Return-path: Received: from he.sipsolutions.net ([78.46.109.217]:53552 "EHLO sipsolutions.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754395Ab0JFLqV (ORCPT ); Wed, 6 Oct 2010 07:46:21 -0400 Subject: Re: [PATCH] mac80211: hoist sta->lock from reorder release timer From: Johannes Berg To: Christian Lamparter Cc: linux-wireless@vger.kernel.org, linville@tuxdriver.com, Ben Greear , Ming Lei In-Reply-To: <201010061343.23463.chunkeey@googlemail.com> References: <201010061200.54364.chunkeey@googlemail.com> <201010061220.50800.chunkeey@googlemail.com> <1286361705.3655.121.camel@jlt3.sipsolutions.net> <201010061343.23463.chunkeey@googlemail.com> Content-Type: text/plain; charset="UTF-8" Date: Wed, 06 Oct 2010 13:46:18 +0200 Message-ID: <1286365578.3655.187.camel@jlt3.sipsolutions.net> Mime-Version: 1.0 Sender: linux-wireless-owner@vger.kernel.org List-ID: On Wed, 2010-10-06 at 13:43 +0200, Christian Lamparter wrote: > > Basically the thing is that until your patch, the data in the struct > > didn't actually need locking because it was accessed by the RX path only > > which is not concurrent. > > > I see. So basically all rx handlers are affected by these rx->sta races. > > John, can you please revert (or at least drop from the upcoming 2.6.37-rcX cycle): > > (mac80211: fix release_reorder_timeout in scan) > mac80211: fix rcu-unsafe pointer dereference > mac80211: AMPDU rx reorder timeout timer > (mac80211: remove unused rate function parameter) > mac80211: put rx handlers into separate functions I think it's probably easier to fix than to revert now? There are only a handful of fields, and it seemed to me that most of them can easily be moved under the reorder lock. johannes