From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bjorn Helgaas Subject: Re: bnx2x machine check in bnx2x_ack_int() Date: Mon, 15 Dec 2008 11:04:00 -0700 Message-ID: <200812151104.00724.bjorn.helgaas@hp.com> References: <200812091204.35296.bjorn.helgaas@hp.com> <200812150921.08668.bjorn.helgaas@hp.com> <1229361480.22295.22.camel@lb-tlvb-eliezer> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-15" Content-Transfer-Encoding: 7bit Cc: "netdev@vger.kernel.org" To: "Eilon Greenstein" Return-path: Received: from g5t0006.atlanta.hp.com ([15.192.0.43]:21457 "EHLO g5t0006.atlanta.hp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754359AbYLOSEE (ORCPT ); Mon, 15 Dec 2008 13:04:04 -0500 In-Reply-To: <1229361480.22295.22.camel@lb-tlvb-eliezer> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-ID: On Monday 15 December 2008 10:18:00 am Eilon Greenstein wrote: > On Mon, 2008-12-15 at 08:21 -0800, Bjorn Helgaas wrote: > > > > > I'm using bnx2x 1.45.23 from RHEL5.3s4 on a prototype ia64 platform, > > > > and I see intermittent machine checks at bnx2x_ack_int+176, which is > > > > just after __ia64_readl() returns. > > > I see that you are using two 57710 chips. What kind of board are you > > > using (which PHY)? > > > > This is a blade with separate PHYs. In this case, there is no PHY > > at all because the module that would contain the PHY is not installed. > > > > > > This is a proto with incomplete firmware, and the driver correctly > > > > complains about that, but it seems like there's still a hole where > > > > things blow up. > > > If you set the right FW on the board, do you still see the problem? I'm > > > asking to determine if this is a problem specific to the "no FW" case or > > > general issue with this configuration > > > > I'm trying to get FW installed, but haven't had any luck yet. > > Hmmm... So it is possible that this is HW related (since we are talking > about preliminary board with missing component and missing FW). I will > ask for some field support - someone will contact you shortly. Yep, it's definitely possible that this is HW related. The proto has 57710, but the product will ultimately use the 57711, so I don't really even care about testing the 57710. After upgrading some pieces on my boards, I can't reproduce the MCA either. So I guess we can just write it off for now. The only reason I pointed out the issue at all is because it really hampered testing of standard distro bits on the rest of the box. Sorry for the distraction. Bjorn