Merge with Linux 2.4.0-test5-pre1. This works again on Origin UP.

The IP22 cache bugs which are plaguing some machines are still unfixed.

1 files changed, 58 insertions, 139 deletions

diff --git a/arch/ppc/kernel/bitops.c b/arch/ppc/kernel/bitops.c
index fb5a19e3a..69e07057a 100644
--- a/arch/ppc/kernel/bitops.c
+++ b/arch/ppc/kernel/bitops.c
@@ -6,60 +6,58 @@
 #include <asm/bitops.h>
 
 /*
- * I left these here since the problems with "cc" make it difficult to keep
- * them in bitops.h -- Cort
+ * If the bitops are not inlined in bitops.h, they are defined here.
+ *  -- paulus
  */
-void set_bit(int nr, volatile void *addr)
+#if !__INLINE_BITOPS
+void set_bit(int nr, volatile void * addr)
 {
-	unsigned int t;
-	unsigned int mask = 1 << (nr & 0x1f);
-	volatile unsigned int *p = ((volatile unsigned int *)addr) + (nr >> 5);
-
-	if ((unsigned long)addr & 3)
-		printk(KERN_ERR "set_bit(%x, %p)\n", nr, addr);
-	__asm__ __volatile__("\n\
-1:	lwarx	%0,0,%2
-	or	%0,%0,%1
-	stwcx.	%0,0,%2
+	unsigned long old;
+	unsigned long mask = 1 << (nr & 0x1f);
+	unsigned long *p = ((unsigned long *)addr) + (nr >> 5);
+	
+	__asm__ __volatile__(SMP_WMB "\
+1:	lwarx	%0,0,%3
+	or	%0,%0,%2
+	stwcx.	%0,0,%3
 	bne	1b"
-	: "=&r" (t)		/*, "=m" (*p)*/
-	: "r" (mask), "r" (p)
-	: "cc");
+	SMP_MB
+	: "=&r" (old), "=m" (*p)
+	: "r" (mask), "r" (p), "m" (*p)
+	: "cc" );
 }
 
 void clear_bit(int nr, volatile void *addr)
 {
-	unsigned int t;
-	unsigned int mask = 1 << (nr & 0x1f);
-	volatile unsigned int *p = ((volatile unsigned int *)addr) + (nr >> 5);
+	unsigned long old;
+	unsigned long mask = 1 << (nr & 0x1f);
+	unsigned long *p = ((unsigned long *)addr) + (nr >> 5);
 
-	if ((unsigned long)addr & 3)
-		printk(KERN_ERR "clear_bit(%x, %p)\n", nr, addr);
-	__asm__ __volatile__("\n\
-1:	lwarx	%0,0,%2
-	andc	%0,%0,%1
-	stwcx.	%0,0,%2
+	__asm__ __volatile__(SMP_WMB "\
+1:	lwarx	%0,0,%3
+	andc	%0,%0,%2
+	stwcx.	%0,0,%3
 	bne	1b"
-	: "=&r" (t)		/*, "=m" (*p)*/
-	: "r" (mask), "r" (p)
+	SMP_MB
+	: "=&r" (old), "=m" (*p)
+	: "r" (mask), "r" (p), "m" (*p)
 	: "cc");
 }
 
 void change_bit(int nr, volatile void *addr)
 {
-	unsigned int t;
-	unsigned int mask = 1 << (nr & 0x1f);
-	volatile unsigned int *p = ((volatile unsigned int *)addr) + (nr >> 5);
+	unsigned long old;
+	unsigned long mask = 1 << (nr & 0x1f);
+	unsigned long *p = ((unsigned long *)addr) + (nr >> 5);
 
-	if ((unsigned long)addr & 3)
-		printk(KERN_ERR "change_bit(%x, %p)\n", nr, addr);
-	__asm__ __volatile__("\n\
-1:	lwarx	%0,0,%2
-	xor	%0,%0,%1
-	stwcx.	%0,0,%2
+	__asm__ __volatile__(SMP_WMB "\
+1:	lwarx	%0,0,%3
+	xor	%0,%0,%2
+	stwcx.	%0,0,%3
 	bne	1b"
-	: "=&r" (t)		/*, "=m" (*p)*/
-	: "r" (mask), "r" (p)
+	SMP_MB
+	: "=&r" (old), "=m" (*p)
+	: "r" (mask), "r" (p), "m" (*p)
 	: "cc");
 }
 
@@ -69,15 +67,14 @@ int test_and_set_bit(int nr, volatile void *addr)
 	unsigned int mask = 1 << (nr & 0x1f);
 	volatile unsigned int *p = ((volatile unsigned int *)addr) + (nr >> 5);
 
-	if ((unsigned long)addr & 3)
-		printk(KERN_ERR "test_and_set_bit(%x, %p)\n", nr, addr);
-	__asm__ __volatile__("\n\
-1:	lwarx	%0,0,%3
-	or	%1,%0,%2
-	stwcx.	%1,0,%3
+	__asm__ __volatile__(SMP_WMB "\
+1:	lwarx	%0,0,%4
+	or	%1,%0,%3
+	stwcx.	%1,0,%4
 	bne	1b"
-	: "=&r" (old), "=&r" (t)	/*, "=m" (*p)*/
-	: "r" (mask), "r" (p)
+	SMP_MB
+	: "=&r" (old), "=&r" (t), "=m" (*p)
+	: "r" (mask), "r" (p), "m" (*p)
 	: "cc");
 
 	return (old & mask) != 0;
@@ -89,15 +86,14 @@ int test_and_clear_bit(int nr, volatile void *addr)
 	unsigned int mask = 1 << (nr & 0x1f);
 	volatile unsigned int *p = ((volatile unsigned int *)addr) + (nr >> 5);
 
-	if ((unsigned long)addr & 3)
-		printk(KERN_ERR "test_and_clear_bit(%x, %p)\n", nr, addr);
-	__asm__ __volatile__("\n\
-1:	lwarx	%0,0,%3
-	andc	%1,%0,%2
-	stwcx.	%1,0,%3
+	__asm__ __volatile__(SMP_WMB "\
+1:	lwarx	%0,0,%4
+	andc	%1,%0,%3
+	stwcx.	%1,0,%4
 	bne	1b"
-	: "=&r" (old), "=&r" (t)	/*, "=m" (*p)*/
-	: "r" (mask), "r" (p)
+	SMP_MB
+	: "=&r" (old), "=&r" (t), "=m" (*p)
+	: "r" (mask), "r" (p), "m" (*p)
 	: "cc");
 
 	return (old & mask) != 0;
@@ -109,93 +105,16 @@ int test_and_change_bit(int nr, volatile void *addr)
 	unsigned int mask = 1 << (nr & 0x1f);
 	volatile unsigned int *p = ((volatile unsigned int *)addr) + (nr >> 5);
 
-	if ((unsigned long)addr & 3)
-		printk(KERN_ERR "test_and_change_bit(%x, %p)\n", nr, addr);
-	__asm__ __volatile__("\n\
-1:	lwarx	%0,0,%3
-	xor	%1,%0,%2
-	stwcx.	%1,0,%3
+	__asm__ __volatile__(SMP_WMB "\
+1:	lwarx	%0,0,%4
+	xor	%1,%0,%3
+	stwcx.	%1,0,%4
 	bne	1b"
-	: "=&r" (old), "=&r" (t)	/*, "=m" (*p)*/
-	: "r" (mask), "r" (p)
+	SMP_MB
+	: "=&r" (old), "=&r" (t), "=m" (*p)
+	: "r" (mask), "r" (p), "m" (*p)
 	: "cc");
 
 	return (old & mask) != 0;
 }
-
-/* I put it in bitops.h -- Cort */
-#if 0
-int ffz(unsigned int x)
-{
-	int n;
-
-	x = ~x & (x+1);		/* set LS zero to 1, other bits to 0 */
-	__asm__ ("cntlzw %0,%1" : "=r" (n) : "r" (x));
-	return 31 - n;
-}
-
-/*
- * This implementation of find_{first,next}_zero_bit was stolen from
- * Linus' asm-alpha/bitops.h.
- */
-
-int find_first_zero_bit(void * addr, int size)
-{
-	unsigned int * p = ((unsigned int *) addr);
-	unsigned int result = 0;
-	unsigned int tmp;
-
-	if (size == 0)
-		return 0;
-	while (size & ~31UL) {
-		if (~(tmp = *(p++)))
-			goto found_middle;
-		result += 32;
-		size -= 32;
-	}
-	if (!size)
-		return result;
-	tmp = *p;
-	tmp |= ~0UL << size;
-found_middle:
-	return result + ffz(tmp);
-}
-
-/*
- * Find next zero bit in a bitmap reasonably efficiently..
- */
-int find_next_zero_bit(void * addr, int size, int offset)
-{
-	unsigned int * p = ((unsigned int *) addr) + (offset >> 5);
-	unsigned int result = offset & ~31UL;
-	unsigned int tmp;
-
-	if (offset >= size)
-		return size;
-	size -= result;
-	offset &= 31UL;
-	if (offset) {
-		tmp = *(p++);
-		tmp |= ~0UL >> (32-offset);
-		if (size < 32)
-			goto found_first;
-		if (~tmp)
-			goto found_middle;
-		size -= 32;
-		result += 32;
-	}
-	while (size & ~31UL) {
-		if (~(tmp = *(p++)))
-			goto found_middle;
-		result += 32;
-		size -= 32;
-	}
-	if (!size)
-		return result;
-	tmp = *p;
-found_first:
-	tmp |= ~0UL << size;
-found_middle:
-	return result + ffz(tmp);
-}
-#endif
+#endif /* !__INLINE_BITOPS */