From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from fg-out-1718.google.com ([72.14.220.158]:33398 "EHLO fg-out-1718.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751297Ab0C3Rdt (ORCPT ); Tue, 30 Mar 2010 13:33:49 -0400 Received: by fg-out-1718.google.com with SMTP id l26so3637045fgb.1 for ; Tue, 30 Mar 2010 10:33:48 -0700 (PDT) Message-ID: <4bb235fb.0baa660a.5b56.ffffa9c2@mx.google.com> Date: Tue, 30 Mar 2010 10:33:47 -0700 (PDT) From: Ben Gamari Subject: Re: Another AR5008 hang To: Johannes Berg Cc: linux-wireless@vger.kernel.org, ath9k@lists.ath9k.org In-Reply-To: <4bb1f940.47c2f10a.3469.609b@mx.google.com> References: <4bb0e2da.04c2f10a.32b2.11ae@mx.google.com> <1269928986.3927.5.camel@jlt3.sipsolutions.net> <4bb1f940.47c2f10a.3469.609b@mx.google.com> Sender: linux-wireless-owner@vger.kernel.org List-ID: On Tue, 30 Mar 2010 06:14:40 -0700 (PDT), Ben Gamari wrote: > On Tue, 30 Mar 2010 08:03:06 +0200, Johannes Berg wrote: > > On Mon, 2010-03-29 at 10:26 -0700, Ben Gamari wrote: > > > New (2.6.33 and greater) kernels seem to result in a selective hang on my > > > machine directly only seconds after association with an 802.11N access point > > > (hosted with hostapd and another ar5008). I've included dmesg output for a > > > failure event, including back traces of hung processes. What ath9k.debug flags > > > would be most helpful for tracing this further? > > > > None, I would think, this seems to be a lockup between > > modprobe/stop_machine and some rtnl things, but I have no idea why that > > would happen. Do you have lockdep enabled? And if you do, please disable > > RCU lockdep as it typically gives you a warning and then turns off > > lockdep... > > > Thank you very much for your response. I do not currently have lockdep enabled, > but I'm building a new kernel as we speak. Thanks again, > Well, it will take me a bit longer than expected to get lockdep results on 2.6.34 due to a recent change[1] in how sysfs attributes are exposed to lockdep. Hopefully this will be fixed soon. That being said, after discovering this I went back and recompiled with lockdep. Unfortunately, I'm now having trouble reproducing the issue on this release. I still can with 2.6.34-rc1 though, so, again, hopefully soon I'll be able to say where the deadlock occurs. I'll let you know as soon as I have news. Thanks for your help, - Ben [1] http://lkml.org/lkml/2010/3/8/365