From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754244AbZHMKsR (ORCPT ); Thu, 13 Aug 2009 06:48:17 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1754223AbZHMKsP (ORCPT ); Thu, 13 Aug 2009 06:48:15 -0400 Received: from casper.infradead.org ([85.118.1.10]:47438 "EHLO casper.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754188AbZHMKsN (ORCPT ); Thu, 13 Aug 2009 06:48:13 -0400 Message-Id: <20090813103655.425428004@chello.nl> References: <20090813094752.433445710@chello.nl> User-Agent: quilt/0.46-1 Date: Thu, 13 Aug 2009 11:47:56 +0200 From: Peter Zijlstra To: Ingo Molnar , Paul Mackerras , stephane eranian Cc: Corey J Ashford , LKML , Peter Zijlstra Subject: [PATCH 4/4] perf_counter: Fix swcounter context invariance Content-Disposition: inline; filename=perf-fix-swcounter-is-counting.patch Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org perf_swcounter_is_counting() uses a lock, which means we cannot use swcounters from NMI or when holding that particular lock, this is unintended. The below removes the lock, this opens up race window, but not worse than the swcounters already experience due to RCU traversal of the context in perf_swcounter_ctx_event(). Cc: Paul Mackerras Signed-off-by: Peter Zijlstra --- kernel/perf_counter.c | 44 ++++++++++++++++++-------------------------- 1 file changed, 18 insertions(+), 26 deletions(-) Index: linux-2.6/kernel/perf_counter.c =================================================================== --- linux-2.6.orig/kernel/perf_counter.c +++ linux-2.6/kernel/perf_counter.c @@ -3519,40 +3519,32 @@ static void perf_swcounter_add(struct pe static int perf_swcounter_is_counting(struct perf_counter *counter) { - struct perf_counter_context *ctx; - unsigned long flags; - int count; - + /* + * The counter is active, we're good! + */ if (counter->state == PERF_COUNTER_STATE_ACTIVE) return 1; + /* + * The counter is off/error, not counting. + */ if (counter->state != PERF_COUNTER_STATE_INACTIVE) return 0; /* - * If the counter is inactive, it could be just because - * its task is scheduled out, or because it's in a group - * which could not go on the PMU. We want to count in - * the first case but not the second. If the context is - * currently active then an inactive software counter must - * be the second case. If it's not currently active then - * we need to know whether the counter was active when the - * context was last active, which we can determine by - * comparing counter->tstamp_stopped with ctx->time. - * - * We are within an RCU read-side critical section, - * which protects the existence of *ctx. + * The counter is inactive, if the context is active + * we're part of a group that didn't make it on the 'pmu', + * not counting. */ - ctx = counter->ctx; - spin_lock_irqsave(&ctx->lock, flags); - count = 1; - /* Re-check state now we have the lock */ - if (counter->state < PERF_COUNTER_STATE_INACTIVE || - counter->ctx->is_active || - counter->tstamp_stopped < ctx->time) - count = 0; - spin_unlock_irqrestore(&ctx->lock, flags); - return count; + if (counter->ctx->is_active) + return 0; + + /* + * We're inactive and the context is too, this means the + * task is scheduled out, we're counting events that happen + * to us, like migration events. + */ + return 1; } static int perf_swcounter_match(struct perf_counter *counter, --