From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Richard W.M. Jones" Subject: Re: intel_pstate divide error with v3.13-rc4-256-gb7000ad Date: Fri, 27 Dec 2013 17:17:38 +0000 Message-ID: <20131227171738.GO28747@redhat.com> References: <20131227122422.62857cec@alan.etchedpixels.co.uk> <20131227124749.GE10961@minantech.com> <52BD8B8B.1050209@redhat.com> <20131227165248.GG10961@minantech.com> <20131227170147.GH10961@minantech.com> Mime-Version: 1.0 Return-path: Content-Disposition: inline In-Reply-To: <20131227170147.GH10961@minantech.com> Sender: cpufreq-owner@vger.kernel.org List-ID: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Gleb Natapov Cc: Kashyap Chamarthy , Josh Boyer , One Thousand Gnomes , Viresh Kumar , Dirk Brandewie , "Rafael J. Wysocki" , "cpufreq@vger.kernel.org" , Linux PM list , "Linux-Kernel@Vger. Kernel. Org" On Fri, Dec 27, 2013 at 07:01:48PM +0200, Gleb Natapov wrote: > On Fri, Dec 27, 2013 at 06:52:48PM +0200, Gleb Natapov wrote: > > On Fri, Dec 27, 2013 at 03:15:39PM +0100, Kashyap Chamarthy wrote: > > > [. . .] > > > > > > >> KVM does not emulate P-states at all. intel_pstate_init() calls > > > >> intel_pstate_msrs_not_valid() before printing "Intel P-state driver > > > >> initializing." which suppose to fail since it checks that two reads of > > > >> MSR_IA32_APERF return different values, but KVM does not emulate this msr > > > >> at all, so both calls should return zero (KVM suppose to inject #GP, all rdmsrl > > > >> are patched to be rdmsrl_safe in a guest). > > > >> > > > >> Anything interesting in host dmesg? > > > > > > Heya Gleb, > > > > > > Here's the relevant dmesg snippet (full dmesg, refer the attachment below): > > That's guest dmesg. What about host one? Can you ftrace the failure? > > > Ugh, it looks like guest dmesg but there are KVM messages there too ("[ > 281.443662] kvm [2452]: vcpu0 unhandled rdmsr: 0xe8" is unhandled access > to MSR_IA32_APERF I was talking about above), so I guess this is nested > guest invocation? Does it happen in non nested guest? FWIW I could not reproduce the reported bug. I am also using a guest for testing, but *not* using nested KVM. Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming blog: http://rwmj.wordpress.com Fedora now supports 80 OCaml packages (the OPEN alternative to F#)