diff --git a/arch/arm/lib/delay.S b/arch/arm/lib/delay.S
index b3fb475b4120..9183b06c0e2f 100644
--- a/arch/arm/lib/delay.S
+++ b/arch/arm/lib/delay.S
@@ -9,28 +9,32 @@
  */
 #include <linux/linkage.h>
 #include <asm/assembler.h>
+#include <asm/param.h>
 		.text
 
 .LC0:		.word	loops_per_jiffy
+.LC1:		.word	(2199023*HZ)>>11
 
 /*
- * 0 <= r0 <= 2000
+ * r0  <= 2000
+ * lpj <= 0x01ffffff (max. 3355 bogomips)
+ * HZ  <= 1000
  */
+
 ENTRY(__udelay)
-		mov	r2,     #0x6800
-		orr	r2, r2, #0x00db
+		ldr	r2, .LC1
 		mul	r0, r2, r0
-ENTRY(__const_udelay)				@ 0 <= r0 <= 0x01ffffff
+ENTRY(__const_udelay)				@ 0 <= r0 <= 0x7fffff06
 		ldr	r2, .LC0
-		ldr	r2, [r2]		@ max = 0x0fffffff
-		mov	r0, r0, lsr #11		@ max = 0x00003fff
-		mov	r2, r2, lsr #11		@ max = 0x0003ffff
+		ldr	r2, [r2]		@ max = 0x01ffffff
+		mov	r0, r0, lsr #14		@ max = 0x0001ffff
+		mov	r2, r2, lsr #10		@ max = 0x00007fff
 		mul	r0, r2, r0		@ max = 2^32-1
 		movs	r0, r0, lsr #6
 		RETINSTR(moveq,pc,lr)
 
 /*
- * loops = (r0 * 0x10c6 * 100 * loops_per_jiffy) / 2^32
+ * loops = r0 * HZ * loops_per_jiffy / 1000000
  *
  * Oh, if only we had a cycle counter...
  */
diff --git a/include/asm-arm/delay.h b/include/asm-arm/delay.h
index 1704360e9699..b2deda181549 100644
--- a/include/asm-arm/delay.h
+++ b/include/asm-arm/delay.h
@@ -6,6 +6,8 @@
 #ifndef __ASM_ARM_DELAY_H
 #define __ASM_ARM_DELAY_H
 
+#include <asm/param.h>	/* HZ */
+
 extern void __delay(int loops);
 
 /*
@@ -13,7 +15,7 @@ extern void __delay(int loops);
  * it, it means that you're calling udelay() with an out of range value.
  *
  * With currently imposed limits, this means that we support a max delay
- * of 2000us and 671 bogomips
+ * of 2000us. Further limits: HZ<=1000 and bogomips<=3355
  */
 extern void __bad_udelay(void);
 
@@ -32,10 +34,10 @@ extern void __const_udelay(unsigned long);
 
 #define MAX_UDELAY_MS 2
 
-#define udelay(n)						\
-	(__builtin_constant_p(n) ?				\
-	  ((n) > (MAX_UDELAY_MS * 1000) ? __bad_udelay() :	\
-			__const_udelay((n) * 0x68dbul)) :	\
+#define udelay(n)							\
+	(__builtin_constant_p(n) ?					\
+	  ((n) > (MAX_UDELAY_MS * 1000) ? __bad_udelay() :		\
+			__const_udelay((n) * ((2199023U*HZ)>>11))) :	\
 	  __udelay(n))
 
 #endif /* defined(_ARM_DELAY_H) */