> Until recently, I was running 1.3.98 and things were working fine.
> I decided that I was too far behind the times and upgraded to
> pre2.0.6. When I did that I started getting LOTS of UDP checksum
> errors (part of message log is below). The only thing that seems to have
> have changed is the driver for the smc-ultra cards (at least that affected
> me!). I played around for a while and put the smc-ultra.c from 1.3.98 into
> the 1.3.99 kernel, and all my problems were solved! Unfortunately, I'm not
> convinced that this is a permanant solution. It seems that the addition
> of PIO support is the culprit here. I must admit that I've really not got the
> background to be of much help in fixing the problem, but I'm willing to
> test things. For the record the system is:
Thanks for the report.
I need the following information:
What does your machine report at boot time?
Is the kernel using the new PIO code?
How much memory does the driver think you have?
How much memory is on your card (16K, 32K, or 64K)?
If you could run a test, could you check:
ping -l 40 -s 56
This will check if "single-page" (<256 byte) packets work.
ping -l 40 -s 500
This will check if "multi-page" (>256 byte) packets work.
If the latter test fails, the automatic wrap-around feature of the chip isn't
working as I expected, perhaps because of a packet buffer memory size the
driver is assuming is wrong.
> This is the tail end of the message log using the stock 1.3.99 kernel:
>
> May 21 13:57:29 bell-209 last message repeated 2 times
> May 21 13:58:29 bell-209 last message repeated 2 times
> May 21 13:58:59 bell-209 kernel: UDP: bad checksum. From 82B84005:520 to 82B840FF:520 ulen 492
Are all error packets larger than 256 (really >~240) bytes?
Donald Becker becker@cesdis.gsfc.nasa.gov
USRA-CESDIS, Center of Excellence in Space Data and Information Sciences.
Code 930.5, Goddard Space Flight Center, Greenbelt, MD. 20771
301-286-0882 http://cesdis.gsfc.nasa.gov/pub/people/becker/whoiam.html