diff --git a/arch/sparc64/kernel/irq.c b/arch/sparc64/kernel/irq.c
index 1f6455503f243bfa04d993f9dd4d9bd6b8527b8b..c5dd6daf127f3338fb43c3c85eda49f77ca566df 100644
--- a/arch/sparc64/kernel/irq.c
+++ b/arch/sparc64/kernel/irq.c
@@ -863,10 +863,10 @@ void init_irqwork_curcpu(void)
 
 static void __cpuinit init_one_mondo(unsigned long *pa_ptr, unsigned long type)
 {
-	register unsigned long func __asm__("%o0");
-	register unsigned long arg0 __asm__("%o1");
-	register unsigned long arg1 __asm__("%o2");
-	register unsigned long arg2 __asm__("%o3");
+	register unsigned long func __asm__("%o5");
+	register unsigned long arg0 __asm__("%o0");
+	register unsigned long arg1 __asm__("%o1");
+	register unsigned long arg2 __asm__("%o2");
 	unsigned long page = get_zeroed_page(GFP_ATOMIC);
 
 	if (!page) {
diff --git a/arch/sparc64/kernel/smp.c b/arch/sparc64/kernel/smp.c
index c10a3a8639e824a6b8d2405dd1b0fc9e4c673f3c..f553264588d6eb293a483429ebe7d2091488a522 100644
--- a/arch/sparc64/kernel/smp.c
+++ b/arch/sparc64/kernel/smp.c
@@ -572,10 +572,10 @@ static void hypervisor_xcall_deliver(u64 data0, u64 data1, u64 data2, cpumask_t
 	retries = 0;
 	cnt = init_cpu_list(cpu_list, mask);
 	do {
-		register unsigned long func __asm__("%o0");
-		register unsigned long arg0 __asm__("%o1");
-		register unsigned long arg1 __asm__("%o2");
-		register unsigned long arg2 __asm__("%o3");
+		register unsigned long func __asm__("%o5");
+		register unsigned long arg0 __asm__("%o0");
+		register unsigned long arg1 __asm__("%o1");
+		register unsigned long arg2 __asm__("%o2");
 
 		func = HV_FAST_CPU_MONDO_SEND;
 		arg0 = cnt;
@@ -624,10 +624,10 @@ static void hypervisor_xcall_deliver(u64 data0, u64 data1, u64 data2, cpumask_t
 		int retries = 0;
 
 		do {
-			register unsigned long func __asm__("%o0");
-			register unsigned long arg0 __asm__("%o1");
-			register unsigned long arg1 __asm__("%o2");
-			register unsigned long arg2 __asm__("%o3");
+			register unsigned long func __asm__("%o5");
+			register unsigned long arg0 __asm__("%o0");
+			register unsigned long arg1 __asm__("%o1");
+			register unsigned long arg2 __asm__("%o2");
 
 			cpu_list[0] = i;
 			func = HV_FAST_CPU_MONDO_SEND;
diff --git a/arch/sparc64/kernel/trampoline.S b/arch/sparc64/kernel/trampoline.S
index ffa8b79632cff4844db85b031f5f883681d6cada..c476f5b321fb481dc852357c71fbfa8dbf295f8f 100644
--- a/arch/sparc64/kernel/trampoline.S
+++ b/arch/sparc64/kernel/trampoline.S
@@ -265,20 +265,20 @@ do_unlock:
 	 nop
 
 niagara_lock_tlb:
-	mov		HV_FAST_MMU_MAP_PERM_ADDR, %o0
-	sethi		%hi(KERNBASE), %o1
-	clr		%o2
-	sethi		%hi(kern_locked_tte_data), %o3
-	ldx		[%o3 + %lo(kern_locked_tte_data)], %o3
-	mov		HV_MMU_IMMU, %o4
+	mov		HV_FAST_MMU_MAP_PERM_ADDR, %o5
+	sethi		%hi(KERNBASE), %o0
+	clr		%o1
+	sethi		%hi(kern_locked_tte_data), %o2
+	ldx		[%o2 + %lo(kern_locked_tte_data)], %o2
+	mov		HV_MMU_IMMU, %o3
 	ta		HV_FAST_TRAP
 
-	mov		HV_FAST_MMU_MAP_PERM_ADDR, %o0
-	sethi		%hi(KERNBASE), %o1
-	clr		%o2
-	sethi		%hi(kern_locked_tte_data), %o3
-	ldx		[%o3 + %lo(kern_locked_tte_data)], %o3
-	mov		HV_MMU_DMMU, %o4
+	mov		HV_FAST_MMU_MAP_PERM_ADDR, %o5
+	sethi		%hi(KERNBASE), %o0
+	clr		%o1
+	sethi		%hi(kern_locked_tte_data), %o2
+	ldx		[%o2 + %lo(kern_locked_tte_data)], %o2
+	mov		HV_MMU_DMMU, %o3
 	ta		HV_FAST_TRAP
 
 	sethi		%hi(bigkernel), %g2
@@ -286,24 +286,24 @@ niagara_lock_tlb:
 	brz,pt		%g2, after_lock_tlb
 	 nop
 
-	mov		HV_FAST_MMU_MAP_PERM_ADDR, %o0
-	sethi		%hi(KERNBASE + 0x400000), %o1
-	clr		%o2
-	sethi		%hi(kern_locked_tte_data), %o3
-	ldx		[%o3 + %lo(kern_locked_tte_data)], %o3
-	sethi		%hi(0x400000), %o4
-	add		%o3, %o4, %o3
-	mov		HV_MMU_IMMU, %o4
+	mov		HV_FAST_MMU_MAP_PERM_ADDR, %o5
+	sethi		%hi(KERNBASE + 0x400000), %o0
+	clr		%o1
+	sethi		%hi(kern_locked_tte_data), %o2
+	ldx		[%o2 + %lo(kern_locked_tte_data)], %o2
+	sethi		%hi(0x400000), %o3
+	add		%o2, %o3, %o2
+	mov		HV_MMU_IMMU, %o3
 	ta		HV_FAST_TRAP
 
-	mov		HV_FAST_MMU_MAP_PERM_ADDR, %o0
-	sethi		%hi(KERNBASE + 0x400000), %o1
-	clr		%o2
-	sethi		%hi(kern_locked_tte_data), %o3
-	ldx		[%o3 + %lo(kern_locked_tte_data)], %o3
-	sethi		%hi(0x400000), %o4
-	add		%o3, %o4, %o3
-	mov		HV_MMU_DMMU, %o4
+	mov		HV_FAST_MMU_MAP_PERM_ADDR, %o5
+	sethi		%hi(KERNBASE + 0x400000), %o0
+	clr		%o1
+	sethi		%hi(kern_locked_tte_data), %o2
+	ldx		[%o2 + %lo(kern_locked_tte_data)], %o2
+	sethi		%hi(0x400000), %o3
+	add		%o2, %o3, %o2
+	mov		HV_MMU_DMMU, %o3
 	ta		HV_FAST_TRAP
 
 after_lock_tlb:
diff --git a/arch/sparc64/kernel/tsb.S b/arch/sparc64/kernel/tsb.S
index a53ec6fb7697669dfa918c0c61bd86adc76e7e83..8a9351258af8829277e19de58ec9a5f49352f5b1 100644
--- a/arch/sparc64/kernel/tsb.S
+++ b/arch/sparc64/kernel/tsb.S
@@ -266,9 +266,9 @@ __tsb_context_switch:
 	mov	SCRATCHPAD_UTSBREG2, %g1
 	stxa	%g2, [%g1] ASI_SCRATCHPAD
 
-	mov	HV_FAST_MMU_TSB_CTXNON0, %o0
-	mov	1, %o1
-	mov	%o4, %o2
+	mov	HV_FAST_MMU_TSB_CTXNON0, %o5
+	mov	1, %o0
+	mov	%o4, %o1
 	ta	HV_FAST_TRAP
 
 	ba,pt	%xcc, 9f
diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index e602b857071acdf30124e0430ef7c3cc1031cd67..7faba33202a9052ebf06960058403a48647ebc75 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -518,11 +518,11 @@ static void __init hypervisor_tlb_lock(unsigned long vaddr,
 				       unsigned long pte,
 				       unsigned long mmu)
 {
-	register unsigned long func asm("%o0");
-	register unsigned long arg0 asm("%o1");
-	register unsigned long arg1 asm("%o2");
-	register unsigned long arg2 asm("%o3");
-	register unsigned long arg3 asm("%o4");
+	register unsigned long func asm("%o5");
+	register unsigned long arg0 asm("%o0");
+	register unsigned long arg1 asm("%o1");
+	register unsigned long arg2 asm("%o2");
+	register unsigned long arg3 asm("%o3");
 
 	func = HV_FAST_MMU_MAP_PERM_ADDR;
 	arg0 = vaddr;
@@ -1112,18 +1112,18 @@ static void __init tsb_phys_patch(void)
 /* Register this cpu's fault status area with the hypervisor.  */
 void __cpuinit sun4v_register_fault_status(void)
 {
+	register unsigned long func asm("%o5");
 	register unsigned long arg0 asm("%o0");
-	register unsigned long arg1 asm("%o1");
 	int cpu = hard_smp_processor_id();
 	struct trap_per_cpu *tb = &trap_block[cpu];
 	unsigned long pa;
 
 	pa = kern_base + ((unsigned long) tb - KERNBASE);
-	arg0 = HV_FAST_MMU_FAULT_AREA_CONF;
-	arg1 = pa;
+	func = HV_FAST_MMU_FAULT_AREA_CONF;
+	arg0 = pa;
 	__asm__ __volatile__("ta	%4"
-			     : "=&r" (arg0), "=&r" (arg1)
-			     : "0" (arg0), "1" (arg1),
+			     : "=&r" (func), "=&r" (arg0)
+			     : "0" (func), "1" (arg0),
 			       "i" (HV_FAST_TRAP));
 }