From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jesse Barnes Subject: Re: Bad sky2 interrupt handling? Date: Thu, 16 Jul 2009 16:57:41 -0700 Message-ID: <20090716165741.2142a679@jbarnes-g45> References: <20090716162552.7f7b3031@jbarnes-g45> <20090716163727.666db0cc@nehalam> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org To: Stephen Hemminger Return-path: Received: from outbound-mail-148.bluehost.com ([67.222.38.38]:55451 "HELO outbound-mail-148.bluehost.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S933628AbZGPX5o (ORCPT ); Thu, 16 Jul 2009 19:57:44 -0400 In-Reply-To: <20090716163727.666db0cc@nehalam> Sender: netdev-owner@vger.kernel.org List-ID: On Thu, 16 Jul 2009 16:37:27 -0700 Stephen Hemminger wrote: > On Thu, 16 Jul 2009 16:25:52 -0700 > Jesse Barnes wrote: > > > I was debugging what I thought was a graphics driver problem today > > (some page flipping code I'm working on was only flipping at about > > 10Hz rather than ~60Hz) but it appears it's actually due to a bad > > interaction with the Marvell 88E8036 controller on this machine. > > > > MSI fails on this box (haven't looked too much at why yet), so the > > i915 driver shares an interrupt with USB, firewire, sdhci and the > > NIC. > > > > Is sky2 the wrong driver to be using here? Or is there some way of > > reducing its overhead when (presumably) it doesn't have anything to > > do? > > No, that is the right driver. If MSI fails, does the driver still > think it can use MSI? No, I think it's falling back correctly... lemme get the log. ... sky2 driver version 1.23 sky2 0000:01:00.0: PCI INT A -> Link[LNKA] -> GSI 11 (level, low) -> IRQ 11 sky2 0000:01:00.0: setting latency timer to 64 sky2 0000:01:00.0: Yukon-2 FE chip revision 1 sky2 0000:01:00.0: irq 18 for MSI/MSI-X sky2 0000:01:00.0: No interrupt generated using MSI, switching to INTx mode. sky2 eth0: addr 00:a0:d1:bb:88:52 ... sky2 skge: enabling interface sky2 skge: Link is up at 100 Mbps, full duplex, flow control both at least it looks like it. -- Jesse Barnes, Intel Open Source Technology Center