Name: Fix cpu_local_* for generic and x86 variants
Status: Untested
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

cpu_local_inc() must disable preemption on non-atomic implementations,
otherwise they can give the wrong results if preempted at the wrong
time.

Index: linux-2.6.10-rc2-bk13-Percpu/include/asm-x86_64/local.h
===================================================================
--- linux-2.6.10-rc2-bk13-Percpu.orig/include/asm-x86_64/local.h	2003-09-22 10:26:46.000000000 +1000
+++ linux-2.6.10-rc2-bk13-Percpu/include/asm-x86_64/local.h	2004-12-02 13:09:06.000000000 +1100
@@ -60,10 +60,14 @@
  */
 #define cpu_local_read(v)	local_read(&__get_cpu_var(v))
 #define cpu_local_set(v, i)	local_set(&__get_cpu_var(v), (i))
-#define cpu_local_inc(v)	local_inc(&__get_cpu_var(v))
-#define cpu_local_dec(v)	local_dec(&__get_cpu_var(v))
-#define cpu_local_add(i, v)	local_add((i), &__get_cpu_var(v))
-#define cpu_local_sub(i, v)	local_sub((i), &__get_cpu_var(v))
+#define cpu_local_inc(v)	\
+	do { local_inc(&get_cpu_var(v)); put_cpu_var(v); } while(0)
+#define cpu_local_dec(v)	\
+	do { local_dec(&get_cpu_var(v)); put_cpu_var(v); } while(0)
+#define cpu_local_add(i, v)	\
+	do { local_add((i), &get_cpu_var(v)); put_cpu_var(v); } while(0)
+#define cpu_local_sub(i, v)	\
+	do { local_sub((i), &get_cpu_var(v)); put_cpu_var(v); } while(0)
 
 #define __cpu_local_inc(v)	cpu_local_inc(v)
 #define __cpu_local_dec(v)	cpu_local_dec(v)
Index: linux-2.6.10-rc2-bk13-Percpu/include/asm-generic/local.h
===================================================================
--- linux-2.6.10-rc2-bk13-Percpu.orig/include/asm-generic/local.h	2004-10-19 14:34:17.000000000 +1000
+++ linux-2.6.10-rc2-bk13-Percpu/include/asm-generic/local.h	2004-12-02 13:09:06.000000000 +1100
@@ -102,10 +102,14 @@
  */
 #define cpu_local_read(v)	local_read(&__get_cpu_var(v))
 #define cpu_local_set(v, i)	local_set(&__get_cpu_var(v), (i))
-#define cpu_local_inc(v)	local_inc(&__get_cpu_var(v))
-#define cpu_local_dec(v)	local_dec(&__get_cpu_var(v))
-#define cpu_local_add(i, v)	local_add((i), &__get_cpu_var(v))
-#define cpu_local_sub(i, v)	local_sub((i), &__get_cpu_var(v))
+#define cpu_local_inc(v)	\
+	do { local_inc(&get_cpu_var(v)); put_cpu_var(v); } while(0)
+#define cpu_local_dec(v)	\
+	do { local_dec(&get_cpu_var(v)); put_cpu_var(v); } while(0)
+#define cpu_local_add(i, v)	\
+	do { local_add((i), &get_cpu_var(v)); put_cpu_var(v); } while(0)
+#define cpu_local_sub(i, v)	\
+	do { local_sub((i), &get_cpu_var(v)); put_cpu_var(v); } while(0)
 
 /* Non-atomic increments, ie. preemption disabled and won't be touched
  * in interrupt, etc.  Some archs can optimize this case well.
Index: linux-2.6.10-rc2-bk13-Percpu/include/asm-i386/local.h
===================================================================
--- linux-2.6.10-rc2-bk13-Percpu.orig/include/asm-i386/local.h	2003-09-22 10:26:12.000000000 +1000
+++ linux-2.6.10-rc2-bk13-Percpu/include/asm-i386/local.h	2004-12-02 13:09:06.000000000 +1100
@@ -57,10 +57,14 @@
  */
 #define cpu_local_read(v)	local_read(&__get_cpu_var(v))
 #define cpu_local_set(v, i)	local_set(&__get_cpu_var(v), (i))
-#define cpu_local_inc(v)	local_inc(&__get_cpu_var(v))
-#define cpu_local_dec(v)	local_dec(&__get_cpu_var(v))
-#define cpu_local_add(i, v)	local_add((i), &__get_cpu_var(v))
-#define cpu_local_sub(i, v)	local_sub((i), &__get_cpu_var(v))
+#define cpu_local_inc(v)	\
+	do { local_inc(&get_cpu_var(v)); put_cpu_var(v); } while(0)
+#define cpu_local_dec(v)	\
+	do { local_dec(&get_cpu_var(v)); put_cpu_var(v); } while(0)
+#define cpu_local_add(i, v)	\
+	do { local_add((i), &get_cpu_var(v)); put_cpu_var(v); } while(0)
+#define cpu_local_sub(i, v)	\
+	do { local_sub((i), &get_cpu_var(v)); put_cpu_var(v); } while(0)
 
 #define __cpu_local_inc(v)	cpu_local_inc(v)
 #define __cpu_local_dec(v)	cpu_local_dec(v)