From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: [PATCH 17/20][BNX2]: Enhance the heartbeat. Date: Wed, 2 May 2007 09:40:13 +0100 Message-ID: <20070502084013.GB24804@infradead.org> References: <1178068656.4820.53.camel@dell> <46383B00.4080605@garzik.org> <20070502.002426.38687430.davem@davemloft.net> <4638400D.5010306@garzik.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: David Miller , mchan@broadcom.com, netdev@vger.kernel.org To: Jeff Garzik Return-path: Received: from pentafluge.infradead.org ([213.146.154.40]:35568 "EHLO pentafluge.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753654AbXEBIkd (ORCPT ); Wed, 2 May 2007 04:40:33 -0400 Content-Disposition: inline In-Reply-To: <4638400D.5010306@garzik.org> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Wed, May 02, 2007 at 03:38:53AM -0400, Jeff Garzik wrote: > David Miller wrote: > >From: Jeff Garzik > >Date: Wed, 02 May 2007 03:17:20 -0400 > > > >>Michael Chan wrote: > >>>[BNX2]: Enhance the heartbeat. > >>> > >>>In addition to the periodic heartbeat, we're adding a heartbeat > >>>request interrupt when the heartbeat is late. This is useful > >>>especially in -rt kernels where the timer frequently runs late. > >>> > >>>Signed-off-by: Michael Chan > >>Should we really be adding code for such a special situation to upstream > >>code? > >> > >>I lean towards "no", but defer to your and DaveM's judgement here. > > > >My understanding of this situation is that if the timer is delayed a > >lot, which can happen with the -rt kernel, we don't send the heartbeat > >ping to the chip within the required margin. > > > >If the margin is not met, the ASF firmware takes this as a signal that > >the host system is down, and does things like reset the network card > >and other things we don't want it to do. > > Thanks for the explanation. > > My main concern was > > * adding code to use a kernel facility > * then, adding code to handle when that kernel facility doesn't work > > and also > > * adding code for a situation that never occurs in the upstream kernel. Yeah, we shouldn't add this. If the rt kernel can't deliver low latency real time guarantees to the hardware it's pretty buggy :)