diff --git a/arch/arm/src/armv7-a/arm_cpuhead.S b/arch/arm/src/armv7-a/arm_cpuhead.S
index 412029bc474..be95869e3db 100644
--- a/arch/arm/src/armv7-a/arm_cpuhead.S
+++ b/arch/arm/src/armv7-a/arm_cpuhead.S
@@ -101,6 +101,12 @@ __cpu1_start:
 
 	b		.Lcpu_start
 
+	.globl	cpu1_start
+	.type	cpu1_start, object
+
+	cpu1_start:
+	.long	__cpu1_start
+
 .Lcpu1_stackpointer:
 	.long	.Lcpu1_stacktop
 	.size	__cpu1_start, .-__cpu1_start
@@ -124,6 +130,12 @@ __cpu2_start:
 
 	b		.Lcpu_start
 
+	.globl	cpu2_start
+	.type	cpu2_start, object
+
+	cpu2_start:
+	.long	__cpu2_start
+
 .Lcpu2_stackpointer:
 	.long	.Lcpu2_stacktop
 	.size	__cpu2_start, .-__cpu2_start
@@ -147,6 +159,12 @@ __cpu3_start:
 
 	b		.Lcpu_start
 
+	.globl	cpu3_start
+	.type	cpu3_start, object
+
+	cpu3_start:
+	.long	__cpu3_start
+
 .Lcpu3_stackpointer:
 	.long	.Lcpu3_stacktop
 	.size	__cpu3_start, .-__cpu3_start
diff --git a/arch/arm/src/armv7-a/arm_head.S b/arch/arm/src/armv7-a/arm_head.S
index fe23be41c0c..6abc2a26e89 100644
--- a/arch/arm/src/armv7-a/arm_head.S
+++ b/arch/arm/src/armv7-a/arm_head.S
@@ -197,19 +197,24 @@ __start:
 	wfe
 #endif
 
+	ldr		r1, =cpu1_start
+	ldr		r1, [r1]
 	cmp		r0, #1
-	beq		__cpu1_start
+	bxeq		r1
 #  if CONFIG_NCPUS > 2
+	ldr		r1, =cpu2_start
+	ldr		r1, [r1]
 	cmp		r0, #2
-	beq		__cpu2_start
+	bxeq		r1
 #  endif
 #  if CONFIG_NCPUS > 3
+	ldr		r1, =cpu3_start
+	ldr		r1, [r1]
 	cmp		r0, #3
-	beq		__cpu3_start
+	bxeq		r1
 #  endif
 #  if CONFIG_NCPUS > 4
-	cmp		r0, #4
-	beq		__cpu4_start
+#    error This logic needs to extended for CONFIG_SMP_NCPUS > 4
 #  endif
 
 __cpu0_start:
diff --git a/arch/arm/src/armv7-r/arm_cpuhead.S b/arch/arm/src/armv7-r/arm_cpuhead.S
index 8044fd11d57..18970ec99bd 100644
--- a/arch/arm/src/armv7-r/arm_cpuhead.S
+++ b/arch/arm/src/armv7-r/arm_cpuhead.S
@@ -127,6 +127,12 @@ __cpu1_start:
 
 	b		.Lcpu_start
 
+	.globl	cpu1_start
+	.type	cpu1_start, object
+
+	cpu1_start:
+	.long	__cpu1_start
+
 .Lcpu1_stackpointer:
 	.long	.Lcpu1_stacktop
 	.size	__cpu1_start, .-__cpu1_start
@@ -150,6 +156,12 @@ __cpu2_start:
 
 	b		.Lcpu_start
 
+	.globl	cpu2_start
+	.type	cpu2_start, object
+
+	cpu2_start:
+	.long	__cpu2_start
+
 .Lcpu2_stackpointer:
 	.long	.Lcpu2_stacktop
 	.size	__cpu2_start, .-__cpu2_start
@@ -173,6 +185,12 @@ __cpu3_start:
 
 	b		.Lcpu_start
 
+	.globl	cpu3_start
+	.type	cpu3_start, object
+
+	cpu3_start:
+	.long	__cpu3_start
+
 .Lcpu3_stackpointer:
 	.long	.Lcpu3_stacktop
 	.size	__cpu3_start, .-__cpu3_start
diff --git a/arch/arm/src/armv7-r/arm_head.S b/arch/arm/src/armv7-r/arm_head.S
index 5d8e254bedf..8d63ad25140 100644
--- a/arch/arm/src/armv7-r/arm_head.S
+++ b/arch/arm/src/armv7-r/arm_head.S
@@ -145,19 +145,24 @@ __start:
 	wfe
 #endif
 
+	ldr		r1, =cpu1_start
+	ldr		r1, [r1]
 	cmp		r0, #1
-	beq		__cpu1_start
+	bxeq		r1
 #  if CONFIG_NCPUS > 2
+	ldr		r1, =cpu2_start
+	ldr		r1, [r1]
 	cmp		r0, #2
-	beq		__cpu2_start
+	bxeq		r1
 #  endif
 #  if CONFIG_NCPUS > 3
+	ldr		r1, =cpu3_start
+	ldr		r1, [r1]
 	cmp		r0, #3
-	beq		__cpu3_start
+	bxeq		r1
 #  endif
 #  if CONFIG_NCPUS > 4
-	cmp		r0, #4
-	beq		__cpu4_start
+#    error This logic needs to extended for CONFIG_SMP_NCPUS > 4
 #  endif
 
 __cpu0_start: