diff --git a/src/arch/i386/include/div64.h b/src/arch/i386/include/div64.h
new file mode 100644
index 0000000000..28ed8b296a
--- /dev/null
+++ b/src/arch/i386/include/div64.h
@@ -0,0 +1,48 @@
+#ifndef __I386_DIV64
+#define __I386_DIV64
+
+/*
+ * do_div() is NOT a C function. It wants to return
+ * two values (the quotient and the remainder), but
+ * since that doesn't work very well in C, what it
+ * does is:
+ *
+ * - modifies the 64-bit dividend _in_place_
+ * - returns the 32-bit remainder
+ *
+ * This ends up being the most efficient "calling
+ * convention" on x86.
+ */
+#define do_div(n,base) ({ \
+	unsigned long __upper, __low, __high, __mod, __base; \
+	__base = (base); \
+	asm("":"=a" (__low), "=d" (__high):"A" (n)); \
+	__upper = __high; \
+	if (__high) { \
+		__upper = __high % (__base); \
+		__high = __high / (__base); \
+	} \
+	asm("divl %2":"=a" (__low), "=d" (__mod):"rm" (__base), "0" (__low), "1" (__upper)); \
+	asm("":"=A" (n):"a" (__low),"d" (__high)); \
+	__mod; \
+})
+
+/*
+ * (long)X = ((long long)divs) / (long)div
+ * (long)rem = ((long long)divs) % (long)div
+ *
+ * Warning, this will do an exception if X overflows.
+ */
+#define div_long_long_rem(a,b,c) div_ll_X_l_rem(a,b,c)
+
+extern inline long
+div_ll_X_l_rem(long long divs, long div, long *rem)
+{
+	long dum2;
+      __asm__("divl %2":"=a"(dum2), "=d"(*rem)
+      :	"rm"(div), "A"(divs));
+
+	return dum2;
+
+}
+#endif
diff --git a/src/arch/i386/smp/mpspec.c b/src/arch/i386/smp/mpspec.c
index c674cf58f5..daeb78435d 100644
--- a/src/arch/i386/smp/mpspec.c
+++ b/src/arch/i386/smp/mpspec.c
@@ -121,7 +121,7 @@ void smp_write_processors(struct mp_config_table *mc)
 		}
 		cpu_flag = MPC_CPU_ENABLED;
 		if (boot_apic_id == cpu->path.u.apic.apic_id) {
-			cpu_flag = MPC_CPU_ENABLED;
+			cpu_flag = MPC_CPU_ENABLED | MPC_CPU_BOOTPROCESSOR;
 		}
 		smp_write_processor(mc, 
 			cpu->path.u.apic.apic_id, apic_version,
diff --git a/src/console/vtxprintf.c b/src/console/vtxprintf.c
index 0f12672a50..26a5d24351 100644
--- a/src/console/vtxprintf.c
+++ b/src/console/vtxprintf.c
@@ -6,6 +6,7 @@
 
 #include <stdarg.h>
 #include <string.h>
+#include <div64.h>
 
 /* haha, don't need ctype.c */
 #define isdigit(c)	((c) >= '0' && (c) <= '9')
@@ -62,14 +63,8 @@ static int skip_atoi(const char **s)
 #define SPECIAL	32		/* 0x */
 #define LARGE	64		/* use 'ABCDEF' instead of 'abcdef' */
 
-#define do_div(n,base) ({ \
-int __res; \
-__res = ((unsigned long) n) % (unsigned) base; \
-n = ((unsigned long) n) / (unsigned) base; \
-__res; })
-
-static int number(void (*tx_byte)(unsigned char byte), long num, int base, int size, int precision
-	,int type)
+static int number(void (*tx_byte)(unsigned char byte), 
+	unsigned long long num, int base, int size, int precision, int type)
 {
 	char c,sign,tmp[66];
 	const char *digits="0123456789abcdefghijklmnopqrstuvwxyz";
@@ -85,7 +80,7 @@ static int number(void (*tx_byte)(unsigned char byte), long num, int base, int s
 	c = (type & ZEROPAD) ? '0' : ' ';
 	sign = 0;
 	if (type & SIGN) {
-		if (num < 0) {
+		if ((signed long long)num < 0) {
 			sign = '-';
 			num = -num;
 			size--;
@@ -140,7 +135,7 @@ static int number(void (*tx_byte)(unsigned char byte), long num, int base, int s
 int vtxprintf(void (*tx_byte)(unsigned char byte), const char *fmt, va_list args)
 {
 	int len;
-	unsigned long num;
+	unsigned long long num;
 	int i, base;
 	const char *s;
 
@@ -205,6 +200,10 @@ int vtxprintf(void (*tx_byte)(unsigned char byte), const char *fmt, va_list args
 		if (*fmt == 'h' || *fmt == 'l' || *fmt == 'L') {
 			qualifier = *fmt;
 			++fmt;
+			if (*fmt == 'l') {
+				qualifier = 'L';
+				++fmt;
+			}
 		}
 
 		/* default base */
@@ -248,7 +247,10 @@ int vtxprintf(void (*tx_byte)(unsigned char byte), const char *fmt, va_list args
 
 
 		case 'n':
-			if (qualifier == 'l') {
+			if (qualifier == 'L') {
+				long long *ip = va_arg(args, long long *);
+				*ip = count;
+			} else if (qualifier == 'l') {
 				long * ip = va_arg(args, long *);
 				*ip = count;
 			} else {
@@ -286,16 +288,19 @@ int vtxprintf(void (*tx_byte)(unsigned char byte), const char *fmt, va_list args
 				--fmt;
 			continue;
 		}
-		if (qualifier == 'l')
+		if (qualifier == 'L') {
+			num = va_arg(args, unsigned long long);
+		} else if (qualifier == 'l') {
 			num = va_arg(args, unsigned long);
-		else if (qualifier == 'h') {
+		} else if (qualifier == 'h') {
 			num = (unsigned short) va_arg(args, int);
 			if (flags & SIGN)
 				num = (short) num;
-		} else if (flags & SIGN)
+		} else if (flags & SIGN) {
 			num = va_arg(args, int);
-		else
+		} else {
 			num = va_arg(args, unsigned int);
+		}
 		count += number(tx_byte, num, base, field_width, precision, flags);
 	}
 	return count;
diff --git a/src/cpu/x86/lapic/lapic_cpu_init.c b/src/cpu/x86/lapic/lapic_cpu_init.c
index db0cecbd06..b96c7b3586 100644
--- a/src/cpu/x86/lapic/lapic_cpu_init.c
+++ b/src/cpu/x86/lapic/lapic_cpu_init.c
@@ -261,7 +261,7 @@ static void initialize_other_cpus(struct bus *cpu_bus)
 	active_count = atomic_read(&active_cpus);
 	while(active_count > 1) {
 		if (active_count != old_active_count) {
-			printk_info("Waiting for %d CPUS to stop\n", active_count);
+			printk_info("Waiting for %d CPUS to stop\n", active_count - 1);
 			old_active_count = active_count;
 		}
 		udelay(10);
diff --git a/src/northbridge/amd/amdk8/northbridge.c b/src/northbridge/amd/amdk8/northbridge.c
index cb2f9fa0a7..ca5dcd82a3 100644
--- a/src/northbridge/amd/amdk8/northbridge.c
+++ b/src/northbridge/amd/amdk8/northbridge.c
@@ -598,7 +598,7 @@ static void pci_domain_set_resources(device_t dev)
 		mem1->base = resource_max(mem1);
 	}
 
-#if 1
+#if 0
 		printk_debug("base1: 0x%08Lx limit1: 0x%08lx size: 0x%08Lx\n",
 			mem1->base, mem1->limit, mem1->size);
 		printk_debug("base2: 0x%08Lx limit2: 0x%08Lx size: 0x%08Lx\n",
@@ -702,7 +702,7 @@ static struct device_operations pci_domain_ops = {
 	.scan_bus         = pci_domain_scan_bus,
 };
 
-static unsigned int scan_cpu_bus(device_t dev, unsigned int max)
+static unsigned int cpu_bus_scan(device_t dev, unsigned int max)
 {
 	struct bus *cpu_bus;
 	unsigned reg;
@@ -750,18 +750,19 @@ static unsigned int scan_cpu_bus(device_t dev, unsigned int max)
 
 static void cpu_bus_init(device_t dev)
 {
-	printk_debug("cpu_bus_init\n");
-#if 0
 	initialize_cpus(&dev->link[0]);
-#endif
+}
+
+static void cpu_bus_noop(device_t dev) 
+{
 }
 
 static struct device_operations cpu_bus_ops = {
-	.read_resources   = 0,
-	.set_resources    = 0,
-	.enable_resources = 0,
+	.read_resources   = cpu_bus_noop,
+	.set_resources    = cpu_bus_noop,
+	.enable_resources = cpu_bus_noop,
 	.init             = cpu_bus_init,	
-	.scan_bus         = scan_cpu_bus,
+	.scan_bus         = cpu_bus_scan,
 };
 
 static void enable_dev(struct device *dev)
diff --git a/util/newconfig/config.g b/util/newconfig/config.g
index 3f9b5a379e..d01f336a24 100644
--- a/util/newconfig/config.g
+++ b/util/newconfig/config.g
@@ -724,6 +724,8 @@ class partobj:
 	def firstsiblingdevice(self):
 		"""Find the first device in the sibling link."""
 		sibling = self.siblings
+		if ((not sibling) and (self.parent.chip_or_device == 'chip')):
+			sibling = self.parent.siblings
 		while(sibling):
 			if (sibling.chip_or_device == 'device'):
 				return sibling
@@ -921,7 +923,7 @@ class partobj:
 			fatal("Invalid pci_domain: 0x%x is out of the range 0 to 0xffff" % pci_domain)
 		self.path = ".type=DEVICE_PATH_PCI_DOMAIN,.u={.pci_domain={ .domain = 0x%x }}" % (pci_domain)
     
-	def addapic_cluster(self, cluster):
+	def addapic_clusterpath(self, cluster):
 		""" Add a pci_domain number to a chip """
 		if ((cluster < 0) or (cluster > 15)):
 			fatal("Invalid apic cluster: %d is out of the range 0 to ff" % cluster)
@@ -1689,10 +1691,10 @@ parser Config:
 			) 			{{ if(C): partstack.tos().set_enabled(val) }}
 
     rule resource<<C>>:				{{ type = "" }}
-	    		(  IO			{{ type = "IORESOURCE_IO" }}
-			|   MEM			{{ type = "IORESOURCE_MEM" }}
-			|   IRQ			{{ type = "IORESOURCE_IRQ" }}
-			|   DRQ			{{ type = "IORESOURCE_DRQ" }}
+	    		(  IO			{{ type = "IORESOURCE_FIXED | IORESOURCE_IO" }}
+			|   MEM			{{ type = "IORESOURCE_FIXED | IORESOURCE_MEM" }}
+			|   IRQ			{{ type = "IORESOURCE_FIXED | IORESOURCE_IRQ" }}
+			|   DRQ			{{ type = "IORESOURCE_FIXED | IORESOURCE_DRQ" }}
 			)
 			term '='		{{ index = term }}
 			term			{{ value = term }}
@@ -1729,7 +1731,7 @@ parser Config:
 
     rule apic_cluster<<C>>: APIC_CLUSTER 	{{ if (C): devicepart('apic_cluster') }}
 			HEX_NUM			{{ cluster = int(HEX_NUM, 16) }}
-						{{ if (C): partstack.tos().addapicpath(cluster) }}
+						{{ if (C): partstack.tos().addapic_clusterpath(cluster) }}
 
     rule dev_path<<C>>:				
 	    		pci<<C>>		{{ return pci }}