From mboxrd@z Thu Jan 1 00:00:00 1970 From: Igor Mammedov Subject: Re: xen-unstable fails to boot on a system with Ivy Bridge stepping C0 cpu Date: Thu, 03 Nov 2011 16:58:40 +0100 Message-ID: <4EB2BA30.1020307@redhat.com> References: <4EB28CAD.2060401@redhat.com> <20111103131043.GB66800@ocelot.phlegethon.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20111103131043.GB66800@ocelot.phlegethon.org> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Tim Deegan Cc: anthony.perard@citrix.com, jinsong.liu@intel.com, xen-devel@lists.xensource.com, Ian.Campbell@citrix.com List-Id: xen-devel@lists.xenproject.org On 11/03/2011 02:10 PM, Tim Deegan wrote: > At 13:44 +0100 on 03 Nov (1320327869), Igor Mammedov wrote: >> Hi, >> >> I need a help with tracking down following issue: >> >> When trying to boot Xen on a system with Ivy Bridge stepping C0 CPU, it is >> stuck on CPU initialization. >> I've added some tracing to apic writes/reads and traced it so far to >> sending INIT IPI. >> >> (XEN) HVM: VMX enabled >> (XEN) HVM: Hardware Assisted Paging detected. >> (XEN) Setting warm reset code and vector. >> (XEN) apic_wrmsr (0x280,0x0) >> (XEN) apic_rdmsr(0x280) = 0x0 >> (XEN) Asserting INIT. >> (XEN) apic_wrmsr (0x300,0x10000c500) >> >> After this ^^^^ write BSP becomes VERY slow and it takes several minutes >> till it gets to sending STARTUP IPI: > > Does reverting 23724:b3434f24b082 help? Looking at it again I'm not ^^^^^ reverting this c/s won't help, because the same problem exists in older xen versions 3.x.x that doesn't have this c/s. Actually I've tried rhel5, suse 4.1 and xen-unstable so far, all of them are affected. If we put printk right after sending INIT|ASSERT IPI it will show that BSP is slowed down by printing 1 character in a several seconds. So deassert probably is not of an issue here, since this happens on the first INIT to the first AP. > sure that the new logic is entirely correct - in particular it no longer > deasserts the INIT on x2apic systems. > > Cheers, > > Tim. > -- Thanks, Igor