Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S964783AbbFJHEe (ORCPT ); Wed, 10 Jun 2015 03:04:34 -0400 Received: from mx1.redhat.com ([209.132.183.28]:47856 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753849AbbFJHEQ (ORCPT ); Wed, 10 Jun 2015 03:04:16 -0400 Date: Wed, 10 Jun 2015 15:04:12 +0800 From: Baoquan He To: Yinghai Lu Cc: Borislav Petkov , Ingo Molnar , LKML Subject: Re: [PATCH 3/3] x86/crash: Allocate enough low memory when crashkernel=high Message-ID: <20150610070412.GA16351@dhcp-128-28.nay.redhat.com> References: <1433749795-5796-1-git-send-email-bp@alien8.de> <1433749795-5796-4-git-send-email-bp@alien8.de> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4647 Lines: 112 On 06/08/15 at 05:20pm, Yinghai Lu wrote: > On Mon, Jun 8, 2015 at 12:49 AM, Borislav Petkov wrote: > > From: Joerg Roedel > > > > When the crash kernel is loaded above 4GiB in memory, the > > first kernel allocates only 72MiB of low-memory for the DMA > > requirements of the second kernel. On systems with many > > devices this is not enough and causes device driver > > initialization errors and failed crash dumps. Testing by > > SUSE and Redhat has shown that 256MiB is a good default > > value for now and the discussion has lead to this value as > > well. So set this default value to 256MiB to make sure there > > is enough memory available for DMA. Hi Yinghai, > > Is the 256MiB too big? > > BTW, the affected system does not support iommu? I guess the affected system should not support hw iommu or turned off hw iommu since hw iommu always screw up kdump. > > on my intel test box with 6T and 16 pcie cards does not hit the 72MiB > limit. In fact in this case, it doesn't matter how much memory the system has since most of them is above 4G and only 72M is reserved for dma/swiotlb in kdump kernel. And it doesn't matter much how many pci/pcie cards the system has since it matters how greedy those devices expects dma allocation. Before Joerg posted this patchset v1, our customer complained about this too. That means this is not individual case. I think it makes sense to increase the default low-memory to make kdump succeed, it's definitely better than letting user get a failed kdump and then know they need specify more low-memory manually. If user clearly know their system doesn't need so much low-memory, specifying crashkernel=xx,low works for them. > > So the affected system is 12T and more cards? In the case of my bug, it's a system with 12T and many cards. Thanks Baoquan > > > > > Signed-off-by: Joerg Roedel > > Acked-by: Baoquan He > > Cc: Baoquan He > > Cc: Dave Young > > Cc: "H. Peter Anvin" > > Cc: Ingo Molnar > > Cc: J?rg R?del > > Cc: Thomas Gleixner > > Cc: Vivek Goyal > > Cc: kexec@lists.infradead.org > > Cc: x86-ml > > Link: http://lkml.kernel.org/r/1433500202-25531-4-git-send-email-joro@8bytes.org > > [ Reflow comment. ] > > Signed-off-by: Borislav Petkov > > --- > > arch/x86/kernel/setup.c | 13 ++++++++----- > > 1 file changed, 8 insertions(+), 5 deletions(-) > > > > diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c > > index d74ac33290ae..5a697e56634c 100644 > > --- a/arch/x86/kernel/setup.c > > +++ b/arch/x86/kernel/setup.c > > @@ -531,12 +531,15 @@ static void __init reserve_crashkernel_low(void) > > if (ret != 0) { > > /* > > * two parts from lib/swiotlb.c: > > - * swiotlb size: user specified with swiotlb= or default. > > - * swiotlb overflow buffer: now is hardcoded to 32k. > > - * We round it to 8M for other buffers that > > - * may need to stay low too. > > + * - swiotlb size: user-specified with swiotlb= or default. > > + * > > + * -swiotlb overflow buffer: now hardcoded to 32k. > > why do you need to treat them differently ? > "- swiotlb" and "-swiotlb" > > > + * > > Why do you need to put extra blank line here? that 8M round up > is for swiotlb overflow buffer. > > > + * We round it to 8M for other buffers that may need to stay low > > + * too. Also make sure we allocate enough extra low memory so > > + * that we don't run out of DMA buffers for 32-bit devices. > > */ > > - low_size = swiotlb_size_or_default() + (8UL<<20); > > + low_size = max(swiotlb_size_or_default() + (8UL<<20), 256UL<<20); > > auto_set = true; > > } else { > > /* passed with crashkernel=0,low ? */ > > -- > -- > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > Please read the FAQ at http://www.tux.org/lkml/ -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/