[eepro100] Re: eth0: can't fill rx buffer (force 1)!

Andrey Savochkin saw@saw.sw.com.sg
Mon, 30 Oct 2000 11:10:54 +0800


Hello,

On Fri, Oct 27, 2000 at 02:05:46PM +0800, NortonNg wrote:
> 
>    I am using eepro100 for Intel 82559 chip NIC (but the driver said that it
> was 82557 ), Redhat 7.0 Kernel 2.2.17 (patched from 2.2.16). After I do some
> modification in dev.c for my packet dispatcher, when i run the nttcp 
> (performance benchmark) several times (about transfered 400mega bits), 
> both the gateway system will crash !!  my configuration is below..
>   
[snip]
> eth0: can't fill rx buffer (force 1)!
> eth0: can't fill rx buffer (force 1)!
> eth0: can't fill rx buffer (force 1)!
> eth0: can't fill rx buffer (force 1)!
[snip]
> VM: do_try_to_free_pages failed for kswapd
> VM: do_try_to_free_pages failed for kswapd
> VM: do_try_to_free_pages failed for kswapd
> VM: do_try_to_free_pages failed for kswapd

It's a typical out-of-memory situation, where some part of the kernel
consumes more memory than is available or can be freed.
So, different kernel subsystems start to fail.

If it starts to occur after your changes, I guess that the out-of-memory
happens because you insert packets in the queue without any limits on the
queue length or source throttling.

Best regards
					Andrey V.
					Savochkin