diff --git a/arch/arm/cpu/armv8/fsl-layerscape/cpu.c b/arch/arm/cpu/armv8/fsl-layerscape/cpu.c
index 374fde6b65..6c87c1b11a 100644
--- a/arch/arm/cpu/armv8/fsl-layerscape/cpu.c
+++ b/arch/arm/cpu/armv8/fsl-layerscape/cpu.c
@@ -5,6 +5,7 @@
  */
 
 #include <common.h>
+#include <cpu_func.h>
 #include <env.h>
 #include <fsl_ddr_sdram.h>
 #include <vsprintf.h>
diff --git a/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch2_speed.c b/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch2_speed.c
index 6d82cfeb58..25e9a495f7 100644
--- a/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch2_speed.c
+++ b/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch2_speed.c
@@ -5,6 +5,7 @@
  */
 
 #include <common.h>
+#include <cpu_func.h>
 #include <linux/compiler.h>
 #include <asm/io.h>
 #include <asm/processor.h>
diff --git a/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch3_speed.c b/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch3_speed.c
index ede96742aa..4b047a39c0 100644
--- a/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch3_speed.c
+++ b/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch3_speed.c
@@ -7,6 +7,7 @@
  */
 
 #include <common.h>
+#include <cpu_func.h>
 #include <linux/compiler.h>
 #include <fsl_ifc.h>
 #include <asm/processor.h>
diff --git a/arch/arm/cpu/armv8/s32v234/cpu.c b/arch/arm/cpu/armv8/s32v234/cpu.c
index b4cb67a66a..b5a9513ead 100644
--- a/arch/arm/cpu/armv8/s32v234/cpu.c
+++ b/arch/arm/cpu/armv8/s32v234/cpu.c
@@ -4,6 +4,7 @@
  */
 
 #include <common.h>
+#include <cpu_func.h>
 #include <asm/io.h>
 #include <asm/system.h>
 #include <asm/armv8/mmu.h>
diff --git a/arch/powerpc/cpu/mpc85xx/speed.c b/arch/powerpc/cpu/mpc85xx/speed.c
index acc2f2bb81..15b05fcc51 100644
--- a/arch/powerpc/cpu/mpc85xx/speed.c
+++ b/arch/powerpc/cpu/mpc85xx/speed.c
@@ -10,6 +10,7 @@
  */
 
 #include <common.h>
+#include <cpu_func.h>
 #include <ppc_asm.tmpl>
 #include <linux/compiler.h>
 #include <asm/processor.h>
diff --git a/arch/powerpc/cpu/mpc86xx/cpu.c b/arch/powerpc/cpu/mpc86xx/cpu.c
index 1c3c78217c..bb14444a2e 100644
--- a/arch/powerpc/cpu/mpc86xx/cpu.c
+++ b/arch/powerpc/cpu/mpc86xx/cpu.c
@@ -6,6 +6,7 @@
  */
 
 #include <common.h>
+#include <cpu_func.h>
 #include <vsprintf.h>
 #include <watchdog.h>
 #include <command.h>
diff --git a/arch/powerpc/cpu/mpc8xxx/cpu.c b/arch/powerpc/cpu/mpc8xxx/cpu.c
index 467eac4a2e..ed482a9c09 100644
--- a/arch/powerpc/cpu/mpc8xxx/cpu.c
+++ b/arch/powerpc/cpu/mpc8xxx/cpu.c
@@ -10,6 +10,7 @@
 #include <config.h>
 #include <common.h>
 #include <command.h>
+#include <cpu_func.h>
 #include <tsec.h>
 #include <fm_eth.h>
 #include <netdev.h>
diff --git a/arch/powerpc/cpu/mpc8xxx/fdt.c b/arch/powerpc/cpu/mpc8xxx/fdt.c
index 0d877c43be..485c2d4feb 100644
--- a/arch/powerpc/cpu/mpc8xxx/fdt.c
+++ b/arch/powerpc/cpu/mpc8xxx/fdt.c
@@ -8,6 +8,7 @@
  */
 
 #include <common.h>
+#include <cpu_func.h>
 #include <linux/libfdt.h>
 #include <fdt_support.h>
 #include <asm/mp.h>
diff --git a/board/freescale/qemu-ppce500/qemu-ppce500.c b/board/freescale/qemu-ppce500/qemu-ppce500.c
index fb36d8366c..4a5ab72b1f 100644
--- a/board/freescale/qemu-ppce500/qemu-ppce500.c
+++ b/board/freescale/qemu-ppce500/qemu-ppce500.c
@@ -5,6 +5,7 @@
 
 #include <common.h>
 #include <command.h>
+#include <cpu_func.h>
 #include <env.h>
 #include <pci.h>
 #include <asm/processor.h>
diff --git a/include/common.h b/include/common.h
index 992675100d..4de05032b2 100644
--- a/include/common.h
+++ b/include/common.h
@@ -203,26 +203,6 @@ void	relocate_code(ulong, gd_t *, ulong) __attribute__ ((noreturn));
 ulong	get_endaddr   (void);
 void	trap_init     (ulong);
 
-/* $(CPU)/cpu.c */
-static inline int cpumask_next(int cpu, unsigned int mask)
-{
-	for (cpu++; !((1 << cpu) & mask); cpu++)
-		;
-
-	return cpu;
-}
-
-#define for_each_cpu(iter, cpu, num_cpus, mask) \
-	for (iter = 0, cpu = cpumask_next(-1, mask); \
-		iter < num_cpus; \
-		iter++, cpu = cpumask_next(cpu, mask)) \
-
-int	cpu_numcores  (void);
-int	cpu_num_dspcores(void);
-u32	cpu_mask      (void);
-u32	cpu_dsp_mask(void);
-int	is_core_valid (unsigned int);
-
 void s_init(void);
 
 int	checkcpu      (void);