Re: [PATCH] firewire: fix some broken hardware NMI interrupt

Re: [PATCH] firewire: fix some broken hardware NMI interrupt

[JiSheng Zhang]

Hi,
after a lot of test today, the problem appears again. Now it also emit a similar
message even applied my patch. Tested on 2.6.25-4 and 2.6.26-rc3.
The via chip is VT6306. How can I dig into this problem more?

CU,
JiSheng

>From: Stefan Richter <stefanr@s5r6.in-berlin.de>
>Reply-To: 
>To: JiSheng Zhang <jszhang3@mail.ustc.edu.cn>
>Subject: Re: [PATCH] firewire: fix some broken hardware NMI interrupt
>Date:Thu, 22 May 2008 15:38:30 +0200
>
>JiSheng Zhang wrote:
> >> From: Stefan Richter <stefanr@s5r6.in-berlin.de>
> >> jszhang3@mail.ustc.edu.cn wrote:
> >>> Some broken pci-1394 card does not clear completely on-chip memory during
> >>> boot, it
> >>> may cause pci parity error, then NMI interrupt.
> >> 
> >> Which card?
> > hmm, one card using VIA chip.
> >> 
> >> Is the problem still present in linux-2.6.26-rc3?
> > It seems that the problem disappeared. Thanks
> 
> Great, thanks for testing.
> 
> I hope this is now permanently fixed.  Should it ever appear again, just
> tell us.
> -- 
> Stefan Richter
> -=====-==--- -=-= =-==-
>

Re: [PATCH] firewire: fix some broken hardware NMI interrupt

[Sean Bruno]

On Thu, 2008-05-22 at 20:07 +0800, JiSheng Zhang wrote:
> Hi,
> 
> 
> >From: Stefan Richter <stefanr@s5r6.in-berlin.de>
> >Reply-To: 
> >To: jszhang3@mail.ustc.edu.cn
> >Subject: Re: [PATCH] firewire: fix some broken hardware NMI interrupt
> >Date:Wed, 21 May 2008 17:28:25 +0200
> >
> >jszhang3@mail.ustc.edu.cn wrote:
> > > Some broken pci-1394 card does not clear completely on-chip memory during
> boot, it
> > > may cause pci parity error, then NMI interrupt.
> > 
> > Which card?
> hmm, one card using VIA chip.
> > 
> > Is the problem still present in linux-2.6.26-rc3?
> It seems that the problem disappeared. Thanks
> > 
> > Or if you don't want to test 2.6.26-rc yet, please test 2.6.25.4 with
> > the firewire patchkit from
> > http://me.in-berlin.de/~s5r6/linux1394/updates/.  There were changes to
> > fw-ohci which influence chip initialization.
> > -- 
> > Stefan Richter
> > -=====-==--- -=-= =-=-=
> > http://arcgraph.de/sr/
> >
> 
> 
I've got a VIA board causing the same error reported here.

If someone at Redhat has the time to roll a test kernel for this issue I
can test it in about 15 minutes.

Right now I'm running RHEL 5, 2.6.18-53.1.21.el5

Se

Re: [PATCH] firewire: fix some broken hardware NMI interrupt

[Stefan Richter]

JiSheng Zhang wrote:
>> From: Stefan Richter <stefanr@s5r6.in-berlin.de>
>> jszhang3@mail.ustc.edu.cn wrote:
>>> Some broken pci-1394 card does not clear completely on-chip memory during
>>> boot, it
>>> may cause pci parity error, then NMI interrupt.
>> 
>> Which card?
> hmm, one card using VIA chip.
>> 
>> Is the problem still present in linux-2.6.26-rc3?
> It seems that the problem disappeared. Thanks

Great, thanks for testing.

I hope this is now permanently fixed.  Should it ever appear again, just
tell us.
-- 
Stefan Richter
-=====-==--- -=-= =-==-
http://arcgraph.de/sr/

Re: [PATCH] firewire: fix some broken hardware NMI interrupt

[JiSheng Zhang]

Hi,


>From: Stefan Richter <stefanr@s5r6.in-berlin.de>
>Reply-To: 
>To: jszhang3@mail.ustc.edu.cn
>Subject: Re: [PATCH] firewire: fix some broken hardware NMI interrupt
>Date:Wed, 21 May 2008 17:28:25 +0200
>
>jszhang3@mail.ustc.edu.cn wrote:
> > Some broken pci-1394 card does not clear completely on-chip memory during
boot, it
> > may cause pci parity error, then NMI interrupt.
> 
> Which card?
hmm, one card using VIA chip.
> 
> Is the problem still present in linux-2.6.26-rc3?
It seems that the problem disappeared. Thanks
> 
> Or if you don't want to test 2.6.26-rc yet, please test 2.6.25.4 with
> the firewire patchkit from
> http://me.in-berlin.de/~s5r6/linux1394/updates/.  There were changes to
> fw-ohci which influence chip initialization.
> -- 
> Stefan Richter
> -=====-==--- -=-= =-=-=
> http://arcgraph.de/sr/
>

Re: [PATCH] firewire: fix some broken hardware NMI interrupt

[Stefan Richter]

Stefan Richter wrote:
> jszhang3@mail.ustc.edu.cn wrote:
>> Some broken pci-1394 card does not clear completely on-chip memory during boot, it
>> may cause pci parity error, then NMI interrupt.
> 
> Which card?
> 
> Is the problem still present in linux-2.6.26-rc3?
> 
> Or if you don't want to test 2.6.26-rc yet, please test 2.6.25.4 with
> the firewire patchkit from
> http://me.in-berlin.de/~s5r6/linux1394/updates/.  There were changes to
> fw-ohci which influence chip initialization.

PS:
And does the parity error happen with the ohci1394 driver too?
-- 
Stefan Richter
-=====-==--- -=-= =-==-
http://arcgraph.de/sr/

Re: [PATCH] firewire: fix some broken hardware NMI interrupt

[Stefan Richter]

jszhang3@mail.ustc.edu.cn wrote:
> Some broken pci-1394 card does not clear completely on-chip memory during boot, it
> may cause pci parity error, then NMI interrupt.

Which card?

Is the problem still present in linux-2.6.26-rc3?

Or if you don't want to test 2.6.26-rc yet, please test 2.6.25.4 with
the firewire patchkit from
http://me.in-berlin.de/~s5r6/linux1394/updates/.  There were changes to
fw-ohci which influence chip initialization.
-- 
Stefan Richter
-=====-==--- -=-= =-=-=
http://arcgraph.de/sr/

[PATCH] firewire: fix some broken hardware NMI interrupt

[jszhang3]

>From JiSheng Zhang <jszhang3@mail.ustc.edu.cn>

Some broken pci-1394 card does not clear completely on-chip memory during boot, it
may cause pci parity error, then NMI interrupt. It will emit kernel message such
as"Uhhuh. NMI received for unknown reason b1 on CPU 0. You have some hardware
problem, likely on the PCI bus. Dazed and confused, but trying to continue" etc.
If panic_on_unrecovered_nmi is set to 1, the system will panic. This patch will
fix it


--- linux-2.6.25.4/drivers/firewire/fw-ohci.c.org	2008-05-21 20:44:26.901066202
+0800
+++ linux-2.6.25.4/drivers/firewire/fw-ohci.c	2008-05-21 20:55:15.290371532 +0800
@@ -2051,6 +2051,7 @@
 	struct fw_ohci *ohci;
 	u32 bus_options, max_receive, link_speed;
 	u64 guid;
+	u16 reg;
 	int err;
 	size_t size;
 
@@ -2082,6 +2083,10 @@
 
 	pci_set_master(dev);
 	pci_write_config_dword(dev, OHCI1394_PCI_HCI_Control, 0);
+	/* some broken hardware */
+	pci_read_config_word(dev, PCI_COMMAND, &reg);
+	reg &= ~(PCI_COMMAND_SERR | PCI_COMMAND_PARITY);
+	pci_write_config_word(dev, PCI_COMMAND, reg);
 	pci_set_drvdata(dev, ohci);
 
 #if defined(CONFIG_PPC_PMAC) && defined(CONFIG_PPC32)