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>
26 files changed:
. = ALIGN(8);
SECURITY_INIT
. = ALIGN(8);
SECURITY_INIT
- . = ALIGN(8192);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
. = ALIGN(2*8192);
__init_end = .;
. = ALIGN(2*8192);
__init_end = .;
. = ALIGN(4096);
__per_cpu_start = .;
*(.data.percpu)
. = ALIGN(4096);
__per_cpu_start = .;
*(.data.percpu)
+ *(.data.percpu.shared_aligned)
__per_cpu_end = .;
#ifndef CONFIG_XIP_KERNEL
__init_begin = _stext;
__per_cpu_end = .;
#ifndef CONFIG_XIP_KERNEL
__init_begin = _stext;
- . = ALIGN (8192);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
#ifdef CONFIG_BLK_DEV_INITRD
.init.ramfs : {
#ifdef CONFIG_BLK_DEV_INITRD
.init.ramfs : {
__alt_instructions_end = .;
.altinstr_replacement : { *(.altinstr_replacement) }
__alt_instructions_end = .;
.altinstr_replacement : { *(.altinstr_replacement) }
- . = ALIGN(4096);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
#ifdef CONFIG_BLK_DEV_INITRD
. = ALIGN(4096);
#ifdef CONFIG_BLK_DEV_INITRD
. = ALIGN(4096);
.data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) {
__per_cpu_start = .;
*(.data.percpu)
.data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) {
__per_cpu_start = .;
*(.data.percpu)
+ *(.data.percpu.shared_aligned)
__per_cpu_end = .;
}
. = ALIGN(4096);
__per_cpu_end = .;
}
. = ALIGN(4096);
{
__per_cpu_start = .;
*(.data.percpu)
{
__per_cpu_start = .;
*(.data.percpu)
+ *(.data.percpu.shared_aligned)
__per_cpu_end = .;
}
. = __phys_per_cpu_start + PERCPU_PAGE_SIZE; /* ensure percpu data fits
__per_cpu_end = .;
}
. = __phys_per_cpu_start + PERCPU_PAGE_SIZE; /* ensure percpu data fits
__initramfs_end = .;
#endif
__initramfs_end = .;
#endif
- . = ALIGN(4096);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
. = ALIGN(4096);
__init_end = .;
/* freed after init ends here */
. = ALIGN(4096);
__init_end = .;
/* freed after init ends here */
.init.ramfs : { *(.init.ramfs) }
__initramfs_end = .;
#endif
.init.ramfs : { *(.init.ramfs) }
__initramfs_end = .;
#endif
- . = ALIGN(_PAGE_SIZE);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
. = ALIGN(_PAGE_SIZE);
__init_end = .;
/* freed after init ends here */
. = ALIGN(_PAGE_SIZE);
__init_end = .;
/* freed after init ends here */
.init.ramfs : { *(.init.ramfs) }
__initramfs_end = .;
#endif
.init.ramfs : { *(.init.ramfs) }
__initramfs_end = .;
#endif
- . = ALIGN(ASM_PAGE_SIZE);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
+
+ PERCPU(ASM_PAGE_SIZE)
+
. = ALIGN(ASM_PAGE_SIZE);
__init_end = .;
/* freed after init ends here */
. = ALIGN(ASM_PAGE_SIZE);
__init_end = .;
/* freed after init ends here */
.data.percpu : {
__per_cpu_start = .;
*(.data.percpu)
.data.percpu : {
__per_cpu_start = .;
*(.data.percpu)
+ *(.data.percpu.shared_aligned)
__ftr_fixup : { *(__ftr_fixup) }
__stop___ftr_fixup = .;
__ftr_fixup : { *(__ftr_fixup) }
__stop___ftr_fixup = .;
- . = ALIGN(4096);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
#ifdef CONFIG_BLK_DEV_INITRD
. = ALIGN(4096);
#ifdef CONFIG_BLK_DEV_INITRD
. = ALIGN(4096);
. = ALIGN(2);
__initramfs_end = .;
#endif
. = ALIGN(2);
__initramfs_end = .;
#endif
- . = ALIGN(4096);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
. = ALIGN(4096);
__init_end = .;
/* freed after init ends here */
. = ALIGN(4096);
__init_end = .;
/* freed after init ends here */
. = ALIGN(PAGE_SIZE);
__nosave_end = .;
. = ALIGN(PAGE_SIZE);
__nosave_end = .;
- . = ALIGN(PAGE_SIZE);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
.data.cacheline_aligned : { *(.data.cacheline_aligned) }
_edata = .; /* End of data section */
.data.cacheline_aligned : { *(.data.cacheline_aligned) }
_edata = .; /* End of data section */
. = ALIGN(PAGE_SIZE);
__per_cpu_start = .;
. = ALIGN(PAGE_SIZE);
__per_cpu_start = .;
- .data.percpu : C_PHYS(.data.percpu) { *(.data.percpu) }
+ .data.percpu : C_PHYS(.data.percpu) {
+ *(.data.percpu)
+ *(.data.percpu.shared_aligned)
+ }
__per_cpu_end = . ;
.data.cacheline_aligned : C_PHYS(.data.cacheline_aligned) { *(.data.cacheline_aligned) }
__per_cpu_end = . ;
.data.cacheline_aligned : C_PHYS(.data.cacheline_aligned) { *(.data.cacheline_aligned) }
__initramfs_end = .;
#endif
__initramfs_end = .;
#endif
- . = ALIGN(4096);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
. = ALIGN(4096);
__init_end = .;
. = ALIGN(32);
. = ALIGN(4096);
__init_end = .;
. = ALIGN(32);
__initramfs_end = .;
#endif
__initramfs_end = .;
#endif
- . = ALIGN(PAGE_SIZE);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
. = ALIGN(PAGE_SIZE);
__init_end = .;
__bss_start = .;
. = ALIGN(PAGE_SIZE);
__init_end = .;
__bss_start = .;
__initramfs_end = .;
#endif
__initramfs_end = .;
#endif
- . = ALIGN(4096);
- __per_cpu_start = .;
- .data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) { *(.data.percpu) }
- __per_cpu_end = .;
. = ALIGN(4096);
__init_end = .;
. = ALIGN(4096);
__init_end = .;
__initramfs_end = .;
#endif
__initramfs_end = .;
#endif
- . = ALIGN(4096);
- __per_cpu_start = .;
- .data.percpu : { *(.data.percpu) }
- __per_cpu_end = .;
/* We need this dummy segment here */
/* We need this dummy segment here */
#define DEFINE_PER_CPU(type, name) \
__attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
#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); \
/* var is in discarded region: offset to particular copy we want */
#define per_cpu(var, cpu) (*({ \
extern int simple_identifier_##var(void); \
#define DEFINE_PER_CPU(type, name) \
__typeof__(type) per_cpu__##name
#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
#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
*(.initcall7.init) \
*(.initcall7s.init)
*(.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 = .;
#define DEFINE_PER_CPU(type, name) \
__attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
#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);
/* We can use this directly for local CPU (faster). */
DECLARE_PER_CPU(unsigned long, this_cpu_off);
__attribute__((__section__(".data.percpu"))) \
__SMALL_ADDR_AREA __typeof__(type) per_cpu__##name
__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.
/*
* Pretty much a literal copy of asm-generic/percpu.h, except that percpu_modcopy() is an
* external routine, to avoid include-hell.
#define DEFINE_PER_CPU(type, name) \
__attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
#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()))
/* 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()))
#define DEFINE_PER_CPU(type, name) \
__typeof__(type) per_cpu__##name
#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 per_cpu(var, cpu) (*((void)(cpu), &per_cpu__##var))
#define __get_cpu_var(var) per_cpu__##var
__attribute__((__section__(".data.percpu"))) \
__typeof__(type) per_cpu__##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
+
#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])
#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])
#define DEFINE_PER_CPU(type, name) \
__typeof__(type) per_cpu__##name
#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)
#define __get_cpu_var(var) __reloc_hide(var,0)
#define __raw_get_cpu_var(var) __reloc_hide(var,0)
#define DEFINE_PER_CPU(type, name) \
__attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
#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 */
register unsigned long __local_per_cpu_offset asm("g5");
/* var is in discarded region: offset to particular copy we want */
#define real_setup_per_cpu_areas() do { } while (0)
#define DEFINE_PER_CPU(type, name) \
__typeof__(type) per_cpu__##name
#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
#define per_cpu(var, cpu) (*((void)cpu, &per_cpu__##var))
#define __get_cpu_var(var) per_cpu__##var
#define DEFINE_PER_CPU(type, name) \
__attribute__((__section__(".data.percpu"))) __typeof__(type) per_cpu__##name
#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); \
/* var is in discarded region: offset to particular copy we want */
#define per_cpu(var, cpu) (*({ \
extern int simple_identifier_##var(void); \
#define DEFINE_PER_CPU(type, name) \
__typeof__(type) per_cpu__##name
#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 per_cpu(var, cpu) (*((void)(cpu), &per_cpu__##var))
#define __get_cpu_var(var) per_cpu__##var