From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754522Ab0HCIBX (ORCPT ); Tue, 3 Aug 2010 04:01:23 -0400 Received: from rcsinet10.oracle.com ([148.87.113.121]:21485 "EHLO rcsinet10.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754069Ab0HCIBU (ORCPT ); Tue, 3 Aug 2010 04:01:20 -0400 Message-ID: <4C57CCBB.1010608@kernel.org> Date: Tue, 03 Aug 2010 01:00:59 -0700 From: Yinghai Lu User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.11) Gecko/20100714 SUSE/3.0.6 Thunderbird/3.0.6 MIME-Version: 1.0 To: "Eric W. Biederman" CC: Dave Airlie , LKML , Ingo Molnar Subject: Re: oops in ioapic_write_entry References: <4C577197.9020003@kernel.org> <4C57723C.1060400@kernel.org> <4C57C319.8070800@kernel.org> In-Reply-To: <4C57C319.8070800@kernel.org> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Source-IP: acsmt354.oracle.com [141.146.40.154] X-Auth-Type: Internal IP X-CT-RefId: str=0001.0A090206.4C57CCC3.0008,ss=1,fgs=0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 08/03/2010 12:19 AM, Yinghai Lu wrote: > On 08/02/2010 08:13 PM, Eric W. Biederman wrote: >> Yinghai Lu writes: >> >>> On 08/02/2010 06:32 PM, Yinghai Lu wrote: >>>> On 08/02/2010 04:17 PM, Dave Airlie wrote: >>>>>> >>>>>> the kernel is using mptable, and the system have mcp55, so how come >>>>>> with irq 35? >>>>>> assume we should only have ioapic irq 0 - 23 ... >>>>>> >>>>>> Can you send out boot log with "debug apic=debug pci=routeirq" with >>>>>> 2.6.32 and 2.6.35? >>>>> >>>>> Okay el6log is from a RHEL6 2.6.32 kernel, but it should give a good >>>>> baseline, the 2.6.35 oops even earlier with all those options and is >>>>> in the second attachment. >>>> >>> >> >> This patch is wrong and there is no reason to even suspect it will >> affect this problem. At best this patch will trade one set of bugs >> for another because at least on some platforms we always did something >> like this. Having an irq 35 is odd and certainly a result of recent >> changes, but in this case it doesn't look like it has anything to do >> with the problem. >> >> Nacked-by: "Eric W. Biederman" >> >>> please use this one instead..., forget to run quilt refresh before sending it. >>> >>> [PATCH -v2] x86: fix pin_2_irq mapping >>> >>> We should not twist gsi to irq mapping if acpi is not used. >>> >>> -v2 remove not used irq_to_gsi() >>> >>> Signed-off-by: Yinghai Lu >>> >>> --- >>> arch/x86/include/asm/io_apic.h | 10 ++++++++++ >>> arch/x86/kernel/acpi/boot.c | 4 ++-- >>> arch/x86/kernel/apic/io_apic.c | 5 +---- >>> 3 files changed, 13 insertions(+), 6 deletions(-) >>> >>> Index: linux-2.6/arch/x86/include/asm/io_apic.h >>> =================================================================== >>> --- linux-2.6.orig/arch/x86/include/asm/io_apic.h >>> +++ linux-2.6/arch/x86/include/asm/io_apic.h >>> @@ -185,6 +185,16 @@ int mp_find_ioapic_pin(int ioapic, u32 g >>> void __init mp_register_ioapic(int id, u32 address, u32 gsi_base); >>> extern void __init pre_init_apic_IRQ0(void); >>> >>> +#ifdef CONFIG_ACPI >>> +unsigned int gsi_to_irq(unsigned int gsi); >>> +u32 irq_to_gsi(int irq); >>> +#else >>> +static inline unsigned int gsi_to_irq(unsigned int gsi) >>> +{ >>> + return gsi; >>> +} >>> +#endif >>> + >>> #else /* !CONFIG_X86_IO_APIC */ >>> >>> #define io_apic_assign_pci_irqs 0 >>> Index: linux-2.6/arch/x86/kernel/acpi/boot.c >>> =================================================================== >>> --- linux-2.6.orig/arch/x86/kernel/acpi/boot.c >>> +++ linux-2.6/arch/x86/kernel/acpi/boot.c >>> @@ -100,7 +100,7 @@ static u32 isa_irq_to_gsi[NR_IRQS_LEGACY >>> 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 >>> }; >>> >>> -static unsigned int gsi_to_irq(unsigned int gsi) >>> +unsigned int gsi_to_irq(unsigned int gsi) >>> { >>> unsigned int irq = gsi + NR_IRQS_LEGACY; >>> unsigned int i; >>> @@ -123,7 +123,7 @@ static unsigned int gsi_to_irq(unsigned >>> return irq; >>> } >>> >>> -static u32 irq_to_gsi(int irq) >>> +u32 irq_to_gsi(int irq) >>> { >>> unsigned int gsi; >>> >>> Index: linux-2.6/arch/x86/kernel/apic/io_apic.c >>> =================================================================== >>> --- linux-2.6.orig/arch/x86/kernel/apic/io_apic.c >>> +++ linux-2.6/arch/x86/kernel/apic/io_apic.c >>> @@ -1029,10 +1029,7 @@ static int pin_2_irq(int idx, int apic, >>> } else { >>> u32 gsi = mp_gsi_routing[apic].gsi_base + pin; >>> >>> - if (gsi >= NR_IRQS_LEGACY) >>> - irq = gsi; >>> - else >>> - irq = gsi_top + gsi; >>> + irq = gsi_to_irq(gsi); >>> } >>> >>> #ifdef CONFIG_X86_32 > > what is the point for making irq = gsi_top + gsi when mptable is used instead of acpi? > just tried those blind shifting gsi cause kernel with acpi crash in virtual box. [ 5.536000] querying PCI -> IRQ mapping bus:0, slot:11, pin:0. [ 5.540000] ehci_hcd 0000:00:0b.0: can't find IRQ for PCI INT A; probably buggy MP table [ and on kvm it got: [ 4.352280] e1000: Intel(R) PRO/1000 Network Driver - version 7.3.21-k6-NAPI [ 4.356012] e1000: Copyright (c) 1999-2006 Intel Corporation. [ 4.360120] querying PCI -> IRQ mapping bus:0, slot:3, pin:0. [ 4.364006] PCI BIOS passed nonexistent PCI bus 0! [ 4.368007] e1000 0000:00:03.0: can't find IRQ for PCI INT A; probably buggy MP table [ 4.372049] e1000 0000:00:03.0: setting latency timer to 64 Yinghai