Re: [cpuops cmpxchg V1 2/4] x86: this_cpu_cmpxchg and this_cpu_xchgoperations

[Christoph Lameter]

Argh. I included __this_cpu_cmpxchg_8 done with cmpxchg8b for 32 bit. That
has some issues....


Subject: Fixup __this_cpu_cmpxchg_8

__this_cpu_cmpxchg_8 can cause compilation failure since it is expanded
for any this_cpu_cmpxchg independent of the size or type of argument.
However. __this_cpu_cmpxchg_8 only generates code for the 8 byte
case. Use a union to get around the compilation failures.

Signed-off-by: Christoph Lameter <cl@xxxxxxxxx>

---
 arch/x86/include/asm/percpu.h |   20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)

Index: linux-2.6/arch/x86/include/asm/percpu.h
===================================================================
--- linux-2.6.orig/arch/x86/include/asm/percpu.h	2010-12-08 14:48:02.000000000 -0600
+++ linux-2.6/arch/x86/include/asm/percpu.h	2010-12-08 16:01:24.000000000 -0600
@@ -391,15 +391,23 @@ do {									\

 #ifndef CONFIG_X86_64
 #ifdef CONFIG_X86_CMPXCHG64
-/* We can support a 8 byte cmpxchg with a special instruction on 32 bit */
+/*
+ * We can support a 8 byte cmpxchg with a special instruction on 32 bit.
+ *
+ * Note that some of the strangeness here with the __new variable is due to
+ * the need to expand and compile this function for arbitary types even
+ * though the code will not be included in the .o file if used for a smaller
+ * type.
+ */
 #define __this_cpu_cmpxchg_8(pcp, oval, nval)				\
 ({									\
-	typeof(var) __ret;						\
-	typeof(var) __old = (oval);					\
-	typeof(var) __new = (nval);					\
+	typeof(pcp) __ret;						\
+	typeof(pcp) __old = (oval);					\
+	union x { typeof(pcp) n;u64 m; } __new;				\
+	__new.n = (nval);						\
 	asm("cmpxchg8b %2, "__percpu_arg(1)				\
-	    : "=A" (__ret), "+m" (&pcp)					\
-	    : "b" (((u32)new), "c" ((u32)(new >> 32)),  "0" (__old)	\
+	    : "=A" (__ret), "+m" (pcp)					\
+	    : "b" ((u32)__new.m), "c" ((u32)(__new.m >> 32)),  "0" (__old)	\
 	    : "memory");						\
 	__ret;								\
 })

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/