From mboxrd@z Thu Jan 1 00:00:00 1970 Received: by rv-out-0910.google.com with SMTP id f1so832882rvb.26 for ; Sat, 16 Feb 2008 22:22:17 -0800 (PST) Message-ID: <86802c440802162222k47f5bebbhf42fef0f11ce3243@mail.gmail.com> Date: Sat, 16 Feb 2008 22:22:17 -0800 From: "Yinghai Lu" Subject: Re: [PATCH 3/4] x86_64: Fold pda into per cpu area In-Reply-To: <20080215201640.GA6200@elte.hu> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Content-Disposition: inline References: <20080201191414.961558000@sgi.com> <20080201191415.450555000@sgi.com> <20080215201640.GA6200@elte.hu> Sender: owner-linux-mm@kvack.org Return-Path: To: Ingo Molnar Cc: travis@sgi.com, Andrew Morton , Andi Kleen , Thomas Gleixner , Jeremy Fitzhardinge , Christoph Lameter , Jack Steiner , linux-mm@kvack.org, linux-kernel@vger.kernel.org List-ID: On Feb 15, 2008 12:16 PM, Ingo Molnar wrote: > > * travis@sgi.com wrote: > > > include/asm-generic/vmlinux.lds.h | 2 + > > include/linux/percpu.h | 9 ++++- > > couldnt these two generic bits be done separately (perhaps a preparatory > but otherwise NOP patch pushed upstream straight away) to make > subsequent patches only touch x86 architecture files? this patch need to apply to mainline asap. or you need revert to the patch about include/asm-x86/percpu.h +#ifdef CONFIG_X86_64 +#include + +/* Same as asm-generic/percpu.h, except that we store the per cpu offset + in the PDA. Longer term the PDA and every per cpu variable + should be just put into a single section and referenced directly + from %gs */ + +#ifdef CONFIG_SMP +#include + +#define __per_cpu_offset(cpu) (cpu_pda(cpu)->data_offset) +#define __my_cpu_offset read_pda(data_offset) + +#define per_cpu_offset(x) (__per_cpu_offset(x)) + #endif +#include + +DECLARE_PER_CPU(struct x8664_pda, pda); + +#else /* CONFIG_X86_64 */ because current tree in setup_per_cpu_areas will have cpu_pda(i)->data_offset = ptr - __per_cpu_start; but at that time all APs will use cpu_pda for boot cpu...,and APs will get their pda in do_boot_cpu() the result is all cpu will have same data_offset, there will share one per_cpu_data..that is totally wrong!! that could explain a lot of strange panic ....recently about NUMA... YH -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org