From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from arrakis.dune.hu ([78.24.191.176]:36409 "EHLO arrakis.dune.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753728Ab2LKRQ6 (ORCPT ); Tue, 11 Dec 2012 12:16:58 -0500 Message-ID: <50C76A88.1030008@openwrt.org> (sfid-20121211_181702_037099_009A67F3) Date: Tue, 11 Dec 2012 18:16:56 +0100 From: Gabor Juhos MIME-Version: 1.0 To: Christian Lamparter CC: "John W. Linville" , linux-wireless@vger.kernel.org, users@rt2x00.serialmonkey.com Subject: Re: [PATCH 3/3] p54: zero-out rx_status References: <1355234105-12013-1-git-send-email-juhosg@openwrt.org> <1355234105-12013-3-git-send-email-juhosg@openwrt.org> <201212111608.43293.chunkeey@googlemail.com> In-Reply-To: <201212111608.43293.chunkeey@googlemail.com> Content-Type: text/plain; charset=ISO-8859-15 Sender: linux-wireless-owner@vger.kernel.org List-ID: 2012.12.11. 16:08 keltezéssel, Christian Lamparter írta: > On Tuesday, December 11, 2012 02:55:05 PM Gabor Juhos wrote: >> In commit 'mac80211: support radiotap vendor namespace RX data' >> new fields were added to 'struct ieee80211_rx_status'. >> The ath5k driver does not initializes those fields and > ^^^^^ p54?! Of course. >> this can cause unexpected behaviour. The patch ensures >> that each field gets initialized with zeroes. > > Actually, when the skb is alloced/initialized by > __alloc_skb, the skb->cb is already zeroed Thank you for the explanation, I was not aware of this. > (which is where the ieee80211_rx_status will be stored). > > And while p54 recycles command response skbs, the > driver does not touch the skb->cb of 802.11 skbs, > until the frame is destined for ieee80211_rx_irqsave. Ok. > If this issue just popped up now, I suspect that > something else is silently corrupting our SKBs > [or can anybody see how rt2x00 hit this issue?] It is popped up now because we have switched to compat-wireless-2012-12-06 in OpenWrt a few days ago. Previously we were using compat-wireless-2012-09-07. -Gabor