X-Git-Url: http://git.ipfire.org/?p=thirdparty%2Fmdadm.git;a=blobdiff_plain;f=super1.c;h=f1732f7b6c751e56842b255702e4bf3ad5b2f729;hp=a0a3a682d85640618fea123ee472080b9d0acdeb;hb=beae1dfe2e5a3f11e6e52a93fbf617d644708415;hpb=f9c25f1d2af73f9494589afab42e07aa9ca2f740

diff --git a/super1.c b/super1.c
index a0a3a682..f1732f7b 100644
--- a/super1.c
+++ b/super1.c
@@ -1,7 +1,7 @@
 /*
  * mdadm - manage Linux "md" devices aka RAID arrays.
  *
- * Copyright (C) 2001-2004 Neil Brown <neilb@cse.unsw.edu.au>
+ * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
  *
  *
  *    This program is free software; you can redistribute it and/or modify
@@ -28,8 +28,6 @@
  */
 
 #include "mdadm.h"
-#include <endian.h>
-#include "asm/byteorder.h"
 /*
  * The version-1 superblock :
  * All numeric fields are little-endian.
@@ -58,7 +56,14 @@ struct mdp_superblock_1 {
 				 * NOTE: signed, so bitmap can be before superblock
 				 * only meaningful of feature_map[0] is set.
 				 */
-	__u8	pad1[128-100];	/* set to 0 when written */
+
+	/* These are only valid with feature bit '4' */
+	__u32	new_level;	/* new level we are reshaping to		*/
+	__u64	reshape_position;	/* next address in array-space for reshape */
+	__u32	delta_disks;	/* change in number of raid_disks		*/
+	__u32	new_layout;	/* new layout					*/
+	__u32	new_chunk;	/* new chunk size (bytes)			*/
+	__u8	pad1[128-124];	/* set to 0 when written */
 
 	/* constant this-device information - 64 bytes */
 	__u64	data_offset;	/* sector start of data, often 0 */
@@ -89,8 +94,21 @@ struct mdp_superblock_1 {
 	__u16	dev_roles[0];	/* role in array, or 0xffff for a spare, or 0xfffe for faulty */
 };
 
+struct misc_dev_info {
+	__u64 device_size;
+};
+
+/* feature_map bits */
+#define MD_FEATURE_BITMAP_OFFSET	1
+#define	MD_FEATURE_RECOVERY_OFFSET	2 /* recovery_offset is present and
+					   * must be honoured
+					   */
+#define	MD_FEATURE_RESHAPE_ACTIVE	4
+
+#define	MD_FEATURE_ALL			(1|2|4)
+
 #ifndef offsetof
-#define offsetof(t,f) ((int)&(((t*)0)->f))
+#define offsetof(t,f) ((size_t)&(((t*)0)->f))
 #endif
 static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb)
 {
@@ -110,19 +128,21 @@ static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb)
 	disk_csum = sb->sb_csum;
 	sb->sb_csum = 0;
 	newcsum = 0;
-	for (i=0; size>=4; size -= 4 )
-		newcsum += __le32_to_cpu(*isuper++);
+	for (i=0; size>=4; size -= 4 ) {
+		newcsum += __le32_to_cpu(*isuper);
+		isuper++;
+	}
 
 	if (size == 2)
 		newcsum += __le16_to_cpu(*(unsigned short*) isuper);
 
 	csum = (newcsum & 0xffffffff) + (newcsum >> 32);
 	sb->sb_csum = disk_csum;
-	return csum;
+	return __cpu_to_le32(csum);
 }
 
 #ifndef MDASSEMBLE
-static void examine_super1(void *sbv)
+static void examine_super1(void *sbv, char *homehost)
 {
 	struct mdp_superblock_1 *sb = sbv;
 	time_t atime;
@@ -130,28 +150,61 @@ static void examine_super1(void *sbv)
 	int faulty;
 	int i;
 	char *c;
+	int l = homehost ? strlen(homehost) : 0;
+	int layout;
 
 	printf("          Magic : %08x\n", __le32_to_cpu(sb->magic));
-	printf("        Version : %02d.%02d\n", 1, __le32_to_cpu(sb->feature_map));
+	printf("        Version : %02d\n", 1);
+	printf("    Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map));
 	printf("     Array UUID : ");
 	for (i=0; i<16; i++) {
 		if ((i&3)==0 && i != 0) printf(":");
 		printf("%02x", sb->set_uuid[i]);
 	}
 	printf("\n");
-	printf("           Name : %.32s\n", sb->set_name);
-
+	printf("           Name : %.32s", sb->set_name);
+	if (l > 0 && l < 32 &&
+	    sb->set_name[l] == ':' &&
+	    strncmp(sb->set_name, homehost, l) == 0)
+		printf("  (local to host %s)", homehost);
+	printf("\n");
 	atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL;
 	printf("  Creation Time : %.24s\n", ctime(&atime));
 	c=map_num(pers, __le32_to_cpu(sb->level));
 	printf("     Raid Level : %s\n", c?c:"-unknown-");
 	printf("   Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks));
 	printf("\n");
-	printf("    Device Size : %llu%s\n", (unsigned long long)sb->data_size, human_size(sb->data_size<<9));
+	printf("  Used Dev Size : %llu%s\n",
+	       (unsigned long long)sb->data_size,
+	       human_size(sb->data_size<<9));
+	if (__le32_to_cpu(sb->level) >= 0) {
+		int ddsks=0;
+		switch(__le32_to_cpu(sb->level)) {
+		case 1: ddsks=1;break;
+		case 4:
+		case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break;
+		case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break;
+		case 10:
+			layout = __le32_to_cpu(sb->layout);
+			ddsks = __le32_to_cpu(sb->raid_disks)
+				 / (layout&255) / ((layout>>8)&255);
+		}
+		if (ddsks)
+			printf("     Array Size : %llu%s\n",
+			       ddsks*(unsigned long long)__le64_to_cpu(sb->size),
+			       human_size(ddsks*__le64_to_cpu(sb->size)<<9));
+		if (sb->size != sb->data_size)
+			printf("      Used Size : %llu%s\n",
+			       (unsigned long long)__le64_to_cpu(sb->size),
+			       human_size(__le64_to_cpu(sb->size)<<9));
+	}
 	if (sb->data_offset)
-		printf("    Data Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->data_offset));
-	if (sb->super_offset)
-		printf("   Super Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->super_offset));
+		printf("    Data Offset : %llu sectors\n",
+		       (unsigned long long)__le64_to_cpu(sb->data_offset));
+	printf("   Super Offset : %llu sectors\n",
+	       (unsigned long long)__le64_to_cpu(sb->super_offset));
+	if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET)
+		printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset));
 	printf("          State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean");
 	printf("    Device UUID : ");
 	for (i=0; i<16; i++) {
@@ -159,6 +212,44 @@ static void examine_super1(void *sbv)
 		printf("%02x", sb->device_uuid[i]);
 	}
 	printf("\n");
+	printf("\n");
+	if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) {
+		printf("Internal Bitmap : %ld sectors from superblock\n",
+		       (long)__le32_to_cpu(sb->bitmap_offset));
+	}
+	if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
+		printf("  Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb->reshape_position)/2,
+		       human_size(__le64_to_cpu(sb->reshape_position)<<9));
+		if (__le32_to_cpu(sb->delta_disks)) {
+			printf("  Delta Devices : %d", __le32_to_cpu(sb->delta_disks));
+			if (__le32_to_cpu(sb->delta_disks))
+				printf(" (%d->%d)\n",
+				       __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks),
+				       __le32_to_cpu(sb->raid_disks));
+			else
+				printf(" (%d->%d)\n", __le32_to_cpu(sb->raid_disks),
+				       __le32_to_cpu(sb->raid_disks)+__le32_to_cpu(sb->delta_disks));
+		}
+		if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) {
+			c = map_num(pers, __le32_to_cpu(sb->new_level));
+			printf("      New Level : %s\n", c?c:"-unknown-");
+		}
+		if (__le32_to_cpu(sb->new_layout) != __le32_to_cpu(sb->layout)) {
+			if (__le32_to_cpu(sb->level) == 5) {
+				c = map_num(r5layout, __le32_to_cpu(sb->new_layout));
+				printf("     New Layout : %s\n", c?c:"-unknown-");
+			}
+			if (__le32_to_cpu(sb->level) == 10) {
+				printf("     New Layout : near=%d, %s=%d\n",
+				       __le32_to_cpu(sb->new_layout)&255,
+				       (__le32_to_cpu(sb->new_layout)&0x10000)?"offset":"far",
+				       (__le32_to_cpu(sb->new_layout)>>8)&255);
+			}
+		}
+		if (__le32_to_cpu(sb->new_chunk) != __le32_to_cpu(sb->chunksize))
+			printf("  New Chunksize : %dK\n", __le32_to_cpu(sb->new_chunk)/2);
+		printf("\n");
+	}
 	if (sb->devflags) {
 		printf("      Flags :");
 		if (sb->devflags & WriteMostly1)
@@ -182,21 +273,37 @@ static void examine_super1(void *sbv)
 	}
 	if (__le32_to_cpu(sb->level) == 10) {
 		int lo = __le32_to_cpu(sb->layout);
-		printf("         Layout : near=%d, far=%d\n",
-		       lo&255, (lo>>8)&255);
+		printf("         Layout : near=%d, %s=%d\n",
+		       lo&255,
+		       (lo&0x10000)?"offset":"far",
+		       (lo>>8)&255);
 	}
 	switch(__le32_to_cpu(sb->level)) {
 	case 0:
 	case 4:
 	case 5:
-		printf("     Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize/2));
+	case 6:
+	case 10:
+		printf("     Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize)/2);
 		break;
 	case -1:
-		printf("       Rounding : %dK\n", __le32_to_cpu(sb->chunksize/2));
+		printf("       Rounding : %dK\n", __le32_to_cpu(sb->chunksize)/2);
 		break;
 	default: break;
 	}
 	printf("\n");
+	printf("    Array Slot : %d (", __le32_to_cpu(sb->dev_number));
+	for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--)
+		if (__le16_to_cpu(sb->dev_roles[i-1]) != 0xffff)
+			break;
+	for (d=0; d < i; d++) {
+		int role = __le16_to_cpu(sb->dev_roles[d]);
+		if (d) printf(", ");
+		if (role == 0xffff) printf("empty");
+		else if(role == 0xfffe) printf("failed");
+		else printf("%d", role);
+	}
+	printf(")\n");
 	printf("   Array State : ");
 	for (d=0; d<__le32_to_cpu(sb->raid_disks); d++) {
 		int cnt = 0;
@@ -230,27 +337,41 @@ static void brief_examine_super1(void *sbv)
 {
 	struct mdp_superblock_1 *sb = sbv;
 	int i;
-
+	char *nm;
 	char *c=map_num(pers, __le32_to_cpu(sb->level));
 
-	printf("ARRAY /dev/?? level=%s metadata=1 num-devices=%d UUID=",
-	       c?c:"-unknown-", sb->raid_disks);
+	nm = strchr(sb->set_name, ':');
+	if (nm)
+		nm++;
+	else if (sb->set_name[0])
+		nm = sb->set_name;
+	else
+		nm = "??";
+
+	printf("ARRAY /dev/md/%s level=%s metadata=1 num-devices=%d UUID=",
+	       nm,
+	       c?c:"-unknown-", __le32_to_cpu(sb->raid_disks));
 	for (i=0; i<16; i++) {
-		printf("%02x", sb->set_uuid[i]);
 		if ((i&3)==0 && i != 0) printf(":");
+		printf("%02x", sb->set_uuid[i]);
 	}
 	if (sb->set_name[0])
 		printf(" name=%.32s", sb->set_name);
 	printf("\n");
 }
 
-static void detail_super1(void *sbv)
+static void detail_super1(void *sbv, char *homehost)
 {
 	struct mdp_superblock_1 *sb = sbv;
 	int i;
-
-	printf("           Name : %.32s\n", sb->set_name);
-	printf("           UUID : ");
+	int l = homehost ? strlen(homehost) : 0;
+
+	printf("           Name : %.32s", sb->set_name);
+	if (l > 0 && l < 32 &&
+	    sb->set_name[l] == ':' &&
+	    strncmp(sb->set_name, homehost, l) == 0)
+		printf("  (local to host %s)", homehost);
+	printf("\n           UUID : ");
 	for (i=0; i<16; i++) {
 		if ((i&3)==0 && i != 0) printf(":");
 		printf("%02x", sb->set_uuid[i]);
@@ -274,6 +395,16 @@ static void brief_detail_super1(void *sbv)
 
 #endif
 
+static int match_home1(void *sbv, char *homehost)
+{
+	struct mdp_superblock_1 *sb = sbv;
+	int l = homehost ? strlen(homehost) : 0;
+
+	return (l > 0 && l < 32 &&
+		sb->set_name[l] == ':' &&
+		strncmp(sb->set_name, homehost, l) == 0);
+}
+
 static void uuid_from_super1(int uuid[4], void * sbv)
 {
 	struct mdp_superblock_1 *super = sbv;
@@ -283,7 +414,7 @@ static void uuid_from_super1(int uuid[4], void * sbv)
 		cuuid[i] = super->set_uuid[i];
 }
 
-static void getinfo_super1(struct mdinfo *info, mddev_ident_t ident, void *sbv)
+static void getinfo_super1(struct mdinfo *info, void *sbv)
 {
 	struct mdp_superblock_1 *sb = sbv;
 	int working = 0;
@@ -298,6 +429,12 @@ static void getinfo_super1(struct mdinfo *info, mddev_ident_t ident, void *sbv)
 	info->array.layout = __le32_to_cpu(sb->layout);
 	info->array.md_minor = -1;
 	info->array.ctime = __le64_to_cpu(sb->ctime);
+	info->array.utime = __le64_to_cpu(sb->utime);
+	info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512;
+	info->array.state = (__le64_to_cpu(sb->resync_offset)+1) ? 0 : 1;
+
+	info->data_offset = __le64_to_cpu(sb->data_offset);
+	info->component_size = __le64_to_cpu(sb->size);
 
 	info->disk.major = 0;
 	info->disk.minor = 0;
@@ -324,8 +461,18 @@ static void getinfo_super1(struct mdinfo *info, mddev_ident_t ident, void *sbv)
 
 	memcpy(info->uuid, sb->set_uuid, 16);
 
-	strncpy(ident->name, sb->set_name, 32);
-	ident->name[32] = 0;
+	strncpy(info->name, sb->set_name, 32);
+	info->name[32] = 0;
+
+	if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
+		info->reshape_active = 1;
+		info->reshape_progress = __le64_to_cpu(sb->reshape_position);
+		info->new_level = __le32_to_cpu(sb->new_level);
+		info->delta_disks = __le32_to_cpu(sb->delta_disks);
+		info->new_layout = __le32_to_cpu(sb->new_layout);
+		info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9;
+	} else
+		info->reshape_active = 0;
 
 	for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
 		role = __le16_to_cpu(sb->dev_roles[i]);
@@ -336,16 +483,33 @@ static void getinfo_super1(struct mdinfo *info, mddev_ident_t ident, void *sbv)
 	info->array.working_disks = working;
 }
 
-static int update_super1(struct mdinfo *info, void *sbv, char *update, char *devname, int verbose)
+static int update_super1(struct mdinfo *info, void *sbv, char *update,
+			 char *devname, int verbose,
+			 int uuid_set, char *homehost)
 {
+	/* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
+	 * For others, the return value is ignored.
+	 */
 	int rv = 0;
 	struct mdp_superblock_1 *sb = sbv;
 
-	if (strcmp(update, "force")==0) {
+	if (strcmp(update, "force-one")==0) {
+		/* Not enough devices for a working array,
+		 * so bring this one up-to-date
+		 */
+		if (sb->events != __cpu_to_le64(info->events))
+			rv = 1;
 		sb->events = __cpu_to_le64(info->events);
+	}
+	if (strcmp(update, "force-array")==0) {
+		/* Degraded array and 'force' requests to
+		 * maybe need to mark it 'clean'.
+		 */
 		switch(__le32_to_cpu(sb->level)) {
 		case 5: case 4: case 6:
 			/* need to force clean */
+			if (sb->resync_offset != ~0ULL)
+				rv = 1;
 			sb->resync_offset = ~0ULL;
 		}
 	}
@@ -361,48 +525,93 @@ static int update_super1(struct mdinfo *info, void *sbv, char *update, char *dev
 			rv = 1;
 		}
 	}
-#if 0
-	if (strcmp(update, "newdev") == 0) {
-		int d = info->disk.number;
-		memset(&sb->disks[d], 0, sizeof(sb->disks[d]));
-		sb->disks[d].number = d;
-		sb->disks[d].major = info->disk.major;
-		sb->disks[d].minor = info->disk.minor;
-		sb->disks[d].raid_disk = info->disk.raid_disk;
-		sb->disks[d].state = info->disk.state;
-		sb->this_disk = sb->disks[d];
-	}
-#endif
 	if (strcmp(update, "grow") == 0) {
 		sb->raid_disks = __cpu_to_le32(info->array.raid_disks);
-		/* FIXME */
+		/* As we are just adding a spare, there is no need to
+		 * make any change to the dev_roles array
+		 */
 	}
 	if (strcmp(update, "resync") == 0) {
 		/* make sure resync happens */
-		sb->resync_offset = ~0ULL;
+		sb->resync_offset = 0ULL;
 	}
+	if (strcmp(update, "uuid") == 0) {
+		if (super1.swapuuid) {
+			unsigned char *ac = (unsigned char *)sb->set_uuid;
+			unsigned char *bc = (unsigned char *)info->uuid;
+			int i;
+			for (i=0; i<16; i+= 4) {
+				ac[i+0] = bc[i+3];
+				ac[i+1] = bc[i+2];
+				ac[i+2] = bc[i+1];
+				ac[i+3] = bc[i+0];
+			}
+		} else
+			memcpy(sb->set_uuid, info->uuid, 16);
+
+		if (__le32_to_cpu(sb->feature_map)&MD_FEATURE_BITMAP_OFFSET) {
+			struct bitmap_super_s *bm;
+			bm = (struct bitmap_super_s*)(sbv+1024);
+			memcpy(bm->uuid, sb->set_uuid, 16);
+		}
+	}
+	if (strcmp(update, "homehost") == 0 &&
+	    homehost) {
+		char *c;
+		update = "name";
+		c = strchr(sb->set_name, ':');
+		if (c)
+			strncpy(info->name, c+1, 31 - (c-sb->set_name));
+		else
+			strncpy(info->name, sb->set_name, 32);
+		info->name[32] = 0;
+	}
+	if (strcmp(update, "name") == 0) {
+		if (info->name[0] == 0)
+			sprintf(info->name, "%d", info->array.md_minor);
+		memset(sb->set_name, 0, sizeof(sb->set_name));
+		if (homehost &&
+		    strchr(info->name, ':') == NULL &&
+		    strlen(homehost)+1+strlen(info->name) < 32) {
+			strcpy(sb->set_name, homehost);
+			strcat(sb->set_name, ":");
+			strcat(sb->set_name, info->name);
+		} else
+			strcpy(sb->set_name, info->name);
+	}
+	if (strcmp(update, "devicesize") == 0 &&
+	    __le64_to_cpu(sb->super_offset) <
+	    __le64_to_cpu(sb->data_offset)) {
+		/* set data_size to device size less data_offset */
+		struct misc_dev_info *misc = (struct misc_dev_info*)
+			(sbv + 1024 + sizeof(struct bitmap_super_s));
+		printf("Size was %llu\n", __le64_to_cpu(sb->data_size));
+		sb->data_size = __cpu_to_le64(
+			misc->device_size - __le64_to_cpu(sb->data_offset));
+		printf("Size is %llu\n", __le64_to_cpu(sb->data_size));
+	}
+	if (strcmp(update, "_reshape_progress")==0)
+		sb->reshape_position = __cpu_to_le64(info->reshape_progress);
 
 	sb->sb_csum = calc_sb_1_csum(sb);
 	return rv;
 }
 
-
-static __u64 event_super1(void *sbv)
-{
-	struct mdp_superblock_1 *sb = sbv;
-	return __le64_to_cpu(sb->events);
-}
-
-static int init_super1(struct supertype *st, void **sbp, mdu_array_info_t *info, char *name)
+static int init_super1(struct supertype *st, void **sbp, mdu_array_info_t *info,
+		       unsigned long long size, char *name, char *homehost)
 {
-	struct mdp_superblock_1 *sb = malloc(1024 + sizeof(bitmap_super_t));
+	struct mdp_superblock_1 *sb = malloc(1024 + sizeof(bitmap_super_t) +
+					     sizeof(struct misc_dev_info));
 	int spares;
 	int rfd;
+	char defname[10];
 	memset(sb, 0, 1024);
 
-	if (info->major_version == -1)
+	if (info->major_version == -1) {
 		/* zeroing superblock */
+		*sbp = sb;
 		return 0;
+	}
 
 	spares = info->working_disks - info->active_disks;
 	if (info->raid_disks + spares  > 384) {
@@ -411,7 +620,6 @@ static int init_super1(struct supertype *st, void **sbp, mdu_array_info_t *info,
 		return 0;
 	}
 
-
 	sb->magic = __cpu_to_le32(MD_SB_MAGIC);
 	sb->major_version = __cpu_to_le32(1);
 	sb->feature_map = 0;
@@ -426,13 +634,24 @@ static int init_super1(struct supertype *st, void **sbp, mdu_array_info_t *info,
 	}
 	if (rfd >= 0) close(rfd);
 
+	if (name == NULL || *name == 0) {
+		sprintf(defname, "%d", info->md_minor);
+		name = defname;
+	}
 	memset(sb->set_name, 0, 32);
-	strcpy(sb->set_name, name);
+	if (homehost &&
+	    strchr(name, ':')== NULL &&
+	    strlen(homehost)+1+strlen(name) < 32) {
+		strcpy(sb->set_name, homehost);
+		strcat(sb->set_name, ":");
+		strcat(sb->set_name, name);
+	} else
+		strcpy(sb->set_name, name);
 
 	sb->ctime = __cpu_to_le64((unsigned long long)time(0));
 	sb->level = __cpu_to_le32(info->level);
 	sb->layout = __cpu_to_le32(info->layout);
-	sb->size = __cpu_to_le64(info->size*2ULL);
+	sb->size = __cpu_to_le64(size*2ULL);
 	sb->chunksize = __cpu_to_le32(info->chunk_size>>9);
 	sb->raid_disks = __cpu_to_le32(info->raid_disks);
 
@@ -470,18 +689,21 @@ static void add_to_super1(void *sbv, mdu_disk_info_t *dk)
 		*rp = 0xfffe;
 }
 
+static void locate_bitmap1(struct supertype *st, int fd, void *sbv);
+
 static int store_super1(struct supertype *st, int fd, void *sbv)
 {
 	struct mdp_superblock_1 *sb = sbv;
 	unsigned long long sb_offset;
 	int sbsize;
-	long size;
+	unsigned long long dsize;
 
-	if (ioctl(fd, BLKGETSIZE, &size))
+	if (!get_dev_size(fd, NULL, &dsize))
 		return 1;
 
+	dsize >>= 9;
 
-	if (size < 24)
+	if (dsize < 24)
 		return 2;
 
 	/*
@@ -494,7 +716,7 @@ static int store_super1(struct supertype *st, int fd, void *sbv)
 	 */
 	switch(st->minor_version) {
 	case 0:
-		sb_offset = size;
+		sb_offset = dsize;
 		sb_offset -= 8*2;
 		sb_offset &= ~(4*2-1);
 		break;
@@ -525,6 +747,15 @@ static int store_super1(struct supertype *st, int fd, void *sbv)
 	if (write(fd, sb, sbsize) != sbsize)
 		return 4;
 
+	if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) {
+		struct bitmap_super_s *bm = (struct bitmap_super_s*)
+			(((char*)sb)+1024);
+		if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) {
+			locate_bitmap1(st, fd, sbv);
+			if (write(fd, bm, sizeof(*bm)) != sizeof(*bm))
+			    return 5;
+		}
+	}
 	fsync(fd);
 	return 0;
 }
@@ -539,8 +770,10 @@ static int write_init_super1(struct supertype *st, void *sbv,
 	int fd = open(devname, O_RDWR | O_EXCL);
 	int rfd;
 	int rv;
+	int bm_space;
 
-	long size;
+	unsigned long space;
+	unsigned long long dsize, array_size;
 	long long sb_offset;
 
 
@@ -552,7 +785,7 @@ static int write_init_super1(struct supertype *st, void *sbv,
 
 	sb->dev_number = __cpu_to_le32(dinfo->number);
 	if (dinfo->state & (1<<MD_DISK_WRITEMOSTLY))
-		sb->devflags |= WriteMostly1;
+		sb->devflags |= __cpu_to_le32(WriteMostly1);
 
 	if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
 	    read(rfd, sb->device_uuid, 16) != 16) {
@@ -571,17 +804,20 @@ static int write_init_super1(struct supertype *st, void *sbv,
 		if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) {
 			/* same array, so preserve events and dev_number */
 			sb->events = refsb->events;
-			sb->dev_number = refsb->dev_number;
+			/* bugs in 2.6.17 and earlier mean the dev_number
+			 * chosen in Manage must be preserved
+			 */
+			if (get_linux_version() >= 2006018)
+				sb->dev_number = refsb->dev_number;
 		}
 		free(refsb);
 	}
 
-	if (ioctl(fd, BLKGETSIZE, &size)) {
-		close(fd);
+	if (!get_dev_size(fd, NULL, &dsize))
 		return 1;
-	}
+	dsize >>= 9;
 
-	if (size < 24) {
+	if (dsize < 24) {
 		close(fd);
 		return 2;
 	}
@@ -594,26 +830,41 @@ static int write_init_super1(struct supertype *st, void *sbv,
 	 * 0: At least 8K, but less than 12K, from end of device
 	 * 1: At start of device
 	 * 2: 4K from start of device.
+	 * Depending on the array size, we might leave extra space
+	 * for a bitmap.
 	 */
+	array_size = __le64_to_cpu(sb->size);
+	/* work out how much space we left of a bitmap */
+	if (array_size >= 200*1024*1024*2)
+		bm_space = 128*2;
+	else if (array_size > 8*1024*1024*2)
+		bm_space = 64*2;
+	else
+		bm_space = 0;
+
 	switch(st->minor_version) {
 	case 0:
-		sb_offset = size;
+		sb_offset = dsize;
 		sb_offset -= 8*2;
 		sb_offset &= ~(4*2-1);
 		sb->super_offset = __cpu_to_le64(sb_offset);
 		sb->data_offset = __cpu_to_le64(0);
-		sb->data_size = __cpu_to_le64(sb_offset);
+		sb->data_size = __cpu_to_le64(sb_offset - bm_space);
 		break;
 	case 1:
 		sb->super_offset = __cpu_to_le64(0);
-		sb->data_offset = __cpu_to_le64(4*2); /* leave 4k for super and bitmap */
-		sb->data_size = __cpu_to_le64(size - 4*2);
+		sb->data_offset = __cpu_to_le64(bm_space + 4*2);
+		sb->data_size = __cpu_to_le64(dsize - bm_space - 4*2);
 		break;
 	case 2:
 		sb_offset = 4*2;
-		sb->super_offset = __cpu_to_le64(sb_offset);
-		sb->data_offset = __cpu_to_le64(sb_offset+4*2);
-		sb->data_size = __cpu_to_le64(size - 4*2 - 4*2);
+		if (dsize - 4*2 - 64*2 >= array_size && array_size > 8*1024*1024*2)
+			space = 64*2;
+		else
+			space = 4*2;
+		sb->super_offset = __cpu_to_le64(4*2);
+		sb->data_offset = __cpu_to_le64(4*2 + 4*2 + bm_space);
+		sb->data_size = __cpu_to_le64(dsize - 4*2 - 4*2 - bm_space );
 		break;
 	default:
 		return -EINVAL;
@@ -649,8 +900,10 @@ static int compare_super1(void **firstp, void *secondv)
 		return 1;
 
 	if (!first) {
-		first = malloc(1024);
-		memcpy(first, second, 1024);
+		first = malloc(1024+sizeof(bitmap_super_t) +
+			       sizeof(struct misc_dev_info));
+		memcpy(first, second, 1024+sizeof(bitmap_super_t) +
+		       sizeof(struct misc_dev_info));
 		*firstp = first;
 		return 0;
 	}
@@ -669,10 +922,12 @@ static int compare_super1(void **firstp, void *secondv)
 
 static int load_super1(struct supertype *st, int fd, void **sbp, char *devname)
 {
-	unsigned long size;
+	unsigned long long dsize;
 	unsigned long long sb_offset;
 	struct mdp_superblock_1 *super;
-
+	int uuid[4];
+	struct bitmap_super_s *bsb;
+	struct misc_dev_info *misc;
 
 
 	if (st->ss == NULL) {
@@ -707,31 +962,28 @@ static int load_super1(struct supertype *st, int fd, void **sbp, char *devname)
 		st->ss = NULL;
 		return 2;
 	}
-	if (ioctl(fd, BLKGETSIZE, &size)) {
-		if (devname)
-			fprintf(stderr, Name ": cannot find device size for %s: %s\n",
-				devname, strerror(errno));
+	if (!get_dev_size(fd, devname, &dsize))
 		return 1;
-	}
+	dsize >>= 9;
 
-	if (size < 24) {
+	if (dsize < 24) {
 		if (devname)
-			fprintf(stderr, Name ": %s is too small for md: size is %lu sectors.\n",
-				devname, size);
+			fprintf(stderr, Name ": %s is too small for md: size is %llu sectors.\n",
+				devname, dsize);
 		return 1;
 	}
 
 	/*
 	 * Calculate the position of the superblock.
 	 * It is always aligned to a 4K boundary and
-	 * depeding on minor_version, it can be:
+	 * depending on minor_version, it can be:
 	 * 0: At least 8K, but less than 12K, from end of device
 	 * 1: At start of device
 	 * 2: 4K from start of device.
 	 */
 	switch(st->minor_version) {
 	case 0:
-		sb_offset = size;
+		sb_offset = dsize;
 		sb_offset -= 8*2;
 		sb_offset &= ~(4*2-1);
 		break;
@@ -755,7 +1007,8 @@ static int load_super1(struct supertype *st, int fd, void **sbp, char *devname)
 		return 1;
 	}
 
-	super = malloc(1024 + sizeof(bitmap_super_t));
+	super = malloc(1024 + sizeof(bitmap_super_t) +
+		       sizeof(struct misc_dev_info));
 
 	if (read(fd, super, 1024) != 1024) {
 		if (devname)
@@ -788,6 +1041,32 @@ static int load_super1(struct supertype *st, int fd, void **sbp, char *devname)
 		return 2;
 	}
 	*sbp = super;
+
+	bsb = (struct bitmap_super_s *)(((char*)super)+1024);
+
+	misc = (struct misc_dev_info*) (bsb+1);
+	misc->device_size = dsize;
+
+	/* Now check on the bitmap superblock */
+	if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0)
+		return 0;
+	/* Read the bitmap superblock and make sure it looks
+	 * valid.  If it doesn't clear the bit.  An --assemble --force
+	 * should get that written out.
+	 */
+	locate_bitmap1(st, fd, super);
+	if (read(fd, ((char*)super)+1024, sizeof(struct bitmap_super_s))
+	    != sizeof(struct bitmap_super_s))
+		goto no_bitmap;
+
+	uuid_from_super1(uuid, super);
+	if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
+	    memcmp(bsb->uuid, uuid, 16) != 0)
+		goto no_bitmap;
+	return 0;
+
+ no_bitmap:
+	super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) & ~1);
 	return 0;
 }
 
@@ -800,7 +1079,8 @@ static struct supertype *match_metadata_desc1(char *arg)
 	st->ss = &super1;
 	st->max_devs = 384;
 	if (strcmp(arg, "1") == 0 ||
-	    strcmp(arg, "1.0") == 0) {
+	    strcmp(arg, "1.0") == 0 ||
+	    strcmp(arg, "default/large") == 0) {
 		st->minor_version = 0;
 		return st;
 	}
@@ -826,6 +1106,14 @@ static __u64 avail_size1(struct supertype *st, __u64 devsize)
 	if (devsize < 24)
 		return 0;
 
+	/* if the device is bigger than 8Gig, save 64k for bitmap usage,
+	 * if bigger than 200Gig, save 128k
+	 */
+	if (devsize-64*2 >= 200*1024*1024*2)
+		devsize -= 128*2;
+	else if (devsize >= 8*1024*1024*2)
+		devsize -= 64*2;
+
 	switch(st->minor_version) {
 	case 0:
 		/* at end */
@@ -842,29 +1130,91 @@ static __u64 avail_size1(struct supertype *st, __u64 devsize)
 
 static int
 add_internal_bitmap1(struct supertype *st, void *sbv,
-		     int chunk, int delay, int write_behind, unsigned long long size, int may_change, int major)
+		     int *chunkp, int delay, int write_behind,
+		     unsigned long long size,
+		     int may_change, int major)
 {
 	/*
 	 * If not may_change, then this is a 'Grow', and the bitmap
 	 * must fit after the superblock.
 	 * If may_change, then this is create, and we can put the bitmap
 	 * before the superblock if we like, or may move the start.
-	 * For now, just squeeze the bitmap into 3k and don't change anything.
+	 * If !may_change, the bitmap MUST live at offset of 1K, until
+	 * we get a sysfs interface.
 	 *
 	 * size is in sectors,  chunk is in bytes !!!
 	 */
 
 	unsigned long long bits;
-	unsigned long long max_bits = (3*512 - sizeof(bitmap_super_t)) * 8;
+	unsigned long long max_bits;
 	unsigned long long min_chunk;
+	long offset;
+	int chunk = *chunkp;
+	int room = 0;
 	struct mdp_superblock_1 *sb = sbv;
 	bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + 1024);
 
-	if (st->minor_version && !may_change &&
-	    __le64_to_cpu(sb->data_offset) - __le64_to_cpu(sb->super_offset) < 8)
-		return 0; /* doesn't fit */
+	switch(st->minor_version) {
+	case 0:
+		/* either 3K after the superblock, or some amount of space
+		 * before.
+		 */
+		if (may_change) {
+			/* We are creating array, so we *know* how much room has
+			 * been left.
+			 */
+			offset = 0;
+			if (__le64_to_cpu(sb->size) >= 200*1024*1024*2)
+				room = 128*2;
+			else if (__le64_to_cpu(sb->size) > 8*1024*1024*2)
+				room = 64*2;
+			else {
+				room = 3*2;
+				offset = 2;
+			}
+		} else {
+			room = __le64_to_cpu(sb->super_offset)
+				- __le64_to_cpu(sb->data_offset)
+				- __le64_to_cpu(sb->data_size);
+			/* remove '1 ||' when we can set offset via sysfs */
+			if (1 || (room < 3*2 &&
+				  __le32_to_cpu(sb->max_dev) <= 384)) {
+				room = 3*2;
+				offset = 1*2;
+			} else {
+				offset = 0; /* means movable offset */
+			}
+		}
+		break;
+	case 1:
+	case 2: /* between superblock and data */
+		if (may_change) {
+			offset = 4*2;
+			if (__le64_to_cpu(sb->size) >= 200*1024*1024*2)
+				room = 128*2;
+			else if (__le64_to_cpu(sb->size) > 8*1024*1024*2)
+				room = 64*2;
+			else
+				room = 3*2;
+		} else {
+			room = __le64_to_cpu(sb->data_offset)
+				- __le64_to_cpu(sb->super_offset);
+			if (1 || __le32_to_cpu(sb->max_dev) <= 384) {
+				room -= 2;
+				offset = 2;
+			} else {
+				room -= 4*2;
+				offset = 4*2;
+			}
+		}
+		break;
+	}
 
+	if (chunk == UnSet && room > 128*2)
+		/* Limit to 128K of bitmap when chunk size not requested */
+		room = 128*2;
 
+	max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8;
 
 	min_chunk = 4096; /* sub-page chunks don't work yet.. */
 	bits = (size*512)/min_chunk +1;
@@ -876,11 +1226,19 @@ add_internal_bitmap1(struct supertype *st, void *sbv,
 		chunk = min_chunk;
 	else if (chunk < min_chunk)
 		return 0; /* chunk size too small */
+	if (chunk == 0) /* rounding problem */
+		return 0;
+
+	if (offset == 0) {
+		bits = (size*512) / chunk + 1;
+		room = ((bits+7)/8 + sizeof(bitmap_super_t) +511)/512;
+		offset = -room;
+	}
 
-	sb->bitmap_offset = __cpu_to_le32(2);
+	sb->bitmap_offset = __cpu_to_le32(offset);
 
 	sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | 1);
-	memset(bms, sizeof(*bms), 0);
+	memset(bms, 0, sizeof(*bms));
 	bms->magic = __cpu_to_le32(BITMAP_MAGIC);
 	bms->version = __cpu_to_le32(major);
 	uuid_from_super1((int*)bms->uuid, sb);
@@ -889,29 +1247,32 @@ add_internal_bitmap1(struct supertype *st, void *sbv,
 	bms->sync_size = __cpu_to_le64(size);
 	bms->write_behind = __cpu_to_le32(write_behind);
 
+	*chunkp = chunk;
 	return 1;
 }
 
 
-void locate_bitmap1(struct supertype *st, int fd, void *sbv)
+static void locate_bitmap1(struct supertype *st, int fd, void *sbv)
 {
 	unsigned long long offset;
 	struct mdp_superblock_1 *sb;
+	int mustfree = 0;
 
-	if (!sbv)
-		if (st->ss->load_super(st, fd, sbv, NULL))
+	if (!sbv) {
+		if (st->ss->load_super(st, fd, &sbv, NULL))
 			return; /* no error I hope... */
-
+		mustfree = 1;
+	}
 	sb = sbv;
 
 	offset = __le64_to_cpu(sb->super_offset);
 	offset += (long) __le32_to_cpu(sb->bitmap_offset);
-	if (!sbv)
+	if (mustfree)
 		free(sb);
 	lseek64(fd, offset<<9, 0);
 }
 
-int write_bitmap1(struct supertype *st, int fd, void *sbv)
+static int write_bitmap1(struct supertype *st, int fd, void *sbv)
 {
 	struct mdp_superblock_1 *sb = sbv;
 	bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+1024);
@@ -922,7 +1283,9 @@ int write_bitmap1(struct supertype *st, int fd, void *sbv)
 
 	locate_bitmap1(st, fd, sbv);
 
-	write(fd, ((char*)sb)+1024, sizeof(bitmap_super_t));
+	if (write(fd, ((char*)sb)+1024, sizeof(bitmap_super_t)) !=
+	    sizeof(bitmap_super_t))
+		return -2;
 	towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9);
 	towrite = (towrite+7) >> 3; /* bits to bytes */
 	memset(buf, 0xff, sizeof(buf));
@@ -950,10 +1313,10 @@ struct superswitch super1 = {
 	.detail_super = detail_super1,
 	.brief_detail_super = brief_detail_super1,
 #endif
+	.match_home = match_home1,
 	.uuid_from_super = uuid_from_super1,
 	.getinfo_super = getinfo_super1,
 	.update_super = update_super1,
-	.event_super = event_super1,
 	.init_super = init_super1,
 	.add_to_super = add_to_super1,
 	.store_super = store_super1,