From mboxrd@z Thu Jan 1 00:00:00 1970 From: Francois Romieu Subject: Re: [RFC] r8169 : why SG / TX checksum are default disabled Date: Wed, 18 Jul 2012 01:40:37 +0200 Message-ID: <20120717234037.GA26972@electric-eye.fr.zoreil.com> References: <1342564781.2626.1264.camel@edumazet-glaptop> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: netdev@vger.kernel.org, Hayes Wang To: Eric Dumazet Return-path: Received: from violet.fr.zoreil.com ([92.243.8.30]:50077 "EHLO violet.fr.zoreil.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753420Ab2GQXxR (ORCPT ); Tue, 17 Jul 2012 19:53:17 -0400 Content-Disposition: inline In-Reply-To: <1342564781.2626.1264.camel@edumazet-glaptop> Sender: netdev-owner@vger.kernel.org List-ID: Eric Dumazet : [...] > I was wondering why Scatter Gather (NETIF_F_SG) and TX checksum > (NETIF_F_IP_CSUM) were disabled by default on r8169. It was not unconditionally stable. For a number of reasons: - incompatibility with jumbo frames. rtl_chip_infos now allows to take care of it - lack of support for chipset dependent Tx descriptor rtl_tx_desc_info handles it. There are yet a few unused ipv6 bits. No idea how useable they are. - some false alarms due to different issues and probably some others I could hunt for in the archives. > Couldnt we enable them by default, maybe for a whitelist of "good" > nics ? Sure. Hayes, should we not add into the kernel driver something similar to the rtl8168_start_xmit::skb_checksum_help stuff in Realtek's 8168 driver ? There seems to be a bug for (skb->len < 60 && RTL_GIGA_MAC_VER_34. > (I found that activating them with ethtool automatically enables GSO, > and performance with GSO is not good) It's still an improvement though, isn't it ? -- Ueimor