diff --git a/libcpu/arm/arm926/start_gcc.S b/libcpu/arm/arm926/start_gcc.S
index aeeb37616b..4924583cb8 100644
--- a/libcpu/arm/arm926/start_gcc.S
+++ b/libcpu/arm/arm926/start_gcc.S
@@ -22,6 +22,7 @@
  * 2011-01-13     weety      first version
  * 2015-04-15     ArdaFu     Split from AT91SAM9260 BSP
  * 2015-04-21     ArdaFu     Remove remap code. Using mmu to map vector table
+ * 2015-06-04     aozima     Align stack address to 8 byte.
  */
  
 #define S_FRAME_SIZE    (18*4)   //72
@@ -61,31 +62,32 @@
     .section .nobss, "w"
 
     .space UND_STK_SIZE
+    .align 3
     .global UND_STACK_START
 UND_STACK_START:
 
     .space ABT_STK_SIZE
-    .align 2
+    .align 3
     .global ABT_STACK_START
 ABT_STACK_START:
 
     .space FIQ_STK_SIZE
-    .align 2
+    .align 3
     .global FIQ_STACK_START
 FIQ_STACK_START:
 
     .space IRQ_STK_SIZE
-    .align 2
+    .align 3
     .global IRQ_STACK_START
 IRQ_STACK_START:
 
     .skip SYS_STK_SIZE
-    .align 2
+    .align 3
     .global SYS_STACK_START
 SYS_STACK_START:
 
     .space SVC_STK_SIZE
-    .align 2
+    .align 3
     .global SVC_STACK_START
 SVC_STACK_START:
 
diff --git a/libcpu/arm/arm926/start_iar.S b/libcpu/arm/arm926/start_iar.S
index 4e689236a5..3494b108d7 100644
--- a/libcpu/arm/arm926/start_iar.S
+++ b/libcpu/arm/arm926/start_iar.S
@@ -22,6 +22,7 @@
 ; * 2011-01-13     weety      first version
 ; * 2015-04-15     ArdaFu     Split from AT91SAM9260 BSP
 ; * 2015-04-21     ArdaFu     Remove remap code. Using mmu to map vector table
+; * 2015-06-04     aozima     Align stack address to 8 byte.
 ; */
  
 #define S_FRAME_SIZE    (18*4)   ;72
@@ -62,31 +63,32 @@
     SECTION .noinit:DATA:NOROOT(3)
     DATA
 
+    ALIGNRAM 3
     DS8 UND_STK_SIZE
     PUBLIC UND_STACK_START
 UND_STACK_START:
 
-    ALIGNRAM 2
+    ALIGNRAM 3
     DS8 ABT_STK_SIZE
     PUBLIC ABT_STACK_START
 ABT_STACK_START:
 
-    ALIGNRAM 2
+    ALIGNRAM 3
     DS8 FIQ_STK_SIZE
     PUBLIC FIQ_STACK_START
 FIQ_STACK_START:
 
-    ALIGNRAM 2
+    ALIGNRAM 3
     DS8 IRQ_STK_SIZE
     PUBLIC IRQ_STACK_START
 IRQ_STACK_START:
 
-    ALIGNRAM 2
+    ALIGNRAM 3
     DS8 SYS_STK_SIZE
     PUBLIC SYS_STACK_START
 SYS_STACK_START:
 
-    ALIGNRAM 2
+    ALIGNRAM 3
     DS8 SVC_STK_SIZE
     PUBLIC SVC_STACK_START
 SVC_STACK_START:
diff --git a/libcpu/arm/arm926/start_rvds.S b/libcpu/arm/arm926/start_rvds.S
index e086e394f1..68fd887d66 100644
--- a/libcpu/arm/arm926/start_rvds.S
+++ b/libcpu/arm/arm926/start_rvds.S
@@ -22,6 +22,7 @@
 ; * 2011-08-14     weety      first version
 ; * 2015-04-15     ArdaFu     Split from AT91SAM9260 BSP
 ; * 2015-04-21     ArdaFu     Remove remap code. Using mmu to map vector table
+; * 2015-06-04     aozima     Align stack address to 8 byte.
 ; */
 
 S_FRAME_SIZE    EQU     (18*4)   ;72
@@ -60,34 +61,34 @@ NOINT           EQU     0xC0
     GET rt_low_level_keil.inc
     
 ;----------------------- Stack and Heap Definitions ----------------------------
-    AREA    STACK, NOINIT, READWRITE, ALIGN=2
+    AREA    STACK, NOINIT, READWRITE, ALIGN=3
 Stack_Mem
 
     SPACE   UND_STK_SIZE
     EXPORT UND_STACK_START
 UND_STACK_START
 
-    ALIGN 4
+    ALIGN 8
     SPACE   ABT_STK_SIZE
     EXPORT ABT_STACK_START
 ABT_STACK_START
 
-    ALIGN 4
+    ALIGN 8
     SPACE   FIQ_STK_SIZE
     EXPORT FIQ_STACK_START
 FIQ_STACK_START
 
-    ALIGN 4
+    ALIGN 8
     SPACE   IRQ_STK_SIZE
     EXPORT IRQ_STACK_START
 IRQ_STACK_START
 
-    ALIGN 4
+    ALIGN 8
     SPACE   SYS_STK_SIZE
     EXPORT SYS_STACK_START
 SYS_STACK_START
 
-    ALIGN 4
+    ALIGN 8
     SPACE   SVC_STK_SIZE
     EXPORT SVC_STACK_START
 SVC_STACK_START