On Sat, 2007-10-06 at 12:48 -0600, Robert Han**** wrote:
> > On Oct 4, 2007, at 11:52 AM, Scott wrote:
> >
> >> I'm having what I think is a PCI bus problem.
> >>
> >> I have a ASUS P5B Intel 965 motherboard and a DVICO Fusion HDTV5 RT
> >> adapter on the PCI bus. When this adapter is recording (anything) I see
> >> Oct 2 21:59:12 htpc cx88[0]: irq mpeg [0x80000] pci_abort*
> >> Oct 2 21:59:12 htpc cx88[0]/2-mpeg: general errors: 0x00080000
> >> Oct 2 21:59:20 htpc cx88[0]: irq mpeg [0x80000] pci_abort*
> >> Oct 2 21:59:20 htpc cx88[0]/2-mpeg: general errors: 0x00080000
> >> Oct 2 21:59:32 htpc cx88[0]: irq mpeg [0x80000] pci_abort*
> >> Oct 2 21:59:32 htpc cx88[0]/2-mpeg: general errors: 0x0008000
> >
> PCI aborts won't have anything to do with interrupts. It's a PCI DMA
> transfer that either the initiator (here, the TV card) or the target
> (the chipset host bridge) decided to puke on for some reason. Possibly
> parity errors?

I assumed it was an interrupt issue based on the 'irq mepg' error, not a
DMA issue. Any suggestions on how to log more information about PCI DMA
events on a system wide basis?

I've seen this since 2.6.18.

--
Scott

-
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/

Scott wrote:
> On Sat, 2007-10-06 at 12:48 -0600, Robert Han**** wrote:
>>>> Oct 2 21:59:32 htpc cx88[0]: irq mpeg [0x80000] pci_abort*
>>>> Oct 2 21:59:32 htpc cx88[0]/2-mpeg: general errors: 0x0008000

> I assumed it was an interrupt issue based on the 'irq mepg' error, not a
> DMA issue.

It's a DMA issue. The chip raised an irq to signal the error condition
to the driver, thats why it is printed by the irq handler. Happened
while streamimg mpeg data.

Havn't looked at the code for years now, but IIRC the 0x800000 is the
raw error code from some status register, pci_abort is the error bit in
clear text, meaning some PCI DMA transfer was aborted. No idea why
though ...

cheers,
Gerd
-
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/

On Oct 8, 2007, at 3:44 AM, Gerd Hoffmann wrote:

> Scott wrote:
>> On Sat, 2007-10-06 at 12:48 -0600, Robert Han**** wrote:
>>>>> Oct 2 21:59:32 htpc cx88[0]: irq mpeg [0x80000] pci_abort*
>>>>> Oct 2 21:59:32 htpc cx88[0]/2-mpeg: general errors: 0x0008000
>
>> I assumed it was an interrupt issue based on the 'irq mepg' error,
>> not a
>> DMA issue.
>
> It's a DMA issue. The chip raised an irq to signal the error
> condition
> to the driver, thats why it is printed by the irq handler. Happened
> while streamimg mpeg data.
>
> Havn't looked at the code for years now, but IIRC the 0x800000 is the
> raw error code from some status register, pci_abort is the error
> bit in
> clear text, meaning some PCI DMA transfer was aborted. No idea why
> though ...

I went through and did the basics. Disabled SMP (it's a core 2 duo
cpu), tried with noapic, nolapic settings, tried with pci=irqroute,
and tried with acpi=off. Poking through the module code I noticd in
cx88-mpeg.c the irq_debug option. I think this is the debug switch I
want so I'll try to gather more info this week. At the very least it
looks like it will help me walk through the code to get a better idea
of what's going on.

--
Scott

-
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/