X-Git-Url: http://git.ipfire.org/?p=thirdparty%2Fmdadm.git;a=blobdiff_plain;f=super-intel.c;h=f5c6b09b93a5b464fa93319316130e3eaa2baf21;hp=618baf68800aa6f6dab39f47d4a120906b507185;hb=f35f25259279573c6274e2783536c0b0a399bdd4;hpb=5802a8118e447833749cbf0fe7b909f3c7d8349d

diff --git a/super-intel.c b/super-intel.c
index 618baf68..f5c6b09b 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -68,8 +68,10 @@ struct imsm_map {
 	__u8  num_members;	/* number of member disks */
 	__u8  reserved[3];
 	__u32 filler[7];	/* expansion area */
+#define IMSM_ORD_REBUILD (1 << 24)
 	__u32 disk_ord_tbl[1];	/* disk_ord_tbl[num_members],
-				   top byte special */
+				 * top byte contains some flags
+				 */
 } __attribute__ ((packed));
 
 struct imsm_vol {
@@ -100,16 +102,43 @@ struct imsm_super {
 	__u32 mpb_size;			/* 0x24 - 0x27 Size of MPB */
 	__u32 family_num;		/* 0x28 - 0x2B Checksum from first time this config was written */
 	__u32 generation_num;		/* 0x2C - 0x2F Incremented each time this array's MPB is written */
-	__u32 reserved[2];		/* 0x30 - 0x37 */
+	__u32 error_log_size;		/* 0x30 - 0x33 in bytes */
+	__u32 attributes;		/* 0x34 - 0x37 */
 	__u8 num_disks;			/* 0x38 Number of configured disks */
 	__u8 num_raid_devs;		/* 0x39 Number of configured volumes */
-	__u8 fill[2];			/* 0x3A - 0x3B */
-#define IMSM_FILLERS 39
-	__u32 filler[IMSM_FILLERS];	/* 0x3C - 0xD7 RAID_MPB_FILLERS */
+	__u8 error_log_pos;		/* 0x3A  */
+	__u8 fill[1];			/* 0x3B */
+	__u32 cache_size;		/* 0x3c - 0x40 in mb */
+	__u32 orig_family_num;		/* 0x40 - 0x43 original family num */
+	__u32 pwr_cycle_count;		/* 0x44 - 0x47 simulated power cycle count for array */
+	__u32 bbm_log_size;		/* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
+#define IMSM_FILLERS 35
+	__u32 filler[IMSM_FILLERS];	/* 0x4C - 0xD7 RAID_MPB_FILLERS */
 	struct imsm_disk disk[1];	/* 0xD8 diskTbl[numDisks] */
 	/* here comes imsm_dev[num_raid_devs] */
+	/* here comes BBM logs */
 } __attribute__ ((packed));
 
+#define BBM_LOG_MAX_ENTRIES 254
+
+struct bbm_log_entry {
+	__u64 defective_block_start;
+#define UNREADABLE 0xFFFFFFFF
+	__u32 spare_block_offset;
+	__u16 remapped_marked_count;
+	__u16 disk_ordinal;
+} __attribute__ ((__packed__));
+
+struct bbm_log {
+	__u32 signature; /* 0xABADB10C */
+	__u32 entry_count;
+	__u32 reserved_spare_block_count; /* 0 */
+	__u32 reserved; /* 0xFFFF */
+	__u64 first_spare_lba;
+	struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES];
+} __attribute__ ((__packed__));
+
+
 #ifndef MDASSEMBLE
 static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
 #endif
@@ -127,20 +156,28 @@ static unsigned int mpb_sectors(struct imsm_super *mpb)
 /* internal representation of IMSM metadata */
 struct intel_super {
 	union {
-		struct imsm_super *mpb;
-		void *buf;
+		void *buf; /* O_DIRECT buffer for reading/writing metadata */
+		struct imsm_super *anchor; /* immovable parameters */
 	};
+	size_t len; /* size of the 'buf' allocation */
+	void *next_buf; /* for realloc'ing buf from the manager */
+	size_t next_len;
 	int updates_pending; /* count of pending updates for mdmon */
 	int creating_imsm; /* flag to indicate container creation */
 	int current_vol; /* index of raid device undergoing creation */
+	#define IMSM_MAX_RAID_DEVS 2
+	struct imsm_dev *dev_tbl[IMSM_MAX_RAID_DEVS];
 	struct dl {
 		struct dl *next;
 		int index;
 		__u8 serial[MAX_RAID_SERIAL_LEN];
 		int major, minor;
 		char *devname;
+		struct imsm_disk disk;
 		int fd;
 	} *disks;
+	struct dl *add; /* list of disks to add while mdmon active */
+	struct bbm_log *bbm_log;
 };
 
 struct extent {
@@ -151,11 +188,12 @@ struct extent {
 enum imsm_update_type {
 	update_activate_spare,
 	update_create_array,
+	update_add_disk,
 };
 
 struct imsm_update_activate_spare {
 	enum imsm_update_type type;
-	int disk_idx;
+	struct dl *dl;
 	int slot;
 	int array;
 	struct imsm_update_activate_spare *next;
@@ -163,10 +201,25 @@ struct imsm_update_activate_spare {
 
 struct imsm_update_create_array {
 	enum imsm_update_type type;
-	struct imsm_dev dev;
 	int dev_idx;
+	struct imsm_dev dev;
+};
+
+struct imsm_update_add_disk {
+	enum imsm_update_type type;
 };
 
+static int imsm_env_devname_as_serial(void)
+{
+	char *val = getenv("IMSM_DEVNAME_AS_SERIAL");
+
+	if (val && atoi(val) == 1)
+		return 1;
+
+	return 0;
+}
+
+
 static struct supertype *match_metadata_desc_imsm(char *arg)
 {
 	struct supertype *st;
@@ -185,19 +238,41 @@ static struct supertype *match_metadata_desc_imsm(char *arg)
 	return st;
 }
 
+#ifndef MDASSEMBLE
 static __u8 *get_imsm_version(struct imsm_super *mpb)
 {
 	return &mpb->sig[MPB_SIG_LEN];
 }
+#endif 
 
-static struct imsm_disk *get_imsm_disk(struct imsm_super *mpb, __u8 index)
+/* retrieve a disk directly from the anchor when the anchor is known to be
+ * up-to-date, currently only at load time
+ */
+static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
 {
-	if (index > mpb->num_disks - 1)
+	if (index >= mpb->num_disks)
 		return NULL;
 	return &mpb->disk[index];
 }
 
-static __u32 gen_imsm_checksum(struct imsm_super *mpb)
+#ifndef MDASSEMBLE
+/* retrieve a disk from the parsed metadata */
+static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
+{
+	struct dl *d;
+
+	for (d = super->disks; d; d = d->next)
+		if (d->index == index)
+			return &d->disk;
+	
+	return NULL;
+}
+#endif
+
+/* generate a checksum directly from the anchor when the anchor is known to be
+ * up-to-date, currently only at load or write_super after coalescing
+ */
+static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
 {
 	__u32 end = mpb->mpb_size / sizeof(end);
 	__u32 *p = (__u32 *) mpb;
@@ -209,29 +284,50 @@ static __u32 gen_imsm_checksum(struct imsm_super *mpb)
         return sum - __le32_to_cpu(mpb->check_sum);
 }
 
-static size_t sizeof_imsm_dev(struct imsm_dev *dev)
+static size_t sizeof_imsm_map(struct imsm_map *map)
 {
-	size_t size = sizeof(*dev);
+	return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
+}
+
+struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
+{
+	struct imsm_map *map = &dev->vol.map[0];
+
+	if (second_map && !dev->vol.migr_state)
+		return NULL;
+	else if (second_map) {
+		void *ptr = map;
+
+		return ptr + sizeof_imsm_map(map);
+	} else
+		return map;
+		
+}
 
-	/* each map has disk_ord_tbl[num_members - 1] additional space */
-	size += sizeof(__u32) * (dev->vol.map[0].num_members - 1);
+/* return the size of the device.
+ * migr_state increases the returned size if map[0] were to be duplicated
+ */
+static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
+{
+	size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
+		      sizeof_imsm_map(get_imsm_map(dev, 0));
 
 	/* migrating means an additional map */
-	if (dev->vol.migr_state) {
-		size += sizeof(struct imsm_map);
-		size += sizeof(__u32) * (dev->vol.map[1].num_members - 1);
-	}
+	if (dev->vol.migr_state)
+		size += sizeof_imsm_map(get_imsm_map(dev, 1));
+	else if (migr_state)
+		size += sizeof_imsm_map(get_imsm_map(dev, 0));
 
 	return size;
 }
 
-static struct imsm_dev *get_imsm_dev(struct imsm_super *mpb, __u8 index)
+static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
 {
 	int offset;
 	int i;
 	void *_mpb = mpb;
 
-	if (index > mpb->num_raid_devs - 1)
+	if (index >= mpb->num_raid_devs)
 		return NULL;
 
 	/* devices start after all disks */
@@ -241,17 +337,42 @@ static struct imsm_dev *get_imsm_dev(struct imsm_super *mpb, __u8 index)
 		if (i == index)
 			return _mpb + offset;
 		else
-			offset += sizeof_imsm_dev(_mpb + offset);
+			offset += sizeof_imsm_dev(_mpb + offset, 0);
 
 	return NULL;
 }
 
-static __u32 get_imsm_disk_idx(struct imsm_map *map, int slot)
+static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
+{
+	if (index >= super->anchor->num_raid_devs)
+		return NULL;
+	return super->dev_tbl[index];
+}
+
+static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, int slot)
+{
+	struct imsm_map *map;
+
+	if (dev->vol.migr_state)
+		map = get_imsm_map(dev, 1);
+	else
+		map = get_imsm_map(dev, 0);
+
+	/* top byte identifies disk under rebuild */
+	return __le32_to_cpu(map->disk_ord_tbl[slot]);
+}
+
+#define ord_to_idx(ord) (((ord) << 8) >> 8)
+static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot)
 {
-	__u32 *ord_tbl = &map->disk_ord_tbl[slot];
+	__u32 ord = get_imsm_ord_tbl_ent(dev, slot);
 
-	/* top byte is 'special' */
-	return __le32_to_cpu(*ord_tbl & ~(0xff << 24));
+	return ord_to_idx(ord);
+}
+
+static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
+{
+	map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
 }
 
 static int get_imsm_raid_level(struct imsm_map *map)
@@ -266,6 +387,7 @@ static int get_imsm_raid_level(struct imsm_map *map)
 	return map->raid_level;
 }
 
+#ifndef MDASSEMBLE
 static int cmp_extent(const void *av, const void *bv)
 {
 	const struct extent *a = av;
@@ -280,22 +402,16 @@ static int cmp_extent(const void *av, const void *bv)
 static struct extent *get_extents(struct intel_super *super, struct dl *dl)
 {
 	/* find a list of used extents on the given physical device */
-	struct imsm_super *mpb = super->mpb;
-	struct imsm_disk *disk;
 	struct extent *rv, *e;
 	int i, j;
 	int memberships = 0;
 
-	disk = get_imsm_disk(mpb, dl->index);
-	if (!disk)
-		return NULL;
-
-	for (i = 0; i < mpb->num_raid_devs; i++) {
-		struct imsm_dev *dev = get_imsm_dev(mpb, i);
-		struct imsm_map *map = dev->vol.map;
+	for (i = 0; i < super->anchor->num_raid_devs; i++) {
+		struct imsm_dev *dev = get_imsm_dev(super, i);
+		struct imsm_map *map = get_imsm_map(dev, 0);
 
 		for (j = 0; j < map->num_members; j++) {
-			__u32 index = get_imsm_disk_idx(map, j);
+			__u32 index = get_imsm_disk_idx(dev, j);
 
 			if (index == dl->index)
 				memberships++;
@@ -306,12 +422,12 @@ static struct extent *get_extents(struct intel_super *super, struct dl *dl)
 		return NULL;
 	e = rv;
 
-	for (i = 0; i < mpb->num_raid_devs; i++) {
-		struct imsm_dev *dev = get_imsm_dev(mpb, i);
-		struct imsm_map *map = dev->vol.map;
+	for (i = 0; i < super->anchor->num_raid_devs; i++) {
+		struct imsm_dev *dev = get_imsm_dev(super, i);
+		struct imsm_map *map = get_imsm_map(dev, 0);
 
 		for (j = 0; j < map->num_members; j++) {
-			__u32 index = get_imsm_disk_idx(map, j);
+			__u32 index = get_imsm_disk_idx(dev, j);
 
 			if (index == dl->index) {
 				e->start = __le32_to_cpu(map->pba_of_lba0);
@@ -322,29 +438,31 @@ static struct extent *get_extents(struct intel_super *super, struct dl *dl)
 	}
 	qsort(rv, memberships, sizeof(*rv), cmp_extent);
 
-	e->start = __le32_to_cpu(disk->total_blocks) -
+	e->start = __le32_to_cpu(dl->disk.total_blocks) -
 		   (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
 	e->size = 0;
 	return rv;
 }
 
-#ifndef MDASSEMBLE
 static void print_imsm_dev(struct imsm_dev *dev, int index)
 {
 	__u64 sz;
 	int slot;
-	struct imsm_map *map = dev->vol.map;
+	struct imsm_map *map = get_imsm_map(dev, 0);
+	__u32 ord;
 
 	printf("\n");
 	printf("[%s]:\n", dev->volume);
 	printf("     RAID Level : %d\n", get_imsm_raid_level(map));
 	printf("        Members : %d\n", map->num_members);
 	for (slot = 0; slot < map->num_members; slot++)
-		if (index == get_imsm_disk_idx(map, slot))
+		if (index == get_imsm_disk_idx(dev, slot))
 			break;
-	if (slot < map->num_members)
-		printf("      This Slot : %d\n", slot);
-	else
+	if (slot < map->num_members) {
+		ord = get_imsm_ord_tbl_ent(dev, slot);
+		printf("      This Slot : %d%s\n", slot,
+		       ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
+	} else
 		printf("      This Slot : ?\n");
 	sz = __le32_to_cpu(dev->size_high);
 	sz <<= 32;
@@ -361,20 +479,31 @@ static void print_imsm_dev(struct imsm_dev *dev, int index)
 	printf("     Chunk Size : %u KiB\n",
 		__le16_to_cpu(map->blocks_per_strip) / 2);
 	printf("       Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
-	printf("  Migrate State : %s\n", dev->vol.migr_state ? "migrating" : "idle");
+	printf("  Migrate State : %s", dev->vol.migr_state ? "migrating" : "idle");
+	if (dev->vol.migr_state)
+		printf(": %s", dev->vol.migr_type ? "rebuilding" : "initializing");
+	printf("\n");
+	printf("      Map State : %s", map_state_str[map->map_state]);
+	if (dev->vol.migr_state) {
+		struct imsm_map *map = get_imsm_map(dev, 1);
+		printf(" <-- %s", map_state_str[map->map_state]);
+	}
+	printf("\n");
 	printf("    Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean");
-	printf("      Map State : %s\n", map_state_str[map->map_state]);
 }
 
 static void print_imsm_disk(struct imsm_super *mpb, int index)
 {
-	struct imsm_disk *disk = get_imsm_disk(mpb, index);
-	char str[MAX_RAID_SERIAL_LEN];
+	struct imsm_disk *disk = __get_imsm_disk(mpb, index);
+	char str[MAX_RAID_SERIAL_LEN + 1];
 	__u32 s;
 	__u64 sz;
 
+	if (index < 0)
+		return;
+
 	printf("\n");
-	snprintf(str, MAX_RAID_SERIAL_LEN, "%s", disk->serial);
+	snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
 	printf("  Disk%02d Serial : %s\n", index, str);
 	s = __le32_to_cpu(disk->status);
 	printf("          State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "",
@@ -391,7 +520,7 @@ static void print_imsm_disk(struct imsm_super *mpb, int index)
 static void examine_super_imsm(struct supertype *st, char *homehost)
 {
 	struct intel_super *super = st->sb;
-	struct imsm_super *mpb = super->mpb;
+	struct imsm_super *mpb = super->anchor;
 	char str[MAX_SIGNATURE_LENGTH];
 	int i;
 	__u32 sum;
@@ -404,13 +533,24 @@ static void examine_super_imsm(struct supertype *st, char *homehost)
 	printf("     Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
 	sum = __le32_to_cpu(mpb->check_sum);
 	printf("       Checksum : %08x %s\n", sum,
-		gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
+		__gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
 	printf("    MPB Sectors : %d\n", mpb_sectors(mpb));
 	printf("          Disks : %d\n", mpb->num_disks);
 	printf("   RAID Devices : %d\n", mpb->num_raid_devs);
 	print_imsm_disk(mpb, super->disks->index);
+	if (super->bbm_log) {
+		struct bbm_log *log = super->bbm_log;
+
+		printf("\n");
+		printf("Bad Block Management Log:\n");
+		printf("       Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
+		printf("      Signature : %x\n", __le32_to_cpu(log->signature));
+		printf("    Entry Count : %d\n", __le32_to_cpu(log->entry_count));
+		printf("   Spare Blocks : %d\n",  __le32_to_cpu(log->reserved_spare_block_count));
+		printf("    First Spare : %llx\n", __le64_to_cpu(log->first_spare_lba));
+	}
 	for (i = 0; i < mpb->num_raid_devs; i++)
-		print_imsm_dev(get_imsm_dev(mpb, i), super->disks->index);
+		print_imsm_dev(__get_imsm_dev(mpb, i), super->disks->index);
 	for (i = 0; i < mpb->num_disks; i++) {
 		if (i == super->disks->index)
 			continue;
@@ -420,11 +560,7 @@ static void examine_super_imsm(struct supertype *st, char *homehost)
 
 static void brief_examine_super_imsm(struct supertype *st)
 {
-	struct intel_super *super = st->sb;
-	struct imsm_super *mpb = super->mpb;
-
-	printf("ARRAY /dev/imsm family=%08x metadata=external:imsm\n",
-		__le32_to_cpu(mpb->family_num));
+	printf("ARRAY /dev/imsm metadata=imsm\n");
 }
 
 static void detail_super_imsm(struct supertype *st, char *homehost)
@@ -447,7 +583,13 @@ static int match_home_imsm(struct supertype *st, char *homehost)
 
 static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
 {
-	printf("%s\n", __FUNCTION__);
+	/* imsm does not track uuid's so just make sure we never return
+	 * the same value twice to break uuid matching in Manage_subdevs
+	 * FIXME what about the use of uuid's with bitmap's?
+	 */
+	static int dummy_id = 0;
+
+	uuid[0] = dummy_id++;
 }
 
 #if 0
@@ -486,7 +628,7 @@ static int imsm_level_to_layout(int level)
 		return 0;
 	case 5:
 	case 6:
-		return ALGORITHM_LEFT_SYMMETRIC;
+		return ALGORITHM_LEFT_ASYMMETRIC;
 	case 10:
 		return 0x102; //FIXME is this correct?
 	}
@@ -496,9 +638,8 @@ static int imsm_level_to_layout(int level)
 static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info)
 {
 	struct intel_super *super = st->sb;
-	struct imsm_super *mpb = super->mpb;
-	struct imsm_dev *dev = get_imsm_dev(mpb, super->current_vol);
-	struct imsm_map *map = &dev->vol.map[0];
+	struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
+	struct imsm_map *map = get_imsm_map(dev, 0);
 
 	info->container_member	  = super->current_vol;
 	info->array.raid_disks    = map->num_members;
@@ -507,24 +648,37 @@ static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info)
 	info->array.md_minor	  = -1;
 	info->array.ctime	  = 0;
 	info->array.utime	  = 0;
-	info->array.chunk_size	  = __le16_to_cpu(map->blocks_per_strip * 512);
+	info->array.chunk_size	  = __le16_to_cpu(map->blocks_per_strip) << 9;
+	info->array.state	  = !dev->vol.dirty;
+
+	info->disk.major = 0;
+	info->disk.minor = 0;
 
 	info->data_offset	  = __le32_to_cpu(map->pba_of_lba0);
 	info->component_size	  = __le32_to_cpu(map->blocks_per_member);
+	memset(info->uuid, 0, sizeof(info->uuid));
 
-	info->disk.major = 0;
-	info->disk.minor = 0;
+	if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
+	    dev->vol.dirty || dev->vol.migr_state)
+		info->resync_start = 0;
+	else
+		info->resync_start = ~0ULL;
 
+	strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
+	info->name[MAX_RAID_SERIAL_LEN] = 0;
+
+	info->array.major_version = -1;
+	info->array.minor_version = -2;
 	sprintf(info->text_version, "/%s/%d",
 		devnum2devname(st->container_dev),
 		info->container_member);
+	info->safe_mode_delay = 4000;  /* 4 secs like the Matrix driver */
 }
 
 
 static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info)
 {
 	struct intel_super *super = st->sb;
-	struct imsm_super *mpb = super->mpb;
 	struct imsm_disk *disk;
 	__u32 s;
 
@@ -532,7 +686,11 @@ static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info)
 		getinfo_super_imsm_volume(st, info);
 		return;
 	}
-	info->array.raid_disks    = mpb->num_disks;
+
+	/* Set raid_disks to zero so that Assemble will always pull in valid
+	 * spares
+	 */
+	info->array.raid_disks    = 0;
 	info->array.level         = LEVEL_CONTAINER;
 	info->array.layout        = 0;
 	info->array.md_minor      = -1;
@@ -544,12 +702,15 @@ static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info)
 	info->disk.minor = 0;
 	info->disk.raid_disk = -1;
 	info->reshape_active = 0;
+	info->array.major_version = -1;
+	info->array.minor_version = -2;
 	strcpy(info->text_version, "imsm");
+	info->safe_mode_delay = 0;
 	info->disk.number = -1;
 	info->disk.state = 0;
 
 	if (super->disks) {
-		disk = get_imsm_disk(mpb, super->disks->index);
+		disk = &super->disks->disk;
 		info->disk.number = super->disks->index;
 		info->disk.raid_disk = super->disks->index;
 		info->data_offset = __le32_to_cpu(disk->total_blocks) -
@@ -648,42 +809,75 @@ static int compare_super_imsm(struct supertype *st, struct supertype *tst)
                 return 0;
         }
 
-	if (memcmp(first->mpb->sig, sec->mpb->sig, MAX_SIGNATURE_LENGTH) != 0)
-		return 3;
-	if (first->mpb->family_num != sec->mpb->family_num)
-		return 3;
-	if (first->mpb->mpb_size != sec->mpb->mpb_size)
-		return 3;
-	if (first->mpb->check_sum != sec->mpb->check_sum)
+	if (memcmp(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH) != 0)
 		return 3;
 
+	/* if an anchor does not have num_raid_devs set then it is a free
+	 * floating spare
+	 */
+	if (first->anchor->num_raid_devs > 0 &&
+	    sec->anchor->num_raid_devs > 0) {
+		if (first->anchor->family_num != sec->anchor->family_num)
+			return 3;
+	}
+
+	/* if 'first' is a spare promote it to a populated mpb with sec's
+	 * family number
+	 */
+	if (first->anchor->num_raid_devs == 0 &&
+	    sec->anchor->num_raid_devs > 0) {
+		first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
+		first->anchor->family_num = sec->anchor->family_num;
+	}
+
 	return 0;
 }
 
+static void fd2devname(int fd, char *name)
+{
+	struct stat st;
+	char path[256];
+	char dname[100];
+	char *nm;
+	int rv;
+
+	name[0] = '\0';
+	if (fstat(fd, &st) != 0)
+		return;
+	sprintf(path, "/sys/dev/block/%d:%d",
+		major(st.st_rdev), minor(st.st_rdev));
+
+	rv = readlink(path, dname, sizeof(dname));
+	if (rv <= 0)
+		return;
+	
+	dname[rv] = '\0';
+	nm = strrchr(dname, '/');
+	nm++;
+	snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
+}
+
+
 extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
 
 static int imsm_read_serial(int fd, char *devname,
 			    __u8 serial[MAX_RAID_SERIAL_LEN])
 {
 	unsigned char scsi_serial[255];
-	int sg_fd;
 	int rv;
 	int rsp_len;
-	int i, cnt;
+	int len;
+	char *c, *rsp_buf;
 
 	memset(scsi_serial, 0, sizeof(scsi_serial));
 
-	sg_fd = sysfs_disk_to_sg(fd);
-	if (sg_fd < 0) {
-		if (devname)
-			fprintf(stderr,
-				Name ": Failed to open sg interface for %s: %s\n",
-				devname, strerror(errno));
-		return 1;
-	}
+	rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial));
 
-	rv = scsi_get_serial(sg_fd, scsi_serial, sizeof(scsi_serial));
-	close(sg_fd);
+	if (rv && imsm_env_devname_as_serial()) {
+		memset(serial, 0, MAX_RAID_SERIAL_LEN);
+		fd2devname(fd, (char *) serial);
+		return 0;
+	}
 
 	if (rv != 0) {
 		if (devname)
@@ -693,30 +887,62 @@ static int imsm_read_serial(int fd, char *devname,
 		return rv;
 	}
 
+	/* trim whitespace */
 	rsp_len = scsi_serial[3];
-	for (i = 0, cnt = 0; i < rsp_len; i++) {
-		if (!isspace(scsi_serial[4 + i]))
-			serial[cnt++] = scsi_serial[4 + i];
-		if (cnt == MAX_RAID_SERIAL_LEN)
-			break;
-	}
-
-	serial[MAX_RAID_SERIAL_LEN - 1] = '\0';
+	rsp_buf = (char *) &scsi_serial[4];
+	c = rsp_buf;
+	while (isspace(*c))
+		c++;
+	if (c + MAX_RAID_SERIAL_LEN > rsp_buf + rsp_len)
+		len = rsp_len - (c - rsp_buf);
+	else
+		len = MAX_RAID_SERIAL_LEN;
+	memcpy(serial, c, len);
+	c = (char *) &serial[len - 1];
+	while (isspace(*c) || *c == '\0')
+		*c-- = '\0';
 
 	return 0;
 }
 
+static int serialcmp(__u8 *s1, __u8 *s2)
+{
+	return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
+}
+
+static void serialcpy(__u8 *dest, __u8 *src)
+{
+	strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
+}
+
 static int
 load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
 {
-	struct imsm_super *mpb = super->mpb;
 	struct dl *dl;
 	struct stat stb;
-	struct imsm_disk *disk;
 	int rv;
 	int i;
+	int alloc = 1;
+	__u8 serial[MAX_RAID_SERIAL_LEN];
+
+	rv = imsm_read_serial(fd, devname, serial);
+
+	if (rv != 0)
+		return 2;
+
+	/* check if this is a disk we have seen before.  it may be a spare in
+	 * super->disks while the current anchor believes it is a raid member,
+	 * check if we need to update dl->index
+	 */
+	for (dl = super->disks; dl; dl = dl->next)
+		if (serialcmp(dl->serial, serial) == 0)
+			break;
+
+	if (!dl)
+		dl = malloc(sizeof(*dl));
+	else
+		alloc = 0;
 
-	dl = malloc(sizeof(*dl));
 	if (!dl) {
 		if (devname)
 			fprintf(stderr,
@@ -724,48 +950,154 @@ load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
 				devname);
 		return 2;
 	}
-	memset(dl, 0, sizeof(*dl));
 
-	fstat(fd, &stb);
-	dl->major = major(stb.st_rdev);
-	dl->minor = minor(stb.st_rdev);
-	dl->next = super->disks;
-	dl->fd = keep_fd ? fd : -1;
-	dl->devname = devname ? strdup(devname) : NULL;
-	dl->index = -1;
-	super->disks = dl;
-	rv = imsm_read_serial(fd, devname, dl->serial);
+	if (alloc) {
+		fstat(fd, &stb);
+		dl->major = major(stb.st_rdev);
+		dl->minor = minor(stb.st_rdev);
+		dl->next = super->disks;
+		dl->fd = keep_fd ? fd : -1;
+		dl->devname = devname ? strdup(devname) : NULL;
+		serialcpy(dl->serial, serial);
+		dl->index = -2;
+	} else if (keep_fd) {
+		close(dl->fd);
+		dl->fd = fd;
+	}
 
-	if (rv != 0)
-		return 2;
+	/* look up this disk's index in the current anchor */
+	for (i = 0; i < super->anchor->num_disks; i++) {
+		struct imsm_disk *disk_iter;
 
-	/* look up this disk's index */
-	for (i = 0; i < mpb->num_disks; i++) {
-		disk = get_imsm_disk(mpb, i);
+		disk_iter = __get_imsm_disk(super->anchor, i);
+
+		if (serialcmp(disk_iter->serial, dl->serial) == 0) {
+			__u32 status;
+
+			dl->disk = *disk_iter;
+			status = __le32_to_cpu(dl->disk.status);
+			/* only set index on disks that are a member of a
+			 * populated contianer, i.e. one with raid_devs
+			 */
+			if (status & FAILED_DISK)
+				dl->index = -2;
+			else if (status & SPARE_DISK)
+				dl->index = -1;
+			else
+				dl->index = i;
 
-		if (memcmp(disk->serial, dl->serial, MAX_RAID_SERIAL_LEN) == 0)
 			break;
+		}
 	}
 
-	if (i > mpb->num_disks)
-		return 2;
+	if (alloc)
+		super->disks = dl;
+
+	return 0;
+}
+
+static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
+{
+	memcpy(dest, src, sizeof_imsm_dev(src, 0));
+}
+
+#ifndef MDASSEMBLE
+/* When migrating map0 contains the 'destination' state while map1
+ * contains the current state.  When not migrating map0 contains the
+ * current state.  This routine assumes that map[0].map_state is set to
+ * the current array state before being called.
+ *
+ * Migration is indicated by one of the following states
+ * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
+ * 2/ Initialize (migr_state=1 migr_type=0 map0state=normal
+ *    map1state=unitialized)
+ * 3/ Verify (Resync) (migr_state=1 migr_type=1 map0state=normal
+ *    map1state=normal)
+ * 4/ Rebuild (migr_state=1 migr_type=1 map0state=normal
+ *    map1state=degraded)
+ */
+static void migrate(struct imsm_dev *dev, __u8 to_state, int rebuild_resync)
+{
+	struct imsm_map *dest;
+	struct imsm_map *src = get_imsm_map(dev, 0);
+
+	dev->vol.migr_state = 1;
+	dev->vol.migr_type = rebuild_resync;
+	dest = get_imsm_map(dev, 1);
 
-	dl->index = i;
+	memcpy(dest, src, sizeof_imsm_map(src));
+	src->map_state = to_state;
+}
+#endif
+
+static int parse_raid_devices(struct intel_super *super)
+{
+	int i;
+	struct imsm_dev *dev_new;
+	size_t len, len_migr;
+	size_t space_needed = 0;
+	struct imsm_super *mpb = super->anchor;
+
+	for (i = 0; i < super->anchor->num_raid_devs; i++) {
+		struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
+
+		len = sizeof_imsm_dev(dev_iter, 0);
+		len_migr = sizeof_imsm_dev(dev_iter, 1);
+		if (len_migr > len)
+			space_needed += len_migr - len;
+		
+		dev_new = malloc(len_migr);
+		if (!dev_new)
+			return 1;
+		imsm_copy_dev(dev_new, dev_iter);
+		super->dev_tbl[i] = dev_new;
+	}
+
+	/* ensure that super->buf is large enough when all raid devices
+	 * are migrating
+	 */
+	if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
+		void *buf;
+
+		len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512);
+		if (posix_memalign(&buf, 512, len) != 0)
+			return 1;
 
+		memcpy(buf, super->buf, len);
+		free(super->buf);
+		super->buf = buf;
+		super->len = len;
+	}
+		
 	return 0;
 }
 
+/* retrieve a pointer to the bbm log which starts after all raid devices */
+struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb)
+{
+	void *ptr = NULL;
+
+	if (__le32_to_cpu(mpb->bbm_log_size)) {
+		ptr = mpb;
+		ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size);
+	} 
+
+	return ptr;
+}
+
+static void __free_imsm(struct intel_super *super, int free_disks);
+
 /* load_imsm_mpb - read matrix metadata
  * allocates super->mpb to be freed by free_super
  */
 static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
 {
 	unsigned long long dsize;
-	size_t len, mpb_size;
 	unsigned long long sectors;
 	struct stat;
 	struct imsm_super *anchor;
 	__u32 check_sum;
+	int rc;
 
 	get_dev_size(fd, NULL, &dsize);
 
@@ -777,15 +1109,14 @@ static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
 		return 1;
 	}
 
-	len = 512;
-	if (posix_memalign((void**)&anchor, 512, len) != 0) {
+	if (posix_memalign((void**)&anchor, 512, 512) != 0) {
 		if (devname)
 			fprintf(stderr,
 				Name ": Failed to allocate imsm anchor buffer"
 				" on %s\n", devname);
 		return 1;
 	}
-	if (read(fd, anchor, len) != len) {
+	if (read(fd, anchor, 512) != 512) {
 		if (devname)
 			fprintf(stderr,
 				Name ": Cannot read anchor block on %s: %s\n",
@@ -802,22 +1133,26 @@ static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
 		return 2;
 	}
 
-	mpb_size = __le32_to_cpu(anchor->mpb_size);
-	mpb_size = ROUND_UP(mpb_size, 512);
-	if (posix_memalign(&super->buf, 512, mpb_size) != 0) {
+	__free_imsm(super, 0);
+	super->len = ROUND_UP(anchor->mpb_size, 512);
+	if (posix_memalign(&super->buf, 512, super->len) != 0) {
 		if (devname)
 			fprintf(stderr,
 				Name ": unable to allocate %zu byte mpb buffer\n",
-				mpb_size);
+				super->len);
 		free(anchor);
 		return 2;
 	}
-	memcpy(super->buf, anchor, len);
+	memcpy(super->buf, anchor, 512);
 
 	sectors = mpb_sectors(anchor) - 1;
 	free(anchor);
-	if (!sectors)
-		return load_imsm_disk(fd, super, devname, 0);
+	if (!sectors) {
+		rc = load_imsm_disk(fd, super, devname, 0);
+		if (rc == 0)
+			rc = parse_raid_devices(super);
+		return rc;
+	}
 
 	/* read the extended mpb */
 	if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) {
@@ -828,8 +1163,7 @@ static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
 		return 1;
 	}
 
-	len = mpb_size - 512;
-	if (read(fd, super->buf + 512, len) != len) {
+	if (read(fd, super->buf + 512, super->len - 512) != super->len - 512) {
 		if (devname)
 			fprintf(stderr,
 				Name ": Cannot read extended mpb on %s: %s\n",
@@ -837,42 +1171,72 @@ static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
 		return 2;
 	}
 
-	check_sum = gen_imsm_checksum(super->mpb);
-	if (check_sum != __le32_to_cpu(super->mpb->check_sum)) {
+	check_sum = __gen_imsm_checksum(super->anchor);
+	if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
 		if (devname)
 			fprintf(stderr,
 				Name ": IMSM checksum %x != %x on %s\n",
-				check_sum, __le32_to_cpu(super->mpb->check_sum),
+				check_sum, __le32_to_cpu(super->anchor->check_sum),
 				devname);
 		return 2;
 	}
 
-	return load_imsm_disk(fd, super, devname, 0);
+	/* FIXME the BBM log is disk specific so we cannot use this global
+	 * buffer for all disks.  Ok for now since we only look at the global
+	 * bbm_log_size parameter to gate assembly
+	 */
+	super->bbm_log = __get_imsm_bbm_log(super->anchor);
+
+	rc = load_imsm_disk(fd, super, devname, 0);
+	if (rc == 0)
+		rc = parse_raid_devices(super);
+
+	return rc;
 }
 
+static void __free_imsm_disk(struct dl *d)
+{
+	if (d->fd >= 0)
+		close(d->fd);
+	if (d->devname)
+		free(d->devname);
+	free(d);
+
+}
 static void free_imsm_disks(struct intel_super *super)
 {
 	while (super->disks) {
 		struct dl *d = super->disks;
 
 		super->disks = d->next;
-		if (d->fd >= 0)
-			close(d->fd);
-		if (d->devname)
-			free(d->devname);
-		free(d);
+		__free_imsm_disk(d);
 	}
 }
 
+/* free all the pieces hanging off of a super pointer */
+static void __free_imsm(struct intel_super *super, int free_disks)
+{
+	int i;
+
+	if (super->buf) {
+		free(super->buf);
+		super->buf = NULL;
+	}
+	if (free_disks)
+		free_imsm_disks(super);
+	for (i = 0; i < IMSM_MAX_RAID_DEVS; i++)
+		if (super->dev_tbl[i]) {
+			free(super->dev_tbl[i]);
+			super->dev_tbl[i] = NULL;
+		}
+}
+
 static void free_imsm(struct intel_super *super)
 {
-	if (super->mpb)
-		free(super->mpb);
-	free_imsm_disks(super);
+	__free_imsm(super, 1);
 	free(super);
 }
 
-
 static void free_super_imsm(struct supertype *st)
 {
 	struct intel_super *super = st->sb;
@@ -924,7 +1288,7 @@ static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
 	if (!super)
 		return 1;
 
-	/* find the most up to date disk in this array */
+	/* find the most up to date disk in this array, skipping spares */
 	for (sd = sra->devs; sd; sd = sd->next) {
 		sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
 		dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY);
@@ -936,7 +1300,10 @@ static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
 		if (!keep_fd)
 			close(dfd);
 		if (rv == 0) {
-			gen = __le32_to_cpu(super->mpb->generation_num);
+			if (super->anchor->num_raid_devs == 0)
+				gen = 0;
+			else
+				gen = __le32_to_cpu(super->anchor->generation_num);
 			if (!best || gen > bestgen) {
 				bestgen = gen;
 				best = sd;
@@ -966,10 +1333,7 @@ static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
 		return 2;
 	}
 
-	/* reset the disk list */
-	free_imsm_disks(super);
-
-	/* populate disk list */
+	/* re-parse the disk list with the current anchor */
 	for (sd = sra->devs ; sd ; sd = sd->next) {
 		sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
 		dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
@@ -983,18 +1347,18 @@ static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
 	}
 
 	if (st->subarray[0]) {
-		if (atoi(st->subarray) <= super->mpb->num_raid_devs)
+		if (atoi(st->subarray) <= super->anchor->num_raid_devs)
 			super->current_vol = atoi(st->subarray);
 		else
 			return 1;
 	}
 
 	*sbp = super;
+	st->container_dev = fd2devnum(fd);
 	if (st->ss == NULL) {
 		st->ss = &super_imsm;
 		st->minor_version = 0;
 		st->max_devs = IMSM_MAX_DEVICES;
-		st->container_dev = fd2devnum(fd);
 	}
 
 	return 0;
@@ -1060,6 +1424,11 @@ static __u32 info_to_num_data_stripes(mdu_array_info_t *info)
 	return num_stripes;
 }
 
+static __u32 info_to_blocks_per_member(mdu_array_info_t *info)
+{
+	return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1);
+}
+
 static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
 				  unsigned long long size, char *name,
 				  char *homehost, int *uuid)
@@ -1068,7 +1437,7 @@ static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
 	 * so st->sb is already set.
 	 */
 	struct intel_super *super = st->sb;
-	struct imsm_super *mpb = super->mpb;
+	struct imsm_super *mpb = super->anchor;
 	struct imsm_dev *dev;
 	struct imsm_vol *vol;
 	struct imsm_map *map;
@@ -1098,14 +1467,23 @@ static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
 		memcpy(mpb_new, mpb, size_old);
 		free(mpb);
 		mpb = mpb_new;
-		super->mpb = mpb_new;
+		super->anchor = mpb_new;
 		mpb->mpb_size = __cpu_to_le32(size_new);
 		memset(mpb_new + size_old, 0, size_round - size_old);
 	}
 	super->current_vol = idx;
+	/* when creating the first raid device in this container set num_disks
+	 * to zero, i.e. delete this spare and add raid member devices in
+	 * add_to_super_imsm_volume()
+	 */
+	if (super->current_vol == 0)
+		mpb->num_disks = 0;
 	sprintf(st->subarray, "%d", idx);
-	mpb->num_raid_devs++;
-	dev = get_imsm_dev(mpb, idx);
+	dev = malloc(sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
+	if (!dev) {
+		fprintf(stderr, Name": could not allocate raid device\n");
+		return 0;
+	}
 	strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
 	array_blocks = calc_array_size(info->level, info->raid_disks,
 				       info->layout, info->chunk_size,
@@ -1119,15 +1497,15 @@ static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
 	vol->migr_type = 0;
 	vol->dirty = 0;
 	for (i = 0; i < idx; i++) {
-		struct imsm_dev *prev = get_imsm_dev(mpb, i);
-		struct imsm_map *pmap = &prev->vol.map[0];
+		struct imsm_dev *prev = get_imsm_dev(super, i);
+		struct imsm_map *pmap = get_imsm_map(prev, 0);
 
 		offset += __le32_to_cpu(pmap->blocks_per_member);
 		offset += IMSM_RESERVED_SECTORS;
 	}
-	map = &vol->map[0];
+	map = get_imsm_map(dev, 0);
 	map->pba_of_lba0 = __cpu_to_le32(offset);
-	map->blocks_per_member = __cpu_to_le32(info->size * 2);
+	map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info));
 	map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
 	map->num_data_stripes = __cpu_to_le32(info_to_num_data_stripes(info));
 	map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED :
@@ -1146,8 +1524,10 @@ static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
 	map->num_members = info->raid_disks;
 	for (i = 0; i < map->num_members; i++) {
 		/* initialized in add_to_super */
-		map->disk_ord_tbl[i] = __cpu_to_le32(0);
+		set_imsm_ord_tbl_ent(map, i, 0);
 	}
+	mpb->num_raid_devs++;
+	super->dev_tbl[super->current_vol] = dev;
 
 	return 1;
 }
@@ -1196,40 +1576,54 @@ static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
 	return 1;
 }
 
+#ifndef MDASSEMBLE
 static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
 				     int fd, char *devname)
 {
 	struct intel_super *super = st->sb;
-	struct imsm_super *mpb = super->mpb;
+	struct imsm_super *mpb = super->anchor;
 	struct dl *dl;
 	struct imsm_dev *dev;
 	struct imsm_map *map;
-	struct imsm_disk *disk;
 	__u32 status;
 
-	dev = get_imsm_dev(mpb, super->current_vol);
-	map = &dev->vol.map[0];
+	dev = get_imsm_dev(super, super->current_vol);
+	map = get_imsm_map(dev, 0);
 
 	for (dl = super->disks; dl ; dl = dl->next)
 		if (dl->major == dk->major &&
 		    dl->minor == dk->minor)
 			break;
+
 	if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
 		return;
 
-	map->disk_ord_tbl[dk->number] = __cpu_to_le32(dl->index);
-
-	disk = get_imsm_disk(mpb, dl->index);
+	/* add a pristine spare to the metadata */
+	if (dl->index < 0) {
+		dl->index = super->anchor->num_disks;
+		super->anchor->num_disks++;
+	}
+	set_imsm_ord_tbl_ent(map, dk->number, dl->index);
 	status = CONFIGURED_DISK | USABLE_DISK;
-	disk->status = __cpu_to_le32(status);
+	dl->disk.status = __cpu_to_le32(status);
+
+	/* if we are creating the first raid device update the family number */
+	if (super->current_vol == 0) {
+		__u32 sum;
+		struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
+		struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index);
+
+		*_dev = *dev;
+		*_disk = dl->disk;
+		sum = __gen_imsm_checksum(mpb);
+		mpb->family_num = __cpu_to_le32(sum);
+	}
 }
 
 static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
 			      int fd, char *devname)
 {
 	struct intel_super *super = st->sb;
-	struct imsm_super *mpb = super->mpb;
-	struct imsm_disk *disk;
 	struct dl *dd;
 	unsigned long long size;
 	__u32 status, id;
@@ -1251,118 +1645,212 @@ static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
 	memset(dd, 0, sizeof(*dd));
 	dd->major = major(stb.st_rdev);
 	dd->minor = minor(stb.st_rdev);
-	dd->index = dk->number;
+	dd->index = -1;
 	dd->devname = devname ? strdup(devname) : NULL;
-	dd->next = super->disks;
 	dd->fd = fd;
 	rv = imsm_read_serial(fd, devname, dd->serial);
 	if (rv) {
 		fprintf(stderr,
-			Name ": failed to retrieve scsi serial "
-			"using \'%s\' instead\n", devname);
-		strcpy((char *) dd->serial, devname);
+			Name ": failed to retrieve scsi serial, aborting\n");
+		free(dd);
+		abort();
 	}
 
-	if (mpb->num_disks <= dk->number)
-		mpb->num_disks = dk->number + 1;
-
-	disk = get_imsm_disk(mpb, dk->number);
 	get_dev_size(fd, NULL, &size);
 	size /= 512;
 	status = USABLE_DISK | SPARE_DISK;
-	strcpy((char *) disk->serial, (char *) dd->serial);
-	disk->total_blocks = __cpu_to_le32(size);
-	disk->status = __cpu_to_le32(status);
+	serialcpy(dd->disk.serial, dd->serial);
+	dd->disk.total_blocks = __cpu_to_le32(size);
+	dd->disk.status = __cpu_to_le32(status);
 	if (sysfs_disk_to_scsi_id(fd, &id) == 0)
-		disk->scsi_id = __cpu_to_le32(id);
+		dd->disk.scsi_id = __cpu_to_le32(id);
 	else
-		disk->scsi_id = __cpu_to_le32(0);
+		dd->disk.scsi_id = __cpu_to_le32(0);
 
-	/* update the family number if we are creating a container */
-	if (super->creating_imsm)
-		mpb->family_num = __cpu_to_le32(gen_imsm_checksum(mpb));
-	
-	super->disks = dd;
+	if (st->update_tail) {
+		dd->next = super->add;
+		super->add = dd;
+	} else {
+		dd->next = super->disks;
+		super->disks = dd;
+	}
 }
 
 static int store_imsm_mpb(int fd, struct intel_super *super);
 
+/* spare records have their own family number and do not have any defined raid
+ * devices
+ */
+static int write_super_imsm_spares(struct intel_super *super, int doclose)
+{
+	struct imsm_super mpb_save;
+	struct imsm_super *mpb = super->anchor;
+	__u32 sum;
+	struct dl *d;
+
+	mpb_save = *mpb;
+	mpb->num_raid_devs = 0;
+	mpb->num_disks = 1;
+	mpb->mpb_size = sizeof(struct imsm_super);
+	mpb->generation_num = __cpu_to_le32(1UL);
+
+	for (d = super->disks; d; d = d->next) {
+		if (d->index != -1)
+			continue;
+
+		mpb->disk[0] = d->disk;
+		sum = __gen_imsm_checksum(mpb);
+		mpb->family_num = __cpu_to_le32(sum);
+		sum = __gen_imsm_checksum(mpb);
+		mpb->check_sum = __cpu_to_le32(sum);
+
+		if (store_imsm_mpb(d->fd, super)) {
+			fprintf(stderr, "%s: failed for device %d:%d %s\n",
+				__func__, d->major, d->minor, strerror(errno));
+			*mpb = mpb_save;
+			return 1;
+		}
+		if (doclose) {
+			close(d->fd);
+			d->fd = -1;
+		}
+	}
+
+	*mpb = mpb_save;
+	return 0;
+}
+
 static int write_super_imsm(struct intel_super *super, int doclose)
 {
-	struct imsm_super *mpb = super->mpb;
+	struct imsm_super *mpb = super->anchor;
 	struct dl *d;
 	__u32 generation;
 	__u32 sum;
+	int spares = 0;
+	int i;
+	__u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
 
 	/* 'generation' is incremented everytime the metadata is written */
 	generation = __le32_to_cpu(mpb->generation_num);
 	generation++;
 	mpb->generation_num = __cpu_to_le32(generation);
 
+	for (d = super->disks; d; d = d->next) {
+		if (d->index == -1)
+			spares++;
+		else {
+			mpb->disk[d->index] = d->disk;
+			mpb_size += sizeof(struct imsm_disk);
+		}
+	}
+
+	for (i = 0; i < mpb->num_raid_devs; i++) {
+		struct imsm_dev *dev = __get_imsm_dev(mpb, i);
+
+		imsm_copy_dev(dev, super->dev_tbl[i]);
+		mpb_size += sizeof_imsm_dev(dev, 0);
+	}
+	mpb_size += __le32_to_cpu(mpb->bbm_log_size);
+	mpb->mpb_size = __cpu_to_le32(mpb_size);
+
 	/* recalculate checksum */
-	sum = gen_imsm_checksum(mpb);
+	sum = __gen_imsm_checksum(mpb);
 	mpb->check_sum = __cpu_to_le32(sum);
 
+	/* write the mpb for disks that compose raid devices */
 	for (d = super->disks; d ; d = d->next) {
-		if (store_imsm_mpb(d->fd, super)) {
+		if (d->index < 0)
+			continue;
+		if (store_imsm_mpb(d->fd, super))
 			fprintf(stderr, "%s: failed for device %d:%d %s\n",
 				__func__, d->major, d->minor, strerror(errno));
-			return 0;
-		}
 		if (doclose) {
 			close(d->fd);
 			d->fd = -1;
 		}
 	}
 
-	return 1;
+	if (spares)
+		return write_super_imsm_spares(super, doclose);
+
+	return 0;
+}
+
+
+static int create_array(struct supertype *st)
+{
+	size_t len;
+	struct imsm_update_create_array *u;
+	struct intel_super *super = st->sb;
+	struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
+
+	len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0);
+	u = malloc(len);
+	if (!u) {
+		fprintf(stderr, "%s: failed to allocate update buffer\n",
+			__func__);
+		return 1;
+	}
+
+	u->type = update_create_array;
+	u->dev_idx = super->current_vol;
+	imsm_copy_dev(&u->dev, dev);
+	append_metadata_update(st, u, len);
+
+	return 0;
+}
+
+static int _add_disk(struct supertype *st)
+{
+	struct intel_super *super = st->sb;
+	size_t len;
+	struct imsm_update_add_disk *u;
+
+	if (!super->add)
+		return 0;
+
+	len = sizeof(*u);
+	u = malloc(len);
+	if (!u) {
+		fprintf(stderr, "%s: failed to allocate update buffer\n",
+			__func__);
+		return 1;
+	}
+
+	u->type = update_add_disk;
+	append_metadata_update(st, u, len);
+
+	return 0;
 }
 
 static int write_init_super_imsm(struct supertype *st)
 {
 	if (st->update_tail) {
-		/* queue the recently created array as a metadata update */
-		size_t len;
-		struct imsm_update_create_array *u;
+		/* queue the recently created array / added disk
+		 * as a metadata update */
 		struct intel_super *super = st->sb;
-		struct imsm_super *mpb = super->mpb;
-		struct imsm_dev *dev;
-		struct imsm_map *map;
 		struct dl *d;
+		int rv;
 
-		if (super->current_vol < 0 ||
-		    !(dev = get_imsm_dev(mpb, super->current_vol))) {
-			fprintf(stderr, "%s: could not determine sub-array\n",
-				__func__);
-			return 1;
-		}
-
-
-		map = &dev->vol.map[0];
-		len = sizeof(*u) + sizeof(__u32) * (map->num_members - 1);
-		u = malloc(len);
-		if (!u) {
-			fprintf(stderr, "%s: failed to allocate update buffer\n",
-				__func__);
-			return 1;
-		}
-
-		u->type = update_create_array;
-		u->dev_idx = super->current_vol;
-		memcpy(&u->dev, dev, sizeof(*dev));
-		memcpy(u->dev.vol.map[0].disk_ord_tbl, map->disk_ord_tbl,
-		       sizeof(__u32) * map->num_members);
-		append_metadata_update(st, u, len);
+		/* determine if we are creating a volume or adding a disk */
+		if (super->current_vol < 0) {
+			/* in the add disk case we are running in mdmon
+			 * context, so don't close fd's
+			 */
+			return _add_disk(st);
+		} else
+			rv = create_array(st);
 
 		for (d = super->disks; d ; d = d->next) {
 			close(d->fd);
 			d->fd = -1;
 		}
 
-		return 0;
+		return rv;
 	} else
 		return write_super_imsm(st->sb, 1);
 }
+#endif
 
 static int store_zero_imsm(struct supertype *st, int fd)
 {
@@ -1378,12 +1866,18 @@ static int store_zero_imsm(struct supertype *st, int fd)
 	if (posix_memalign(&buf, 512, 512) != 0)
 		return 1;
 
-	memset(buf, 0, sizeof(buf));
-	if (write(fd, buf, sizeof(buf)) != sizeof(buf))
+	memset(buf, 0, 512);
+	if (write(fd, buf, 512) != 512)
 		return 1;
 	return 0;
 }
 
+static int imsm_bbm_log_size(struct imsm_super *mpb)
+{
+	return __le32_to_cpu(mpb->bbm_log_size);
+}
+
+#ifndef MDASSEMBLE
 static int validate_geometry_imsm_container(struct supertype *st, int level,
 					    int layout, int raiddisks, int chunk,
 					    unsigned long long size, char *dev,
@@ -1612,6 +2106,7 @@ static int validate_geometry_imsm(struct supertype *st, int level, int layout,
 
 	return 1;
 }
+#endif /* MDASSEMBLE */
 
 static struct mdinfo *container_content_imsm(struct supertype *st)
 {
@@ -1624,84 +2119,82 @@ static struct mdinfo *container_content_imsm(struct supertype *st)
 	 *  and create appropriate device mdinfo.
 	 */
 	struct intel_super *super = st->sb;
-	struct imsm_super *mpb = super->mpb;
+	struct imsm_super *mpb = super->anchor;
 	struct mdinfo *rest = NULL;
 	int i;
 
+	/* do not assemble arrays that might have bad blocks */
+	if (imsm_bbm_log_size(super->anchor)) {
+		fprintf(stderr, Name ": BBM log found in metadata. "
+				"Cannot activate array(s).\n");
+		return NULL;
+	}
+
 	for (i = 0; i < mpb->num_raid_devs; i++) {
-		struct imsm_dev *dev = get_imsm_dev(mpb, i);
-		struct imsm_vol *vol = &dev->vol;
-		struct imsm_map *map = vol->map;
+		struct imsm_dev *dev = get_imsm_dev(super, i);
+		struct imsm_map *map = get_imsm_map(dev, 0);
 		struct mdinfo *this;
-		__u64 sz;
 		int slot;
 
 		this = malloc(sizeof(*this));
 		memset(this, 0, sizeof(*this));
 		this->next = rest;
-		rest = this;
-
-		this->array.level = get_imsm_raid_level(map);
-		this->array.raid_disks = map->num_members;
-		this->array.layout = imsm_level_to_layout(this->array.level);
-		this->array.md_minor = -1;
-		this->array.ctime = 0;
-		this->array.utime = 0;
-		this->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9;
-		this->array.state = !vol->dirty;
-		this->container_member = i;
-		if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty)
-			this->resync_start = 0;
-		else
-			this->resync_start = ~0ULL;
-
-		strncpy(this->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
-		this->name[MAX_RAID_SERIAL_LEN] = 0;
-
-		sprintf(this->text_version, "/%s/%d",
-			devnum2devname(st->container_dev),
-			this->container_member);
-
-		memset(this->uuid, 0, sizeof(this->uuid));
-
-		sz = __le32_to_cpu(dev->size_high);
-		sz <<= 32;
-		sz += __le32_to_cpu(dev->size_low);
-		this->component_size = sz;
-		this->array.size = this->component_size / 2;
 
+		super->current_vol = i;
+		getinfo_super_imsm_volume(st, this);
 		for (slot = 0 ; slot <  map->num_members; slot++) {
-			struct imsm_disk *disk;
 			struct mdinfo *info_d;
 			struct dl *d;
 			int idx;
+			int skip;
 			__u32 s;
+			__u32 ord;
 
-			idx = __le32_to_cpu(map->disk_ord_tbl[slot] & ~(0xff << 24));
+			skip = 0;
+			idx = get_imsm_disk_idx(dev, slot);
+			ord = get_imsm_ord_tbl_ent(dev, slot); 
 			for (d = super->disks; d ; d = d->next)
 				if (d->index == idx)
                                         break;
 
 			if (d == NULL)
-				break; /* shouldn't this be continue ?? */
+				skip = 1;
+
+			s = d ? __le32_to_cpu(d->disk.status) : 0;
+			if (s & FAILED_DISK)
+				skip = 1;
+			if (!(s & USABLE_DISK))
+				skip = 1;
+			if (ord & IMSM_ORD_REBUILD)
+				skip = 1;
+
+			/* 
+			 * if we skip some disks the array will be assmebled degraded;
+			 * reset resync start to avoid a dirty-degraded situation
+			 *
+			 * FIXME handle dirty degraded
+			 */
+			if (skip && !dev->vol.dirty)
+				this->resync_start = ~0ULL;
+			if (skip)
+				continue;
 
 			info_d = malloc(sizeof(*info_d));
-			if (!info_d)
-				break; /* ditto ?? */
+			if (!info_d) {
+				fprintf(stderr, Name ": failed to allocate disk"
+					" for volume %s\n", (char *) dev->volume);
+				free(this);
+				this = rest;
+				break;
+			}
 			memset(info_d, 0, sizeof(*info_d));
 			info_d->next = this->devs;
 			this->devs = info_d;
 
-			disk = get_imsm_disk(mpb, idx);
-			s = __le32_to_cpu(disk->status);
-
 			info_d->disk.number = d->index;
 			info_d->disk.major = d->major;
 			info_d->disk.minor = d->minor;
 			info_d->disk.raid_disk = slot;
-			info_d->disk.state  = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
-			info_d->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
-			info_d->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0;
 
 			this->array.working_disks++;
 
@@ -1711,19 +2204,21 @@ static struct mdinfo *container_content_imsm(struct supertype *st)
 			if (d->devname)
 				strcpy(info_d->name, d->devname);
 		}
+		rest = this;
 	}
 
 	return rest;
 }
 
 
+#ifndef MDASSEMBLE
 static int imsm_open_new(struct supertype *c, struct active_array *a,
 			 char *inst)
 {
 	struct intel_super *super = c->sb;
-	struct imsm_super *mpb = super->mpb;
+	struct imsm_super *mpb = super->anchor;
 	
-	if (atoi(inst) + 1 > mpb->num_raid_devs) {
+	if (atoi(inst) >= mpb->num_raid_devs) {
 		fprintf(stderr, "%s: subarry index %d, out of range\n",
 			__func__, atoi(inst));
 		return -ENODEV;
@@ -1734,13 +2229,13 @@ static int imsm_open_new(struct supertype *c, struct active_array *a,
 	return 0;
 }
 
-static __u8 imsm_check_degraded(struct imsm_super *mpb, int n, int failed)
+static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed)
 {
-	struct imsm_dev *dev = get_imsm_dev(mpb, n);
-	struct imsm_map *map = dev->vol.map;
+	struct imsm_map *map = get_imsm_map(dev, 0);
 
 	if (!failed)
-		return map->map_state;
+		return map->map_state == IMSM_T_STATE_UNINITIALIZED ? 
+			IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
 
 	switch (get_imsm_raid_level(map)) {
 	case 0:
@@ -1761,22 +2256,24 @@ static __u8 imsm_check_degraded(struct imsm_super *mpb, int n, int failed)
 		int device_per_mirror = 2; /* FIXME is this always the case?
 					    * and are they always adjacent?
 					    */
-		int failed = 0;
+		int r10fail = 0;
 		int i;
 
 		for (i = 0; i < map->num_members; i++) {
-			int idx = get_imsm_disk_idx(map, i);
-			struct imsm_disk *disk = get_imsm_disk(mpb, idx);
+			int idx = get_imsm_disk_idx(dev, i);
+			struct imsm_disk *disk = get_imsm_disk(super, idx);
 
-			if (__le32_to_cpu(disk->status) & FAILED_DISK)
-				failed++;
+			if (!disk)
+				r10fail++;
+			else if (__le32_to_cpu(disk->status) & FAILED_DISK)
+				r10fail++;
 
-			if (failed >= device_per_mirror)
+			if (r10fail >= device_per_mirror)
 				return IMSM_T_STATE_FAILED;
 
-			/* reset 'failed' for next mirror set */
+			/* reset 'r10fail' for next mirror set */
 			if (!((i + 1) % device_per_mirror))
-				failed = 0;
+				r10fail = 0;
 		}
 
 		return IMSM_T_STATE_DEGRADED;
@@ -1794,65 +2291,126 @@ static __u8 imsm_check_degraded(struct imsm_super *mpb, int n, int failed)
 	return map->map_state;
 }
 
-static int imsm_count_failed(struct imsm_super *mpb, struct imsm_map *map)
+static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev)
 {
 	int i;
 	int failed = 0;
 	struct imsm_disk *disk;
+	struct imsm_map *map = get_imsm_map(dev, 0);
 
 	for (i = 0; i < map->num_members; i++) {
-		int idx = get_imsm_disk_idx(map, i);
+		__u32 ord = get_imsm_ord_tbl_ent(dev, i);
+		int idx = ord_to_idx(ord);
 
-		disk = get_imsm_disk(mpb, idx);
-		if (__le32_to_cpu(disk->status) & FAILED_DISK)
+		disk = get_imsm_disk(super, idx);
+		if (!disk ||
+		    __le32_to_cpu(disk->status) & FAILED_DISK ||
+		    ord & IMSM_ORD_REBUILD)
 			failed++;
 	}
 
 	return failed;
 }
 
-static void imsm_set_array_state(struct active_array *a, int consistent)
+static int is_resyncing(struct imsm_dev *dev)
+{
+	struct imsm_map *migr_map;
+
+	if (!dev->vol.migr_state)
+		return 0;
+
+	if (dev->vol.migr_type == 0)
+		return 1;
+
+	migr_map = get_imsm_map(dev, 1);
+
+	if (migr_map->map_state == IMSM_T_STATE_NORMAL)
+		return 1;
+	else
+		return 0;
+}
+
+static int is_rebuilding(struct imsm_dev *dev)
+{
+	struct imsm_map *migr_map;
+
+	if (!dev->vol.migr_state)
+		return 0;
+
+	if (dev->vol.migr_type == 0)
+		return 0;
+
+	migr_map = get_imsm_map(dev, 1);
+
+	if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
+		return 1;
+	else
+		return 0;
+}
+
+/* Handle dirty -> clean transititions and resync.  Degraded and rebuild
+ * states are handled in imsm_set_disk() with one exception, when a
+ * resync is stopped due to a new failure this routine will set the
+ * 'degraded' state for the array.
+ */
+static int imsm_set_array_state(struct active_array *a, int consistent)
 {
 	int inst = a->info.container_member;
 	struct intel_super *super = a->container->sb;
-	struct imsm_dev *dev = get_imsm_dev(super->mpb, inst);
-	struct imsm_map *map = &dev->vol.map[0];
-	int dirty = !consistent;
-	int failed;
-	__u8 map_state;
+	struct imsm_dev *dev = get_imsm_dev(super, inst);
+	struct imsm_map *map = get_imsm_map(dev, 0);
+	int failed = imsm_count_failed(super, dev);
+	__u8 map_state = imsm_check_degraded(super, dev, failed);
+
+	if (consistent == 2 &&
+	    (a->resync_start != ~0ULL ||
+	     map_state != IMSM_T_STATE_NORMAL ||
+	     dev->vol.migr_state))
+		consistent = 0;
 
 	if (a->resync_start == ~0ULL) {
-		failed = imsm_count_failed(super->mpb, map);
-		map_state = imsm_check_degraded(super->mpb, inst, failed);
-		if (!failed)
-			map_state = IMSM_T_STATE_NORMAL;
-		if (map->map_state != map_state) {
-			dprintf("imsm: map_state %d: %d\n",
-				inst, map_state);
+		/* complete intialization / resync,
+		 * recovery is completed in ->set_disk
+		 */
+		if (is_resyncing(dev)) {
+			dprintf("imsm: mark resync done\n");
+			dev->vol.migr_state = 0;
 			map->map_state = map_state;
 			super->updates_pending++;
 		}
+	} else if (!is_resyncing(dev) && !failed) {
+		/* mark the start of the init process if nothing is failed */
+		dprintf("imsm: mark resync start (%llu)\n", a->resync_start);
+		map->map_state = map_state;
+		migrate(dev, IMSM_T_STATE_NORMAL,
+			map->map_state == IMSM_T_STATE_NORMAL);
+		super->updates_pending++;
 	}
 
-	if (dev->vol.dirty != dirty) {
+	/* mark dirty / clean */
+	if (dev->vol.dirty != !consistent) {
 		dprintf("imsm: mark '%s' (%llu)\n",
-			dirty?"dirty":"clean", a->resync_start);
-
-		dev->vol.dirty = dirty;
+			consistent ? "clean" : "dirty", a->resync_start);
+		if (consistent)
+			dev->vol.dirty = 0;
+		else
+			dev->vol.dirty = 1;
 		super->updates_pending++;
 	}
+	return consistent;
 }
 
 static void imsm_set_disk(struct active_array *a, int n, int state)
 {
 	int inst = a->info.container_member;
 	struct intel_super *super = a->container->sb;
-	struct imsm_dev *dev = get_imsm_dev(super->mpb, inst);
-	struct imsm_map *map = dev->vol.map;
+	struct imsm_dev *dev = get_imsm_dev(super, inst);
+	struct imsm_map *map = get_imsm_map(dev, 0);
 	struct imsm_disk *disk;
+	int failed;
 	__u32 status;
-	int failed = 0;
-	int new_failure = 0;
+	__u32 ord;
+	__u8 map_state;
 
 	if (n > map->num_members)
 		fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n",
@@ -1863,44 +2421,52 @@ static void imsm_set_disk(struct active_array *a, int n, int state)
 
 	dprintf("imsm: set_disk %d:%x\n", n, state);
 
-	disk = get_imsm_disk(super->mpb, get_imsm_disk_idx(map, n));
+	ord = get_imsm_ord_tbl_ent(dev, n);
+	disk = get_imsm_disk(super, ord_to_idx(ord));
 
 	/* check for new failures */
 	status = __le32_to_cpu(disk->status);
 	if ((state & DS_FAULTY) && !(status & FAILED_DISK)) {
 		status |= FAILED_DISK;
 		disk->status = __cpu_to_le32(status);
-		new_failure = 1;
+		disk->scsi_id = __cpu_to_le32(~(__u32)0);
+		memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
 		super->updates_pending++;
 	}
+	/* check if in_sync */
+	if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD) {
+		struct imsm_map *migr_map = get_imsm_map(dev, 1);
 
-	/* the number of failures have changed, count up 'failed' to determine
-	 * degraded / failed status
-	 */
-	if (new_failure && map->map_state != IMSM_T_STATE_FAILED)
-		failed = imsm_count_failed(super->mpb, map);
-
-	/* determine map_state based on failed or in_sync count */
-	if (failed)
-		map->map_state = imsm_check_degraded(super->mpb, inst, failed);
-	else if (map->map_state == IMSM_T_STATE_DEGRADED) {
-		struct mdinfo *d;
-		int working = 0;
+		set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
+		super->updates_pending++;
+	}
 
-		for (d = a->info.devs ; d ; d = d->next)
-			if (d->curr_state & DS_INSYNC)
-				working++;
+	failed = imsm_count_failed(super, dev);
+	map_state = imsm_check_degraded(super, dev, failed);
 
-		if (working == a->info.array.raid_disks) {
-			map->map_state = IMSM_T_STATE_NORMAL;
-			super->updates_pending++;
-		}
+	/* check if recovery complete, newly degraded, or failed */
+	if (map_state == IMSM_T_STATE_NORMAL && is_rebuilding(dev)) {
+		map->map_state = map_state;
+		dev->vol.migr_state = 0;
+		super->updates_pending++;
+	} else if (map_state == IMSM_T_STATE_DEGRADED &&
+		   map->map_state != map_state &&
+		   !dev->vol.migr_state) {
+		dprintf("imsm: mark degraded\n");
+		map->map_state = map_state;
+		super->updates_pending++;
+	} else if (map_state == IMSM_T_STATE_FAILED &&
+		   map->map_state != map_state) {
+		dprintf("imsm: mark failed\n");
+		dev->vol.migr_state = 0;
+		map->map_state = map_state;
+		super->updates_pending++;
 	}
 }
 
 static int store_imsm_mpb(int fd, struct intel_super *super)
 {
-	struct imsm_super *mpb = super->mpb;
+	struct imsm_super *mpb = super->anchor;
 	__u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
 	unsigned long long dsize;
 	unsigned long long sectors;
@@ -1941,25 +2507,123 @@ static void imsm_sync_metadata(struct supertype *container)
 	super->updates_pending = 0;
 }
 
+static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
+{
+	struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
+	int i = get_imsm_disk_idx(dev, idx);
+	struct dl *dl;
+
+	for (dl = super->disks; dl; dl = dl->next)
+		if (dl->index == i)
+			break;
+
+	if (dl && __le32_to_cpu(dl->disk.status) & FAILED_DISK)
+		dl = NULL;
+
+	if (dl)
+		dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor);
+
+	return dl;
+}
+
+static struct dl *imsm_add_spare(struct intel_super *super, int slot, struct active_array *a)
+{
+	struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
+	int idx = get_imsm_disk_idx(dev, slot);
+	struct imsm_map *map = get_imsm_map(dev, 0);
+	unsigned long long esize;
+	unsigned long long pos;
+	struct mdinfo *d;
+	struct extent *ex;
+	int j;
+	int found;
+	__u32 array_start;
+	__u32 status;
+	struct dl *dl;
+
+	for (dl = super->disks; dl; dl = dl->next) {
+		/* If in this array, skip */
+		for (d = a->info.devs ; d ; d = d->next)
+			if (d->state_fd >= 0 &&
+			    d->disk.major == dl->major &&
+			    d->disk.minor == dl->minor) {
+				dprintf("%x:%x already in array\n", dl->major, dl->minor);
+				break;
+			}
+		if (d)
+			continue;
+
+		/* skip in use or failed drives */
+		status = __le32_to_cpu(dl->disk.status);
+		if (status & FAILED_DISK || idx == dl->index) {
+			dprintf("%x:%x status ( %s%s)\n",
+			dl->major, dl->minor,
+			status & FAILED_DISK ? "failed " : "",
+			idx == dl->index ? "in use " : "");
+			continue;
+		}
+
+		/* Does this unused device have the requisite free space?
+		 * We need a->info.component_size sectors
+		 */
+		ex = get_extents(super, dl);
+		if (!ex) {
+			dprintf("cannot get extents\n");
+			continue;
+		}
+		found = 0;
+		j = 0;
+		pos = 0;
+		array_start = __le32_to_cpu(map->pba_of_lba0);
+
+		do {
+			/* check that we can start at pba_of_lba0 with
+			 * a->info.component_size of space
+			 */
+			esize = ex[j].start - pos;
+			if (array_start >= pos &&
+			    array_start + a->info.component_size < ex[j].start) {
+				found = 1;
+				break;
+			}
+			pos = ex[j].start + ex[j].size;
+			j++;
+			    
+		} while (ex[j-1].size);
+
+		free(ex);
+		if (!found) {
+			dprintf("%x:%x does not have %llu at %d\n",
+				dl->major, dl->minor,
+				a->info.component_size,
+				__le32_to_cpu(map->pba_of_lba0));
+			/* No room */
+			continue;
+		} else
+			break;
+	}
+
+	return dl;
+}
+
 static struct mdinfo *imsm_activate_spare(struct active_array *a,
 					  struct metadata_update **updates)
 {
 	/**
-	 * Take a device that is marked spare in the metadata and use it to
-	 * replace a failed/vacant slot in an array.  There may be a case where
-	 * a device is failed in one array but active in a second.
-	 * imsm_process_update catches this case and does not clear the SPARE_DISK
-	 * flag, allowing the second array to start using the device on failure.
-	 * SPARE_DISK is cleared when all arrays are using a device.
+	 * Find a device with unused free space and use it to replace a
+	 * failed/vacant region in an array.  We replace failed regions one a
+	 * array at a time.  The result is that a new spare disk will be added
+	 * to the first failed array and after the monitor has finished
+	 * propagating failures the remainder will be consumed.
 	 *
-	 * FIXME: is this a valid use of SPARE_DISK?
+	 * FIXME add a capability for mdmon to request spares from another
+	 * container.
 	 */
 
 	struct intel_super *super = a->container->sb;
-	struct imsm_super *mpb = super->mpb;
 	int inst = a->info.container_member;
-	struct imsm_dev *dev = get_imsm_dev(mpb, inst);
-	struct imsm_map *map = dev->vol.map;
+	struct imsm_dev *dev = get_imsm_dev(super, inst);
+	struct imsm_map *map = get_imsm_map(dev, 0);
 	int failed = a->info.array.raid_disks;
 	struct mdinfo *rv = NULL;
 	struct mdinfo *d;
@@ -1981,11 +2645,10 @@ static struct mdinfo *imsm_activate_spare(struct active_array *a,
 
 	dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
 		inst, failed, a->info.array.raid_disks, a->info.array.level);
-	if (imsm_check_degraded(mpb, inst, failed) != IMSM_T_STATE_DEGRADED)
+	if (imsm_check_degraded(super, dev, failed) != IMSM_T_STATE_DEGRADED)
 		return NULL;
 
 	/* For each slot, if it is not working, find a spare */
-	dl = super->disks;
 	for (i = 0; i < a->info.array.raid_disks; i++) {
 		for (d = a->info.devs ; d ; d = d->next)
 			if (d->disk.raid_disk == i)
@@ -1994,88 +2657,47 @@ static struct mdinfo *imsm_activate_spare(struct active_array *a,
 		if (d && (d->state_fd >= 0))
 			continue;
 
-		/* OK, this device needs recovery.  Find a spare */
-		for ( ; dl ; dl = dl->next) {
-			unsigned long long esize;
-			unsigned long long pos;
-			struct mdinfo *d2;
-			struct extent *ex;
-			struct imsm_disk *disk;
-			int j;
-			int found;
-			__u32 array_start;
-
-			/* If in this array, skip */
-			for (d2 = a->info.devs ; d2 ; d2 = d2->next)
-				if (d2->disk.major == dl->major &&
-				    d2->disk.minor == dl->minor) {
-					dprintf("%x:%x already in array\n", dl->major, dl->minor);
-					break;
-				}
-			if (d2)
-				continue;
-
-			/* is this unused device marked as a spare? */
-			disk = get_imsm_disk(mpb, dl->index);
-			if (!(__le32_to_cpu(disk->status) & SPARE_DISK))
-				continue;
-
-			/* We are allowed to use this device - is there space?
-			 * We need a->info.component_size sectors */
-			ex = get_extents(super, dl);
-			if (!ex) {
-				dprintf("cannot get extents\n");
-				continue;
-			}
-			found = 0;
-			j = 0;
-			pos = 0;
-			array_start = __le32_to_cpu(map->pba_of_lba0);
-
-			do {
-				/* check that we can start at pba_of_lba0 with
-				 * a->info.component_size of space
-				 */
-				esize = ex[j].start - pos;
-				if (array_start >= pos &&
-				    array_start + a->info.component_size < ex[j].start) {
-			    		found = 1;
-					break;
-				}
-				pos = ex[j].start + ex[j].size;
-				j++;
-				    
-			} while (ex[j-1].size);
-
-			free(ex);
-			if (!found) {
-				dprintf("%x:%x does not have %llu at %d\n",
-					dl->major, dl->minor,
-					a->info.component_size,
-					__le32_to_cpu(map->pba_of_lba0));
-				/* No room */
-				continue;
-			}
+		/*
+		 * OK, this device needs recovery.  Try to re-add the previous
+		 * occupant of this slot, if this fails add a new spare
+		 */
+		dl = imsm_readd(super, i, a);
+		if (!dl)
+			dl = imsm_add_spare(super, i, a);
+		if (!dl)
+			continue;
+ 
+		/* found a usable disk with enough space */
+		di = malloc(sizeof(*di));
+		memset(di, 0, sizeof(*di));
+
+		/* dl->index will be -1 in the case we are activating a
+		 * pristine spare.  imsm_process_update() will create a
+		 * new index in this case.  Once a disk is found to be
+		 * failed in all member arrays it is kicked from the
+		 * metadata
+		 */
+		di->disk.number = dl->index;
 
-			/* found a usable disk with enough space */
-			di = malloc(sizeof(*di));
-			memset(di, 0, sizeof(*di));
-			di->disk.number = dl->index;
-			di->disk.raid_disk = i;
-			di->disk.major = dl->major;
-			di->disk.minor = dl->minor;
-			di->disk.state = 0;
-			di->data_offset = array_start;
-			di->component_size = a->info.component_size;
-			di->container_member = inst;
-			di->next = rv;
-			rv = di;
-			num_spares++;
-			dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
-				i, pos);
+		/* (ab)use di->devs to store a pointer to the device
+		 * we chose
+		 */
+		di->devs = (struct mdinfo *) dl;
+
+		di->disk.raid_disk = i;
+		di->disk.major = dl->major;
+		di->disk.minor = dl->minor;
+		di->disk.state = 0;
+		di->data_offset = __le32_to_cpu(map->pba_of_lba0);
+		di->component_size = a->info.component_size;
+		di->container_member = inst;
+		di->next = rv;
+		rv = di;
+		num_spares++;
+		dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
+			i, di->data_offset);
 
-			break;
-		}
+		break;
 	}
 
 	if (!rv)
@@ -2094,7 +2716,8 @@ static struct mdinfo *imsm_activate_spare(struct active_array *a,
 
 	for (di = rv ; di ; di = di->next) {
 		u->type = update_activate_spare;
-		u->disk_idx = di->disk.number; 
+		u->dl = (struct dl *) di->devs;
+		di->devs = NULL;
 		u->slot = di->disk.raid_disk;
 		u->array = inst;
 		u->next = u + 1;
@@ -2106,32 +2729,26 @@ static struct mdinfo *imsm_activate_spare(struct active_array *a,
 	return rv;
 }
 
-static int weight(unsigned int field)
-{
-	int weight;
-
-	for (weight = 0; field; weight++)
-		field &= field - 1;
-
-	return weight;
-}
-
-static int disks_overlap(struct imsm_map *m1, struct imsm_map *m2)
+static int disks_overlap(struct imsm_dev *d1, struct imsm_dev *d2)
 {
+	struct imsm_map *m1 = get_imsm_map(d1, 0);
+	struct imsm_map *m2 = get_imsm_map(d2, 0);
 	int i;
 	int j;
 	int idx;
 
 	for (i = 0; i < m1->num_members; i++) {
-		idx = get_imsm_disk_idx(m1, i);
+		idx = get_imsm_disk_idx(d1, i);
 		for (j = 0; j < m2->num_members; j++)
-			if (idx == get_imsm_disk_idx(m2, j))
+			if (idx == get_imsm_disk_idx(d2, j))
 				return 1;
 	}
 
 	return 0;
 }
 
+static void imsm_delete(struct intel_super *super, struct dl **dlp, int index);
+
 static void imsm_process_update(struct supertype *st,
 			        struct metadata_update *update)
 {
@@ -2144,86 +2761,110 @@ static void imsm_process_update(struct supertype *st,
 	 * 	flag
 	 */
 	struct intel_super *super = st->sb;
-	struct imsm_super *mpb = super->mpb;
+	struct imsm_super *mpb;
 	enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
 
+	/* update requires a larger buf but the allocation failed */
+	if (super->next_len && !super->next_buf) {
+		super->next_len = 0;
+		return;
+	}
+
+	if (super->next_buf) {
+		memcpy(super->next_buf, super->buf, super->len);
+		free(super->buf);
+		super->len = super->next_len;
+		super->buf = super->next_buf;
+
+		super->next_len = 0;
+		super->next_buf = NULL;
+	}
+
+	mpb = super->anchor;
+
 	switch (type) {
 	case update_activate_spare: {
 		struct imsm_update_activate_spare *u = (void *) update->buf; 
-		struct imsm_dev *dev = get_imsm_dev(mpb, u->array);
-		struct imsm_map *map = &dev->vol.map[0];
+		struct imsm_dev *dev = get_imsm_dev(super, u->array);
+		struct imsm_map *map = get_imsm_map(dev, 0);
+		struct imsm_map *migr_map;
 		struct active_array *a;
 		struct imsm_disk *disk;
 		__u32 status;
+		__u8 to_state;
 		struct dl *dl;
-		struct mdinfo *d;
-		unsigned int members;
 		unsigned int found;
-		int victim;
+		int failed;
+		int victim = get_imsm_disk_idx(dev, u->slot);
 		int i;
 
 		for (dl = super->disks; dl; dl = dl->next)
-			if (dl->index == u->disk_idx)
+			if (dl == u->dl)
 				break;
 
 		if (!dl) {
 			fprintf(stderr, "error: imsm_activate_spare passed "
-				"an unknown disk_idx: %d\n", u->disk_idx);
+				"an unknown disk (index: %d)\n",
+				u->dl->index);
 			return;
 		}
 
 		super->updates_pending++;
 
-		victim = get_imsm_disk_idx(map, u->slot);
-		map->disk_ord_tbl[u->slot] = __cpu_to_le32(u->disk_idx);
-		disk = get_imsm_disk(mpb, u->disk_idx);
+		/* count failures (excluding rebuilds and the victim)
+		 * to determine map[0] state
+		 */
+		failed = 0;
+		for (i = 0; i < map->num_members; i++) {
+			if (i == u->slot)
+				continue;
+			disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i));
+			if (!disk ||
+			    __le32_to_cpu(disk->status) & FAILED_DISK)
+				failed++;
+		}
+
+		/* adding a pristine spare, assign a new index */
+		if (dl->index < 0) {
+			dl->index = super->anchor->num_disks;
+			super->anchor->num_disks++;
+		}
+		disk = &dl->disk;
 		status = __le32_to_cpu(disk->status);
 		status |= CONFIGURED_DISK;
+		status &= ~SPARE_DISK;
 		disk->status = __cpu_to_le32(status);
 
-		/* map unique/live arrays using the spare */
-		members = 0;
-		found = 0;
-		for (a = st->arrays; a; a = a->next) {
-			int inst = a->info.container_member;
-
-			dev = get_imsm_dev(mpb, inst);
-			map = &dev->vol.map[0];
-			if (map->raid_level > 0)
-				members |= 1 << inst;
-			for (d = a->info.devs; d; d = d->next)
-				if (d->disk.major == dl->major &&
-				    d->disk.minor == dl->minor)
-					found |= 1 << inst;
-		}
-
-		/* until all arrays that can absorb this disk have absorbed
-		 * this disk it can still be considered a spare
-		 */
-		if (weight(found) >= weight(members)) {
-			status = __le32_to_cpu(disk->status);
-			status &= ~SPARE_DISK;
-			disk->status = __cpu_to_le32(status);
-		}
+		/* mark rebuild */
+		to_state = imsm_check_degraded(super, dev, failed);
+		map->map_state = IMSM_T_STATE_DEGRADED;
+		migrate(dev, to_state, 1);
+		migr_map = get_imsm_map(dev, 1);
+		set_imsm_ord_tbl_ent(map, u->slot, dl->index);
+		set_imsm_ord_tbl_ent(migr_map, u->slot, dl->index | IMSM_ORD_REBUILD);
 
 		/* count arrays using the victim in the metadata */
 		found = 0;
 		for (a = st->arrays; a ; a = a->next) {
-			dev = get_imsm_dev(mpb, a->info.container_member);
-			map = &dev->vol.map[0];
+			dev = get_imsm_dev(super, a->info.container_member);
 			for (i = 0; i < map->num_members; i++)
-				if (victim == get_imsm_disk_idx(map, i))
+				if (victim == get_imsm_disk_idx(dev, i))
 					found++;
 		}
 
-		/* clear some flags if the victim is no longer being
+		/* delete the victim if it is no longer being
 		 * utilized anywhere
 		 */
-		disk = get_imsm_disk(mpb, victim);
 		if (!found) {
-			status = __le32_to_cpu(disk->status);
-			status &= ~(CONFIGURED_DISK | USABLE_DISK);
-			disk->status = __cpu_to_le32(status);
+			struct dl **dlp;
+
+			for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
+				if ((*dlp)->index == victim)
+					break;
+			/* We know that 'manager' isn't touching anything,
+			 * so it is safe to:
+			 */
+			imsm_delete(super, dlp, victim);
 		}
 		break;
 	}
@@ -2254,12 +2895,12 @@ static void imsm_process_update(struct supertype *st,
 
 		/* check update is next in sequence */
 		if (u->dev_idx != mpb->num_raid_devs) {
-			dprintf("%s: can not create arrays out of sequence\n",
-				__func__);
+			dprintf("%s: can not create array %d expected index %d\n",
+				__func__, u->dev_idx, mpb->num_raid_devs);
 			return;
 		}
 
-		new_map = &u->dev.vol.map[0];
+		new_map = get_imsm_map(&u->dev, 0);
 		new_start = __le32_to_cpu(new_map->pba_of_lba0);
 		new_end = new_start + __le32_to_cpu(new_map->blocks_per_member);
 
@@ -2268,14 +2909,14 @@ static void imsm_process_update(struct supertype *st,
 		 * overalpping disks
 		 */
 		for (i = 0; i < mpb->num_raid_devs; i++) {
-			dev = get_imsm_dev(mpb, i);
-			map = &dev->vol.map[0];
+			dev = get_imsm_dev(super, i);
+			map = get_imsm_map(dev, 0);
 			start = __le32_to_cpu(map->pba_of_lba0);
 			end = start + __le32_to_cpu(map->blocks_per_member);
 			if ((new_start >= start && new_start <= end) ||
 			    (start >= new_start && start <= new_end))
 				overlap = 1;
-			if (overlap && disks_overlap(map, new_map)) {
+			if (overlap && disks_overlap(dev, &u->dev)) {
 				dprintf("%s: arrays overlap\n", __func__);
 				return;
 			}
@@ -2286,51 +2927,158 @@ static void imsm_process_update(struct supertype *st,
 			return;
 		}
 
+		/* check that prepare update was successful */
+		if (!update->space) {
+			dprintf("%s: prepare update failed\n", __func__);
+			return;
+		}
+
 		super->updates_pending++;
+		dev = update->space;
+		map = get_imsm_map(dev, 0);
+		update->space = NULL;
+		imsm_copy_dev(dev, &u->dev);
+		map = get_imsm_map(dev, 0);
+		super->dev_tbl[u->dev_idx] = dev;
 		mpb->num_raid_devs++;
-		dev = get_imsm_dev(mpb, u->dev_idx);
-		memcpy(dev, &u->dev, sizeof(*dev));
-		map = &dev->vol.map[0];
-		memcpy(map->disk_ord_tbl, new_map->disk_ord_tbl,
-		       sizeof(__u32) * new_map->num_members);
-
-		/* fix up flags, if arrays overlap then the drives can not be
-		 * spares
-		 */
+
+		/* fix up flags */
 		for (i = 0; i < map->num_members; i++) {
 			struct imsm_disk *disk;
 			__u32 status;
 
-			disk = get_imsm_disk(mpb, get_imsm_disk_idx(map, i));
+			disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i));
 			status = __le32_to_cpu(disk->status);
 			status |= CONFIGURED_DISK;
-			if (overlap)
-				status &= ~SPARE_DISK;
+			status &= ~SPARE_DISK;
 			disk->status = __cpu_to_le32(status);
 		}
 		break;
 	}
+	case update_add_disk:
+
+		/* we may be able to repair some arrays if disks are
+		 * being added */
+		if (super->add) {
+			struct active_array *a;
+ 			for (a = st->arrays; a; a = a->next)
+				a->check_degraded = 1;
+		}
+		/* add some spares to the metadata */
+		while (super->add) {
+			struct dl *al;
+
+			al = super->add;
+			super->add = al->next;
+			al->next = super->disks;
+			super->disks = al;
+			dprintf("%s: added %x:%x\n",
+				__func__, al->major, al->minor);
+		}
+
+		break;
 	}
 }
 
 static void imsm_prepare_update(struct supertype *st,
 				struct metadata_update *update)
 {
-	/* Allocate space to hold a new mpb if necessary.  We currently
-	 * allocate enough to hold 2 subarrays for the given number of disks.
-	 * This may not be sufficient iff reshaping.
-	 *
-	 * FIX ME handle the reshape case.
-	 *
-	 * The monitor will be able to safely change super->mpb by arranging
-	 * for it to be freed in check_update_queue().  I.e. the monitor thread
-	 * will start using the new pointer and the manager can continue to use
-	 * the old value until check_update_queue() runs.
+	/**
+	 * Allocate space to hold new disk entries, raid-device entries or a new
+	 * mpb if necessary.  The manager synchronously waits for updates to
+	 * complete in the monitor, so new mpb buffers allocated here can be
+	 * integrated by the monitor thread without worrying about live pointers
+	 * in the manager thread.
 	 */
+	enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
+	struct intel_super *super = st->sb;
+	struct imsm_super *mpb = super->anchor;
+	size_t buf_len;
+	size_t len = 0;
+
+	switch (type) {
+	case update_create_array: {
+		struct imsm_update_create_array *u = (void *) update->buf;
 
-	return;
+		len = sizeof_imsm_dev(&u->dev, 1);
+		update->space = malloc(len);
+		break;
+	default:
+		break;
+	}
+	}
+
+	/* check if we need a larger metadata buffer */
+	if (super->next_buf)
+		buf_len = super->next_len;
+	else
+		buf_len = super->len;
+
+	if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) {
+		/* ok we need a larger buf than what is currently allocated
+		 * if this allocation fails process_update will notice that
+		 * ->next_len is set and ->next_buf is NULL
+		 */
+		buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512);
+		if (super->next_buf)
+			free(super->next_buf);
+
+		super->next_len = buf_len;
+		if (posix_memalign(&super->next_buf, buf_len, 512) != 0)
+			super->next_buf = NULL;
+	}
 }
 
+/* must be called while manager is quiesced */
+static void imsm_delete(struct intel_super *super, struct dl **dlp, int index)
+{
+	struct imsm_super *mpb = super->anchor;
+	struct dl *iter;
+	struct imsm_dev *dev;
+	struct imsm_map *map;
+	int i, j, num_members;
+	__u32 ord;
+
+	dprintf("%s: deleting device[%d] from imsm_super\n",
+		__func__, index);
+
+	/* shift all indexes down one */
+	for (iter = super->disks; iter; iter = iter->next)
+		if (iter->index > index)
+			iter->index--;
+
+	for (i = 0; i < mpb->num_raid_devs; i++) {
+		dev = get_imsm_dev(super, i);
+		map = get_imsm_map(dev, 0);
+		num_members = map->num_members;
+		for (j = 0; j < num_members; j++) {
+			/* update ord entries being careful not to propagate
+			 * ord-flags to the first map
+			 */
+			ord = get_imsm_ord_tbl_ent(dev, j);
+
+			if (ord_to_idx(ord) <= index)
+				continue;
+
+			map = get_imsm_map(dev, 0);
+			set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1));
+			map = get_imsm_map(dev, 1);
+			if (map)
+				set_imsm_ord_tbl_ent(map, j, ord - 1);
+		}
+	}
+
+	mpb->num_disks--;
+	super->updates_pending++;
+	if (*dlp) {
+		struct dl *dl = *dlp;
+
+		*dlp = (*dlp)->next;
+		__free_imsm_disk(dl);
+	}
+}
+#endif /* MDASSEMBLE */
+
 struct superswitch super_imsm = {
 #ifndef	MDASSEMBLE
 	.examine_super	= examine_super_imsm,
@@ -2338,6 +3086,8 @@ struct superswitch super_imsm = {
 	.detail_super	= detail_super_imsm,
 	.brief_detail_super = brief_detail_super_imsm,
 	.write_init_super = write_init_super_imsm,
+	.validate_geometry = validate_geometry_imsm,
+	.add_to_super	= add_to_super_imsm,
 #endif
 	.match_home	= match_home_imsm,
 	.uuid_from_super= uuid_from_super_imsm,
@@ -2350,15 +3100,14 @@ struct superswitch super_imsm = {
 
 	.load_super	= load_super_imsm,
 	.init_super	= init_super_imsm,
-	.add_to_super	= add_to_super_imsm,
 	.store_super	= store_zero_imsm,
 	.free_super	= free_super_imsm,
 	.match_metadata_desc = match_metadata_desc_imsm,
 	.container_content = container_content_imsm,
 
-	.validate_geometry = validate_geometry_imsm,
 	.external	= 1,
 
+#ifndef MDASSEMBLE
 /* for mdmon */
 	.open_new	= imsm_open_new,
 	.load_super	= load_super_imsm,
@@ -2368,4 +3117,5 @@ struct superswitch super_imsm = {
 	.activate_spare = imsm_activate_spare,
 	.process_update = imsm_process_update,
 	.prepare_update = imsm_prepare_update,
+#endif /* MDASSEMBLE */
 };