From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH net-next v5 0/5] Introduce NETIF_F_GRO_HW Date: Tue, 19 Dec 2017 14:24:31 -0500 (EST) Message-ID: <20171219.142431.1940142843749180845.davem@davemloft.net> References: <1513411784-17653-1-git-send-email-michael.chan@broadcom.com> <20171219.105024.1047087594002892419.davem@davemloft.net> <20171219190426.GF6122@localhost.localdomain> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: michael.chan@broadcom.com, netdev@vger.kernel.org, andrew.gospodarek@broadcom.com To: marcelo.leitner@gmail.com Return-path: Received: from shards.monkeyblade.net ([184.105.139.130]:55472 "EHLO shards.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752922AbdLSTYd (ORCPT ); Tue, 19 Dec 2017 14:24:33 -0500 In-Reply-To: <20171219190426.GF6122@localhost.localdomain> Sender: netdev-owner@vger.kernel.org List-ID: From: Marcelo Ricardo Leitner Date: Tue, 19 Dec 2017 17:04:27 -0200 > I understand that for all that matters, the hardware operations > involved on GRO_HW are really for only 1 packet, so it would make > sense to count it as 1. OTOH, this bump may cause additional pressure > in other places as in fact we are allowing more packets in in a given > cycle. More data, but not more packets or (realistically) "work". The stack is going to parse only one SKB, one set of networking headers, do one route lookup, one socket demux, etc. In that view, counting the HW GRO packet as only one frame is appropriate.