From mboxrd@z Thu Jan 1 00:00:00 1970 From: Paolo Bonzini Subject: Re: [PATCH V5 1/2] perf ignore LBR and extra_regs Date: Mon, 14 Jul 2014 14:40:33 +0200 Message-ID: <53C3CFC1.9070800@redhat.com> References: <1404989984-3068-1-git-send-email-kan.liang@intel.com> <53C3C517.4000500@redhat.com> <20140714120922.GV9918@twins.programming.kicks-ass.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: kan.liang@intel.com, andi@firstfloor.org, linux-kernel@vger.kernel.org, kvm@vger.kernel.org To: Peter Zijlstra Return-path: In-Reply-To: <20140714120922.GV9918@twins.programming.kicks-ass.net> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org Il 14/07/2014 14:09, Peter Zijlstra ha scritto: > On Mon, Jul 14, 2014 at 01:55:03PM +0200, Paolo Bonzini wrote: >> Il 10/07/2014 12:59, kan.liang@intel.com ha scritto: >>> From: Kan Liang >>> >>> x86, perf: Protect LBR and extra_regs against KVM lying >>> >>> With -cpu host, KVM reports LBR and extra_regs support, if the host has support. >>> When the guest perf driver tries to access LBR or extra_regs MSR, >>> it #GPs all MSR accesses,since KVM doesn't handle LBR and extra_regs support. >>> So check the related MSRs access right once at initialization time to avoid the error access at runtime. >>> >>> For reproducing the issue, please build the kernel with CONFIG_KVM_INTEL = y (for host kernel). >>> And CONFIG_PARAVIRT = n and CONFIG_KVM_GUEST = n (for guest kernel). >> >> I'm not sure this is a useful patch. >> >> This is #GP'ing just because of a limitation in the PMU; just compile the >> kernel with CONFIG_PARAVIRT > > How's that going to help? If you run kvm -host the VM is lying through > its teeth is the kernel is going to assume all those MSRs present, > PARAVIRT isn't going to help with this. > >> , or split the "rdmsr is always rdmsr_safe" >> behavior out of CONFIG_PARAVIRT and into a new Kconfig symbol. > > That's not useful either, because non of these code-paths are going to > check the return value. Hmmm, I thought rdmsr_safe was going to return zero, but it just returns whatever happened to be in edx:eax which maybe should also be fixed. Kan Liang, what happens if CONFIG_PARAVIRT=y? Do you get garbage or just no events reported? >> In fact there's no reason why LBR cannot be virtualized (though it does need >> support from the processor), and it may even be possible to support >> OFFCORE_RSP_X in the KVM virtual PMU. > > But its not, so something needs to be done, right? A first thing that could be done, is to have a way for the kernel to detect absence of LBR, for example an all-1s setting of the LBR format field of IA32_PERF_CAPABILITIES. If Intel can tell us "all 1s will never be used", we can have KVM set the field that way. The kernel then should disable LBR. Paolo