percpu: introduce DEFINE_PER_CPU_PAGE_ALIGNED() macro
authorEric Dumazet <dada1@cosmosbay.com>
Mon, 12 May 2008 13:44:40 +0000 (15:44 +0200)
committerThomas Gleixner <tglx@linutronix.de>
Sun, 25 May 2008 05:03:46 +0000 (07:03 +0200)
commit63cc8c75156462d4b42cbdd76c293b7eee7ddbfe
treee014b4c61399f8f74347352e6ea76ddf1502dcfa
parent75d3bce2fc0a80f435fe12f2c9ed2632c8ac29e4
percpu: introduce DEFINE_PER_CPU_PAGE_ALIGNED() macro

While examining holes in percpu section I found this :

c05f5000 D per_cpu__current_task
c05f5000 D __per_cpu_start
c05f5004 D per_cpu__cpu_number
c05f5008 D per_cpu__irq_regs
c05f500c d per_cpu__cpu_devices
c05f5040 D per_cpu__cyc2ns

<Big Hole of about 4000 bytes>

c05f6000 d per_cpu__cpuid4_info
c05f6004 d per_cpu__cache_kobject
c05f6008 d per_cpu__index_kobject

<Big Hole of about 4000 bytes>

c05f7000 D per_cpu__gdt_page

This is because gdt_page is a percpu variable, defined with
a page alignement, and linker is doing its job, two times because of .o
nesting in the build process.

I introduced a new macro DEFINE_PER_CPU_PAGE_ALIGNED() to avoid
wasting this space. All page aligned variables (only one at this time)
are put in a separate
subsection .data.percpu.page_aligned, at the very begining of percpu zone.

Before patch , on a x86_32 machine :

.data.percpu                30232   3227471872
.data.percpu                22168   3227471872

Thats 8064 bytes saved for each CPU.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
arch/x86/kernel/cpu/common.c
arch/x86/kernel/vmlinux_32.lds.S
include/asm-generic/vmlinux.lds.h
include/linux/percpu.h