From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: unexpected GRO/veth behavior Date: Tue, 11 Sep 2018 08:42:42 -0700 Message-ID: References: <4106d3f7eee7f0186fcfdd0331cdafeecd3240c0.camel@redhat.com> <3ab5755e75cfe2bfaeeb133b53512da43cfdb8b0.camel@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Cc: Toshiaki Makita To: Paolo Abeni , Eric Dumazet , netdev@vger.kernel.org Return-path: Received: from mail-wr1-f48.google.com ([209.85.221.48]:46718 "EHLO mail-wr1-f48.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726622AbeIKUmi (ORCPT ); Tue, 11 Sep 2018 16:42:38 -0400 Received: by mail-wr1-f48.google.com with SMTP id a108-v6so26533283wrc.13 for ; Tue, 11 Sep 2018 08:42:45 -0700 (PDT) In-Reply-To: <3ab5755e75cfe2bfaeeb133b53512da43cfdb8b0.camel@redhat.com> Content-Language: en-US Sender: netdev-owner@vger.kernel.org List-ID: On 09/11/2018 04:44 AM, Paolo Abeni wrote: > On Tue, 2018-09-11 at 03:27 -0700, Eric Dumazet wrote: >> On 09/10/2018 11:54 PM, Paolo Abeni wrote: >>> I *think* we can hit a similar condition with a tun device in IFF_NAPI >>> mode. >> >> Why ? >> >> tun_get_user() does not attach skb to a socket, that would be quite useless since skb >> > > I think we can hit this code path: > > tun_get_user() -> tun_alloc_skb() -> sock_alloc_send_pskb() -> > skb_set_owner_w() > > Than later napi and GRO. > No we can not. Please look again.