From mboxrd@z Thu Jan 1 00:00:00 1970 From: Denys Fedoryschenko Subject: Re: r8169, enabling TX checksumming breaks things? Date: Sat, 26 Sep 2009 02:18:54 +0300 Message-ID: <200909260218.54899.denys@visp.net.lb> References: <200909230915.27854.denys@visp.net.lb> <200909231724.23132.denys@visp.net.lb> <1253754752.13176.19.camel@obelisk.thedillows.org> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: romieu@fr.zoreil.com, netdev@vger.kernel.org To: David Dillow Return-path: Received: from hosting.visp.net.lb ([194.146.153.11]:37885 "EHLO hosting.visp.net.lb" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751945AbZIYXPH (ORCPT ); Fri, 25 Sep 2009 19:15:07 -0400 In-Reply-To: <1253754752.13176.19.camel@obelisk.thedillows.org> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-ID: > > This is the patch I have against HEAD, which Works For Me (TM). If I > recall correctly the list in rtl8169_set_tso_csum() is from combing the > vendor drivers, but I'm not 100% on that. Too much time has passed. > > I added the MAC version to the driver banner, which should make it > easier to figure out which MAC you have, and to change that function if > it does not work for you out of the box. It seems that some chips don't > work with this at all, based on testing by Michael Riepe (XID 3c4000c0). > > As should be obvious, this isn't ready to go anywhere near upstream yet > I think. > Applied to 2.6.31.1 (i cannot run latest git, it is loaded proxy, too much risk) (one reject only, fixed by hand). [ 0.793643] r8169 Gigabit Ethernet driver 2.3LK-NAPI-TSO loaded [ 0.793734] r8169 0000:03:00.0: PCI INT A -> GSI 17 (level, low) -> IRQ 17 [ 0.793842] r8169 0000:03:00.0: setting latency timer to 64 [ 0.793898] r8169 0000:03:00.0: irq 27 for MSI/MSI-X [ 0.794188] eth0: RTL8168d/8111d (25) at 0xf8420000, 00:1c:c0:c2:4d:df, XID 281000c0 IRQ 27 [ 2.110317] r8169: eth0: link down [ 2.110396] r8169: eth0: link down [ 3.872590] r8169: eth0: link up Working fine now, now when i enable TX checksumming it doesn't fail. I will run for a while with tx checksumming, and then i will run the rest and will do some performance measurements (by perf tool) and i will try on a little bit other model of motherboard. Thanks a lot for your patch!