From mboxrd@z Thu Jan 1 00:00:00 1970 From: Weidong Han Subject: Re: Xen 4.0.0-rc7 problem/hang with vt-d DMAR parsing Date: Wed, 07 Apr 2010 09:43:39 +0800 Message-ID: <4BBBE34B.8090007@intel.com> References: <20100323193748.GW1878@reaktio.net> <20100323200515.GZ1878@reaktio.net> <4BA9DA400200007800036ABB@vpn.id2.novell.com> <4BA9D512.9090902@intel.com> <4BA9ED8D0200007800036B3F@vpn.id2.novell.com> <4BA9F0C1.7080809@intel.com> <8115AF16522A3D4383C1FF753036713F9B3AB02E@cosmail01.lsi.com> <4BAAA89C.9040308@intel.com> <8115AF16522A3D4383C1FF753036713F9B49CCBE@cosmail01.lsi.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <8115AF16522A3D4383C1FF753036713F9B49CCBE@cosmail01.lsi.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: "Nadolski, Ed" Cc: "xen-devel@lists.xensource.com" , Keir Fraser , Jan Beulich , "Cui, Dexuan" List-Id: xen-devel@lists.xenproject.org Nadolski, Ed wrote: >>> I'm also seeing a hang on boot with 4.0.0-rc7, but I am not clear if >>> >> this is the same issue. Here is the tail end of the trace: >> >>> [ 4.802862] NetLabel: unlabeled traffic allowed by default >>> [ 4.808653] DMAR:Host address width 40 >>> [ 4.812310] DMAR:DRHD base: 0x000000dfffe000 flags: 0x0 >>> [ 4.817604] IOMMU dfffe000: ver 1:0 cap c90780106f0462 ecap f020fe >>> [ 4.823816] DMAR:DRHD base: 0x000000fedc0000 flags: 0x1 >>> [ 4.829105] IOMMU fedc0000: ver 1:0 cap c90780106f0462 ecap f020f6 >>> [ 4.835325] DMAR:RMRR base: 0x000000dbe58000 end: 0x000000dbe6ffff >>> [ 4.841556] DMAR:ATSR flags: 0x0 >>> [ 4.844844] DMAR:ATSR flags: 0x0 >>> [ 4.848133] Not all IO-APIC's listed under remapping hardware >>> [ 4.854028] IOMMU 0xdfffe000: using Queued invalidation >>> [ 4.859217] IOMMU 0xfedc0000: using Queued invalidation >>> [ 4.864493] IOMMU: Setting RMRR: >>> [ 4.867778] IOMMU: Setting identity map for device 0000:00:1d.0 >>> >> [0xdbe58000 - 0xdbe70000] >> >>> [ 4.876053] IOMMU: Setting identity map for device 0000:00:1d.1 >>> >> [0xdbe58000 - 0xdbe70000] >> >>> [ 4.884264] IOMMU: Setting identity map for device 0000:00:1d.2 >>> >> [0xdbe58000 - 0xdbe70000] >> >>> [ 4.892485] IOMMU: Setting identity map for device 0000:00:1d.7 >>> >> [0xdbe58000 - 0xdbe70000] >> >>> [ 4.900709] IOMMU: Setting identity map for device 0000:00:1a.0 >>> >> [0xdbe58000 - 0xdbe70000] >> >>> [ 4.908928] IOMMU: Setting identity map for device 0000:00:1a.1 >>> >> [0xdbe58000 - 0xdbe70000] >> >>> [ 4.917150] IOMMU: Setting identity map for device 0000:00:1a.2 >>> >> [0xdbe58000 - 0xdbe70000] >> >>> [ 4.925390] IOMMU: Setting identity map for device 0000:00:1a.7 >>> >> [0xdbe58000 - 0xdbe70000] >> >>> [ 4.933612] IOMMU: Prepare 0-16MiB unity mapping for LPC >>> [ 4.938938] IOMMU: Setting identity map for device 0000:00:1f.0 >>> >> [0x0 - 0x1000000] >> >>> [ 4.946651] VT-d detected invalid descriptor: low=11, high=0 >>> >>> Then it hangs. This is on a Dell T7500 quad-core Xeon with the latest >>> >> Dell BIOS (rev. A03). Should I try the suggested patch, or any >> particular command line or build options? Should I dump the DMAR >> tables? >> >> Your issue is different. It's caused by enabling VT-d in pv-ops dom0. >> but pv-ops dom0 should not see DMAR table because its signature is >> zapped after xen enables it. Pls have a try with turn off VT-d in >> pvops config (# CONFIG_DMAR is not set) , it should solve your issue. But we >> need to find why pv-ops dom0 still sees DMAR table in ACPI. Pls dump >> DMAR tables. >> > > Changing CONFIG_DMAR as you say fixes the hang. I've also attached the DMAR table dump as you requested, does this look OK? > > Thanks again, > Ed > > When use "iasl" to disassemble attached file, it prompted: Loading Acpi table from file DellT7500_BIOS_A03-dmar.bin Table signature [DMAR] is invalid or not supported Could not get table from the file Did you dump it on dom0 (not native Linux)? if so, that means DMAR signature is already zapped, dom0 should not detect it and won't go to enable it. Regards, Weidong > > > > >