RE: Rx overruns with e1000 card

From: Feldman, Scott (scott.feldman@intel.com)
Date: Wed Jul 30 2003 - 02:21:06 EST


> I am running a Pentium 4 machine 2.53 Ghz machine with 1Ghz
> RAM. Kernel is 2.4.18. I am running it in Polling mode. The
> problem is i am having some occasional packet loss due to
> receive overruns as shown by e1000 statistics at the packet
> rate above 10000 (64 bytes) ...

10kpps 64-byte packets is really low. You should be 20x (or more) at
1GbE. Something isn't right. Are you getting any other Rx errors? It
sounds like you're having trouble getting packets out of the Rx FIFO and
DMA'ed across the bus into host buffers. We need to know if the FIFO is
gummed up because we're receiving faster than we can DMA out of the
FIFO, we're not getting Rx buffers (unlikely), or the hardware
descriptor ring is full and we're being throttle by the polling rate of
NAPI (also unlikely at this rate). The Rx errors should tell us where
the bottleneck is. What's that other device doing?

-scott
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html



This archive was generated by hypermail 2b29 : Thu Jul 31 2003 - 22:00:02 EST