Date: Wed, 8 Dec 2010 12:21:36 -0500
From: Neil Horman <nhorman@redhat.com>
To: Vivek Goyal <vgoyal@redhat.com>
Cc: Neil Horman <nhorman@tuxdriver.com>, linux-pci@vger.kernel.org,
        kexec@lists.infradead.org, linux-kernel@vger.kernel.org,
        Jesse Barnes <jbarnes@virtuousgeek.org>
Subject: Re: [PATCH] Update MCP55 quirk to not affect non HyperTransport
 variants
Message-ID: <20101208172136.GK11454@hmsreliant.think-freely.org>
References: <1291819668-15624-1-git-send-email-nhorman@tuxdriver.com>
 <20101208164423.GH31703@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20101208164423.GH31703@redhat.com>
User-Agent: Mutt/1.5.20 (2009-08-17)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 3060
Lines: 72

On Wed, Dec 08, 2010 at 11:44:23AM -0500, Vivek Goyal wrote:
> On Wed, Dec 08, 2010 at 09:47:48AM -0500, Neil Horman wrote:
> > I wrote this quirk awhile ago to properly setup MCP55 chips on hypertransport
> > busses so that interrupts reached whatever cpu happend to boot the kdump kernel.
> > while that works well, it was recently shown to me that a a non-hypertransport
> > variant of the MCP55 exists, and on those system the register that this quirk
> > manipulates causes hangs if you write to it.  Since the quirk was only meant to
> > handle errors found on MCP55 chips that have a HT interface, this patch adds a
> > filter to make sure the chip is an HT capable before making the needed register
> > adjustment.  This lets the broken MCP55s work with kdump while not breaking the
> > non-HT variants.
> > 
> 
> So Neil, with non hypertransport MCP55s, interrupts are delivered to
> all the cpus and seond kernel still boots?
> 
I cannot guarantee that.  All that I can guarantee is that, on MCP55 chips
without hypertransport capability, the register which is used to configure
interrupt delivery is either non-existant, or different in such a way that
writing to it results in a hang on the boot of the initial kernel.  So this
quirk has no effect on kdumps behavior regardless of the interrupt delivery
behavior on such variants of the chip.  I would presume however, given that this
register is so different that interrupts would be delivered to all cpus, as one
would expect.  Mathieu, since you have the affected hardware, could you attempt
to preform a kexec operation and tell us for certain?  Thanks!
Neil

> Acked-by: Vivek Goyal <vgoyal@redhat.com>
> 
> Thanks
> Vivek
> 
> > Resolves https://bugzilla.kernel.org/show_bug.cgi?id=23952
> > 
> > Tested successfully by the reporter and myself.
> > 
> > Reported-by: Mathieu B?rard <mathieu@mberard.eu>
> > CC: linux-pci@vger.kernel.org
> > CC: kexec@lists.infradead.org
> > CC: Vivek Goyal <vgoyal@redhat.com>
> > CC: Jesse Barnes <jbarnes@virtuousgeek.org>
> > Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
> > ---
> >  drivers/pci/quirks.c |    3 +++
> >  1 files changed, 3 insertions(+), 0 deletions(-)
> > 
> > diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> > index 6f9350c..313c0bd 100644
> > --- a/drivers/pci/quirks.c
> > +++ b/drivers/pci/quirks.c
> > @@ -2329,6 +2329,9 @@ static void __devinit nvbridge_check_legacy_irq_routing(struct pci_dev *dev)
> >  {
> >  	u32 cfg;
> >  
> > +	if (!pci_find_capability(dev, PCI_CAP_ID_HT))
> > +		return;
> > +
> >  	pci_read_config_dword(dev, 0x74, &cfg);
> >  
> >  	if (cfg & ((1 << 2) | (1 << 15))) {
> > -- 
> > 1.7.2.3
> 
> _______________________________________________
> kexec mailing list
> kexec@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/kexec
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/