Subject: Re: de(4) and transmit underflow
To: None <email@example.com>
From: Manuel Bouyer <firstname.lastname@example.org>
Date: 04/20/1998 20:51:07
On Apr 20, Al Urbaitis wrote
> > Your system seems to be too slow to fill the FIFOs fast enough.
> > Replacing my 486DX133 with a Pentium 100 solved the problem for me.
> I'm running a P133 ....
> The task that seems to get this going is a large copy task. The box
> is an nfs server and if you copy a large file to one of its
> filesystems (200-300Mbytes).
> looking at top there is nothing inparticular going on (2% to 3%)
> cpu utilization and 4 nfsd's mostly asleep.
> Why would the cpu be a reasonable guess...
Hum, wait a minute ... Why would the CPU speed get involed here ?
If I understand it correctly, the FIFOs are filled with bus-master DMA
transfers, aren't they ? If this is correct, this means that the DMA operation
was not quick enougth. So there is something else that blocks DMA operation,
and I don't think this can be the CPU ...
I've just seen these kind of message on my PPro-200 with a 100Mb board,
doing a large file copy. This host also have a aha2940 with an UW drive.
Ethernet DMA may have been slowed down because of aha2940 tranfers.
Manuel Bouyer, LIP6, Universite Paris VI. Manuel.Bouyer@lip6.fr