diff --git a/arch/ia64/include/asm/sal.h b/arch/ia64/include/asm/sal.h
index 89594b442f8..ea310c0812a 100644
--- a/arch/ia64/include/asm/sal.h
+++ b/arch/ia64/include/asm/sal.h
@@ -236,7 +236,7 @@ extern struct ia64_sal_desc_ptc *ia64_ptc_domain_info;
 
 extern unsigned short sal_revision;	/* supported SAL spec revision */
 extern unsigned short sal_version;	/* SAL version; OEM dependent */
-#define SAL_VERSION_CODE(major, minor) ((BIN2BCD(major) << 8) | BIN2BCD(minor))
+#define SAL_VERSION_CODE(major, minor) ((bin2bcd(major) << 8) | bin2bcd(minor))
 
 extern const char *ia64_sal_strerror (long status);
 extern void ia64_sal_init (struct ia64_sal_systab *sal_systab);
diff --git a/arch/ia64/kernel/head.S b/arch/ia64/kernel/head.S
index 41c712917ff..8bdea8eb62e 100644
--- a/arch/ia64/kernel/head.S
+++ b/arch/ia64/kernel/head.S
@@ -359,7 +359,31 @@ start_ap:
 	mov ar.rsc=0		// place RSE in enforced lazy mode
 	;;
 	loadrs			// clear the dirty partition
-	mov IA64_KR(PER_CPU_DATA)=r0	// clear physical per-CPU base
+	movl r19=__phys_per_cpu_start
+	mov r18=PERCPU_PAGE_SIZE
+	;;
+#ifndef CONFIG_SMP
+	add r19=r19,r18
+	;;
+#else
+(isAP)	br.few 2f
+	mov r20=r19
+	sub r19=r19,r18
+	;;
+	shr.u r18=r18,3
+1:
+	ld8 r21=[r20],8;;
+	st8[r19]=r21,8
+	adds r18=-1,r18;;
+	cmp4.lt p7,p6=0,r18
+(p7)	br.cond.dptk.few 1b
+2:
+#endif
+	tpa r19=r19
+	;;
+	.pred.rel.mutex isBP,isAP
+(isBP)	mov IA64_KR(PER_CPU_DATA)=r19	// per-CPU base for cpu0
+(isAP)	mov IA64_KR(PER_CPU_DATA)=r0	// clear physical per-CPU base
 	;;
 	mov ar.bspstore=r2	// establish the new RSE stack
 	;;
diff --git a/arch/ia64/kernel/setup.c b/arch/ia64/kernel/setup.c
index 593279f33e9..c27d5b2c182 100644
--- a/arch/ia64/kernel/setup.c
+++ b/arch/ia64/kernel/setup.c
@@ -927,17 +927,19 @@ cpu_init (void)
 	if (smp_processor_id() == 0) {
 		cpu_set(0, per_cpu(cpu_sibling_map, 0));
 		cpu_set(0, cpu_core_map[0]);
+	} else {
+		/*
+		 * Set ar.k3 so that assembly code in MCA handler can compute
+		 * physical addresses of per cpu variables with a simple:
+		 *   phys = ar.k3 + &per_cpu_var
+		 * and the alt-dtlb-miss handler can set per-cpu mapping into
+		 * the TLB when needed. head.S already did this for cpu0.
+		 */
+		ia64_set_kr(IA64_KR_PER_CPU_DATA,
+			    ia64_tpa(cpu_data) - (long) __per_cpu_start);
 	}
 #endif
 
-	/*
-	 * We set ar.k3 so that assembly code in MCA handler can compute
-	 * physical addresses of per cpu variables with a simple:
-	 *   phys = ar.k3 + &per_cpu_var
-	 */
-	ia64_set_kr(IA64_KR_PER_CPU_DATA,
-		    ia64_tpa(cpu_data) - (long) __per_cpu_start);
-
 	get_max_cacheline_size();
 
 	/*
diff --git a/arch/ia64/kernel/smpboot.c b/arch/ia64/kernel/smpboot.c
index 03f1a9908af..b39853a292d 100644
--- a/arch/ia64/kernel/smpboot.c
+++ b/arch/ia64/kernel/smpboot.c
@@ -467,7 +467,9 @@ start_secondary (void *unused)
 {
 	/* Early console may use I/O ports */
 	ia64_set_kr(IA64_KR_IO_BASE, __pa(ia64_iobase));
+#ifndef CONFIG_PRINTK_TIME
 	Dprintk("start_secondary: starting CPU 0x%x\n", hard_smp_processor_id());
+#endif
 	efi_map_pal_code();
 	cpu_init();
 	preempt_disable();
diff --git a/arch/ia64/kernel/vmlinux.lds.S b/arch/ia64/kernel/vmlinux.lds.S
index 5a77206c249..de71da811cd 100644
--- a/arch/ia64/kernel/vmlinux.lds.S
+++ b/arch/ia64/kernel/vmlinux.lds.S
@@ -215,6 +215,9 @@ SECTIONS
   /* Per-cpu data: */
   percpu : { } :percpu
   . = ALIGN(PERCPU_PAGE_SIZE);
+#ifdef	CONFIG_SMP
+  . = . + PERCPU_PAGE_SIZE;	/* cpu0 per-cpu space */
+#endif
   __phys_per_cpu_start = .;
   .data.percpu PERCPU_ADDR : AT(__phys_per_cpu_start - LOAD_OFFSET)
 	{
diff --git a/arch/ia64/mm/contig.c b/arch/ia64/mm/contig.c
index 798bf9835a5..e566ff43884 100644
--- a/arch/ia64/mm/contig.c
+++ b/arch/ia64/mm/contig.c
@@ -163,8 +163,14 @@ per_cpu_init (void)
 	 * get_zeroed_page().
 	 */
 	if (first_time) {
+		void *cpu0_data = __phys_per_cpu_start - PERCPU_PAGE_SIZE;
+
 		first_time=0;
-		for (cpu = 0; cpu < NR_CPUS; cpu++) {
+
+		__per_cpu_offset[0] = (char *) cpu0_data - __per_cpu_start;
+		per_cpu(local_per_cpu_offset, 0) = __per_cpu_offset[0];
+
+		for (cpu = 1; cpu < NR_CPUS; cpu++) {
 			memcpy(cpu_data, __phys_per_cpu_start, __per_cpu_end - __per_cpu_start);
 			__per_cpu_offset[cpu] = (char *) cpu_data - __per_cpu_start;
 			cpu_data += PERCPU_PAGE_SIZE;
@@ -177,7 +183,7 @@ per_cpu_init (void)
 static inline void
 alloc_per_cpu_data(void)
 {
-	cpu_data = __alloc_bootmem(PERCPU_PAGE_SIZE * NR_CPUS,
+	cpu_data = __alloc_bootmem(PERCPU_PAGE_SIZE * NR_CPUS-1,
 				   PERCPU_PAGE_SIZE, __pa(MAX_DMA_ADDRESS));
 }
 #else
diff --git a/arch/ia64/mm/discontig.c b/arch/ia64/mm/discontig.c
index d83125e1ed2..78026aabaa7 100644
--- a/arch/ia64/mm/discontig.c
+++ b/arch/ia64/mm/discontig.c
@@ -143,7 +143,11 @@ static void *per_cpu_node_setup(void *cpu_data, int node)
 	int cpu;
 
 	for_each_possible_early_cpu(cpu) {
-		if (node == node_cpuid[cpu].nid) {
+		if (cpu == 0) {
+			void *cpu0_data = __phys_per_cpu_start - PERCPU_PAGE_SIZE;
+			__per_cpu_offset[cpu] = (char*)cpu0_data -
+				__per_cpu_start;
+		} else if (node == node_cpuid[cpu].nid) {
 			memcpy(__va(cpu_data), __phys_per_cpu_start,
 			       __per_cpu_end - __per_cpu_start);
 			__per_cpu_offset[cpu] = (char*)__va(cpu_data) -