From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH net-next 0/2] net: Fix "hw csum failure" message flood for ppp tunnel Date: Mon, 20 Apr 2015 22:01:44 -0400 (EDT) Message-ID: <20150420.220144.784138484983756987.davem@davemloft.net> References: <1429564205-1661247-1-git-send-email-tom@herbertland.com> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org To: tom@herbertland.com Return-path: Received: from shards.monkeyblade.net ([149.20.54.216]:48761 "EHLO shards.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752021AbbDUCBs (ORCPT ); Mon, 20 Apr 2015 22:01:48 -0400 In-Reply-To: <1429564205-1661247-1-git-send-email-tom@herbertland.com> Sender: netdev-owner@vger.kernel.org List-ID: From: Tom Herbert Date: Mon, 20 Apr 2015 14:10:03 -0700 > This patch set addresses bug "Bug 95171 - "hw csum failure" message > flood for ppp tunnel since upgrade to 3.16". The problem is that pppoe > is being used over UDP with UDP checksusm enabled. On receive > checksum conversion turns checksum-unnecessary in checksum- > complete. The PPP receive functions do no properly pull > the checksum over its headers, so that when an encapsulated > checksums is considered the checksum-complete value is incorrect. > > This patch adds skb_checksum_complete_unset which can be called > in the receive path in lieu of pulling checksum complete in > layer. This is useful when the packet is being modified (e.g. > decompressed) and the checksum-complete value is no longer > relevant. > > In the ppp_receive_frame we call skb_checksum_complete_unset to toss > out checksum-complete. This should eliminate the reported messages. > Alternatively, we could add skb_postpull_rcsum and probably > special case handling for VJ compression if maintaining the > checksum-complete is needed (not clear to me this is worth the > effort). > > I haven't tested this since setting up the failure scenario doesn't > seem trivial to configure. I'm preemptively applying this, but it's really important for folks to give this some good testing. Thanks Tom.