3 files changed, 84 insertions, 32 deletions
diff --git a/arch/parisc/include/asm/delay.h b/arch/parisc/include/asm/delay.h
index 912ee7e6a579..08e58e679e3e 100644
--- a/arch/parisc/include/asm/delay.h
+++ b/arch/parisc/include/asm/delay.h
@@ -1,15 +1,5 @@
-#ifndef _PARISC_DELAY_H
-#define _PARISC_DELAY_H
-
-#include <asm/special_insns.h>    /* for mfctl() */
-#include <asm/processor.h> /* for boot_cpu_data */
-
-
-/*
- * Copyright (C) 1993 Linus Torvalds
- *
- * Delay routines
- */
+#ifndef _ASM_PARISC_DELAY_H
+#define _ASM_PARISC_DELAY_H
 
 static __inline__ void __delay(unsigned long loops) {
 	asm volatile(
@@ -19,25 +9,14 @@ static __inline__ void __delay(unsigned long loops) {
 		: "=r" (loops) : "0" (loops));
 }
 
-static __inline__ void __cr16_delay(unsigned long clocks) {
-	unsigned long start;
-
-	/*
-	 * Note: Due to unsigned math, cr16 rollovers shouldn't be
-	 * a problem here. However, on 32 bit, we need to make sure
-	 * we don't pass in too big a value. The current default
-	 * value of MAX_UDELAY_MS should help prevent this.
-	 */
+extern void __udelay(unsigned long usecs);
+extern void __udelay_bad(unsigned long usecs);
 
-	start = mfctl(16);
-	while ((mfctl(16) - start) < clocks)
-	    ;
+static inline void udelay(unsigned long usecs)
+{
+	if (__builtin_constant_p(usecs) && (usecs) > 20000)
+		__udelay_bad(usecs);
+	__udelay(usecs);
 }
 
-static __inline__ void __udelay(unsigned long usecs) {
-	__cr16_delay(usecs * ((unsigned long)boot_cpu_data.cpu_hz / 1000000UL));
-}
-
-#define udelay(n) __udelay(n)
-
-#endif /* defined(_PARISC_DELAY_H) */
+#endif /* _ASM_PARISC_DELAY_H */
diff --git a/arch/parisc/lib/Makefile b/arch/parisc/lib/Makefile
index 5651536ac733..8fa92b8d839a 100644
--- a/arch/parisc/lib/Makefile
+++ b/arch/parisc/lib/Makefile
@@ -3,6 +3,6 @@
 #
 
 lib-y	:= lusercopy.o bitops.o checksum.o io.o memset.o fixup.o memcpy.o \
-	   ucmpdi2.o
+	   ucmpdi2.o delay.o
 
 obj-y	:= iomap.o
diff --git a/arch/parisc/lib/delay.c b/arch/parisc/lib/delay.c
new file mode 100644
index 000000000000..ec9255f27a81
--- /dev/null
+++ b/arch/parisc/lib/delay.c
@@ -0,0 +1,73 @@
+/*
+ *	Precise Delay Loops for parisc
+ *
+ *	based on code by:
+ *	Copyright (C) 1993 Linus Torvalds
+ *	Copyright (C) 1997 Martin Mares <mj@atrey.karlin.mff.cuni.cz>
+ *	Copyright (C) 2008 Jiri Hladky <hladky _dot_ jiri _at_ gmail _dot_ com>
+ *
+ *	parisc implementation:
+ *	Copyright (C) 2013 Helge Deller <deller@gmx.de>
+ */
+
+
+#include <linux/module.h>
+#include <linux/preempt.h>
+#include <linux/init.h>
+
+#include <asm/processor.h>
+#include <asm/delay.h>
+
+#include <asm/special_insns.h>    /* for mfctl() */
+#include <asm/processor.h> /* for boot_cpu_data */
+
+/* CR16 based delay: */
+static void __cr16_delay(unsigned long __loops)
+{
+	/*
+	 * Note: Due to unsigned math, cr16 rollovers shouldn't be
+	 * a problem here. However, on 32 bit, we need to make sure
+	 * we don't pass in too big a value. The current default
+	 * value of MAX_UDELAY_MS should help prevent this.
+	 */
+	u32 bclock, now, loops = __loops;
+	int cpu;
+
+	preempt_disable();
+	cpu = smp_processor_id();
+	bclock = mfctl(16);
+	for (;;) {
+		now = mfctl(16);
+		if ((now - bclock) >= loops)
+			break;
+
+		/* Allow RT tasks to run */
+		preempt_enable();
+		asm volatile("	nop\n");
+		barrier();
+		preempt_disable();
+
+		/*
+		 * It is possible that we moved to another CPU, and
+		 * since CR16's are per-cpu we need to calculate
+		 * that. The delay must guarantee that we wait "at
+		 * least" the amount of time. Being moved to another
+		 * CPU could make the wait longer but we just need to
+		 * make sure we waited long enough. Rebalance the
+		 * counter for this CPU.
+		 */
+		if (unlikely(cpu != smp_processor_id())) {
+			loops -= (now - bclock);
+			cpu = smp_processor_id();
+			bclock = mfctl(16);
+		}
+	}
+	preempt_enable();
+}
+
+
+void __udelay(unsigned long usecs)
+{
+	__cr16_delay(usecs * ((unsigned long)boot_cpu_data.cpu_hz / 1000000UL));
+}
+EXPORT_SYMBOL(__udelay);