From mboxrd@z Thu Jan 1 00:00:00 1970 From: Wei Huang Subject: Re: Panic:: handle_irq_event_percpu Date: Wed, 26 Oct 2011 09:21:49 -0500 Message-ID: <4EA8177D.40707@amd.com> References: <1904679511.20111018205519@eikelenboom.it> <20111019160801.GA22918@phenom.dumpdata.com> <707354928.20111019193003@eikelenboom.it> <884563847.20111022165748@eikelenboom.it> <4EA5DDEB.90308@amd.com> <11510207729.20111026161232@eikelenboom.it> Reply-To: wei.huang2@amd.com Mime-Version: 1.0 Content-Type: text/plain; charset="ISO-8859-15"; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <11510207729.20111026161232@eikelenboom.it> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Sander Eikelenboom Cc: "xen-devel@lists.xensource.com" , Rzeszutek Wilk , Konrad List-Id: xen-devel@lists.xenproject.org That is good. We are preparing a series of patches for xen-4.1.3. They should be out in next few days. Thanks for your help on testing. -Wei On 10/26/2011 09:12 AM, Sander Eikelenboom wrote: > Hello Wei, > > Sorry you were right, i'm a bit accustomed that patches apply to the root of the project, this one was for /xen specifically. > I have done a lot of reboots with the patch applied on: > - xen 4.1.3-rc1-pre > - dom0 kernel from Konrad's linux-next tree + devel/acpi-cpufreq.v3 branch + patch from liang tang + linus 3.1 tree pulled in. > - iommu specific boot options used for xen in grub: iommu=on,amd-iommu-debug amd_iommu=on amd-iommu-debug > > I haven't seen kernel panics on boot so far, is this output that was expected (from serial console): > > > (XEN) AMD-Vi: Add device table entry: device id = 0x0a07, interupt table = 0x24e004000 > (XEN) AMD-Vi: Add device table entry: device id = 0x0b00, interupt table = 0x24e004000 > (XEN) AMD-Vi: IOMMU 0 Enabled. > (XEN) AMD-Vi: Enabling global vector map > (XEN) I/O virtualisation enabled > (XEN) - Dom0 mode: Relaxed > (XEN) Getting VERSION: 80050010 > > > Complete serial console output is attached > > > Is this patch a candidate for 4.1.3 ? > > -- > > Sander > > > > Monday, October 24, 2011, 11:51:39 PM, you wrote: > >> Yes, it is for xen-4.1-testing tree. I just applied it to the latest tip (23174:d4253bc5418b). There were some hunks; but it worked in general. Could you try it and replace your xen.gz file with the new version? >> (Stripping trailing CRs from patch.) >> patching file arch/x86/io_apic.c >> Hunk #1 succeeded at 548 (offset -21 lines). >> (Stripping trailing CRs from patch.) >> patching file arch/x86/irq.c >> Hunk #5 succeeded at 192 (offset 1 line). >> Hunk #6 succeeded at 225 (offset 1 line). >> Hunk #7 succeeded at 236 (offset 1 line). >> Hunk #8 succeeded at 302 (offset 1 line). >> Hunk #9 succeeded at 372 (offset 1 line). >> Hunk #10 succeeded at 425 (offset 1 line). >> Hunk #11 succeeded at 440 (offset 1 line). >> Hunk #12 succeeded at 476 (offset 1 line). >> Hunk #13 succeeded at 497 (offset 1 line). >> Hunk #14 succeeded at 1599 (offset -11 lines). >> Hunk #15 succeeded at 1647 (offset -11 lines). >> (Stripping trailing CRs from patch.) >> patching file drivers/passthrough/amd/pci_amd_iommu.c >> (Stripping trailing CRs from patch.) >> patching file include/asm-x86/irq.h >> (Stripping trailing CRs from patch.) >> patching file include/xen/pci.h > > >> -Wei >> On 10/22/2011 09:57 AM, Sander Eikelenboom wrote: >>> Hello Wei, >>> >>> Is this for xen 4.1.2 ? >>> All hunks of the patch seem to be rejected. >>> >>> -- >>> Sander >>> >>> >>> Wednesday, October 19, 2011, 8:27:55 PM, you wrote: >>> >>>> This reminds me of 23768:09595fdf3638, which reverted WeiWang's patch of forcing perdev-intremap as default. Could you try the attached patch on your Xen 4.1 tree and test with patched xen.gz? >>>> Thanks, >>>> -Wei >>>> -----Original Message----- >>>> From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of Sander Eikelenboom >>>> Sent: Wednesday, October 19, 2011 12:30 PM >>>> To: Huang2, Wei >>>> Cc: xen-devel@lists.xensource.com; Konrad Rzeszutek Wilk >>>> Subject: Re: [Xen-devel] Panic:: handle_irq_event_percpu >>>> Hello Wei, >>>> Result is that using only "1" or "1" in combination with "2" always result in SATA I/O errors, so that doesn't boot. >>>> Haven't tried rebooting a lot of times without iommu yet. >>>> (the reason i bought the 890fx board in the first place) >>>> Will try that a few times when i have some more time. >>>> -- >>>> Sander >>>> Wednesday, October 19, 2011, 6:38:38 PM, you wrote: >>>>> It looks similar. >>>>> Hi Sander, could you please test your system with the following options? >>>>> 1. iommu=amd-iommu-perdev-intremap in xen.gz GRUB entry >>>>> 2. pci=nosmi in pvops GRUB entry >>>>> 3. iommu=0 in xen.gz GRUB entry [optional, only if (1) doesn't work for you] >>>>> We were seeing an issue which was caused by messed-up interrupt remap table. George Dunlap's perdev interrupt map solved our problem. >>>>> Thanks, >>>>> -Wei >>>>> -----Original Message----- >>>>> From: Konrad Rzeszutek Wilk [mailto:konrad.wilk@oracle.com] >>>>> Sent: Wednesday, October 19, 2011 11:08 AM >>>>> To: Sander Eikelenboom; Huang2, Wei >>>>> Cc: xen-devel@lists.xensource.com >>>>> Subject: Re: [Xen-devel] Panic:: handle_irq_event_percpu >>>>> On Tue, Oct 18, 2011 at 08:55:19PM +0200, Sander Eikelenboom wrote: >>>>>> Hi Konrad, >>>>>> >>>>>> On my system (AND phenom x6, 8gb mem) running xen 4.1.2-rc3 >>>>>> and a custom kernel based on: >>>>>> >>>>>> - your linux-next branch >>>>>> - pulled in your apci/cpufreq branch >>>>>> - pulled in latest patches from linuses tree >>>>> Wei, >>>>> Are those similar to what you had been seeing? >>>>>> I'm experiencing infrequent panics at boot (say 1 out of 10 or 20 boots), doesn't seem to matter if it's a cold or a warm boot. >>>>>> It always seem to happen during or shortly after booting. When it doesn't occur then, it can run for at least several days/weeks (haven't seen it panic then). >>>>>> It doesn't seem to happen at a particular point at or shortly after booting, but it happens before starting any guests >>>>>> >>>>>> Unfortunately i haven't had serial console enabled at these times, but i have made some photo's. >>>>>> The first 3 times it locked completely before being able to print a stacktrace, the 4th time it did (although the photo is a bit unsharp). >>>>>> >>>>>> >>>>>> So all in all, it kind of strange, i hope you can figure something out from the stacktrace, if not, i will see if i can try with serial console attached. >>>>>> >>>>>> -- >>>>>> Sander >>> >>> >>> >>> >>>>>> _______________________________________________ >>>>>> Xen-devel mailing list >>>>>> Xen-devel@lists.xensource.com >>>>>> http://lists.xensource.com/xen-devel >>> >>> >>> >>> >>> >>> >>> >>> > > > >