Re: SiS900: NULL pointer encountered in Rx ring, skipping

[Jeff Garzik <jgarzik@pobox.com>]

Dominik Karall wrote:
> On Tuesday 27 July 2004 17:26, Jeff Garzik wrote:
> 
>>Dominik Karall wrote:
>>
>>>After a few hours my network doesn't work on my laptop. There appear a
>>>lot of those messages:
>>>
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>eth0: NULL pointer encountered in Rx ring, skipping
>>>
>>>It works again after restarting network. I'm using 2.6.8-rc2 now. It was
>>>the same problem in 2.6.7, but I didn't test it with earlier kernels.
>>
>>A NULL appears when the machine is temporarily unable to allocate room
>>for a new skb.  Your machine's atomic memory pools are getting too low...
>>
>>	Jeff
> 
> 
> Yes, I took a look at the code and found the debug message. But isn't there 
> any way to avoid network stop working? Because after a network restart it 
> works again, maybe there could be used any "soft reset" to make network 
> working again after such an error.

The OOM problem is completely unrelated to the network, therefore no 
reset should ever be considered for this condition.

The driver should properly handle the 'NULL in rx ring' condition as a 
normal occurence.  It should skip to the next available skb in the ring. 
  If no skbs are remain, it should drop the skb.

See natsemi.c for additional -- and optional -- OOM handling techniques.

	Jeff