From owner-freebsd-current@FreeBSD.ORG Wed Jun 4 07:16:53 2008 Return-Path: Delivered-To: current@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 18E201065672; Wed, 4 Jun 2008 07:16:53 +0000 (UTC) (envelope-from imp@bsdimp.com) Received: from harmony.bsdimp.com (bsdimp.com [199.45.160.85]) by mx1.freebsd.org (Postfix) with ESMTP id CA3608FC4D; Wed, 4 Jun 2008 07:16:52 +0000 (UTC) (envelope-from imp@bsdimp.com) Received: from localhost (localhost [127.0.0.1]) by harmony.bsdimp.com (8.14.2/8.14.1) with ESMTP id m547DL1x087557; Wed, 4 Jun 2008 01:13:23 -0600 (MDT) (envelope-from imp@bsdimp.com) Date: Wed, 04 Jun 2008 01:14:45 -0600 (MDT) Message-Id: <20080604.011445.-957832680.imp@bsdimp.com> To: ianf@clue.co.za From: "M. Warner Losh" In-Reply-To: References: X-Mailer: Mew version 5.2 on Emacs 21.3 / Mule 5.0 (SAKAKI) Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: current@freebsd.org Subject: Re: xe(4) MPSAFE patch X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 04 Jun 2008 07:16:53 -0000 In message: Ian FREISLICH writes: : John Baldwin wrote: : > Cool, thanks. If you are still up for testing, try out : > http://www.FreeBSD.org/~jhb/patches/xe_intr.patch : : I'm always up for testing :) : : > It uses the bus_read API instead of bus_space and splits the interrupt : > handler up into sub-functions for tx/rx/MAC events. : : And this, unsurprisingly, also works. :) : : This card however must be a pile of rubbish. The interrupt usage : suggests it's just PIO and uses the CPU to bit-bang the line to : generate the ethernet timing and signals. 73% interrupt on a : PIII-500 to move data at 100Mbit/s. PC Card devices generally were PIO only. DMA was technically in the standard, but in reality it was never implemented. The lack of a standardized DMA interface in the host controller is mostly to blame, and FreeBSD doesn't support the few cards that did support DMA. I don't think it is big-banging the line to generate the ethernet timing, but there are a number of busy loops in the driver that keep it from peak efficiency. Warner