From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Walker Subject: RE: [PATCH 1/3] acpi: add real mutex function calls Date: Thu, 24 Jul 2008 08:44:21 -0400 Message-ID: <1216903461.29701.11.camel@localhost.localdomain> References: <1216491411-24080-1-git-send-email-dwalker@mvista.com> <1216491411-24080-2-git-send-email-dwalker@mvista.com> <1216605078.4135.23.camel@yakui_zhao.sh.intel.com> <1216648574.2294.9.camel@dhcp32.mvista.com> <4884E177.5000809@linux.intel.com> <1216669184.2294.44.camel@dhcp32.mvista.com> <9D39833986E69849A2A8E74C1078B6B3B6925F@orsmsx415.amr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from homer.mvista.com ([63.81.120.158]:62351 "EHLO gateway-1237.mvista.com" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1753696AbYGXMpD (ORCPT ); Thu, 24 Jul 2008 08:45:03 -0400 In-Reply-To: <9D39833986E69849A2A8E74C1078B6B3B6925F@orsmsx415.amr.corp.intel.com> Sender: linux-acpi-owner@vger.kernel.org List-Id: linux-acpi@vger.kernel.org To: "Moore, Robert" Cc: Andi Kleen , "Zhao, Yakui" , linux-acpi@vger.kernel.org, Ingo Molnar , Peter Zijlstra , "Brown, Len" (Added back the CC list.) I don't think the problem is in the machine specific ACPI info .. As far as I can tell the problem is related to how the interpreter lock is interleaved with the AML mutexes . For example, a series of AML locks can be A->B->C and at any point the interpreter lock can be released, and reacquired. So you have a possible lock sequence of I->A->B->C , then in another path you have A->I->B->C. That creates a circular lock scenario and with multiple threads in those different paths you could end up with a deadlock (or a lockdep warning). I was also reading the ACPI spec on how the ASL compiler does checking on mutex usage similar to what lockdep is doing. I would hope there are no lockdep warning coming from that code. Daniel