From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756123AbaEIB3h (ORCPT ); Thu, 8 May 2014 21:29:37 -0400 Received: from mail.anarazel.de ([217.115.131.40]:33634 "EHLO mail.anarazel.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755569AbaEIB3e (ORCPT ); Thu, 8 May 2014 21:29:34 -0400 From: Andres Freund To: Borislav Petkov Cc: x86@kernel.org, linux-kernel@vger.kernel.org Subject: Date: Fri, 9 May 2014 03:29:15 +0200 Message-Id: <1399598957-7011-1-git-send-email-andres@anarazel.de> X-Mailer: git-send-email 2.0.0.rc2.4.g1dc51c6.dirty Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, After 3.14 I noticed that perf started to only support software events. The reason - found after a some staring deathmatches because the patches look innocent - is a typo in __flip_bit() introduced by 22085a66c2fab6cf9b9393c056a3600a6b4735de which only takes effect after c0a639ad0bc6b178b46996bd1f821a04643e2bde. Due to the typo no msr writes via either msr_clear/set_bit actually write back the new value. That then causes cpuid limits not to be lifted causing 'my' perf and very likely some other problems. There's also a typo in the former commit which makes it impossible to actually use the MSR_IA32_MISC_ENABLE_LIMIT_CPUID macro correctly. Regards, Andres Freund