Re: Firewire interface -> dropouts, IRQ?

Dave Platt wrote:
In article <fp3vd4$b6b$1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
Boris Lau <boris.lau@xxxxxxxxxxxxxxxxxx> wrote:

Is there a way to change the IRQ on the mobo?

That depends on the chipset and the BIOS.

In the case of your P5K board, the information on the PCI interrupt
mappings is on page 2-20 of the manual.

Here's how it appears to look to me, based on the manual. Each of the
PCI slots (1, 2, and 3) is assigned to a single PCI interrupt channel
(A, B, and C).

The PCI interrupt channels are then mapped to IRQ numbers.

All of the PCI interrupt channels are shared - interrupt A is shared
with the SATA controller, both PCI-E slots (I assume you have your
video board in one of these?), and two of the USB controllers. None
of the three slots has a dedicated interrupt.

It looks to me as if interrupt B is the least-busy of the 3 - it seems
to be shared only with the Ethernet controller, and not with USB or
disks.

If not, would it help if I buy a new firewire adaptor? Does it make
sense to get a PCI-E firewire thingy, since that bus does not use IRQ?

PCI-E does appear to use interrupts. Both PCIE slots on this board
are on Interrupt A.

If IRQ contention is in fact your problem, putting the Firewire board
into PCI slot 2 (on PCI interrupt B) would probably reduce its
contention.

I'm sceptical, though, as to whether low-level IRQ contention is at
the root of your problem. OS or driver issues (e.g. insufficient
buffer sizes, drivers that disable interrupts for long periods of
time, etc.) seem more of a problem.


The interrupts on PCI Express, are message based.
http://en.wikipedia.org/wiki/PCIe

"PCIe sends all control messages, including interrupts, over
the same links used for data."

There is no interrupt pin evident, at least in this pinout picture.
(I cannot afford to buy a copy of the spec, so this picture is the
best info I have.)

http://web.archive.org/web/20070322211528/http://images.tomshardware.com/2004/11/22/sli_is_coming/pcie-slot-big.gif

I don't expect that changes anything, from a practical perspective.

I think I'd look elsewhere for an explanation as to what's up with
Firewire. But to do that, you'd need some means to monitor the
health of the hardware and driver software, to see if some
significant event is happening, involving a timeout or
similar exception condition. Windows isn't exactly known
for making stuff like that easy to do.

Paul
.

Relevant Pages

  • Re: DUAL Lan Adapter:No Network Connection
    ... You may want to check the interrupt line, if the DLL is loaded and the ... NDIS is trying to bind the adapter with an IP address, ... This could be a PCI interrupt routing problem on the ... > IPNotifyClientsIPEvent: PendingIPGetIPEventRequest 5a0a0 ...
    (microsoft.public.windowsce.platbuilder)
  • RE: PCI Interrupt Latency
    ... Conversation: PCI Interrupt Latency ... Mark the Irp pending via IoMarkIrpPending, ...
    (microsoft.public.development.device.drivers)
  • Re: Why PCI9054 fails to assert pci interrupt when local interrupt input is pulled down? Any advice
    ... I meet a problem of interrupt. ... I wanna assert a PCI interrupt using ... when I set INTCSR as 0X0f000800 (PCI interrupt is not ... I dont know its the reason of WinDriver diagnostic program ...
    (comp.arch.fpga)
  • Re: PCI Interrupts
    ... Maxim Shatskih, Windows DDK MVP ... interrupt, making all the PCI interrupt handling transparent. ... model, where you send an IOCTL to the driver, and it is pended until the ... The application has to know when this PCI interrupt occurs. ...
    (microsoft.public.development.device.drivers)
  • Re: New MSI support in sata_sil24 still broken in 2.6.33-rc3
    ... ncq 524288 out ... register bit 30: "MSI Acknowledge. ... acknowledges a Message Signaled Interrupt and permits generation of another ... And it's also not enabled on the bus which has the Silicon Image controller: ...
    (Linux-Kernel)