diff --git a/arch/alpha/include/asm/percpu.h b/arch/alpha/include/asm/percpu.h
index 3495e8e00d70fd2ca96531e702b1956a3aa5f5de..e9e0bb5a23bf5687096cbd5460b8d138ed5f5534 100644
--- a/arch/alpha/include/asm/percpu.h
+++ b/arch/alpha/include/asm/percpu.h
@@ -73,6 +73,6 @@ extern unsigned long __per_cpu_offset[NR_CPUS];
 
 #endif /* SMP */
 
-#define DECLARE_PER_CPU(type, name) extern __typeof__(type) per_cpu_var(name)
+#include <asm-generic/percpu.h>
 
 #endif /* __ALPHA_PERCPU_H */
diff --git a/arch/ia64/include/asm/smp.h b/arch/ia64/include/asm/smp.h
index 59840833625165fc72c75b40c24e94a1ba7db5ae..d217d1d4e05197351c6bc51c53bb946b3eae85ab 100644
--- a/arch/ia64/include/asm/smp.h
+++ b/arch/ia64/include/asm/smp.h
@@ -58,7 +58,7 @@ extern struct smp_boot_data {
 extern char no_int_routing __devinitdata;
 
 extern cpumask_t cpu_core_map[NR_CPUS];
-DECLARE_PER_CPU(cpumask_t, cpu_sibling_map);
+DECLARE_PER_CPU_SHARED_ALIGNED(cpumask_t, cpu_sibling_map);
 extern int smp_num_siblings;
 extern void __iomem *ipi_base_addr;
 extern unsigned char smp_int_redirect;
diff --git a/arch/x86/include/asm/desc.h b/arch/x86/include/asm/desc.h
index 5623c50d67b268a00c10dc60f8a5013f04bb28aa..c45f415ce315984c109ecbb0df763ae0e19866cd 100644
--- a/arch/x86/include/asm/desc.h
+++ b/arch/x86/include/asm/desc.h
@@ -37,7 +37,7 @@ extern gate_desc idt_table[];
 struct gdt_page {
 	struct desc_struct gdt[GDT_ENTRIES];
 } __attribute__((aligned(PAGE_SIZE)));
-DECLARE_PER_CPU(struct gdt_page, gdt_page);
+DECLARE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page);
 
 static inline struct desc_struct *get_cpu_gdt_table(unsigned int cpu)
 {
diff --git a/arch/x86/include/asm/hardirq.h b/arch/x86/include/asm/hardirq.h
index 039db6aa8e0271951706ee20bd9b819098f8002a..37555e52f980ac170e779dbe2aaa1e5e0c8ca07a 100644
--- a/arch/x86/include/asm/hardirq.h
+++ b/arch/x86/include/asm/hardirq.h
@@ -26,7 +26,7 @@ typedef struct {
 #endif
 } ____cacheline_aligned irq_cpustat_t;
 
-DECLARE_PER_CPU(irq_cpustat_t, irq_stat);
+DECLARE_PER_CPU_SHARED_ALIGNED(irq_cpustat_t, irq_stat);
 
 /* We can have at most NR_VECTORS irqs routed to a cpu at a time */
 #define MAX_HARDIRQS_PER_CPU NR_VECTORS
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index fcf4d92e7e04d27c44393563528eda97cf4c5769..c2cceae709c8655338894c15b84c361e65caf727 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -138,7 +138,7 @@ extern struct tss_struct	doublefault_tss;
 extern __u32			cleared_cpu_caps[NCAPINTS];
 
 #ifdef CONFIG_SMP
-DECLARE_PER_CPU(struct cpuinfo_x86, cpu_info);
+DECLARE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
 #define cpu_data(cpu)		per_cpu(cpu_info, cpu)
 #define current_cpu_data	__get_cpu_var(cpu_info)
 #else
@@ -270,7 +270,7 @@ struct tss_struct {
 
 } ____cacheline_aligned;
 
-DECLARE_PER_CPU(struct tss_struct, init_tss);
+DECLARE_PER_CPU_SHARED_ALIGNED(struct tss_struct, init_tss);
 
 /*
  * Save the original ist values for checking stack pointers during debugging
@@ -393,7 +393,7 @@ union irq_stack_union {
 	};
 };
 
-DECLARE_PER_CPU(union irq_stack_union, irq_stack_union);
+DECLARE_PER_CPU_FIRST(union irq_stack_union, irq_stack_union);
 DECLARE_INIT_PER_CPU(irq_stack_union);
 
 DECLARE_PER_CPU(char *, irq_stack_ptr);
diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index d3539f998f88c71dd6c9a784628c7d084d7521bc..16a5c84b032997f264f704ac68537dc65ba1dbe6 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -152,7 +152,7 @@ struct tlb_state {
 	struct mm_struct *active_mm;
 	int state;
 };
-DECLARE_PER_CPU(struct tlb_state, cpu_tlbstate);
+DECLARE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate);
 
 static inline void reset_lazy_tlbstate(void)
 {
diff --git a/include/asm-generic/percpu.h b/include/asm-generic/percpu.h
index b0e63c672ebdf5a065ddea3374558a9638bc43c1..af47b9e1006432c82a8140dc52999fc055132128 100644
--- a/include/asm-generic/percpu.h
+++ b/include/asm-generic/percpu.h
@@ -73,11 +73,50 @@ extern void setup_per_cpu_areas(void);
 
 #endif	/* SMP */
 
+#ifndef PER_CPU_BASE_SECTION
+#ifdef CONFIG_SMP
+#define PER_CPU_BASE_SECTION ".data.percpu"
+#else
+#define PER_CPU_BASE_SECTION ".data"
+#endif
+#endif
+
+#ifdef CONFIG_SMP
+
+#ifdef MODULE
+#define PER_CPU_SHARED_ALIGNED_SECTION ""
+#else
+#define PER_CPU_SHARED_ALIGNED_SECTION ".shared_aligned"
+#endif
+#define PER_CPU_FIRST_SECTION ".first"
+
+#else
+
+#define PER_CPU_SHARED_ALIGNED_SECTION ""
+#define PER_CPU_FIRST_SECTION ""
+
+#endif
+
 #ifndef PER_CPU_ATTRIBUTES
 #define PER_CPU_ATTRIBUTES
 #endif
 
-#define DECLARE_PER_CPU(type, name) extern PER_CPU_ATTRIBUTES \
-					__typeof__(type) per_cpu_var(name)
+#define DECLARE_PER_CPU_SECTION(type, name, section)			\
+	extern \
+	__attribute__((__section__(PER_CPU_BASE_SECTION section)))	\
+	PER_CPU_ATTRIBUTES __typeof__(type) per_cpu__##name
+
+#define DECLARE_PER_CPU(type, name)					\
+	DECLARE_PER_CPU_SECTION(type, name, "")
+
+#define DECLARE_PER_CPU_SHARED_ALIGNED(type, name)			\
+	DECLARE_PER_CPU_SECTION(type, name, PER_CPU_SHARED_ALIGNED_SECTION) \
+	____cacheline_aligned_in_smp
+
+#define DECLARE_PER_CPU_PAGE_ALIGNED(type, name)				\
+	DECLARE_PER_CPU_SECTION(type, name, ".page_aligned")
+
+#define DECLARE_PER_CPU_FIRST(type, name)				\
+	DECLARE_PER_CPU_SECTION(type, name, PER_CPU_FIRST_SECTION)
 
 #endif /* _ASM_GENERIC_PERCPU_H_ */
diff --git a/include/linux/percpu.h b/include/linux/percpu.h
index cfda2d5ad319e75c8bf88b262d1b95bae8f3a2a6..f052d818499348ef24817335c44720454624b669 100644
--- a/include/linux/percpu.h
+++ b/include/linux/percpu.h
@@ -9,30 +9,6 @@
 
 #include <asm/percpu.h>
 
-#ifndef PER_CPU_BASE_SECTION
-#ifdef CONFIG_SMP
-#define PER_CPU_BASE_SECTION ".data.percpu"
-#else
-#define PER_CPU_BASE_SECTION ".data"
-#endif
-#endif
-
-#ifdef CONFIG_SMP
-
-#ifdef MODULE
-#define PER_CPU_SHARED_ALIGNED_SECTION ""
-#else
-#define PER_CPU_SHARED_ALIGNED_SECTION ".shared_aligned"
-#endif
-#define PER_CPU_FIRST_SECTION ".first"
-
-#else
-
-#define PER_CPU_SHARED_ALIGNED_SECTION ""
-#define PER_CPU_FIRST_SECTION ""
-
-#endif
-
 #define DEFINE_PER_CPU_SECTION(type, name, section)			\
 	__attribute__((__section__(PER_CPU_BASE_SECTION section)))	\
 	PER_CPU_ATTRIBUTES __typeof__(type) per_cpu__##name
diff --git a/net/rds/rds.h b/net/rds/rds.h
index 619f0a30a4e566952642e27e1d2b49f5f546ad11..71794449ca4e58ffdbada8ac274a85b87418bf38 100644
--- a/net/rds/rds.h
+++ b/net/rds/rds.h
@@ -638,7 +638,7 @@ struct rds_message *rds_send_get_message(struct rds_connection *,
 void rds_rdma_unuse(struct rds_sock *rs, u32 r_key, int force);
 
 /* stats.c */
-DECLARE_PER_CPU(struct rds_statistics, rds_stats);
+DECLARE_PER_CPU_SHARED_ALIGNED(struct rds_statistics, rds_stats);
 #define rds_stats_inc_which(which, member) do {		\
 	per_cpu(which, get_cpu()).member++;		\
 	put_cpu();					\