X-Git-Url: http://git.ipfire.org/?p=thirdparty%2Fmdadm.git;a=blobdiff_plain;f=Create.c;h=66a88d2008bd03a950971048437c753207df4bb6;hp=ac776c289f303ce49b76b1dbcf76c9539cdc1dd2;hb=d7eaf49f65ecb99e24255026331b0af7a9245cc2;hpb=682c705194a869b882cd710d5f996142912db390

diff --git a/Create.c b/Create.c
index ac776c28..66a88d20 100644
--- a/Create.c
+++ b/Create.c
@@ -1,7 +1,7 @@
 /*
- * mdctl - manage Linux "md" devices aka RAID arrays.
+ * mdadm - manage Linux "md" devices aka RAID arrays.
  *
- * Copyright (C) 2001 Neil Brown <neilb@cse.unsw.edu.au>
+ * Copyright (C) 2001-2002 Neil Brown <neilb@cse.unsw.edu.au>
  *
  *
  *    This program is free software; you can redistribute it and/or modify
@@ -27,14 +27,14 @@
  *           Australia
  */
 
-#include "mdctl.h"
+#include "mdadm.h"
 #include	"md_u.h"
 #include	"md_p.h"
 
 int Create(char *mddev, int mdfd,
-	   int chunk, int level, int layout, int size, int raiddisks, int sparedisks,
-	   int subdevs, char *subdev[],
-	   int runstop, int verbose)
+	   int chunk, int level, int layout, unsigned long size, int raiddisks, int sparedisks,
+	   int subdevs, mddev_dev_t devlist,
+	   int runstop, int verbose, int force)
 {
 	/*
 	 * Create a new raid array.
@@ -52,70 +52,153 @@ int Create(char *mddev, int mdfd,
 	 * if runstop==run, or raiddisks diskswere used,
 	 * RUN_ARRAY
 	 */
-	int minsize, maxsize;
-	int maxdisc= -1, mindisc = -1;
-	int i;
+	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 = MD_SB_DISKS*2;
+	int missing_disks = 0;
+	int insert_point = MD_SB_DISKS*2; /* where to insert a missing drive */
+	mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */
 
 	mdu_array_info_t array;
-	mdu_param_t param;
 	
 
 	if (md_get_version(mdfd) < 9000) {
-	    fprintf(stderr, Name ": Create requires md driver verison 0.90.0 or later\n");
-	    return 1;
+		fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
+		return 1;
 	}
-	if (level == -10) {
+	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-disks must be given to create an array\n");
+			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 (raiddisks+sparedisks > MD_SB_DISKS) {
 		fprintf(stderr,
-			Name ": too many discs requested: %d+%d > %d\n",
+			Name ": too many devices requested: %d+%d > %d\n",
 			raiddisks, sparedisks, MD_SB_DISKS);
 		return 1;
 	}
 	if (subdevs > raiddisks+sparedisks) {
-	    fprintf(stderr, Name ": You have listed more disks (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks);
-	    return 1;
+		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 == -1)
+	if (layout == UnSet)
 		switch(level) {
 		default: /* no layout */
 			layout = 0;
 			break;
+		case 10:
+			layout = 0x102; /* near=2, far=1 */
+			if (verbose)
+				fprintf(stderr,
+					Name ": layout defaults to n1\n");
+			break;
 		case 5:
+		case 6:
 			layout = map_name(r5layout, "default");
 			if (verbose)
 				fprintf(stderr,
 					Name ": layout defaults to %s\n", map_num(r5layout, layout));
 			break;
+		case LEVEL_FAULTY:
+			layout = map_name(faultylayout, "default");
+
+			if (verbose)
+				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;
 		}
 
-	if (chunk == 0) {
-		chunk = 64;
-		if (verbose)
-			fprintf(stderr, Name ": chunk size defaults to 64K\n");
+	switch(level) {
+	case 4:
+	case 5:
+	case 10:
+	case 6:
+	case 0:
+	case -1: /* linear */
+		if (chunk == 0) {
+			chunk = 64;
+			if (verbose)
+				fprintf(stderr, Name ": chunk size defaults to 64K\n");
+		}
+		break;
+	default: /* raid1, multipath */
+		if (chunk) {
+			chunk = 0;
+			if (verbose)
+				fprintf(stderr, Name ": chunk size ignored for this level\n");
+		}
+		break;
 	}
 
 	/* now look at the subdevs */
-	for (i=0; i<subdevs; i++) {
-		char *dname = subdev[i];
-		int dsize, freesize;
-		int fd = open(dname, O_RDONLY, 0);
+	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));
@@ -123,30 +206,34 @@ int Create(char *mddev, int mdfd,
 			close(fd);
 			continue;
 		}
-		if (dsize < MD_RESERVED_SECTORS*2) {
-		    fprintf(stderr, Name ": %s is too small: %dK\n",
-			    dname, dsize/2);
-		    fail = 1;
-		    close(fd);
-		    continue;
+		else {
+			ldsize = dsize;
+			dsize <<= 9;
+		}
+		if (ldsize < MD_RESERVED_SECTORS*2LL*512LL) {
+			fprintf(stderr, Name ": %s is too small: %luK\n",
+				dname, (unsigned long)(ldsize>>10));
+			fail = 1;
+			close(fd);
+			continue;
 		}
-		freesize = MD_NEW_SIZE_SECTORS(dsize);
+		freesize = MD_NEW_SIZE_SECTORS((ldsize>>9));
 		freesize /= 2;
 
 		if (size && freesize < size) {
-		    fprintf(stderr, Name ": %s is smaller that given size."
-			    " %dK < %dK + superblock\n", dname, freesize, size);
-		    fail = 1;
-		    close(fd);
-		    continue;
+			fprintf(stderr, Name ": %s is smaller that given size."
+				" %lluK < %luK + superblock\n", dname, freesize, size);
+			fail = 1;
+			close(fd);
+			continue;
 		}
-		if (maxdisc< 0 || (maxdisc>=0 && freesize > maxsize)) {
-		    maxdisc = i;
-		    maxsize = freesize;
+		if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
+			maxdisc = dname;
+			maxsize = freesize;
 		}
-		if (mindisc < 0 || (mindisc >=0 && freesize < minsize)) {
-		    mindisc = i;
-		    minsize = freesize;
+		if (mindisc ==NULL || (mindisc && freesize < minsize)) {
+			mindisc = dname;
+			minsize = freesize;
 		}
 		warn |= check_ext2(fd, dname);
 		warn |= check_reiser(fd, dname);
@@ -154,91 +241,164 @@ int Create(char *mddev, int mdfd,
 		close(fd);
 	}
 	if (fail) {
-	    fprintf(stderr, Name ": create aborted\n");
-	    return 1;
+		fprintf(stderr, Name ": create aborted\n");
+		return 1;
 	}
 	if (size == 0) {
-	    if (mindisc == -1) {
-		fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
-		return 1;
-	    }
-	    size = minsize;
-	    if (verbose)
-		fprintf(stderr, Name ": size set to %dK\n", size);
+		if (mindisc == NULL) {
+			fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
+			return 1;
+		}
+		if (level > 0) {
+			/* size is meaningful */
+			if (minsize > 0x100000000ULL) {
+				fprintf(stderr, Name ": devices too large for RAID level %d\n", level);	
+				return 1;
+			}
+			size = minsize;
+			if (verbose)
+				fprintf(stderr, Name ": size set to %luK\n", size);
+		}
 	}
-	if ((maxsize-size)*100 > maxsize) {
-	    fprintf(stderr, Name ": largest drive (%s) exceed size (%dK) by more than 1%\n",
-		    subdev[maxdisc], size);
-	    warn = 1;
+	if (level > 0 && ((maxsize-size)*100 > maxsize)) {
+		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;
+		if (runstop!= 1) {
+			if (!ask("Continue creating array? ")) {
+				fprintf(stderr, Name ": create aborted.\n");
+				return 1;
+			}
+		} else {
+			if (verbose)
+				fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
 		}
-	    } else {
-		if (verbose)
-		    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 (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.nr_disks = raiddisks+sparedisks;
 	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;
-	array.state = 0; /* not clean, but no errors */
-	array.active_disks=0;
-	array.working_disks=0;
-	array.spare_disks=0;
-	array.failed_disks=0;
+	/*** FIX: Need to do something about RAID-6 here ***/
+	if ( ( (level == 5) &&
+	       (insert_point < raiddisks || first_missing < raiddisks) )
+	     ||
+	     ( level == 6 && missing_disks == 2)
+		)
+		array.state = 1; /* clean, but one+ drive will be missing */
+	else
+		array.state = 0; /* not clean, but no errors */
+
+	/* 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 (ioctl(mdfd, SET_ARRAY_INFO, &array)) {
-	    fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
-		    mddev, strerror(errno));
-	    return 1;
+		fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
+			mddev, strerror(errno));
+		return 1;
 	}
 	
-	for (i=0; i<subdevs; i++) {
-	    int fd = open(subdev[i], O_RDONLY, 0);
-	    struct stat stb;
-	    mdu_disk_info_t disk;
-	    if (fd < 0) {
-		fprintf(stderr, Name ": failed to open %s after earlier success - aborting\n",
-			subdev[i]);
-		return 1;
-	    }
-	    fstat(fd, &stb);
-	    disk.number = i;
-	    disk.raid_disk = i;
-	    disk.state = 6; /* active and in sync */
-	    disk.major = MAJOR(stb.st_rdev);
-	    disk.minor = MINOR(stb.st_rdev);
-	    close(fd);
-	    if (ioctl(mdfd, ADD_NEW_DISK, &disk)) {
-		fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\b",
-			subdev[i], strerror(errno));
-		return 1;
-	    }
+	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 = 6; /* active and in sync */
+		else
+			disk.state = 0;
+		if (dnum == insert_point ||
+		    strcasecmp(dv->devname, "missing")==0) {
+			disk.major = 0;
+			disk.minor = 0;
+			disk.state = 1; /* 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);
+		}
+		if (ioctl(mdfd, ADD_NEW_DISK, &disk)) {
+			fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n",
+				dv->devname, strerror(errno));
+			return 1;
+		}
+		if (dv == moved_disk && dnum != insert_point) break;
 	}
 
 	/* 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);
 			return 1;
 		}
 		fprintf(stderr, Name ": array %s started.\n", mddev);
 	} else {
-		fprintf(stderr, Name ": not starting array - not enough discs.\n");
+		fprintf(stderr, Name ": not starting array - not enough devices.\n");
 	}
 	return 0;
 }