define new percpu interface for shared data

per cpu data section contains two types of data.  One set which is
exclusively accessed by the local cpu and the other set which is per cpu,
but also shared by remote cpus.  In the current kernel, these two sets are
not clearely separated out.  This can potentially cause the same data
cacheline shared between the two sets of data, which will result in
unnecessary bouncing of the cacheline between cpus.

One way to fix the problem is to cacheline align the remotely accessed per
cpu data, both at the beginning and at the end.  Because of the padding at
both ends, this will likely cause some memory wastage and also the
interface to achieve this is not clean.

This patch:

Moves the remotely accessed per cpu data (which is currently marked
as ____cacheline_aligned_in_smp) into a different section, where all the data
elements are cacheline aligned. And as such, this differentiates the local
only data and remotely accessed data cleanly.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Acked-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: <linux-arch@vger.kernel.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 files changed, 59 insertions, 0 deletions

diff --git a/include/asm-generic/percpu.h b/include/asm-generic/percpu.h
index d984a904143..d85172e9ed4 100644
--- a/include/asm-generic/percpu.h
+++ b/include/asm-generic/percpu.h
@@ -14,6 +14,11 @@ extern unsigned long __per_cpu_offset[NR_CPUS];
 #define DEFINE_PER_CPU(type, name) \
     __attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
 
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)		\
+    __attribute__((__section__(".data.percpu.shared_aligned"))) \
+    __typeof__(type) per_cpu__##name				\
+    ____cacheline_aligned_in_smp
+
 /* var is in discarded region: offset to particular copy we want */
 #define per_cpu(var, cpu) (*({				\
 	extern int simple_identifier_##var(void);	\
@@ -34,6 +39,9 @@ do {								\
 #define DEFINE_PER_CPU(type, name) \
     __typeof__(type) per_cpu__##name
 
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)	\
+    DEFINE_PER_CPU(type, name)
+
 #define per_cpu(var, cpu)			(*((void)(cpu), &per_cpu__##var))
 #define __get_cpu_var(var)			per_cpu__##var
 #define __raw_get_cpu_var(var)			per_cpu__##var
diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index 84155eb67f1..a2b09ed852a 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -245,3 +245,11 @@
   	*(.initcall7.init)						\
   	*(.initcall7s.init)
 
+#define PERCPU(align)							\
+	. = ALIGN(align);						\
+	__per_cpu_start = .;						\
+	.data.percpu  : AT(ADDR(.data.percpu) - LOAD_OFFSET) {		\
+		*(.data.percpu)						\
+		*(.data.percpu.shared_aligned)				\
+	}								\
+	__per_cpu_end = .;
diff --git a/include/asm-i386/percpu.h b/include/asm-i386/percpu.h
index f54830b5d5a..a7ebd436f3c 100644
--- a/include/asm-i386/percpu.h
+++ b/include/asm-i386/percpu.h
@@ -54,6 +54,11 @@ extern unsigned long __per_cpu_offset[];
 #define DEFINE_PER_CPU(type, name) \
     __attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
 
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)		\
+    __attribute__((__section__(".data.percpu.shared_aligned"))) \
+    __typeof__(type) per_cpu__##name				\
+    ____cacheline_aligned_in_smp
+
 /* We can use this directly for local CPU (faster). */
 DECLARE_PER_CPU(unsigned long, this_cpu_off);
 
diff --git a/include/asm-ia64/percpu.h b/include/asm-ia64/percpu.h
index fbe5cf3ab8d..43a7aac414e 100644
--- a/include/asm-ia64/percpu.h
+++ b/include/asm-ia64/percpu.h
@@ -29,6 +29,16 @@
 	__attribute__((__section__(".data.percpu")))		\
 	__SMALL_ADDR_AREA __typeof__(type) per_cpu__##name
 
+#ifdef CONFIG_SMP
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)			\
+	__attribute__((__section__(".data.percpu.shared_aligned")))	\
+	__SMALL_ADDR_AREA __typeof__(type) per_cpu__##name		\
+	____cacheline_aligned_in_smp
+#else
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)	\
+	DEFINE_PER_CPU(type, name)
+#endif
+
 /*
  * Pretty much a literal copy of asm-generic/percpu.h, except that percpu_modcopy() is an
  * external routine, to avoid include-hell.
diff --git a/include/asm-powerpc/percpu.h b/include/asm-powerpc/percpu.h
index 2f2e3024fa6..73dc8ba4010 100644
--- a/include/asm-powerpc/percpu.h
+++ b/include/asm-powerpc/percpu.h
@@ -20,6 +20,11 @@
 #define DEFINE_PER_CPU(type, name) \
     __attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
 
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)		\
+    __attribute__((__section__(".data.percpu.shared_aligned"))) \
+    __typeof__(type) per_cpu__##name				\
+    ____cacheline_aligned_in_smp
+
 /* var is in discarded region: offset to particular copy we want */
 #define per_cpu(var, cpu) (*RELOC_HIDE(&per_cpu__##var, __per_cpu_offset(cpu)))
 #define __get_cpu_var(var) (*RELOC_HIDE(&per_cpu__##var, __my_cpu_offset()))
@@ -40,6 +45,8 @@ extern void setup_per_cpu_areas(void);
 
 #define DEFINE_PER_CPU(type, name) \
     __typeof__(type) per_cpu__##name
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)	\
+    DEFINE_PER_CPU(type, name)
 
 #define per_cpu(var, cpu)			(*((void)(cpu), &per_cpu__##var))
 #define __get_cpu_var(var)			per_cpu__##var
diff --git a/include/asm-s390/percpu.h b/include/asm-s390/percpu.h
index 9ea7f1023e5..545857e6444 100644
--- a/include/asm-s390/percpu.h
+++ b/include/asm-s390/percpu.h
@@ -41,6 +41,11 @@ extern unsigned long __per_cpu_offset[NR_CPUS];
     __attribute__((__section__(".data.percpu"))) \
     __typeof__(type) per_cpu__##name
 
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)		\
+    __attribute__((__section__(".data.percpu.shared_aligned"))) \
+    __typeof__(type) per_cpu__##name				\
+    ____cacheline_aligned_in_smp
+
 #define __get_cpu_var(var) __reloc_hide(var,S390_lowcore.percpu_offset)
 #define __raw_get_cpu_var(var) __reloc_hide(var,S390_lowcore.percpu_offset)
 #define per_cpu(var,cpu) __reloc_hide(var,__per_cpu_offset[cpu])
@@ -59,6 +64,8 @@ do {								\
 
 #define DEFINE_PER_CPU(type, name) \
     __typeof__(type) per_cpu__##name
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)	\
+    DEFINE_PER_CPU(type, name)
 
 #define __get_cpu_var(var) __reloc_hide(var,0)
 #define __raw_get_cpu_var(var) __reloc_hide(var,0)
diff --git a/include/asm-sparc64/percpu.h b/include/asm-sparc64/percpu.h
index 88db872ce2f..caf8750792f 100644
--- a/include/asm-sparc64/percpu.h
+++ b/include/asm-sparc64/percpu.h
@@ -18,6 +18,11 @@ extern unsigned long __per_cpu_shift;
 #define DEFINE_PER_CPU(type, name) \
     __attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
 
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)		\
+    __attribute__((__section__(".data.percpu.shared_aligned"))) \
+    __typeof__(type) per_cpu__##name				\
+    ____cacheline_aligned_in_smp
+
 register unsigned long __local_per_cpu_offset asm("g5");
 
 /* var is in discarded region: offset to particular copy we want */
@@ -38,6 +43,8 @@ do {								\
 #define real_setup_per_cpu_areas()		do { } while (0)
 #define DEFINE_PER_CPU(type, name) \
     __typeof__(type) per_cpu__##name
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)	\
+    DEFINE_PER_CPU(type, name)
 
 #define per_cpu(var, cpu)			(*((void)cpu, &per_cpu__##var))
 #define __get_cpu_var(var)			per_cpu__##var
diff --git a/include/asm-x86_64/percpu.h b/include/asm-x86_64/percpu.h
index c6fbb67eac9..5abd4827010 100644
--- a/include/asm-x86_64/percpu.h
+++ b/include/asm-x86_64/percpu.h
@@ -20,6 +20,11 @@
 #define DEFINE_PER_CPU(type, name) \
     __attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
 
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)		\
+    __attribute__((__section__(".data.percpu.shared_aligned"))) \
+    __typeof__(type) per_cpu__##name				\
+    ____cacheline_internodealigned_in_smp
+
 /* var is in discarded region: offset to particular copy we want */
 #define per_cpu(var, cpu) (*({				\
 	extern int simple_identifier_##var(void);	\
@@ -46,6 +51,8 @@ extern void setup_per_cpu_areas(void);
 
 #define DEFINE_PER_CPU(type, name) \
     __typeof__(type) per_cpu__##name
+#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name)	\
+    DEFINE_PER_CPU(type, name)
 
 #define per_cpu(var, cpu)			(*((void)(cpu), &per_cpu__##var))
 #define __get_cpu_var(var)			per_cpu__##var