From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joerg Roedel Subject: Re: [PATCH 1/3] perf: add context field to perf_event Date: Tue, 12 Jul 2011 12:24:17 +0200 Message-ID: <20110712102417.GD29812@8bytes.org> References: <4E1C10F8.6010300@redhat.com> <1310462335.14978.12.camel@twins> <4E1C1373.5080500@redhat.com> <1310463060.14978.17.camel@twins> <20110712094131.GA29812@8bytes.org> <4E1C1771.9010300@redhat.com> <20110712094822.GB29812@8bytes.org> <4E1C1A0D.8000707@redhat.com> <20110712100354.GC29812@8bytes.org> <4E1C1CDD.9040005@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Peter Zijlstra , Will Deacon , Frederic Weisbecker , "linux-kernel@vger.kernel.org" , "kvm@vger.kernel.org" , Ingo Molnar , "acme@ghostprotocols.net" , Jason Wessel To: Avi Kivity Return-path: Content-Disposition: inline In-Reply-To: <4E1C1CDD.9040005@redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On Tue, Jul 12, 2011 at 01:07:25PM +0300, Avi Kivity wrote: > On 07/12/2011 01:03 PM, Joerg Roedel wrote: >> Ingo made perf-integration a merge-requirement for LWP. It is not really >> well-suited for being integrated into perf because the design goal was >> easy and efficient self-profiling of tasks (like you stated). So >> integrating it into perf causes some pain. But lets see how it works >> out. > > I don't think it's workable. Having do_mmap() called in the task's > context can change how it works. According to Ingo that is already done at other places in the kernel and should not be an issue. > And the task being able to kill/modify the profile, and not able to > use LWP for itself, is a show stopper IMO. I agree, that is a problem we have no solution for so far. Joerg