From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andi Kleen Subject: Re: restricting counted/sampled events Date: Thu, 12 Nov 2015 14:58:45 -0800 Message-ID: <87fv0a3jve.fsf@tassilo.jf.intel.com> References: <5644D7DA.7050200@suse.de> Mime-Version: 1.0 Content-Type: text/plain Return-path: Received: from mga01.intel.com ([192.55.52.88]:40451 "EHLO mga01.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754270AbbKLW6q (ORCPT ); Thu, 12 Nov 2015 17:58:46 -0500 In-Reply-To: <5644D7DA.7050200@suse.de> (Tony Jones's message of "Thu, 12 Nov 2015 10:18:02 -0800") Sender: linux-perf-users-owner@vger.kernel.org List-ID: To: Tony Jones Cc: linux-perf-users@vger.kernel.org Tony Jones writes: > I don't believe there is current support to do this, but maybe I'm overlooking something? It may be possible in systemtap, but not in the builtin tracers yes. > > I'd like to be able to restrict the pmu (or filter results) to report counting/sampling between two arbitrary code points, such as a tracepoint or kprobe. > > I believe it will require core perf changes and was curious on peoples thoughts? I'm not sure about messing with sampling (you can always filter that later). But able to collect PMU counter values are kprobes/tracepoints would be fairly useful yes. It shouldn't be that difficult to do. -Andi -- ak@linux.intel.com -- Speaking for myself only