[PATCH] powerpc/64: per cpu data optimisations

The current ppc64 per cpu data implementation is quite slow. eg:

        lhz 11,18(13)           /* smp_processor_id() */
        ld 9,.LC63-.LCTOC1(30)  /* per_cpu__variable_name */
        ld 8,.LC61-.LCTOC1(30)  /* __per_cpu_offset */
        sldi 11,11,3            /* form index into __per_cpu_offset */
        mr 10,9
        ldx 9,11,8              /* __per_cpu_offset[smp_processor_id()] */
        ldx 0,10,9              /* load per cpu data */

5 loads for something that is supposed to be fast, pretty awful. One
reason for the large number of loads is that we have to synthesize 2
64bit constants (per_cpu__variable_name and __per_cpu_offset).

By putting __per_cpu_offset into the paca we can avoid the 2 loads
associated with it:

        ld 11,56(13)            /* paca->data_offset */
        ld 9,.LC59-.LCTOC1(30)  /* per_cpu__variable_name */
        ldx 0,9,11              /* load per cpu data

Longer term we can should be able to do even better than 3 loads.
If per_cpu__variable_name wasnt a 64bit constant and paca->data_offset
was in a register we could cut it down to one load. A suggestion from
Rusty is to use gcc's __thread extension here. In order to do this we
would need to free up r13 (the __thread register and where the paca
currently is). So far Ive had a few unsuccessful attempts at doing that :)

The patch also allocates per cpu memory node local on NUMA machines.
This patch from Rusty has been sitting in my queue _forever_ but stalled
when I hit the compiler bug. Sorry about that.

Finally I also only allocate per cpu data for possible cpus, which comes
straight out of the x86-64 port. On a pseries kernel (with NR_CPUS == 128)
and 4 possible cpus we see some nice gains:

             total       used       free     shared    buffers cached
Mem:       4012228     212860    3799368          0          0 162424

             total       used       free     shared    buffers cached
Mem:       4016200     212984    3803216          0          0 162424

A saving of 3.75MB. Quite nice for smaller machines. Note: we now have
to be careful of per cpu users that touch data for !possible cpus.

At this stage it might be worth making the NUMA and possible cpu
optimisations generic, but per cpu init is done so early we have to be
careful that all architectures have their possible map setup correctly.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>

arch/powerpc/kernel/setup_64.c

@@ -33,6 +33,7 @@
 #include <linux/unistd.h>
 #include <linux/serial.h>
 #include <linux/serial_8250.h>
+#include <linux/bootmem.h>
 #include <asm/io.h>
 #include <asm/kdump.h>
 #include <asm/prom.h>
@@ -654,3 +655,28 @@ void cpu_die(void)
 	if (ppc_md.cpu_die)
 		ppc_md.cpu_die();
 }
+
+#ifdef CONFIG_SMP
+void __init setup_per_cpu_areas(void)
+{
+	int i;
+	unsigned long size;
+	char *ptr;
+
+	/* Copy section for each CPU (we discard the original) */
+	size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
+#ifdef CONFIG_MODULES
+	if (size < PERCPU_ENOUGH_ROOM)
+		size = PERCPU_ENOUGH_ROOM;
+#endif
+
+	for_each_cpu(i) {
+		ptr = alloc_bootmem_node(NODE_DATA(cpu_to_node(i)), size);
+		if (!ptr)
+			panic("Cannot allocate cpu data for CPU %d\n", i);
+
+		paca[i].data_offset = ptr - __per_cpu_start;
+		memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
+	}
+}
+#endif

include/asm-powerpc/paca.h

@@ -65,6 +65,7 @@ struct paca_struct {
 	u64 stab_real;			/* Absolute address of segment table */
 	u64 stab_addr;			/* Virtual address of segment table */
 	void *emergency_sp;		/* pointer to emergency stack */
+	u64 data_offset;		/* per cpu data offset */
 	s16 hw_cpu_id;			/* Physical processor number */
 	u8 cpu_start;			/* At startup, processor spins until */
 					/* this becomes non-zero. */

include/asm-powerpc/percpu.h

+#ifndef _ASM_POWERPC_PERCPU_H_
+#define _ASM_POWERPC_PERCPU_H_
+#ifdef __powerpc64__
+#include <linux/compiler.h>
+
+/*
+ * Same as asm-generic/percpu.h, except that we store the per cpu offset
+ * in the paca. Based on the x86-64 implementation.
+ */
+
+#ifdef CONFIG_SMP
+
+#include <asm/paca.h>
+
+#define __per_cpu_offset(cpu) (paca[cpu].data_offset)
+#define __my_cpu_offset() get_paca()->data_offset
+
+/* Separate out the type, so (int[3], foo) works. */
+#define DEFINE_PER_CPU(type, name) \
+    __attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
+
+/* 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()))
+
+/* A macro to avoid #include hell... */
+#define percpu_modcopy(pcpudst, src, size)			\
+do {								\
+	unsigned int __i;					\
+	for (__i = 0; __i < NR_CPUS; __i++)			\
+		if (cpu_possible(__i))				\
+			memcpy((pcpudst)+__per_cpu_offset(__i),	\
+			       (src), (size));			\
+} while (0)
+
+extern void setup_per_cpu_areas(void);
+
+#else /* ! SMP */
+
+#define DEFINE_PER_CPU(type, name) \
+    __typeof__(type) per_cpu__##name
+
+#define per_cpu(var, cpu)			(*((void)(cpu), &per_cpu__##var))
+#define __get_cpu_var(var)			per_cpu__##var
+
+#endif	/* SMP */
+
+#define DECLARE_PER_CPU(type, name) extern __typeof__(type) per_cpu__##name
+
+#define EXPORT_PER_CPU_SYMBOL(var) EXPORT_SYMBOL(per_cpu__##var)
+#define EXPORT_PER_CPU_SYMBOL_GPL(var) EXPORT_SYMBOL_GPL(per_cpu__##var)
+
+#else
 #include <asm-generic/percpu.h>
+#endif
+
+#endif /* _ASM_POWERPC_PERCPU_H_ */