From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1762016Ab0J2UxQ (ORCPT ); Fri, 29 Oct 2010 16:53:16 -0400 Received: from mail-wy0-f174.google.com ([74.125.82.174]:51491 "EHLO mail-wy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1762005Ab0J2UxN (ORCPT ); Fri, 29 Oct 2010 16:53:13 -0400 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=subject:from:to:cc:in-reply-to:references:content-type:date :message-id:mime-version:x-mailer:content-transfer-encoding; b=v26jDg2t8WUbANcKGH+mGdjAOSN5bHiv2SzLSaB7D8KB5wXJcs6u27a/3ndQpAgwyJ jTqZOCeEaXriXWBHScDojqYk/IoV9DXLR1Z/Mo4t4uZvsF1zBCHAgph/rJmkH/t0pUaz fwls/20RQfreQlDDlv6dzhGiYV7mM2k3bRDn8= Subject: Re: [tip:x86/urgent] x86-32: Restore irq stacks NUMA-aware allocations From: Eric Dumazet To: Cyrill Gorcunov Cc: Peter Zijlstra , mingo@redhat.com, hpa@zytor.com, linux-kernel@vger.kernel.org, brgerst@gmail.com, tglx@linutronix.de, mingo@elte.hu, linux-tip-commits@vger.kernel.org In-Reply-To: <20101029202809.GB6130@lenovo> References: <1288276854.2649.607.camel@edumazet-laptop> <1288377146.1988.0.camel@laptop> <20101029202809.GB6130@lenovo> Content-Type: text/plain; charset="UTF-8" Date: Fri, 29 Oct 2010 22:53:07 +0200 Message-ID: <1288385587.2680.13.camel@edumazet-laptop> Mime-Version: 1.0 X-Mailer: Evolution 2.30.3 Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Le samedi 30 octobre 2010 à 00:28 +0400, Cyrill Gorcunov a écrit : > On Fri, Oct 29, 2010 at 08:32:26PM +0200, Peter Zijlstra wrote: > > On Fri, 2010-10-29 at 06:43 +0000, tip-bot for Eric Dumazet wrote: > > > + irqctx = page_address(alloc_pages_node(cpu_to_node(cpu), > > > + THREAD_FLAGS, > > > + THREAD_ORDER)); > > > > Shouldn't we be checking for a NULL return from alloc_pages_node() > > before calling page_address() on it? > > -- > I didnt check for NULL because original code was not either. If you cannot allocate memory for the IRQ stack, only choice is to crash anyway. Adding BUG_ON() is not that helpful in this respect. > Something like below I guess, but probably we could try to allocate > on appropriate NUMA node first and if it fails -- via old alloc_pages > and if it fail in turn -- then we panic. > > Cyrill > --- > arch/x86/kernel/irq_32.c | 17 +++++++++++------ > 1 file changed, 11 insertions(+), 6 deletions(-) > > Index: linux-2.6.git/arch/x86/kernel/irq_32.c > ===================================================================== > --- linux-2.6.git.orig/arch/x86/kernel/irq_32.c > +++ linux-2.6.git/arch/x86/kernel/irq_32.c > @@ -122,13 +122,16 @@ execute_on_irq_stack(int overflow, struc > void __cpuinit irq_ctx_init(int cpu) > { > union irq_ctx *irqctx; > + struct page *page; > > if (per_cpu(hardirq_ctx, cpu)) > return; > > - irqctx = page_address(alloc_pages_node(cpu_to_node(cpu), > - THREAD_FLAGS, > - THREAD_ORDER)); > + page = alloc_pages_node(cpu_to_node(cpu), > + THREAD_FLAGS, THREAD_ORDER); > + BUG_ON(!page); > + > + irqctx = page_address(page); > irqctx->tinfo.task = NULL; > irqctx->tinfo.exec_domain = NULL; > irqctx->tinfo.cpu = cpu; > @@ -137,9 +140,11 @@ void __cpuinit irq_ctx_init(int cpu) > > per_cpu(hardirq_ctx, cpu) = irqctx; > > - irqctx = page_address(alloc_pages_node(cpu_to_node(cpu), > - THREAD_FLAGS, > - THREAD_ORDER)); > + page = alloc_pages_node(cpu_to_node(cpu), > + THREAD_FLAGS, THREAD_ORDER); > + BUG_ON(!page); > + > + irqctx = page_address(page); > irqctx->tinfo.task = NULL; > irqctx->tinfo.exec_domain = NULL; > irqctx->tinfo.cpu = cpu;