From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from phoenix.szarvasnet.hu ([87.101.127.3]:49469 "EHLO phoenix.szarvas.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757744AbZBLNdu (ORCPT ); Thu, 12 Feb 2009 08:33:50 -0500 Message-ID: <49942516.2090607@openwrt.org> (sfid-20090212_143354_358923_01172224) Date: Thu, 12 Feb 2009 14:33:10 +0100 From: Gabor Juhos MIME-Version: 1.0 To: Johannes Berg CC: "linux-wireless@vger.kernel.org" Subject: Re: soft-safe -> soft-unsafe lock order detected References: <4991AC67.90104@openwrt.org> (sfid-20090210_173417_726855_7504F887) <1234433807.16004.88.camel@johannes.local> (sfid-20090212_111736_123131_223B74EE) <1234434246.16004.90.camel@johannes.local> <49941C7D.6090600@openwrt.org> <1234444655.1327.3.camel@johannes.local> In-Reply-To: <1234444655.1327.3.camel@johannes.local> Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: Johannes Berg =C3=ADrta: > On Thu, 2009-02-12 at 13:56 +0100, Gabor Juhos wrote: >> Johannes Berg =C3=ADrta: >>> On Thu, 2009-02-12 at 11:16 +0100, Johannes Berg wrote: >>>>> to a soft-irq-unsafe lock: >>>>> (todo_lock){--..} >>>>> ... which became soft-irq-unsafe at: >>>>> ... [<800b2248>] __lock_acquire+0x624/0x844 >>>>> [<800b24c4>] lock_acquire+0x5c/0x84 >>>>> [<8006b8c4>] _spin_lock+0x34/0x48 >>>>> [] ieee80211_set_default_key+0x4b8/0x4f0 [mac80211] >>>> That seems to be incorrect. ieee80211_set_default_key will have >>>> _irqsave-locked the key lock, so the todo lock is here always lock= ed in >>>> an irq-excluded section. >> Yes, seems to be incorrect. >> >>>> The lock is, however, possibly used that way in ieee80211_key_link= , >>>> which can be fixed easily and we can remove the todo lock too. >>> Not true, but the todo_lock should be made to _irqsave in some plac= es, >>> it seems. >=20 > Actually that doesn't seem to be true either... >=20 >> Should we handle this as a false positive for now, because I have se= en this only >> once, and the transfer has not been interrupted? I will do more test= s on >> different platforms, and i will come back if this happens again. >=20 > Yes, I think we should, something seems to be messed up with this > report. All code paths are regularly executed on all platforms, so we > should be seeing this a lot more if there was a problem. Ok. Thank you for the analysis. Gabor -- To unsubscribe from this list: send the line "unsubscribe linux-wireles= s" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html