From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932546Ab1EXOMX (ORCPT ); Tue, 24 May 2011 10:12:23 -0400 Received: from bombadil.infradead.org ([18.85.46.34]:37605 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932326Ab1EXOMV convert rfc822-to-8bit (ORCPT ); Tue, 24 May 2011 10:12:21 -0400 Subject: Re: perf: regression -- missing /sys/devices/system/cpu/perf_events From: Peter Zijlstra To: Vince Weaver Cc: linux-kernel@vger.kernel.org, fbuihuu@gmail.com, mingo@elte.hu, paulus@samba.org, acme@redhat.com In-Reply-To: References: Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 8BIT Date: Tue, 24 May 2011 16:11:46 +0200 Message-ID: <1306246306.18455.36.camel@twins> Mime-Version: 1.0 X-Mailer: Evolution 2.30.3 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, 2011-05-24 at 09:59 -0400, Vince Weaver wrote: > I know this is a bit late, but for some reason our users sit on these > things and then they aren't willing to take it up with linux-kernel > themselves > > > > > commit 15ac9a395a753cb28c674e7ea80386ffdff21785 > > > > Author: Peter Zijlstra > > > > Date: Mon Sep 6 15:51:45 2010 +0200 > > > > > > > > perf: Remove the sysfs bits > > removed the /sys/devices/system/cpu/perf_events > I thought things in /sys were stable ABI? if only.. that stuff changes way too often (not to say that's a good thing). > Apparently it's common for people to have scripts to check if > perf_events is available in a kernel by checking the existence of > /sys/devices/system/cpu/perf_events, even if they didn't use the > files within. A much more reliable way is simply doing the syscall and seeing what happens. But if you want to poke around in sysfs, /sys/bus/event_source/ is the new location. > Now that 2.6.38 kernels are starting to hit the distros we're getting > complaints that it's missing. Urgh, they'd been broken long before.. and I hadn't received any complaints from people about that. I didn't see the point in keeping broken interfaces around, esp. since with moving to multiple-pmu they don't make any sense at all. Ingo, any idea what to do here?