From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeff Garzik Subject: [PATCH 0/6] forcedeth interrupt and task overhaul, v2 Date: Wed, 17 Oct 2007 01:52:51 -0400 Message-ID: <20071017055251.GA19972@havoc.gtf.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Manfred Spraul , Ingo Molnar , Ayaz Abdulla To: netdev@vger.kernel.org, LKML Return-path: Received: from havoc.gtf.org ([69.61.125.42]:33823 "EHLO havoc.gtf.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750844AbXJQFwx (ORCPT ); Wed, 17 Oct 2007 01:52:53 -0400 Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org These six changes can be found in the 'fe' branch of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/netdev-2.6.git [netdrvr] forcedeth: internal simplifications; changelog removal [netdrvr] forcedeth: timer overhaul [netdrvr] forcedeth: eliminate some duplicate irq handling code [netdrvr] forcedeth: unconditionally enable NAPI [netdrvr] forcedeth: use NAPI for TX completion [netdrvr] interrupt handling overhaul I rewrote the previous forcedeth patchset, even though the direction and methods are quite similar to the last patchset. This hammers out bugs, problems, inconsistencies, mainly with interrupt handling and locking. This moves the driver towards removing the disable_irq() usage, but does not actually take that step (yet). A few changes just went upstream, too: Ingo Molnar (1): forcedeth: fix NAPI rx poll function Jeff Garzik (2): [netdrvr] forcedeth: improved probe info; dev_printk() cleanups [netdrvr] forcedeth: remove in-driver copy of net_device_stats This patchset is diff'd against the latest linux-2.6.git as of an hour ago or so. It applies on top of the three changes listed immediately above this paragraph. At this point, it has been tested on nVidia Corporation CK804 Ethernet Controller (rev a3) but I'm definitely interested in other testing, particularly performance numbers (wire speed? or below?), CPU usage and memory usage info. Most particularly CPU usage info. And maybe interrupt count with/without this patchset, too. If you send feedback, 'dmesg' and 'lspci' output is helpful. The current forcedeth driver in upstream (0.61) prints out useful diagnostic information about your chip at bootup.