Re: eepro100 troubles

From: Andrey Savochkin (saw@saw.sw.com.sg)
Date: Thu Jun 15 2000 - 06:51:04 EST


Hello Brian,

On Thu, Jun 15, 2000 at 04:14:47AM -0400, Brian wrote:
> Day 3 and still going.

Fine!

> I guess it's worth noting that I have a PCI card while Scott's (I gather) is
> integrated into the motherboard.
>
> His comes up as:
> eth0: Intel PCI EtherExpress Pro100 82557, 00:90:27:8D:36:FA, IRQ 21.
>   Board assembly 000000-000, Physical connectors present: RJ45
>   Primary interface chip i82555 PHY #1.

My own card is integrated into mainboard, too, and produce output same as
Scott's. Works fine, BTW :-)

> Mine comes up as:
> eth1: OEM i82557/i82558 10/100 Ethernet, 00:80:29:66:1A:82, IRQ 11.
> Board assembly 650990-001, Physical connectors present: RJ45
> Primary interface chip DP83840 PHY #1.
> DP83840 specific setup, setting register 23 to 8462.
>
>
> I do, however, get a few huge streaks of stuff like this:
> eth1: can't fill rx buffer (force 0)!
> eth1: Tx ring dump, Tx queue 105081587 / 105081587:
[snip]
> I'm not sure what's useful in that. This morning's incident produced about

It's useful if the driver doesn't recover after that.
In your case it did, and more than once :-)

> 300k worth of it in the span of 4 seconds. Force is 1 or 0 with no apparent
> pattern and the l, *, and = usually advance one or two places between dumps.

It's how it should look like.
 
> I don't have any traffic reports fine-grained enough to say if it stopped
> responding during the incident, but it doesn't seem to have any long term
> effect.

It must drop input packets during the incident.
The absence of long term effect is what I was concerned about.

So, thank you for the positive feedback.

I suspect that hangs experienced by Scott may be caused by something like
killing of init because of OOM. But I haven't tried the described experiment
of two interfaces with different speed and large device queue.
When I do it, I'll report the results.

> This is a 256MB Celeron/450 that gets pretty hammered (probably around 7-8
> Mbit at that time of day).
>
> On Mon, 12 Jun 2000, Brian wrote:
> > Meanwhile, I'm going to try 1.20.2.10.
> > On my system, it took the older drivers 1.5 days to hang, so I'll let you
> > know Tuesday.

Best regards
                                        Andrey V.
                                        Savochkin
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.rutgers.edu



This archive was generated by hypermail 2b29 : Thu Jun 15 2000 - 21:00:41 EST