[mythtv-users] Network Device Failure.....

[Brian Wood]

On Jan 29, 2006, at 9:33 AM, John Nelson wrote:

> Mr. Wood,

I'm Brian, BTW :-)
>
> Well, I ran both lspci and the cat /proc/interrupts commands and the
> outputs are copied below.  I was thinking a USB ethernet adapter would
> be a bad idea.  As for PCMCIA, I have the ability to get an IDE  
> adapter
> that goes into a 5.25 slot on a case.  I have no idea about its specs,
> if Linux wil leven see it, etc.  Plus, this would be a very certain  
> kludge.
>
> The First cut/paste is the thernet type.  The second is the  
> interrupts.
>
> I do notice there are quit a few on the 10: 499501 row.  Being a  
> bit of
> a wet behind the ears Linux newbie, is the 10 or the 499501 the
> interrupt?  Or both?  The fact that there is one of my capture cards,
> two USB ports and my Ethernet on whatever that row, do you think that
> might be part of the problem?

Absolutely. "10" is the interrupt BTW. I believe the second number is  
the activity.
>
> As for whether my network device support is modular or compiled in, I
> will have to look into that.
>
> Running cat /proc/interrupts
> _____________________________________________________
>
> [root at mbackend mythtv]# cat /proc/interrupts
>            CPU0
>   0:    2675869          XT-PIC  timer
>   1:        113          XT-PIC  i8042
>   2:          0          XT-PIC  cascade
>   7:          1          XT-PIC  parport0
>   8:          1          XT-PIC  rtc
>   9:          0          XT-PIC  acpi
>  10:     499501          XT-PIC  ivtv0, uhci_hcd:usb2,  
> uhci_hcd:usb3, eth0
>  11:      11623          XT-PIC  libata, VIA8237, ivtv1, ivtv2,
> ehci_hcd:usb1, uhci_hcd:usb4, uhci_hcd:usb5
>  12:       4061          XT-PIC  i8042
>  14:       3755          XT-PIC  ide0
>  15:     100673          XT-PIC  ide1
> NMI:          0
> ERR:          0
> _____________________________________________________
>
>
As you pointed out, INTs 10 and 11 are certainly busy, it would not  
surprise me if the sharing of the ethernet interrupt is helping to  
cause your problems.

If you're not using USB for anything I would not load those drivers,  
same for the parallel port. I don't know if ivtv and eth0 can share  
INTs well, but getting rid of as much conflict as possible certainly  
couldn't hurt.

There's not much you can do with moving PCI cards around, which is  
the traditional method of trying to arrange INTs in a better way. I  
don't know if your mobo has much if any control over INT assignment  
but the best solution would be to get the eth0 on an INT by itself in  
some way, or at least sharing with something little-used. I have the  
very same Rhine net interface but it has its very own INT.

I've never run into an IDE ethernet device. I do have a SCSI net  
device I used to use with a very old Mac and it worked OK, but as you  
pointed out I do not know if Linux would support such a thing. I  
suspect a USB "solution" would make matters here worse.

So get rid of the USB drivers if you can, also the parport (or go to  
a polling driver for it).

If you need help doing this feel free to mail me privately, we can  
post a summary to the list if we get it resolved but let's try to not  
bore the list with kernel config issues.

Meanwhile, if anyone else has any ideas they are most welcome :-)