Return-path: Received: from static-ip-62-75-166-246.inaddr.intergenia.de ([62.75.166.246]:47715 "EHLO vs166246.vserver.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S965524AbXC1LdN (ORCPT ); Wed, 28 Mar 2007 07:33:13 -0400 From: Michael Buesch To: Johannes Berg Subject: Re: rx racing against removing interfaces? Date: Wed, 28 Mar 2007 13:32:40 +0200 Cc: Jiri Benc , linux-wireless References: <1174950322.25887.59.camel@johannes.berg> <200703281259.55764.mb@bu3sch.de> <1175080054.5151.50.camel@johannes.berg> In-Reply-To: <1175080054.5151.50.camel@johannes.berg> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-15" Message-Id: <200703281332.40879.mb@bu3sch.de> Sender: linux-wireless-owner@vger.kernel.org List-ID: On Wednesday 28 March 2007 13:07, Johannes Berg wrote: > On Wed, 2007-03-28 at 12:59 +0200, Michael Buesch wrote: > > > Oh, you mean virtual interfaces.. yeah, there's probably a race. I don't > > know the locking there. > > Yes, I mean virtual interfaces, when the thing is in progress while we > iterate through. > > > But my comment about the non-irqsafe-variants-should-be-hidden still applies :) > > I'll reread it ;) > > But I'm not sure I fully agree. If the driver calls __ieee80211_rx in > softirq context that should be good without requiring to reschedule in > mac80211. I'm not sure there is an easy and non-racy way to drop the driver IRQ lock while calling __ieee80211_rx synchronously. -- Greetings Michael.