[eepro100] About eepro "out of resources" bug

Kallol Biswas kallol@bagula.fpk.hp.com
Tue, 04 Sep 2001 9:33:52 EDT

  The driver reporting "card out of receive resources"
is not a bug, rather it is a design issue. Let me explain,
correct me if I am wrong:

Both CPU and the card work on the same receive 
resource at the same time, called RFD(receive descriptor list). The card
keeps consuming the resources, and the cpu keeps replenishing
them. The cpu tells the card "this" is the last RFD you can
consume, by setting a bit called EL(end of list) bit on the
last RFD. Consider the scenario, the cpu is too busy in
a heavily loaded system, and does not have much time to
pay attention to card's frame received interrupt and can't
replenish a new buffer and so can not move the EL bit
from the last entry. The card consumes the RFDs at a rate
faster than the cpu supplies them, as a result the card hits
the RFD with EL bit and the driver reports the card is in 
"out of receive resource".  Correct me if I am wrong, as
I don't know how linux driver works well.

I think you can ignore the messages. I have a eepr100.c driver (version
v1.10a 4/15/00) that does not have any out of resource message.


> Hello all,
> I  discovered this list by doing a web search for info related
> to the "out of resources" bug on Linux with the eepro100
> adapter.   What I gathered from reading the messages
> was that the cause of the bug has not been identified,
> and it was hard to reproduce, but some people thought
> it might be related to heavy network use.   I wanted to
> communicate the fact that I seem to be able to reliably
> reproduce this bug on my machine in an odd and
> unexpected way:   by playing the game Myth2 from 
> Loki Games in solo player mode for a few hours and 
> not doing any explicity, user initiated, network 
> access (the game may be doing  something I don't know 
> about, but I don't know of any  reason for it to do so).   
> I am running a Linux 2.4.9 kernel with the driver compiled 
> into the kernel itself.  I've set up interrupts so the card is 
> not sharing.  As another poster mentioned,  I can clear 
> the adapter by restarting networking on my Debian 
> system.  I have not experienced any other bugs related 
> to this card, or managed to trigger the bug in any other 
> way.   I do have a Sounblaster Live Value card installed
> on this machine. I have read that this card tends to use
> up a lot of PCI resources and generate a lot of traffic.
> I may try to play the game without the card installed and
> see whether this setup fails to trigger the bug.
> Just wanted to communicate this info in case it provides
> some different insight into possible causes of the problem.
>  I'm not subscribed to this list, so please copy
> me at jstern@citilink.com on any follow-up e-mail that
> required a response.
> Sincerely,
> -= Josh Stern
> _______________________________________________
> eepro100 mailing list
> eepro100@scyld.com
> http://www.scyld.com/mailman/listinfo/eepro100

Phone: 973-443-7469         |  Design Engineer
Telnet: 1-443-7469          |  EIAL UNIX OS LAB
www.kallolbiswas.com        |  Hewlett Packard Company
kallol_biswas@hp.com        |  Florham Park, NJ