Hi Ingo
Here is a patch against linux-2.6-x86 git
Thank you
[PATCH] X86 : Introduce DEFINE_PER_CPU_PAGE_ALIGNED() macro for x86 arch to
shrink percpu section
While examining holes in percpu section I found this :
c0659008 D per_cpu__cpu_number
c065900c D per_cpu__irq_regs
<BIG HOLE of about 4000 bytes>
c065a000 D per_cpu__gdt_page
c065b000 d per_cpu__next_check
This is because gdt_page is a percpu variable, defined with
a page alignement, and linker is doing its job.
I introduced a new macro DEFINE_PER_CPU_PAGE_ALIGNED() to avoid
wasting this space. All page aligned variables are put in a separate
subsection .data.percpu.page_aligned
This macro is defined for x86/generic architectures only
but could be added to other arches if needed.
Before patch :
# size -A vmlinux | grep percpu
.data.percpu 26144 3227328512
After patch :
# size -A vmlinux | grep percpu
.data.percpu 22048 3227328512
Signed-off-by: Eric Dumazet <[EMAIL PROTECTED]>
arch/x86/kernel/cpu/common.c | 2 +-
arch/x86/kernel/vmlinux_32.lds.S | 1 +
include/asm-generic/percpu.h | 4 ++++
include/asm-generic/vmlinux.lds.h | 1 +
include/asm-x86/percpu_32.h | 4 ++++
include/asm-x86/percpu_64.h | 4 ++++
6 files changed, 15 insertions(+), 1 deletion(-)
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index e2fcf20..a262d4c 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -21,7 +21,7 @@
#include "cpu.h"
-DEFINE_PER_CPU(struct gdt_page, gdt_page) = { .gdt = {
+DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = {
[GDT_ENTRY_KERNEL_CS] = { 0x0000ffff, 0x00cf9a00 },
[GDT_ENTRY_KERNEL_DS] = { 0x0000ffff, 0x00cf9200 },
[GDT_ENTRY_DEFAULT_USER_CS] = { 0x0000ffff, 0x00cffa00 },
diff --git a/arch/x86/kernel/vmlinux_32.lds.S b/arch/x86/kernel/vmlinux_32.lds.S
index 7d72cce..b579e8b 100644
--- a/arch/x86/kernel/vmlinux_32.lds.S
+++ b/arch/x86/kernel/vmlinux_32.lds.S
@@ -182,6 +182,7 @@ SECTIONS
. = ALIGN(4096);
.data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) {
__per_cpu_start = .;
+ *(.data.percpu.page_aligned)
*(.data.percpu)
*(.data.percpu.shared_aligned)
__per_cpu_end = .;
diff --git a/include/asm-generic/percpu.h b/include/asm-generic/percpu.h
index d85172e..6c89cda 100644
--- a/include/asm-generic/percpu.h
+++ b/include/asm-generic/percpu.h
@@ -14,6 +14,10 @@ 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_PAGE_ALIGNED(type, name) \
+ __attribute__((__section__(".data.percpu.page_aligned"))) \
+ __typeof__(type) per_cpu__##name
+
#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name) \
__attribute__((__section__(".data.percpu.shared_aligned"))) \
__typeof__(type) per_cpu__##name \
diff --git a/include/asm-generic/vmlinux.lds.h
b/include/asm-generic/vmlinux.lds.h
index 9f584cc..68aaf68 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -259,6 +259,7 @@
. = ALIGN(align); \
__per_cpu_start = .; \
.data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) { \
+ *(.data.percpu.page_aligned) \
*(.data.percpu) \
*(.data.percpu.shared_aligned) \
} \
diff --git a/include/asm-x86/percpu_32.h b/include/asm-x86/percpu_32.h
index a7ebd43..a515897 100644
--- a/include/asm-x86/percpu_32.h
+++ b/include/asm-x86/percpu_32.h
@@ -54,6 +54,10 @@ 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_PAGE_ALIGNED(type, name) \
+ __attribute__((__section__(".data.percpu.page_aligned"))) \
+ __typeof__(type) per_cpu__##name
+
#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name) \
__attribute__((__section__(".data.percpu.shared_aligned"))) \
__typeof__(type) per_cpu__##name \
diff --git a/include/asm-x86/percpu_64.h b/include/asm-x86/percpu_64.h
index 5abd482..9949008 100644
--- a/include/asm-x86/percpu_64.h
+++ b/include/asm-x86/percpu_64.h
@@ -20,6 +20,10 @@
#define DEFINE_PER_CPU(type, name) \
__attribute__((__section__(".data.percpu"))) __typeof__(type)
per_cpu__##name
+#define DEFINE_PER_CPU_PAGE_ALIGNED(type, name) \
+ __attribute__((__section__(".data.percpu.page_aligned"))) \
+ __typeof__(type) per_cpu__##name
+
#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name) \
__attribute__((__section__(".data.percpu.shared_aligned"))) \
__typeof__(type) per_cpu__##name \
