From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757579Ab0CKLsJ (ORCPT ); Thu, 11 Mar 2010 06:48:09 -0500 Received: from bombadil.infradead.org ([18.85.46.34]:44985 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757508Ab0CKLsG (ORCPT ); Thu, 11 Mar 2010 06:48:06 -0500 Subject: Re: [PATCH 0/9] oprofile, perf, x86: introduce new functions to reserve perfctrs From: Peter Zijlstra To: Robert Richter Cc: Ingo Molnar , LKML , oprofile-list In-Reply-To: <1267725559.25158.208.camel@laptop> References: <1267716131-17908-1-git-send-email-robert.richter@amd.com> <1267725559.25158.208.camel@laptop> Content-Type: text/plain; charset="UTF-8" Date: Thu, 11 Mar 2010 12:48:01 +0100 Message-ID: <1268308081.5037.14.camel@laptop> Mime-Version: 1.0 X-Mailer: Evolution 2.28.3 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, 2010-03-04 at 18:59 +0100, Peter Zijlstra wrote: > On Thu, 2010-03-04 at 16:22 +0100, Robert Richter wrote: > > This patch set improves the perfctr reservation code. New functions > > are available to reserve a counter by its index only. It is no longer > > necessary to allocate both msrs of a counter which also improves the > > code and makes it easier. > > > > For oprofile a handler is implemented that returns an error now if a > > counter is already reserved by a different subsystem such as perf or > > watchdog. Before, oprofile silently ignored that counter. Finally the > > new reservation functions can be used to allocate special parts of the > > pmu such as IBS, which is necessary to use IBS with perf too. > > > > The patches are available in the oprofile tree: > > > > git://git.kernel.org/pub/scm/linux/kernel/git/rric/oprofile.git core > > > > If there are no objections, I suggest to merge it into the > > tip/perf/core too, maybe after pending patches went in. If there are > > already conflicts, I will do the merge for this. > > Right, so cleaning up that reservation code is nice, but wouldn't it be > much nicer to simply do away with all that and make everything use the > (low level) perf code? Alternatively, could we maybe further simplify this reservation into: int reserve_pmu(void); void release_pmu(void); And not bother with anything finer grained.