From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754792AbaHKTaJ (ORCPT ); Mon, 11 Aug 2014 15:30:09 -0400 Received: from david.siemens.de ([192.35.17.14]:56637 "EHLO david.siemens.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752712AbaHKTaH (ORCPT ); Mon, 11 Aug 2014 15:30:07 -0400 Message-ID: <53E919AF.1090607@siemens.com> Date: Mon, 11 Aug 2014 21:29:51 +0200 From: Jan Kiszka User-Agent: Mozilla/5.0 (X11; U; Linux i686 (x86_64); de; rv:1.8.1.12) Gecko/20080226 SUSE/2.0.0.12-1.1 Thunderbird/2.0.0.12 Mnenhy/0.7.5.666 MIME-Version: 1.0 To: Thomas Gleixner , Ingo Molnar , "H. Peter Anvin" , x86 , Linux Kernel Mailing List Subject: x86: HPET as CPU-local timer + interrupt remapping = lockup Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi all, just noticed an apparent regression of x86-64 on CPUs without ARAT but a sufficient number if HPET timers: If you have interrupt remapping enabled, the kernel will lock up during boot, apparently waiting for some never-arriving interrupts. I'm currently adding VT-d interrupt emulation to QEMU and stumbled over this behavior. It didn't turn out to be an emulation issue, I just reproduced on real hardware (ARAT patched out) and found some old 3.0 kernel booting fine inside my QEMU version. FWIW, some further details I found out under QEMU: the HPET timers do not seem to be switched to MSI mode yet when the lockup happens. I suppose the issue is uncritical as the combination of hardware feature (or their absence) is probably untypical, correct? I can't invest much into bisecting or debugging right now unfortunately. But maybe someone has an idea what could case this. Jan -- Siemens AG, Corporate Technology, CT RTC ITP SES-DE Corporate Competence Center Embedded Linux