				if (disk.state == 1) break;
				st->ss->write_init_super(st, super, &disk, dv->devname);

				if (ioctl(mdfd, ADD_NEW_DISK, &disk)) {
					fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n",
						dv->devname, strerror(errno));
					free(super);
					return 1;
				}

				break;
			}
			if (dv == moved_disk && dnum != insert_point) break;
		}
	}
	free(super);

	/* param is not actually used */
	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
64c4757e	1	/*
9a9dab36	2	* mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e	3	*
cd29a5c8	4	* Copyright (C) 2001-2002 Neil Brown <neilb@cse.unsw.edu.au>
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
82d9eba6	34	int Create(struct supertype st, char mddev, int mdfd,
aa88f531	35	int chunk, int level, int layout, unsigned long size, int raiddisks, int sparedisks,
947fd4dd	36	char *name,
cd29a5c8	37	int subdevs, mddev_dev_t devlist,
47d79ef8	38	int runstop, int verbose, int force, int assume_clean,
dfd4d8ee	39	char *bitmap_file, int bitmap_chunk, int write_behind, int delay)
64c4757e	40	{
682c7051 NB	41	/*
	42	* Create a new raid array.
	43	*
	44	* First check that necessary details are available
	45	* (i.e. level, raid-disks)
	46	*
	47	* Then check each disk to see what might be on it
	48	* and report anything interesting.
	49	*
	50	* If anything looks odd, and runstop not set,
	51	* abort.
	52	*
	53	* SET_ARRAY_INFO and ADD_NEW_DISK, and
82d9eba6	54	* if runstop==run, or raiddisks disks were used,
682c7051 NB	55	* RUN_ARRAY
682c7051 NB	56	*/
b5e64645	57	unsigned long long minsize=0, maxsize=0;
cd29a5c8 NB	58	char *mindisc = NULL;
cd29a5c8 NB	59	char *maxdisc = NULL;
c913b90e	60	int dnum;
cd29a5c8	61	mddev_dev_t dv;
682c7051	62	int fail=0, warn=0;
82b27616	63	struct stat stb;
82d9eba6	64	int first_missing = subdevs * 2;
52826846	65	int missing_disks = 0;
82d9eba6	66	int insert_point = subdevs * 2; /* where to insert a missing drive */
4b1ac34b NB	67	void *super;
4b1ac34b NB	68	int pass;
82d9eba6 NB	69	int vers;
82d9eba6 NB	70	int rv;
c82f047c	71	int bitmap_fd;
f9c25f1d	72	unsigned long long bitmapsize;
682c7051 NB	73
682c7051 NB	74	mdu_array_info_t array;
dcec9ee5 NB	75	int major = BITMAP_MAJOR_HI;
dcec9ee5 NB	76
82d9eba6 NB	77	vers = md_get_version(mdfd);
82d9eba6 NB	78	if (vers < 9000) {
dd0781e5	79	fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
52826846	80	return 1;
682c7051	81	}
98c6faba	82	if (level == UnSet) {
682c7051 NB	83	fprintf(stderr,
	84	Name ": a RAID level is needed to create an array.\n");
	85	return 1;
	86	}
	87	if (raiddisks < 1) {
	88	fprintf(stderr,
b83d95f3	89	Name ": a number of --raid-devices must be given to create an array\n");
682c7051 NB	90	return 1;
682c7051 NB	91	}
98c6faba NB	92	if (raiddisks < 4 && level == 6) {
	93	fprintf(stderr,
	94	Name ": at least 4 raid-devices needed for level 6\n");
	95	return 1;
	96	}
	97	if (raiddisks > 256 && level == 6) {
	98	fprintf(stderr,
	99	Name ": no more than 256 raid-devices supported for level 6\n");
	100	return 1;
	101	}
52826846 NB	102	if (raiddisks < 2 && level >= 4) {
52826846 NB	103	fprintf(stderr,
98c6faba	104	Name ": at least 2 raid-devices needed for level 4 or 5\n");
52826846 NB	105	return 1;
52826846 NB	106	}
682c7051	107	if (subdevs > raiddisks+sparedisks) {
b83d95f3	108	fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks);
52826846	109	return 1;
682c7051	110	}
cd29a5c8	111	if (subdevs < raiddisks+sparedisks) {
52826846 NB	112	fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n");
	113	return 1;
	114	}
	115
682c7051	116	/* now set some defaults */
98c6faba	117	if (layout == UnSet)
682c7051 NB	118	switch(level) {
	119	default: /* no layout */
	120	layout = 0;
	121	break;
e5329c37 NB	122	case 10:
e5329c37 NB	123	layout = 0x102; /* near=2, far=1 */
dab6685f	124	if (verbose > 0)
e5329c37 NB	125	fprintf(stderr,
	126	Name ": layout defaults to n1\n");
	127	break;
682c7051	128	case 5:
98c6faba	129	case 6:
682c7051	130	layout = map_name(r5layout, "default");
dab6685f	131	if (verbose > 0)
682c7051 NB	132	fprintf(stderr,
	133	Name ": layout defaults to %s\n", map_num(r5layout, layout));
	134	break;
b5e64645 NB	135	case LEVEL_FAULTY:
	136	layout = map_name(faultylayout, "default");
	137
dab6685f	138	if (verbose > 0)
b5e64645 NB	139	fprintf(stderr,
	140	Name ": layout defaults to %s\n", map_num(faultylayout, layout));
	141	break;
682c7051 NB	142	}
682c7051 NB	143
e5329c37 NB	144	if (level == 10)
	145	/* check layout fits in array*/
	146	if ((layout&255) * ((layout>>8)&255) > raiddisks) {
	147	fprintf(stderr, Name ": that layout requires at least %d devices\n",
	148	(layout&255) * ((layout>>8)&255));
	149	return 1;
	150	}
	151
aa88f531 NB	152	switch(level) {
	153	case 4:
	154	case 5:
e5329c37	155	case 10:
98c6faba	156	case 6:
aa88f531 NB	157	case 0:
	158	case -1: /* linear */
	159	if (chunk == 0) {
	160	chunk = 64;
dab6685f	161	if (verbose > 0)
aa88f531 NB	162	fprintf(stderr, Name ": chunk size defaults to 64K\n");
	163	}
	164	break;
	165	default: /* raid1, multipath */
	166	if (chunk) {
	167	chunk = 0;
dab6685f	168	if (verbose > 0)
aa88f531 NB	169	fprintf(stderr, Name ": chunk size ignored for this level\n");
	170	}
	171	break;
682c7051 NB	172	}
	173
	174	/* now look at the subdevs */
52826846 NB	175	array.active_disks = 0;
52826846 NB	176	array.working_disks = 0;
c913b90e NB	177	dnum = 0;
c913b90e NB	178	for (dv=devlist; dv; dv=dv->next, dnum++) {
cd29a5c8	179	char *dname = dv->devname;
b5e64645 NB	180	unsigned long dsize;
b5e64645 NB	181	unsigned long long ldsize, freesize;
52826846	182	int fd;
cd29a5c8	183	if (strcasecmp(dname, "missing")==0) {
c913b90e NB	184	if (first_missing > dnum)
c913b90e NB	185	first_missing = dnum;
52826846 NB	186	missing_disks ++;
	187	continue;
	188	}
	189	array.working_disks++;
c913b90e	190	if (dnum < raiddisks)
52826846	191	array.active_disks++;
d7eaf49f	192	fd = open(dname, O_RDONLY\|O_EXCL, 0);
682c7051 NB	193	if (fd <0 ) {
	194	fprintf(stderr, Name ": Cannot open %s: %s\n",
	195	dname, strerror(errno));
	196	fail=1;
	197	continue;
	198	}
b5e64645 NB	199	#ifdef BLKGETSIZE64
	200	if (ioctl(fd, BLKGETSIZE64, &ldsize)==0)
	201	;
	202	else
	203	#endif
682c7051 NB	204	if (ioctl(fd, BLKGETSIZE, &dsize)) {
	205	fprintf(stderr, Name ": Cannot get size of %s: %s\n",
	206	dname, strerror(errno));
	207	fail = 1;
	208	close(fd);
	209	continue;
	210	}
b5e64645 NB	211	else {
b5e64645 NB	212	ldsize = dsize;
570c0542	213	ldsize <<= 9;
b5e64645	214	}
1bf4e2d9	215	freesize = st->ss->avail_size(st, ldsize >> 9);
82d9eba6	216	if (freesize == 0) {
aa88f531	217	fprintf(stderr, Name ": %s is too small: %luK\n",
b5e64645	218	dname, (unsigned long)(ldsize>>10));
52826846 NB	219	fail = 1;
	220	close(fd);
	221	continue;
682c7051	222	}
82d9eba6 NB	223
82d9eba6 NB	224	freesize /= 2; /* convert to K */
682c7051 NB	225
682c7051 NB	226	if (size && freesize < size) {
52826846	227	fprintf(stderr, Name ": %s is smaller that given size."
b5e64645	228	" %lluK < %luK + superblock\n", dname, freesize, size);
52826846 NB	229	fail = 1;
	230	close(fd);
	231	continue;
682c7051	232	}
cd29a5c8 NB	233	if (maxdisc == NULL \|\| (maxdisc && freesize > maxsize)) {
cd29a5c8 NB	234	maxdisc = dname;
52826846	235	maxsize = freesize;
682c7051	236	}
cd29a5c8 NB	237	if (mindisc ==NULL \|\| (mindisc && freesize < minsize)) {
cd29a5c8 NB	238	mindisc = dname;
52826846	239	minsize = freesize;
682c7051	240	}
dab6685f NB	241	if (runstop != 1 \|\| verbose >= 0) {
	242	warn \|= check_ext2(fd, dname);
	243	warn \|= check_reiser(fd, dname);
	244	warn \|= check_raid(fd, dname);
	245	}
682c7051 NB	246	close(fd);
	247	}
	248	if (fail) {
52826846 NB	249	fprintf(stderr, Name ": create aborted\n");
52826846 NB	250	return 1;
682c7051 NB	251	}
682c7051 NB	252	if (size == 0) {
cd29a5c8	253	if (mindisc == NULL) {
52826846 NB	254	fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
	255	return 1;
	256	}
c13c45e9	257	if (level > 0 \|\| level == LEVEL_MULTIPATH \|\| level == LEVEL_FAULTY) {
b5e64645 NB	258	/* size is meaningful */
	259	if (minsize > 0x100000000ULL) {
	260	fprintf(stderr, Name ": devices too large for RAID level %d\n", level);
	261	return 1;
	262	}
	263	size = minsize;
dab6685f	264	if (verbose > 0)
b5e64645 NB	265	fprintf(stderr, Name ": size set to %luK\n", size);
b5e64645 NB	266	}
682c7051	267	}
b5e64645	268	if (level > 0 && ((maxsize-size)*100 > maxsize)) {
dab6685f NB	269	if (runstop != 1 \|\| verbose >= 0)
	270	fprintf(stderr, Name ": largest drive (%s) exceed size (%luK) by more than 1%%\n",
	271	maxdisc, size);
52826846	272	warn = 1;
682c7051 NB	273	}
	274
	275	if (warn) {
52826846 NB	276	if (runstop!= 1) {
	277	if (!ask("Continue creating array? ")) {
	278	fprintf(stderr, Name ": create aborted.\n");
	279	return 1;
	280	}
	281	} else {
dab6685f	282	if (verbose > 0)
52826846	283	fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
682c7051	284	}
682c7051 NB	285	}
682c7051 NB	286
52826846 NB	287	/* If this is raid5, we want to configure the last active slot
52826846 NB	288	* as missing, so that a reconstruct happens (faster than re-parity)
98c6faba	289	* FIX: Can we do this for raid6 as well?
52826846	290	*/
47d79ef8	291	if (assume_clean==0 && force == 0 && first_missing >= raiddisks) {
98c6faba NB	292	switch ( level ) {
	293	case 5:
	294	insert_point = raiddisks-1;
	295	sparedisks++;
	296	array.active_disks--;
	297	missing_disks++;
	298	break;
	299	default:
	300	break;
	301	}
52826846 NB	302	}
52826846 NB	303
682c7051 NB	304	/* Ok, lets try some ioctls */
	305
	306	array.level = level;
	307	array.size = size;
682c7051	308	array.raid_disks = raiddisks;
82b27616 NB	309	/* The kernel should know what md_minor we are dealing
	310	* with, but it chooses to trust me instead. Sigh
	311	*/
682c7051	312	array.md_minor = 0;
82b27616	313	if (fstat(mdfd, &stb)==0)
0df46c2a	314	array.md_minor = minor(stb.st_rdev);
682c7051	315	array.not_persistent = 0;
98c6faba	316	/* FIX: Need to do something about RAID-6 here */
b5e64645 NB	317	if ( ( (level == 5) &&
	318	(insert_point < raiddisks \|\| first_missing < raiddisks) )
	319	\|\|
	320	( level == 6 && missing_disks == 2)
47d79ef8 NB	321	\|\|
47d79ef8 NB	322	assume_clean
b5e64645	323	)
98c6faba	324	array.state = 1; /* clean, but one+ drive will be missing */
82b27616	325	else
52826846	326	array.state = 0; /* not clean, but no errors */
82b27616	327
f9c25f1d NB	328	if (level == 10) {
	329	/* for raid10, the bitmap size is the capacity of the array,
	330	* which is array.size * raid_disks / ncopies;
	331	* .. but convert to sectors.
	332	*/
	333	int ncopies = (layout>>8) * (layout & 255);
	334	bitmapsize = (unsigned long long)array.size * raiddisks / ncopies * 2;
	335	printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, array.size, raiddisks, ncopies);
	336	} else
	337	bitmapsize = (unsigned long long)array.size * 2;
	338
82b27616 NB	339	/* There is lots of redundancy in these disk counts,
	340	* raid_disks is the most meaningful value
	341	* it describes the geometry of the array
	342	* it is constant
	343	* nr_disks is total number of used slots.
	344	* it should be raid_disks+spare_disks
	345	* spare_disks is the number of extra disks present
	346	* see above
	347	* active_disks is the number of working disks in
	348	* active slots. (With raid_disks)
	349	* working_disks is the total number of working disks,
	350	* including spares
	351	* failed_disks is the number of disks marked failed
	352	*
	353	* Ideally, the kernel would keep these (except raid_disks)
	354	* up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
	355	* So for now, we assume that all raid and spare
	356	* devices will be given.
	357	*/
52826846 NB	358	array.spare_disks=sparedisks;
	359	array.failed_disks=missing_disks;
	360	array.nr_disks = array.working_disks + array.failed_disks;
682c7051 NB	361	array.layout = layout;
682c7051 NB	362	array.chunk_size = chunk*1024;
55935d51	363
82d9eba6	364
947fd4dd	365	if (!st->ss->init_super(st, &super, &array, name))
82d9eba6	366	return 1;
682c7051	367
dcec9ee5 NB	368	if (bitmap_file && vers < 9003) {
	369	major = BITMAP_MAJOR_HOSTENDIAN;
	370	#ifdef __BIG_ENDIAN
	371	fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n"
	372	" between different architectured. Consider upgrading the Linux kernel.\n");
	373	#endif
	374	}
	375
55935d51 NB	376	if (bitmap_file && strcmp(bitmap_file, "internal")==0) {
	377	if ((vers%100) < 2) {
	378	fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n");
	379	return 1;
	380	}
34163fc7	381	if (!st->ss->add_internal_bitmap(st, super, bitmap_chunk, delay, write_behind,
f9c25f1d	382	bitmapsize, 1, major)) {
55935d51 NB	383	fprintf(stderr, Name ": Given bitmap chunk size not supported.\n");
	384	return 1;
	385	}
	386	bitmap_file = NULL;
	387	}
	388
	389
	390
82d9eba6 NB	391	if ((vers % 100) >= 1) { /* can use different versions */
	392	mdu_array_info_t inf;
	393	memset(&inf, 0, sizeof(inf));
	394	inf.major_version = st->ss->major;
	395	inf.minor_version = st->minor_version;
	396	rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
	397	} else
	398	rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
	399	if (rv) {
52826846 NB	400	fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
	401	mddev, strerror(errno));
	402	return 1;
682c7051	403	}
52826846	404
c82f047c NB	405	if (bitmap_file) {
c82f047c NB	406	int uuid[4];
55935d51 NB	407
	408	if (bitmap_chunk == UnSet)
	409	bitmap_chunk = DEFAULT_BITMAP_CHUNK;
	410
c82f047c	411	st->ss->uuid_from_super(uuid, super);
dfd4d8ee	412	if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk,
dcec9ee5	413	delay, write_behind,
f9c25f1d	414	bitmapsize,
dcec9ee5	415	major)) {
c82f047c NB	416	return 1;
	417	}
	418	bitmap_fd = open(bitmap_file, O_RDWR);
	419	if (bitmap_fd < 0) {
	420	fprintf(stderr, Name ": weird: %s cannot be openned\n",
dcec9ee5	421	bitmap_file);
c82f047c NB	422	return 1;
	423	}
	424	if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) {
	425	fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n",
	426	mddev, strerror(errno));
	427	return 1;
	428	}
	429	}
	430
4b1ac34b NB	431
	432
	433	for (pass=1; pass <=2 ; pass++) {
	434	mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */
	435
	436	for (dnum=0, dv = devlist ; dv ;
	437	dv=(dv->next)?(dv->next):moved_disk, dnum++) {
	438	int fd;
	439	struct stat stb;
	440	mdu_disk_info_t disk;
	441
	442	disk.number = dnum;
	443	if (dnum == insert_point) {
	444	moved_disk = dv;
52826846	445	}
4b1ac34b NB	446	disk.raid_disk = disk.number;
4b1ac34b NB	447	if (disk.raid_disk < raiddisks)
dfd4d8ee NB	448	disk.state = (1<<MD_DISK_ACTIVE) \|
dfd4d8ee NB	449	(1<<MD_DISK_SYNC);
4b1ac34b NB	450	else
4b1ac34b NB	451	disk.state = 0;
dfd4d8ee NB	452	if (dv->writemostly)
	453	disk.state \|= (1<<MD_DISK_WRITEMOSTLY);
	454
4b1ac34b NB	455	if (dnum == insert_point \|\|
	456	strcasecmp(dv->devname, "missing")==0) {
	457	disk.major = 0;
	458	disk.minor = 0;
dfd4d8ee	459	disk.state = (1<<MD_DISK_FAULTY);
4b1ac34b NB	460	} else {
	461	fd = open(dv->devname, O_RDONLY\|O_EXCL, 0);
	462	if (fd < 0) {
	463	fprintf(stderr, Name ": failed to open %s after earlier success - aborting\n",
	464	dv->devname);
	465	return 1;
	466	}
	467	fstat(fd, &stb);
	468	disk.major = major(stb.st_rdev);
	469	disk.minor = minor(stb.st_rdev);
	470	close(fd);
	471	}
	472	switch(pass){
	473	case 1:
82d9eba6	474	st->ss->add_to_super(super, &disk);
4b1ac34b NB	475	break;
4b1ac34b NB	476	case 2:
eaac7dde	477	if (disk.state == 1) break;
1bf4e2d9	478	st->ss->write_init_super(st, super, &disk, dv->devname);
4b1ac34b NB	479
	480	if (ioctl(mdfd, ADD_NEW_DISK, &disk)) {
	481	fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n",
	482	dv->devname, strerror(errno));
	483	free(super);
	484	return 1;
	485	}
	486
	487	break;
	488	}
	489	if (dv == moved_disk && dnum != insert_point) break;
52826846	490	}
5e52ae9e	491	}
4b1ac34b	492	free(super);
5e52ae9e	493
682c7051 NB	494	/* param is not actually used */
682c7051 NB	495	if (runstop == 1 \|\| subdevs >= raiddisks) {
82b27616	496	mdu_param_t param;
682c7051 NB	497	if (ioctl(mdfd, RUN_ARRAY, &param)) {
	498	fprintf(stderr, Name ": RUN_ARRAY failed: %s\n",
	499	strerror(errno));
91f068bf	500	Manage_runstop(mddev, mdfd, -1, 0);
682c7051 NB	501	return 1;
682c7051 NB	502	}
dab6685f NB	503	if (verbose >= 0)
dab6685f NB	504	fprintf(stderr, Name ": array %s started.\n", mddev);
682c7051	505	} else {
b83d95f3	506	fprintf(stderr, Name ": not starting array - not enough devices.\n");
682c7051 NB	507	}
682c7051 NB	508	return 0;
64c4757e	509	}