X-Git-Url: http://git.ipfire.org/?p=thirdparty%2Fmdadm.git;a=blobdiff_plain;f=super-ddf.c;h=aefe25ef25cb22f07e951fc2b7dbb37712894ffb;hp=2918b716c2e0a4d0a24c468a86cc7cd647fe5403;hb=edd8d13c0247b8df5876a94c2d61d74d2cf62c7e;hpb=2e735d198233a67f305862f72e3a5d0f0c3c548c

diff --git a/super-ddf.c b/super-ddf.c
index 2918b716..aefe25ef 100644
--- a/super-ddf.c
+++ b/super-ddf.c
@@ -71,7 +71,7 @@ unsigned long crc32(
 #define	DDF_CONCAT	0x1f
 #define	DDF_RAID5E	0x15
 #define	DDF_RAID5EE	0x25
-#define	DDF_RAID6	0x16	/* Vendor unique layout */
+#define	DDF_RAID6	0x06
 
 /* Raid Level Qualifier (RLQ) */
 #define	DDF_RAID0_SIMPLE	0x00
@@ -83,6 +83,7 @@ unsigned long crc32(
 #define	DDF_RAID4_N		0x01 /* parity in last extent */
 /* these apply to raid5e and raid5ee as well */
 #define	DDF_RAID5_0_RESTART	0x00 /* same as 'right asymmetric' - layout 1 */
+#define	DDF_RAID6_0_RESTART	0x01 /* raid6 different from raid5 here!!! */
 #define	DDF_RAID5_N_RESTART	0x02 /* same as 'left asymmetric' - layout 0 */
 #define	DDF_RAID5_N_CONTINUE	0x03 /* same as 'left symmetric' - layout 2 */
 
@@ -108,13 +109,14 @@ unsigned long crc32(
 #define	DDF_BBM_LOG_MAGIC	__cpu_to_be32(0xABADB10C)
 
 #define	DDF_GUID_LEN	24
-#define DDF_REVISION	"01.00.00"
+#define DDF_REVISION_0	"01.00.00"
+#define DDF_REVISION_2	"01.02.00"
 
 struct ddf_header {
-	__u32	magic;
+	__u32	magic;		/* DDF_HEADER_MAGIC */
 	__u32	crc;
 	char	guid[DDF_GUID_LEN];
-	char	revision[8];	/* 01.00.00 */
+	char	revision[8];	/* 01.02.00 */
 	__u32	seq;		/* starts at '1' */
 	__u32	timestamp;
 	__u8	openflag;
@@ -167,7 +169,7 @@ struct ddf_header {
 
 /* The content of the 'controller section' - global scope */
 struct ddf_controller_data {
-	__u32	magic;
+	__u32	magic;			/* DDF_CONTROLLER_MAGIC */
 	__u32	crc;
 	char	guid[DDF_GUID_LEN];
 	struct controller_type {
@@ -183,7 +185,7 @@ struct ddf_controller_data {
 
 /* The content of phys_section - global scope */
 struct phys_disk {
-	__u32	magic;
+	__u32	magic;		/* DDF_PHYS_RECORDS_MAGIC */
 	__u32	crc;
 	__u16	used_pdes;
 	__u16	max_pdes;
@@ -202,7 +204,7 @@ struct phys_disk {
 /* phys_disk_entry.type is a bitmap - bigendian remember */
 #define	DDF_Forced_PD_GUID		1
 #define	DDF_Active_in_VD		2
-#define	DDF_Global_Spare		4
+#define	DDF_Global_Spare		4 /* VD_CONF records are ignored */
 #define	DDF_Spare			8 /* overrides Global_spare */
 #define	DDF_Foreign			16
 #define	DDF_Legacy			32 /* no DDF on this device */
@@ -224,7 +226,7 @@ struct phys_disk {
 
 /* The content of the virt_section global scope */
 struct virtual_disk {
-	__u32	magic;
+	__u32	magic;		/* DDF_VIRT_RECORDS_MAGIC */
 	__u32	crc;
 	__u16	populated_vdes;
 	__u16	max_vdes;
@@ -278,7 +280,7 @@ struct virtual_disk {
  */
 
 struct vd_config {
-	__u32	magic;
+	__u32	magic;		/* DDF_VD_CONF_MAGIC */
 	__u32	crc;
 	char	guid[DDF_GUID_LEN];
 	__u32	timestamp;
@@ -322,7 +324,7 @@ struct vd_config {
 #define	DDF_cache_rallowed	64	/* enable read caching */
 
 struct spare_assign {
-	__u32	magic;
+	__u32	magic;		/* DDF_SPARE_ASSIGN_MAGIC */
 	__u32	crc;
 	__u32	timestamp;
 	__u8	reserved[7];
@@ -344,7 +346,7 @@ struct spare_assign {
 
 /* The data_section contents - local scope */
 struct disk_data {
-	__u32	magic;
+	__u32	magic;		/* DDF_PHYS_DATA_MAGIC */
 	__u32	crc;
 	char	guid[DDF_GUID_LEN];
 	__u32	refnum;		/* crc of some magic drive data ... */
@@ -399,19 +401,23 @@ struct ddf_super {
 	struct virtual_disk	*virt;
 	int pdsize, vdsize;
 	int max_part, mppe, conf_rec_len;
+	int currentdev;
 	struct vcl {
 		struct vcl	*next;
 		__u64		*lba_offset; /* location in 'conf' of
 					      * the lba table */
+		int	vcnum; /* index into ->virt */
+		__u64		*block_sizes; /* NULL if all the same */
 		struct vd_config conf;
-	} *conflist, *newconf;
-	int conf_num; /* Index into 'virt' of entry matching 'newconf' */
+	} *conflist, *currentconf;
 	struct dl {
 		struct dl	*next;
 		struct disk_data disk;
 		int major, minor;
 		char *devname;
 		int fd;
+		unsigned long long size; /* sectors */
+		int pdnum;	/* index in ->phys */
 		struct spare_assign *spare;
 		struct vcl *vlist[0]; /* max_part in size */
 	} *dlist;
@@ -421,7 +427,6 @@ struct ddf_super {
 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
 #endif
 
-extern struct superswitch super_ddf_container, super_ddf_bvd, super_ddf;
 
 static int calc_crc(void *buf, int len)
 {
@@ -546,11 +551,12 @@ static int load_ddf_headers(int fd, struct ddf_super *super, char *devname)
 				devname);
 		return 2;
 	}
-	if (memcmp(super->anchor.revision, DDF_REVISION, 8) != 0) {
+	if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 &&
+	    memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) {
 		if (devname)
 			fprintf(stderr, Name ": can only support super revision"
-				" %.8s, not %.8s on %s\n",
-				DDF_REVISION, super->anchor.revision, devname);
+				" %.8s and earlier, not %.8s on %s\n",
+				DDF_REVISION_2, super->anchor.revision,devname);
 		return 2;
 	}
 	if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba),
@@ -622,6 +628,8 @@ static int load_ddf_local(int fd, struct ddf_super *super,
 	char *conf;
 	int i;
 	int vnum;
+	int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries);
+	unsigned long long dsize;
 
 	/* First the local disk info */
 	dl = malloc(sizeof(*dl) +
@@ -638,10 +646,19 @@ static int load_ddf_local(int fd, struct ddf_super *super,
 	dl->minor = minor(stb.st_rdev);
 	dl->next = super->dlist;
 	dl->fd = keep ? fd : -1;
+
+	dl->size = 0;
+	if (get_dev_size(fd, devname, &dsize))
+		dl->size = dsize >> 9;
 	dl->spare = NULL;
 	for (i=0 ; i < super->max_part ; i++)
 		dl->vlist[i] = NULL;
 	super->dlist = dl;
+	dl->pdnum = -1;
+	for (i=0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
+		if (memcmp(super->phys->entries[i].guid,
+			   dl->disk.guid, DDF_GUID_LEN) == 0)
+			dl->pdnum = i;
 
 	/* Now the config list. */
 	/* 'conf' is an array of config entries, some of which are
@@ -682,16 +699,24 @@ static int load_ddf_local(int fd, struct ddf_super *super,
 			if (__be32_to_cpu(vd->seqnum) <=
 			    __be32_to_cpu(vcl->conf.seqnum))
 				continue;
- 		} else {
+		} else {
 			vcl = malloc(super->conf_rec_len*512 +
 				     offsetof(struct vcl, conf));
 			vcl->next = super->conflist;
+			vcl->block_sizes = NULL; /* FIXME not for CONCAT */
 			super->conflist = vcl;
 			dl->vlist[vnum++] = vcl;
 		}
 		memcpy(&vcl->conf, vd, super->conf_rec_len*512);
 		vcl->lba_offset = (__u64*)
 			&vcl->conf.phys_refnum[super->mppe];
+
+		for (i=0; i < max_virt_disks ; i++)
+			if (memcmp(super->virt->entries[i].guid,
+				   vcl->conf.guid, DDF_GUID_LEN)==0)
+				break;
+		if (i < max_virt_disks)
+			vcl->vcnum = i;
 	}
 	free(conf);
 
@@ -710,9 +735,12 @@ static int load_super_ddf(struct supertype *st, int fd,
 	int rv;
 
 #ifndef MDASSEMBLE
+	/* if 'fd' is a container, load metadata from all the devices */
 	if (load_super_ddf_all(st, fd, &st->sb, devname, 1) == 0)
 		return 0;
 #endif
+	if (st->subarray[0])
+		return 1; /* FIXME Is this correct */
 
 	if (get_dev_size(fd, devname, &dsize) == 0)
 		return 1;
@@ -788,6 +816,8 @@ static void free_super_ddf(struct supertype *st)
 	while (ddf->conflist) {
 		struct vcl *v = ddf->conflist;
 		ddf->conflist = v->next;
+		if (v->block_sizes)
+			free(v->block_sizes);
 		free(v);
 	}
 	while (ddf->dlist) {
@@ -813,6 +843,7 @@ static struct supertype *match_metadata_desc_ddf(char *arg)
 		return NULL;
 
 	st = malloc(sizeof(*st));
+	memset(st, 0, sizeof(*st));
 	st->ss = &super_ddf;
 	st->max_devs = 512;
 	st->minor_version = 0;
@@ -820,38 +851,6 @@ static struct supertype *match_metadata_desc_ddf(char *arg)
 	return st;
 }
 
-static struct supertype *match_metadata_desc_ddf_bvd(char *arg)
-{
-	struct supertype *st;
-	if (strcmp(arg, "ddf/bvd") != 0 &&
-	    strcmp(arg, "bvd") != 0 &&
-	    strcmp(arg, "default") != 0
-		)
-		return NULL;
-
-	st = malloc(sizeof(*st));
-	st->ss = &super_ddf_bvd;
-	st->max_devs = 512;
-	st->minor_version = 0;
-	st->sb = NULL;
-	return st;
-}
-static struct supertype *match_metadata_desc_ddf_svd(char *arg)
-{
-	struct supertype *st;
-	if (strcmp(arg, "ddf/svd") != 0 &&
-	    strcmp(arg, "svd") != 0 &&
-	    strcmp(arg, "default") != 0
-		)
-		return NULL;
-
-	st = malloc(sizeof(*st));
-	st->ss = &super_ddf_svd;
-	st->max_devs = 512;
-	st->minor_version = 0;
-	st->sb = NULL;
-	return st;
-}
 
 #ifndef MDASSEMBLE
 
@@ -937,29 +936,36 @@ static void print_guid(char *guid, int tstamp)
 {
 	/* A GUIDs are part (or all) ASCII and part binary.
 	 * They tend to be space padded.
-	 * We ignore trailing spaces and print numbers
-	 * <0x20 and >=0x7f as \xXX
-	 * Some GUIDs have a time stamp in bytes 16-19.
-	 * We print that if appropriate
+	 * We print the GUID in HEX, then in parentheses add
+	 * any initial ASCII sequence, and a possible
+	 * time stamp from bytes 16-19
 	 */
 	int l = DDF_GUID_LEN;
 	int i;
+
+	for (i=0 ; i<DDF_GUID_LEN ; i++) {
+		if ((i&3)==0 && i != 0) printf(":");
+		printf("%02X", guid[i]&255);
+	}
+
+	printf(" (");
 	while (l && guid[l-1] == ' ')
 		l--;
 	for (i=0 ; i<l ; i++) {
 		if (guid[i] >= 0x20 && guid[i] < 0x7f)
 			fputc(guid[i], stdout);
 		else
-			fprintf(stdout, "\\x%02x", guid[i]&255);
+			break;
 	}
 	if (tstamp) {
 		time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
 		char tbuf[100];
 		struct tm *tm;
 		tm = localtime(&then);
-		strftime(tbuf, 100, " (%D %T)",tm);
+		strftime(tbuf, 100, " %D %T",tm);
 		fputs(tbuf, stdout);
 	}
+	printf(")");
 }
 
 static void examine_vd(int n, struct ddf_super *sb, char *guid)
@@ -1088,11 +1094,11 @@ static void brief_examine_super_ddf(struct supertype *st)
 	 */
 	struct ddf_super *ddf = st->sb;
 	int i;
-	printf("ARRAY /dev/ddf UUID=");
+	printf("ARRAY /dev/ddf metadata=ddf UUID=");
 	for (i = 0; i < DDF_GUID_LEN; i++) {
-		printf("%02x", ddf->anchor.guid[i]);
 		if ((i&3) == 0 && i != 0)
 			printf(":");
+		printf("%02X", 255&ddf->anchor.guid[i]);
 	}
 	printf("\n");
 }
@@ -1114,8 +1120,6 @@ static void brief_detail_super_ddf(struct supertype *st)
 	 */
 //	struct ddf_super *ddf = st->sb;
 }
-
-
 #endif
 
 static int match_home_ddf(struct supertype *st, char *homehost)
@@ -1136,12 +1140,9 @@ static int match_home_ddf(struct supertype *st, char *homehost)
 static struct vd_config *find_vdcr(struct ddf_super *ddf, int inst)
 {
 	struct vcl *v;
-	if (inst < 0 || inst > __be16_to_cpu(ddf->virt->populated_vdes))
-		return NULL;
+
 	for (v = ddf->conflist; v; v = v->next)
-		if (memcmp(v->conf.guid,
-			   ddf->virt->entries[inst].guid,
-			   DDF_GUID_LEN) == 0)
+		if (inst == v->vcnum)
 			return &v->conf;
 	return NULL;
 }
@@ -1177,30 +1178,33 @@ static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
 	 * The first 16 bytes of the sha1 of these is used.
 	 */
 	struct ddf_super *ddf = st->sb;
-	struct vd_config *vd = find_vdcr(ddf, ddf->conf_num);
+	struct vcl *vcl = ddf->currentconf;
 
-	if (!vd)
+	if (!vcl)
 		memset(uuid, 0, sizeof (uuid));
 	else {
 		char buf[20];
 		struct sha1_ctx ctx;
 		sha1_init_ctx(&ctx);
-		sha1_process_bytes(&vd->guid, DDF_GUID_LEN, &ctx);
-		if (vd->sec_elmnt_count > 1)
-			sha1_process_bytes(&vd->sec_elmnt_seq, 1, &ctx);
+		sha1_process_bytes(&vcl->conf.guid, DDF_GUID_LEN, &ctx);
+		if (vcl->conf.sec_elmnt_count > 1)
+			sha1_process_bytes(&vcl->conf.sec_elmnt_seq, 1, &ctx);
 		sha1_finish_ctx(&ctx, buf);
 		memcpy(uuid, buf, sizeof(uuid));
 	}
 }
 
+static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info);
+
 static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
 {
 	struct ddf_super *ddf = st->sb;
-	int i;
 
-	info->array.major_version = 1000;
-	info->array.minor_version = 0; /* FIXME use ddf->revision somehow */
-	info->array.patch_version = 0;
+	if (ddf->currentconf) {
+		getinfo_super_ddf_bvd(st, info);
+		return;
+	}
+
 	info->array.raid_disks    = __be16_to_cpu(ddf->phys->used_pdes);
 	info->array.level	  = LEVEL_CONTAINER;
 	info->array.layout	  = 0;
@@ -1210,26 +1214,24 @@ static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
 	info->array.utime	  = 0;
 	info->array.chunk_size	  = 0;
 
-//	info->data_offset	  = ???;
-//	info->component_size	  = ???;
 
 	info->disk.major = 0;
 	info->disk.minor = 0;
 	if (ddf->dlist) {
 		info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
-		info->disk.raid_disk = -1;
-		for (i = 0; i < __be16_to_cpu(ddf->phys->max_pdes) ; i++)
-			if (ddf->phys->entries[i].refnum ==
-			    ddf->dlist->disk.refnum) {
-				info->disk.raid_disk = i;
-				break;
-			}
+		info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum);
+
+		info->data_offset = __be64_to_cpu(ddf->phys->
+					  entries[info->disk.raid_disk].
+					  config_size);
+		info->component_size = ddf->dlist->size - info->data_offset;
 	} else {
 		info->disk.number = -1;
 //		info->disk.raid_disk = find refnum in the table and use index;
 	}
 	info->disk.state = (1 << MD_DISK_SYNC);
 
+
 	info->reshape_active = 0;
 
 	strcpy(info->text_version, "ddf");
@@ -1239,39 +1241,33 @@ static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
 //	info->name[] ?? ;
 }
 
-static void getinfo_super_n_container(struct supertype *st, struct mdinfo *info)
-{
-	/* just need offset and size */
-	struct ddf_super *ddf = st->sb;
-	int n = info->disk.number;
-
-	info->data_offset = __be64_to_cpu(ddf->phys->entries[n].config_size);
-	info->component_size = 32*1024*1024 / 512;
-}
-
 static int rlq_to_layout(int rlq, int prl, int raiddisks);
 
 static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
 {
 	struct ddf_super *ddf = st->sb;
-	struct vd_config *vd = find_vdcr(ddf, info->container_member);
+	struct vcl *vc = ddf->currentconf;
+	int cd = ddf->currentdev;
 
 	/* FIXME this returns BVD info - what if we want SVD ?? */
 
-	info->array.major_version = 1000;
-	info->array.minor_version = 0; /* FIXME use ddf->revision somehow */
-	info->array.patch_version = 0;
-	info->array.raid_disks    = __be16_to_cpu(vd->prim_elmnt_count);
-	info->array.level	  = map_num1(ddf_level_num, vd->prl);
-	info->array.layout	  = rlq_to_layout(vd->rlq, vd->prl,
+	info->array.raid_disks    = __be16_to_cpu(vc->conf.prim_elmnt_count);
+	info->array.level	  = map_num1(ddf_level_num, vc->conf.prl);
+	info->array.layout	  = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
 						  info->array.raid_disks);
 	info->array.md_minor	  = -1;
-	info->array.ctime	  = DECADE + __be32_to_cpu(*(__u32*)(vd->guid+16));
-	info->array.utime	  = DECADE + __be32_to_cpu(vd->timestamp);
-	info->array.chunk_size	  = 512 << vd->chunk_shift;
-
-//	info->data_offset	  = ???;
-//	info->component_size	  = ???;
+	info->array.ctime	  = DECADE +
+		__be32_to_cpu(*(__u32*)(vc->conf.guid+16));
+	info->array.utime	  = DECADE + __be32_to_cpu(vc->conf.timestamp);
+	info->array.chunk_size	  = 512 << vc->conf.chunk_shift;
+
+	if (cd >= 0 && cd < ddf->mppe) {
+		info->data_offset	  = __be64_to_cpu(vc->lba_offset[cd]);
+		if (vc->block_sizes)
+			info->component_size = vc->block_sizes[cd];
+		else
+			info->component_size = __be64_to_cpu(vc->conf.blocks);
+	}
 
 	info->disk.major = 0;
 	info->disk.minor = 0;
@@ -1279,6 +1275,14 @@ static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
 //	info->disk.raid_disk = find refnum in the table and use index;
 //	info->disk.state = ???;
 
+	info->resync_start = 0;
+	if (!(ddf->virt->entries[info->container_member].state
+	      & DDF_state_inconsistent)  &&
+	    (ddf->virt->entries[info->container_member].init_state
+	     & DDF_initstate_mask)
+	    == DDF_init_full)
+		info->resync_start = ~0ULL;
+
 	uuid_from_super_ddf(st, info->uuid);
 
 	sprintf(info->text_version, "/%s/%d",
@@ -1288,17 +1292,6 @@ static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
 //	info->name[] ?? ;
 }
 
-static void getinfo_super_n_bvd(struct supertype *st, struct mdinfo *info)
-{
-	/* Find the particular details for info->disk.raid_disk.
-	 * This includes data_offset, component_size,
-	 */
-	struct ddf_super *ddf = st->sb;
-	__u64 *lba_offset = ddf->newconf->lba_offset;
-	struct vd_config *conf = &ddf->newconf->conf;
-	info->data_offset = __be64_to_cpu(lba_offset[info->disk.raid_disk]);
-	info->component_size = __be64_to_cpu(conf->blocks);
-}
 
 static int update_super_ddf(struct supertype *st, struct mdinfo *info,
 			    char *update,
@@ -1319,7 +1312,7 @@ static int update_super_ddf(struct supertype *st, struct mdinfo *info,
 	 *  grow:  Array has gained a new device - this is currently for
 	 *		linear only
 	 *  resync: mark as dirty so a resync will happen.
-	 *  uuid:  Change the uuid of the array to match watch is given
+	 *  uuid:  Change the uuid of the array to match what is given
 	 *  homehost:  update the recorded homehost
 	 *  name:  update the name - preserving the homehost
 	 *  _reshape_progress: record new reshape_progress position.
@@ -1334,7 +1327,6 @@ static int update_super_ddf(struct supertype *st, struct mdinfo *info,
 //	struct vd_config *vd = find_vdcr(ddf, info->container_member);
 //	struct virtual_entry *ve = find_ve(ddf);
 
-
 	/* we don't need to handle "force-*" or "assemble" as
 	 * there is no need to 'trick' the kernel.  We the metadata is
 	 * first updated to activate the array, all the implied modifications
@@ -1395,6 +1387,13 @@ static void make_header_guid(char *guid)
 	memcpy(guid+20, &stamp, 4);
 	if (rfd >= 0) close(rfd);
 }
+
+static int init_super_ddf_bvd(struct supertype *st,
+			      mdu_array_info_t *info,
+			      unsigned long long size,
+			      char *name, char *homehost,
+			      int *uuid);
+
 static int init_super_ddf(struct supertype *st,
 			  mdu_array_info_t *info,
 			  unsigned long long size, char *name, char *homehost,
@@ -1434,7 +1433,16 @@ static int init_super_ddf(struct supertype *st,
 	struct phys_disk *pd;
 	struct virtual_disk *vd;
 
+	if (!info) {
+		st->sb = NULL;
+		return 0;
+	}
+	if (st->sb)
+		return init_super_ddf_bvd(st, info, size, name, homehost,
+					  uuid);
+
 	ddf = malloc(sizeof(*ddf));
+	memset(ddf, 0, sizeof(*ddf));
 	ddf->dlist = NULL; /* no physical disks yet */
 	ddf->conflist = NULL; /* No virtual disks yet */
 
@@ -1449,7 +1457,7 @@ static int init_super_ddf(struct supertype *st,
 	ddf->anchor.magic = DDF_HEADER_MAGIC;
 	make_header_guid(ddf->anchor.guid);
 
-	memcpy(ddf->anchor.revision, DDF_REVISION, 8);
+	memcpy(ddf->anchor.revision, DDF_REVISION_2, 8);
 	ddf->anchor.seq = __cpu_to_be32(1);
 	ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
 	ddf->anchor.openflag = 0xFF;
@@ -1471,12 +1479,11 @@ static int init_super_ddf(struct supertype *st,
 	ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
 	ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
 	ddf->max_part = 64;
-	ddf->conf_rec_len = 1 + 256 * 12 / 512;
-	ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
-	ddf->anchor.max_primary_element_entries = __cpu_to_be16(256);
 	ddf->mppe = 256;
+	ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512;
+	ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
+	ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe);
 	memset(ddf->anchor.pad3, 0xff, 54);
-
 	/* controller sections is one sector long immediately
 	 * after the ddf header */
 	sector = 1;
@@ -1510,7 +1517,7 @@ static int init_super_ddf(struct supertype *st,
 		__cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
 	sector += vdsize/512;
 
-	clen = (1 + 256*12/512) * (64+1);
+	clen = ddf->conf_rec_len * (ddf->max_part+1);
 	ddf->anchor.config_section_offset = __cpu_to_be32(sector);
 	ddf->anchor.config_section_length = __cpu_to_be32(clen);
 	sector += clen;
@@ -1546,8 +1553,8 @@ static int init_super_ddf(struct supertype *st,
 	 * Remaining 16 are serial number.... maybe a hostname would do?
 	 */
 	memcpy(ddf->controller.guid, T10, sizeof(T10));
-	gethostname(hostname, 17);
-	hostname[17] = 0;
+	gethostname(hostname, sizeof(hostname));
+	hostname[sizeof(hostname) - 1] = 0;
 	hostlen = strlen(hostname);
 	memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
 	for (i = strlen(T10) ; i+hostlen < 24; i++)
@@ -1633,7 +1640,10 @@ static int layout_to_rlq(int level, int layout, int raiddisks)
 		case ALGORITHM_LEFT_ASYMMETRIC:
 			return DDF_RAID5_N_RESTART;
 		case ALGORITHM_RIGHT_ASYMMETRIC:
-			return DDF_RAID5_0_RESTART;
+			if (level == 5)
+				return DDF_RAID5_0_RESTART;
+			else
+				return DDF_RAID6_0_RESTART;
 		case ALGORITHM_LEFT_SYMMETRIC:
 			return DDF_RAID5_N_CONTINUE;
 		case ALGORITHM_RIGHT_SYMMETRIC:
@@ -1660,7 +1670,6 @@ static int rlq_to_layout(int rlq, int prl, int raiddisks)
 			return -1; /* FIXME this isn't checked */
 		}
 	case DDF_RAID5:
-	case DDF_RAID6:
 		switch(rlq) {
 		case DDF_RAID5_N_RESTART:
 			return ALGORITHM_LEFT_ASYMMETRIC;
@@ -1671,10 +1680,72 @@ static int rlq_to_layout(int rlq, int prl, int raiddisks)
 		default:
 			return -1;
 		}
+	case DDF_RAID6:
+		switch(rlq) {
+		case DDF_RAID5_N_RESTART:
+			return ALGORITHM_LEFT_ASYMMETRIC;
+		case DDF_RAID6_0_RESTART:
+			return ALGORITHM_RIGHT_ASYMMETRIC;
+		case DDF_RAID5_N_CONTINUE:
+			return ALGORITHM_LEFT_SYMMETRIC;
+		default:
+			return -1;
+		}
 	}
 	return -1;
 }
 
+struct extent {
+	unsigned long long start, size;
+};
+static int cmp_extent(const void *av, const void *bv)
+{
+	const struct extent *a = av;
+	const struct extent *b = bv;
+	if (a->start < b->start)
+		return -1;
+	if (a->start > b->start)
+		return 1;
+	return 0;
+}
+
+static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
+{
+	/* find a list of used extents on the give physical device
+	 * (dnum) of the given ddf.
+	 * Return a malloced array of 'struct extent'
+
+FIXME ignore DDF_Legacy devices?
+
+	 */
+	struct extent *rv;
+	int n = 0;
+	int i, j;
+
+	rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
+	if (!rv)
+		return NULL;
+
+	for (i = 0; i < ddf->max_part; i++) {
+		struct vcl *v = dl->vlist[i];
+		if (v == NULL)
+			continue;
+		for (j=0; j < v->conf.prim_elmnt_count; j++)
+			if (v->conf.phys_refnum[j] == dl->disk.refnum) {
+				/* This device plays role 'j' in  'v'. */
+				rv[n].start = __be64_to_cpu(v->lba_offset[j]);
+				rv[n].size = __be64_to_cpu(v->conf.blocks);
+				n++;
+				break;
+			}
+	}
+	qsort(rv, n, sizeof(*rv), cmp_extent);
+
+	rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size);
+	rv[n].size = 0;
+	return rv;
+}
+
 static int init_super_ddf_bvd(struct supertype *st,
 			      mdu_array_info_t *info,
 			      unsigned long long size,
@@ -1708,7 +1779,6 @@ static int init_super_ddf_bvd(struct supertype *st,
 		return 0;
 	}
 	ve = &ddf->virt->entries[venum];
-	ddf->conf_num = venum;
 
 	/* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
 	 * timestamp, random number
@@ -1734,6 +1804,9 @@ static int init_super_ddf_bvd(struct supertype *st,
 	/* Now create a new vd_config */
 	vcl = malloc(offsetof(struct vcl, conf) + ddf->conf_rec_len * 512);
 	vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->mppe];
+	vcl->vcnum = venum;
+	sprintf(st->subarray, "%d", venum);
+	vcl->block_sizes = NULL; /* FIXME not for CONCAT */
 
 	vc = &vcl->conf;
 
@@ -1778,7 +1851,7 @@ static int init_super_ddf_bvd(struct supertype *st,
 
 	vcl->next = ddf->conflist;
 	ddf->conflist = vcl;
-	ddf->newconf = vcl;
+	ddf->currentconf = vcl;
 	return 1;
 }
 
@@ -1790,13 +1863,16 @@ static void add_to_super_ddf_bvd(struct supertype *st,
 	 * We need to find suitable free space in that device and update
 	 * the phys_refnum and lba_offset for the newly created vd_config.
 	 * We might also want to update the type in the phys_disk
-	 * section. FIXME
+	 * section.
 	 */
 	struct dl *dl;
 	struct ddf_super *ddf = st->sb;
 	struct vd_config *vc;
 	__u64 *lba_offset;
 	int working;
+	int i;
+	unsigned long long blocks, pos, esize;
+	struct extent *ex;
 
 	for (dl = ddf->dlist; dl ; dl = dl->next)
 		if (dl->major == dk->major &&
@@ -1805,12 +1881,40 @@ static void add_to_super_ddf_bvd(struct supertype *st,
 	if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
 		return;
 
-	vc = &ddf->newconf->conf;
+	vc = &ddf->currentconf->conf;
+	lba_offset = ddf->currentconf->lba_offset;
+
+	ex = get_extents(ddf, dl);
+	if (!ex)
+		return;
+
+	i = 0; pos = 0;
+	blocks = __be64_to_cpu(vc->blocks);
+	if (ddf->currentconf->block_sizes)
+		blocks = ddf->currentconf->block_sizes[dk->raid_disk];
+
+	do {
+		esize = ex[i].start - pos;
+		if (esize >= blocks)
+			break;
+		pos = ex[i].start + ex[i].size;
+		i++;
+	} while (ex[i-1].size);
+
+	free(ex);
+	if (esize < blocks)
+		return;
+
+	ddf->currentdev = dk->raid_disk;
 	vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
-	lba_offset = (__u64*)(vc->phys_refnum + ddf->mppe);
-	lba_offset[dk->raid_disk] = 0; /* FIXME */
+	lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
 
-	dl->vlist[0] = ddf->newconf; /* FIXME */
+	for (i=0; i < ddf->max_part ; i++)
+		if (dl->vlist[i] == NULL)
+			break;
+	if (i == ddf->max_part)
+		return;
+	dl->vlist[i] = ddf->currentconf;
 
 	dl->fd = fd;
 	dl->devname = devname;
@@ -1819,19 +1923,26 @@ static void add_to_super_ddf_bvd(struct supertype *st,
 	 * array as optimal yet
 	 */
 	working = 0;
-#if 0
+
 	for (i=0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
 		if (vc->phys_refnum[i] != 0xffffffff)
 			working++;
+
+	/* Find which virtual_entry */
+	i = ddf->currentconf->vcnum;
 	if (working == __be16_to_cpu(vc->prim_elmnt_count))
-		->entries[xx].state = (->entries[xx].state & ~DDF_state_mask)
+		ddf->virt->entries[i].state =
+			(ddf->virt->entries[i].state & ~DDF_state_mask)
 			| DDF_state_optimal;
 
 	if (vc->prl == DDF_RAID6 &&
 	    working+1 == __be16_to_cpu(vc->prim_elmnt_count))
-		->entries[xx].state = (->entries[xx].state & ~DDF_state_mask)
+		ddf->virt->entries[i].state =
+			(ddf->virt->entries[i].state & ~DDF_state_mask)
 			| DDF_state_part_optimal;
-#endif
+
+	ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare);
+	ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD);
 }
 
 /* add a device to a container, either while creating it or while
@@ -1849,6 +1960,11 @@ static void add_to_super_ddf(struct supertype *st,
 	int n, i;
 	struct stat stb;
 
+	if (ddf->currentconf) {
+		add_to_super_ddf_bvd(st, dk, fd, devname);
+		return;
+	}
+
 	/* This is device numbered dk->number.  We need to create
 	 * a phys_disk entry and a more detailed disk_data entry.
 	 */
@@ -1869,7 +1985,15 @@ static void add_to_super_ddf(struct supertype *st,
 	*(__u32*)(dd->disk.guid + 16) = random();
 	*(__u32*)(dd->disk.guid + 20) = random();
 
-	dd->disk.refnum = random(); /* and hope for the best FIXME check this is unique!!*/
+	do {
+		/* Cannot be bothered finding a CRC of some irrelevant details*/
+		dd->disk.refnum = random();
+		for (i = __be16_to_cpu(ddf->active->max_pd_entries) - 1;
+		     i >= 0; i--)
+			if (ddf->phys->entries[i].refnum == dd->disk.refnum)
+				break;
+	} while (i >= 0);
+
 	dd->disk.forced_ref = 1;
 	dd->disk.forced_guid = 1;
 	memset(dd->disk.vendor, ' ', 32);
@@ -1880,12 +2004,14 @@ static void add_to_super_ddf(struct supertype *st,
 
 	n = __be16_to_cpu(ddf->phys->used_pdes);
 	pde = &ddf->phys->entries[n];
+	dd->pdnum = n;
+
 	n++;
 	ddf->phys->used_pdes = __cpu_to_be16(n);
 
 	memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
 	pde->refnum = dd->disk.refnum;
-	pde->type = __cpu_to_be16(DDF_Forced_PD_GUID |DDF_Global_Spare);
+	pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare);
 	pde->state = __cpu_to_be16(DDF_Online);
 	get_dev_size(fd, NULL, &size);
 	/* We are required to reserve 32Meg, and record the size in sectors */
@@ -1893,6 +2019,7 @@ static void add_to_super_ddf(struct supertype *st,
 	sprintf(pde->path, "%17.17s","Information: nil") ;
 	memset(pde->pad, 0xff, 6);
 
+	dd->size = size >> 9;
 	ddf->dlist = dd;
 }
 
@@ -1995,7 +2122,32 @@ static int __write_init_super_ddf(struct supertype *st, int do_close)
 
 static int write_init_super_ddf(struct supertype *st)
 {
-	return __write_init_super_ddf(st, 1);
+
+	if (st->update_tail) {
+		/* queue the virtual_disk and vd_config as metadata updates */
+		struct virtual_disk *vd;
+		struct vd_config *vc;
+		struct ddf_super *ddf = st->sb;
+		int len;
+
+		/* First the virtual disk.  We have a slightly fake header */
+		len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
+		vd = malloc(len);
+		*vd = *ddf->virt;
+		vd->entries[0] = ddf->virt->entries[ddf->currentconf->vcnum];
+		vd->populated_vdes = __cpu_to_be16(ddf->currentconf->vcnum);
+		append_metadata_update(st, vd, len);
+
+		/* Then the vd_config */
+		len = ddf->conf_rec_len * 512;
+		vc = malloc(len);
+		memcpy(vc, &ddf->currentconf->conf, len);
+		append_metadata_update(st, vc, len);
+
+		/* FIXME I need to close the fds! */
+		return 0;
+	} else 
+		return __write_init_super_ddf(st, 1);
 }
 
 #endif
@@ -2009,7 +2161,17 @@ static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
 }
 
 #ifndef MDASSEMBLE
-int validate_geometry_ddf(struct supertype *st,
+static int validate_geometry_ddf_container(struct supertype *st,
+				    int level, int layout, int raiddisks,
+				    int chunk, unsigned long long size,
+				    char *dev, unsigned long long *freesize);
+
+static int validate_geometry_ddf_bvd(struct supertype *st,
+				     int level, int layout, int raiddisks,
+				     int chunk, unsigned long long size,
+				     char *dev, unsigned long long *freesize);
+
+static int validate_geometry_ddf(struct supertype *st,
 			  int level, int layout, int raiddisks,
 			  int chunk, unsigned long long size,
 			  char *dev, unsigned long long *freesize)
@@ -2026,53 +2188,50 @@ int validate_geometry_ddf(struct supertype *st,
 	 */
 
 	if (level == LEVEL_CONTAINER) {
-		st->ss = &super_ddf_container;
-		if (dev) {
-			int rv =st->ss->validate_geometry(st, level, layout,
-							  raiddisks, chunk,
-							  size,
-							  NULL, freesize);
-			if (rv)
-				return rv;
-		}
-		return st->ss->validate_geometry(st, level, layout, raiddisks,
-						 chunk, size, dev, freesize);
+		/* Must be a fresh device to add to a container */
+		return validate_geometry_ddf_container(st, level, layout,
+					       raiddisks,
+					       chunk, size, dev, freesize);
 	}
 
 	if (st->sb) {
-		/* creating in a given container */
-		st->ss = &super_ddf_bvd;
-		if (dev) {
-			int rv =st->ss->validate_geometry(st, level, layout,
-							  raiddisks, chunk,
-							  size,
-							  NULL, freesize);
-			if (rv)
-				return rv;
-		}
-		return st->ss->validate_geometry(st, level, layout, raiddisks,
+		/* A container has already been opened, so we are
+		 * creating in there.  Maybe a BVD, maybe an SVD.
+		 * Should make a distinction one day.
+		 */
+		return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
 						 chunk, size, dev, freesize);
 	}
-	/* FIXME should exclude MULTIPATH, or more appropriately, allow
-	 * only known levels.
-	 */
-	if (!dev)
+	if (!dev) {
+		/* Initial sanity check.  Exclude illegal levels. */
+		int i;
+		for (i=0; ddf_level_num[i].num1 != MAXINT; i++)
+			if (ddf_level_num[i].num2 == level)
+				break;
+		if (ddf_level_num[i].num1 == MAXINT)
+			return 0;
+		/* Should check layout? etc */
 		return 1;
+	}
 
-	/* This device needs to be either a device in a 'ddf' container,
-	 * or it needs to be a 'ddf-bvd' array.
+	/* This is the first device for the array.
+	 * If it is a container, we read it in and do automagic allocations,
+	 * no other devices should be given.
+	 * Otherwise it must be a member device of a container, and we
+	 * do manual allocation.
+	 * Later we should check for a BVD and make an SVD.
 	 */
-
 	fd = open(dev, O_RDONLY|O_EXCL, 0);
 	if (fd >= 0) {
 		sra = sysfs_read(fd, 0, GET_VERSION);
 		close(fd);
 		if (sra && sra->array.major_version == -1 &&
-		    strcmp(sra->text_version, "ddf-bvd") == 0) {
-			st->ss = &super_ddf_svd;
-			return st->ss->validate_geometry(st, level, layout,
-							 raiddisks, chunk, size,
-							 dev, freesize);
+		    strcmp(sra->text_version, "ddf") == 0) {
+
+			/* load super */
+			/* find space for 'n' devices. */
+			/* remember the devices */
+			/* Somehow return the fact that we have enough */
 		}
 
 		fprintf(stderr,
@@ -2101,26 +2260,25 @@ int validate_geometry_ddf(struct supertype *st,
 		 * and try to create a bvd
 		 */
 		struct ddf_super *ddf;
-		st->ss = &super_ddf_bvd;
 		if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL, 1) == 0) {
 			st->sb = ddf;
 			st->container_dev = fd2devnum(cfd);
 			close(cfd);
-			return st->ss->validate_geometry(st, level, layout,
+			return validate_geometry_ddf_bvd(st, level, layout,
 							 raiddisks, chunk, size,
 							 dev, freesize);
 		}
 		close(cfd);
-	}
-	fprintf(stderr, Name ": Cannot use %s: Already in use\n",
-		dev);
+	} else /* device may belong to a different container */
+		return 0;
+
 	return 1;
 }
 
-int validate_geometry_ddf_container(struct supertype *st,
-				    int level, int layout, int raiddisks,
-				    int chunk, unsigned long long size,
-				    char *dev, unsigned long long *freesize)
+static int validate_geometry_ddf_container(struct supertype *st,
+				   int level, int layout, int raiddisks,
+				   int chunk, unsigned long long size,
+				   char *dev, unsigned long long *freesize)
 {
 	int fd;
 	unsigned long long ldsize;
@@ -2142,76 +2300,15 @@ int validate_geometry_ddf_container(struct supertype *st,
 	}
 	close(fd);
 
-	*freesize = avail_size_ddf(st, ldsize);
+	*freesize = avail_size_ddf(st, ldsize >> 9);
 
 	return 1;
 }
 
-struct extent {
-	unsigned long long start, size;
-};
-int cmp_extent(const void *av, const void *bv)
-{
-	const struct extent *a = av;
-	const struct extent *b = bv;
-	if (a->start < b->start)
-		return -1;
-	if (a->start > b->start)
-		return 1;
-	return 0;
-}
-
-struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
-{
-	/* find a list of used extents on the give physical device
-	 * (dnum) or the given ddf.
-	 * Return a malloced array of 'struct extent'
-
-FIXME ignore DDF_Legacy devices?
-
-	 */
-	struct extent *rv;
-	int n = 0;
-	int dnum;
-	int i, j;
-
-	for (dnum = 0; dnum < ddf->phys->used_pdes; dnum++)
-		if (memcmp(dl->disk.guid,
-			   ddf->phys->entries[dnum].guid,
-			   DDF_GUID_LEN) == 0)
-			break;
-
-	if (dnum == ddf->phys->used_pdes)
-		return NULL;
-
-	rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
-	if (!rv)
-		return NULL;
-
-	for (i = 0; i < ddf->max_part; i++) {
-		struct vcl *v = dl->vlist[i];
-		if (v == NULL)
-			continue;
-		for (j=0; j < v->conf.prim_elmnt_count; j++)
-			if (v->conf.phys_refnum[j] == dl->disk.refnum) {
-				/* This device plays role 'j' in  'v'. */
-				rv[n].start = __be64_to_cpu(v->lba_offset[j]);
-				rv[n].size = __be64_to_cpu(v->conf.blocks);
-				n++;
-				break;
-			}
-	}
-	qsort(rv, n, sizeof(*rv), cmp_extent);
-
-	rv[n].start = __be64_to_cpu(ddf->phys->entries[dnum].config_size);
-	rv[n].size = 0;
-	return rv;
-}
-
-int validate_geometry_ddf_bvd(struct supertype *st,
-			      int level, int layout, int raiddisks,
-			      int chunk, unsigned long long size,
-			      char *dev, unsigned long long *freesize)
+static int validate_geometry_ddf_bvd(struct supertype *st,
+				     int level, int layout, int raiddisks,
+				     int chunk, unsigned long long size,
+				     char *dev, unsigned long long *freesize)
 {
 	struct stat stb;
 	struct ddf_super *ddf = st->sb;
@@ -2238,6 +2335,7 @@ int validate_geometry_ddf_bvd(struct supertype *st,
 		for (dl = ddf->dlist; dl ; dl = dl->next)
 		{
 			int found = 0;
+			pos = 0;
 
 			i = 0;
 			e = get_extents(ddf, dl);
@@ -2293,19 +2391,6 @@ int validate_geometry_ddf_bvd(struct supertype *st,
 
 	return 1;
 }
-int validate_geometry_ddf_svd(struct supertype *st,
-			      int level, int layout, int raiddisks,
-			      int chunk, unsigned long long size,
-			      char *dev, unsigned long long *freesize)
-{
-	/* dd/svd only supports striped, mirrored, concat, spanned... */
-	if (level != LEVEL_LINEAR &&
-	    level != 0 &&
-	    level != 1)
-		return 0;
-	return 1;
-}
-
 
 static int load_super_ddf_all(struct supertype *st, int fd,
 			      void **sbp, char *devname, int keep_fd)
@@ -2369,9 +2454,18 @@ static int load_super_ddf_all(struct supertype *st, int fd,
 		seq = load_ddf_local(dfd, super, NULL, keep_fd);
 		if (!keep_fd) close(dfd);
 	}
+	if (st->subarray[0]) {
+		struct vcl *v;
+
+		for (v = super->conflist; v; v = v->next)
+			if (v->vcnum == atoi(st->subarray))
+				super->currentconf = v;
+		if (!super->currentconf)
+			return 1;
+	}
 	*sbp = super;
 	if (st->ss == NULL) {
-		st->ss = &super_ddf_container;
+		st->ss = &super_ddf;
 		st->minor_version = 0;
 		st->max_devs = 512;
 		st->container_dev = fd2devnum(fd);
@@ -2380,8 +2474,6 @@ static int load_super_ddf_all(struct supertype *st, int fd,
 }
 #endif
 
-
-
 static struct mdinfo *container_content_ddf(struct supertype *st)
 {
 	/* Given a container loaded by load_super_ddf_all,
@@ -2405,9 +2497,6 @@ static struct mdinfo *container_content_ddf(struct supertype *st)
 		this->next = rest;
 		rest = this;
 
-		this->array.major_version = 1000;
-		this->array.minor_version = 0;
-		this->array.patch_version = 0;
 		this->array.level = map_num1(ddf_level_num, vc->conf.prl);
 		this->array.raid_disks =
 			__be16_to_cpu(vc->conf.prim_elmnt_count);
@@ -2420,10 +2509,7 @@ static struct mdinfo *container_content_ddf(struct supertype *st)
 			__be32_to_cpu(vc->conf.timestamp);
 		this->array.chunk_size	  = 512 << vc->conf.chunk_shift;
 
-		for (i=0; i < __be16_to_cpu(ddf->virt->populated_vdes); i++)
-			if (memcmp(ddf->virt->entries[i].guid,
-				   vc->conf.guid, DDF_GUID_LEN) == 0)
-				break;
+		i = vc->vcnum;
 		if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
 		    (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
 		    DDF_init_full) {
@@ -2445,7 +2531,6 @@ static struct mdinfo *container_content_ddf(struct supertype *st)
 			devnum2devname(st->container_dev),
 			this->container_member);
 
-
 		for (i=0 ; i < ddf->mppe ; i++) {
 			struct mdinfo *dev;
 			struct dl *d;
@@ -2482,22 +2567,12 @@ static struct mdinfo *container_content_ddf(struct supertype *st)
 	return rest;
 }
 
-static int init_zero_ddf(struct supertype *st,
-			 mdu_array_info_t *info,
-			 unsigned long long size, char *name,
-			 char *homehost, int *uuid)
-{
-	st->sb = NULL;
-	return 0;
-}
-
 static int store_zero_ddf(struct supertype *st, int fd)
 {
 	unsigned long long dsize;
 	char buf[512];
 	memset(buf, 0, 512);
 
-
 	if (!get_dev_size(fd, NULL, &dsize))
 		return 1;
 
@@ -2618,6 +2693,8 @@ static void ddf_set_disk(struct active_array *a, int n, int state)
 		}
 	}
 
+	fprintf(stderr, "ddf: set_disk %d to %x\n", n, state);
+
 	/* Now we need to check the state of the array and update
 	 * virtual_disk.entries[n].state.
 	 * It needs to be one of "optimal", "degraded", "failed".
@@ -2628,8 +2705,8 @@ static void ddf_set_disk(struct active_array *a, int n, int state)
 		pd = find_phys(ddf, vc->phys_refnum[i]);
 		if (pd < 0)
 			continue;
-		st = ddf->phys->entries[pd].state;
-		if ((state & (DDF_Online|DDF_Failed|DDF_Rebuilding))
+		st = __be16_to_cpu(ddf->phys->entries[pd].state);
+		if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
 		    == DDF_Online)
 			working++;
 	}
@@ -2663,7 +2740,6 @@ static void ddf_set_disk(struct active_array *a, int n, int state)
 		(ddf->virt->entries[inst].state & ~DDF_state_mask)
 		| state;
 
-	fprintf(stderr, "ddf: set_disk %d\n", n);
 }
 
 static void ddf_sync_metadata(struct supertype *st)
@@ -2680,6 +2756,361 @@ static void ddf_sync_metadata(struct supertype *st)
 	fprintf(stderr, "ddf: sync_metadata\n");
 }
 
+static void ddf_process_update(struct supertype *st,
+			       struct metadata_update *update)
+{
+	/* Apply this update to the metadata.
+	 * The first 4 bytes are a DDF_*_MAGIC which guides
+	 * our actions.
+	 * Possible update are:
+	 *  DDF_PHYS_RECORDS_MAGIC
+	 *    Add a new physical device.  Changes to this record
+	 *    only happen implicitly.
+	 *    used_pdes is the device number.
+	 *  DDF_VIRT_RECORDS_MAGIC
+	 *    Add a new VD.  Possibly also change the 'access' bits.
+	 *    populated_vdes is the entry number.
+	 *  DDF_VD_CONF_MAGIC
+	 *    New or updated VD.  the VIRT_RECORD must already
+	 *    exist.  For an update, phys_refnum and lba_offset
+	 *    (at least) are updated, and the VD_CONF must
+	 *    be written to precisely those devices listed with
+	 *    a phys_refnum.
+	 *  DDF_SPARE_ASSIGN_MAGIC
+	 *    replacement Spare Assignment Record... but for which device?
+	 *
+	 * So, e.g.:
+	 *  - to create a new array, we send a VIRT_RECORD and
+	 *    a VD_CONF.  Then assemble and start the array.
+	 *  - to activate a spare we send a VD_CONF to add the phys_refnum
+	 *    and offset.  This will also mark the spare as active with
+	 *    a spare-assignment record.
+	 */
+	struct ddf_super *ddf = st->sb;
+	__u32 *magic = (__u32*)update->buf;
+	struct phys_disk *pd;
+	struct virtual_disk *vd;
+	struct vd_config *vc;
+	struct vcl *vcl;
+	struct dl *dl;
+	int mppe;
+	int ent;
+
+	printf("Process update %x\n", *magic);
+
+	switch (*magic) {
+	case DDF_PHYS_RECORDS_MAGIC:
+
+		if (update->len != (sizeof(struct phys_disk) +
+				    sizeof(struct phys_disk_entry)))
+			return;
+		pd = (struct phys_disk*)update->buf;
+
+		ent = __be16_to_cpu(pd->used_pdes);
+		if (ent >= __be16_to_cpu(ddf->phys->max_pdes))
+			return;
+		if (!all_ff(ddf->phys->entries[ent].guid))
+			return;
+		ddf->phys->entries[ent] = pd->entries[0];
+		ddf->phys->used_pdes = __cpu_to_be16(1 +
+					   __be16_to_cpu(ddf->phys->used_pdes));
+		break;
+
+	case DDF_VIRT_RECORDS_MAGIC:
+
+		if (update->len != (sizeof(struct virtual_disk) +
+				    sizeof(struct virtual_entry)))
+			return;
+		vd = (struct virtual_disk*)update->buf;
+
+		ent = __be16_to_cpu(vd->populated_vdes);
+		if (ent >= __be16_to_cpu(ddf->virt->max_vdes))
+			return;
+		if (!all_ff(ddf->virt->entries[ent].guid))
+			return;
+		ddf->virt->entries[ent] = vd->entries[0];
+		ddf->virt->populated_vdes = __cpu_to_be16(1 +
+			      __be16_to_cpu(ddf->virt->populated_vdes));
+		break;
+
+	case DDF_VD_CONF_MAGIC:
+		printf("len %d %d\n", update->len, ddf->conf_rec_len);
+
+		mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
+		if (update->len != ddf->conf_rec_len * 512)
+			return;
+		vc = (struct vd_config*)update->buf;
+		for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
+			if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
+				break;
+		printf("vcl = %p\n", vcl);
+		if (vcl) {
+			/* An update, just copy the phys_refnum and lba_offset
+			 * fields
+			 */
+			memcpy(vcl->conf.phys_refnum, vc->phys_refnum,
+			       mppe * (sizeof(__u32) + sizeof(__u64)));
+		} else {
+			/* A new VD_CONF */
+			vcl = update->space;
+			update->space = NULL;
+			vcl->next = ddf->conflist;
+			memcpy(&vcl->conf, vc, update->len);
+			vcl->lba_offset = (__u64*)
+				&vcl->conf.phys_refnum[mppe];
+			ddf->conflist = vcl;
+		}
+		/* Now make sure vlist is correct for each dl. */
+		for (dl = ddf->dlist; dl; dl = dl->next) {
+			int dn;
+			int vn = 0;
+			for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
+				for (dn=0; dn < ddf->mppe ; dn++)
+					if (vcl->conf.phys_refnum[dn] ==
+					    dl->disk.refnum) {
+						printf("dev %d has %p at %d\n",
+						       dl->pdnum, vcl, vn);
+						dl->vlist[vn++] = vcl;
+						break;
+					}
+			while (vn < ddf->max_part)
+				dl->vlist[vn++] = NULL;
+			if (dl->vlist[0]) {
+				ddf->phys->entries[dl->pdnum].type &=
+					~__cpu_to_be16(DDF_Global_Spare);
+				ddf->phys->entries[dl->pdnum].type |=
+					__cpu_to_be16(DDF_Active_in_VD);
+			}
+			if (dl->spare) {
+				ddf->phys->entries[dl->pdnum].type &=
+					~__cpu_to_be16(DDF_Global_Spare);
+				ddf->phys->entries[dl->pdnum].type |=
+					__cpu_to_be16(DDF_Spare);
+			}
+			if (!dl->vlist[0] && !dl->spare) {
+				ddf->phys->entries[dl->pdnum].type |=
+					__cpu_to_be16(DDF_Global_Spare);
+				ddf->phys->entries[dl->pdnum].type &=
+					~__cpu_to_be16(DDF_Spare |
+						       DDF_Active_in_VD);
+			}
+		}
+		break;
+	case DDF_SPARE_ASSIGN_MAGIC:
+	default: break;
+	}
+}
+
+static void ddf_prepare_update(struct supertype *st,
+			       struct metadata_update *update)
+{
+	/* This update arrived at managemon.
+	 * We are about to pass it to monitor.
+	 * If a malloc is needed, do it here.
+	 */
+	struct ddf_super *ddf = st->sb;
+	__u32 *magic = (__u32*)update->buf;
+	if (*magic == DDF_VD_CONF_MAGIC)
+		update->space = malloc(offsetof(struct vcl, conf)
+				       + ddf->conf_rec_len * 512);
+}
+
+/*
+ * Check if the array 'a' is degraded but not failed.
+ * If it is, find as many spares as are available and needed and
+ * arrange for their inclusion.
+ * We only choose devices which are not already in the array,
+ * and prefer those with a spare-assignment to this array.
+ * otherwise we choose global spares - assuming always that
+ * there is enough room.
+ * For each spare that we assign, we return an 'mdinfo' which
+ * describes the position for the device in the array.
+ * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
+ * the new phys_refnum and lba_offset values.
+ *
+ * Only worry about BVDs at the moment.
+ */
+static struct mdinfo *ddf_activate_spare(struct active_array *a,
+					 struct metadata_update **updates)
+{
+	int working = 0;
+	struct mdinfo *d;
+	struct ddf_super *ddf = a->container->sb;
+	int global_ok = 0;
+	struct mdinfo *rv = NULL;
+	struct mdinfo *di;
+	struct metadata_update *mu;
+	struct dl *dl;
+	int i;
+	struct vd_config *vc;
+	__u64 *lba;
+
+	for (d = a->info.devs ; d ; d = d->next) {
+		if ((d->curr_state & DS_FAULTY) &&
+			d->state_fd >= 0)
+			/* wait for Removal to happen */
+			return NULL;
+		if (d->state_fd >= 0)
+			working ++;
+	}
+
+	printf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks,
+	       a->info.array.level);
+	if (working == a->info.array.raid_disks)
+		return NULL; /* array not degraded */
+	switch (a->info.array.level) {
+	case 1:
+		if (working == 0)
+			return NULL; /* failed */
+		break;
+	case 4:
+	case 5:
+		if (working < a->info.array.raid_disks - 1)
+			return NULL; /* failed */
+		break;
+	case 6:
+		if (working < a->info.array.raid_disks - 2)
+			return NULL; /* failed */
+		break;
+	default: /* concat or stripe */
+		return NULL; /* failed */
+	}
+
+	/* For each slot, if it is not working, find a spare */
+	dl = ddf->dlist;
+	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)
+				break;
+		printf("found %d: %p %x\n", i, d, d?d->curr_state:0);
+		if (d && (d->state_fd >= 0))
+			continue;
+
+		/* OK, this device needs recovery.  Find a spare */
+	again:
+		for ( ; dl ; dl = dl->next) {
+			unsigned long long esize;
+			unsigned long long pos;
+			struct mdinfo *d2;
+			int is_global = 0;
+			int is_dedicated = 0;
+			struct extent *ex;
+			int j;
+			/* 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) {
+					printf("%x:%x already in array\n", dl->major, dl->minor);
+					break;
+				}
+			if (d2)
+				continue;
+			if (ddf->phys->entries[dl->pdnum].type &
+			    __cpu_to_be16(DDF_Spare)) {
+				/* Check spare assign record */
+				if (dl->spare) {
+					if (dl->spare->type & DDF_spare_dedicated) {
+						/* check spare_ents for guid */
+						for (j = 0 ;
+						     j < __be16_to_cpu(dl->spare->populated);
+						     j++) {
+							if (memcmp(dl->spare->spare_ents[j].guid,
+								   ddf->virt->entries[a->info.container_member].guid,
+								   DDF_GUID_LEN) == 0)
+								is_dedicated = 1;
+						}
+					} else
+						is_global = 1;
+				}
+			} else if (ddf->phys->entries[dl->pdnum].type &
+				   __cpu_to_be16(DDF_Global_Spare)) {
+				is_global = 1;
+			}
+			if ( ! (is_dedicated ||
+				(is_global && global_ok))) {
+				printf("%x:%x not suitable: %d %d\n", dl->major, dl->minor,
+				       is_dedicated, is_global);
+				continue;
+			}
+
+			/* We are allowed to use this device - is there space?
+			 * We need a->info.component_size sectors */
+			ex = get_extents(ddf, dl);
+			if (!ex) {
+				printf("cannot get extents\n");
+				continue;
+			}
+			j = 0; pos = 0;
+			esize = 0;
+
+			do {
+				esize = ex[j].start - pos;
+				if (esize >= a->info.component_size)
+					break;
+				pos = ex[i].start + ex[i].size;
+				i++;
+			} while (ex[i-1].size);
+
+			free(ex);
+			if (esize < a->info.component_size) {
+				printf("%x:%x has no room: %llu %llu\n", dl->major, dl->minor,
+				       esize, a->info.component_size);
+				/* No room */
+				continue;
+			}
+
+			/* Cool, we have a device with some space at pos */
+			di = malloc(sizeof(*di));
+			memset(di, 0, sizeof(*di));
+			di->disk.number = i;
+			di->disk.raid_disk = i;
+			di->disk.major = dl->major;
+			di->disk.minor = dl->minor;
+			di->disk.state = 0;
+			di->data_offset = pos;
+			di->component_size = a->info.component_size;
+			di->container_member = dl->pdnum;
+			di->next = rv;
+			rv = di;
+			printf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
+			       i, pos);
+
+			break;
+		}
+		if (!dl && ! global_ok) {
+			/* not enough dedicated spares, try global */
+			global_ok = 1;
+			dl = ddf->dlist;
+			goto again;
+		}
+	}
+
+	if (!rv)
+		/* No spares found */
+		return rv;
+	/* Now 'rv' has a list of devices to return.
+	 * Create a metadata_update record to update the
+	 * phys_refnum and lba_offset values
+	 */
+	mu = malloc(sizeof(*mu));
+	mu->buf = malloc(ddf->conf_rec_len * 512);
+	mu->space = malloc(sizeof(struct vcl));
+	mu->len = ddf->conf_rec_len;
+	mu->next = *updates;
+	vc = find_vdcr(ddf, a->info.container_member);
+	memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
+
+	vc = (struct vd_config*)mu->buf;
+	lba = (__u64*)&vc->phys_refnum[ddf->mppe];
+	for (di = rv ; di ; di = di->next) {
+		vc->phys_refnum[di->disk.raid_disk] =
+			ddf->phys->entries[dl->pdnum].refnum;
+		lba[di->disk.raid_disk] = di->data_offset;
+	}
+	*updates = mu;
+	return rv;
+}
+
 struct superswitch super_ddf = {
 #ifndef	MDASSEMBLE
 	.examine_super	= examine_super_ddf,
@@ -2687,6 +3118,7 @@ struct superswitch super_ddf = {
 	.detail_super	= detail_super_ddf,
 	.brief_detail_super = brief_detail_super_ddf,
 	.validate_geometry = validate_geometry_ddf,
+	.write_init_super = write_init_super_ddf,
 #endif
 	.match_home	= match_home_ddf,
 	.uuid_from_super= uuid_from_super_ddf,
@@ -2698,15 +3130,13 @@ struct superswitch super_ddf = {
 	.compare_super	= compare_super_ddf,
 
 	.load_super	= load_super_ddf,
-	.init_super	= init_zero_ddf,
+	.init_super	= init_super_ddf,
 	.store_super	= store_zero_ddf,
 	.free_super	= free_super_ddf,
 	.match_metadata_desc = match_metadata_desc_ddf,
-	.getinfo_super_n  = getinfo_super_n_container,
-
+	.add_to_super	= add_to_super_ddf,
+	.container_content = container_content_ddf,
 
-	.major		= 1000,
-	.swapuuid	= 0,
 	.external	= 1,
 
 /* for mdmon */
@@ -2714,71 +3144,8 @@ struct superswitch super_ddf = {
 	.set_array_state= ddf_set_array_state,
 	.set_disk       = ddf_set_disk,
 	.sync_metadata  = ddf_sync_metadata,
+	.process_update	= ddf_process_update,
+	.prepare_update	= ddf_prepare_update,
+	.activate_spare = ddf_activate_spare,
 
-
-};
-
-/* Super_ddf_container is set by validate_geometry_ddf when given a
- * device that is not part of any array
- */
-struct superswitch super_ddf_container = {
-#ifndef MDASSEMBLE
-	.validate_geometry = validate_geometry_ddf_container,
-	.write_init_super = write_init_super_ddf,
-#endif
-
-	.load_super	= load_super_ddf,
-	.init_super	= init_super_ddf,
-	.add_to_super	= add_to_super_ddf,
-	.getinfo_super  = getinfo_super_ddf,
-
-	.free_super	= free_super_ddf,
-
-	.container_content = container_content_ddf,
-	.getinfo_super_n  = getinfo_super_n_container,
-
-	.major		= 1000,
-	.swapuuid	= 0,
-	.external	= 1,
-};
-
-struct superswitch super_ddf_bvd = {
-#ifndef	MDASSEMBLE
-//	.detail_super	= detail_super_ddf_bvd,
-//	.brief_detail_super = brief_detail_super_ddf_bvd,
-	.validate_geometry = validate_geometry_ddf_bvd,
-	.write_init_super = write_init_super_ddf,
-#endif
-	.update_super	= update_super_ddf,
-	.init_super	= init_super_ddf_bvd,
-	.add_to_super	= add_to_super_ddf_bvd,
-	.getinfo_super  = getinfo_super_ddf_bvd,
-	.getinfo_super_n  = getinfo_super_n_bvd,
-
-	.load_super	= load_super_ddf,
-	.free_super	= free_super_ddf,
-	.match_metadata_desc = match_metadata_desc_ddf_bvd,
-
-
-	.major		= 1001,
-	.swapuuid	= 0,
-	.external	= 2,
-};
-
-struct superswitch super_ddf_svd = {
-#ifndef	MDASSEMBLE
-//	.detail_super	= detail_super_ddf_svd,
-//	.brief_detail_super = brief_detail_super_ddf_svd,
-	.validate_geometry = validate_geometry_ddf_svd,
-#endif
-	.update_super	= update_super_ddf,
-	.init_super	= init_super_ddf,
-
-	.load_super	= load_super_ddf,
-	.free_super	= free_super_ddf,
-	.match_metadata_desc = match_metadata_desc_ddf_svd,
-
-	.major		= 1002,
-	.swapuuid	= 0,
-	.external	= 2,
 };