network interface zombies randomly [was Re: interface dies under network load on SMP machines]

[david]

Reply to mail from Mike Simons about interface dies under network load on SMP machines
-----------------
>>    The interface goes dead from time to time without leaving any log
>> messages - a simple ifconfig down/up brings them back to live. To
>> stabilise the systems I've written a small program that checks the
>> network and restarts the interface if required. So everything is 
>> nearly perfectly fine...
[some parts snipped]

stipulations:

 - it may be the tulip driver
 - it may be the [xxx] driver
 - it may be the kernel
    it may be a subtle bug somewhere in the network/device path that's not
    directly related

I used to feel that it was the tulip driver.  Now I am not sure.  The
default version of tulip.c that comes in the 2.1 tarball works fine.  If I
get any of the most recent tulip.c files from cesdis, then the above
anomally appears.

I.e., random dying of the interface.  It can happen within minutes, it can
take days.  I have not yet found a pattern.

As is mentioned, ifconfig down/up fixes it.

If I recall correctly, packets aren't lost by doing ifconfig down/up.  For
example, ping <host> will flood your screen with all the latent packets
finally getting through.

comments?
-d
-- 
Look, look, see Windows 98.  Buy, lemmings, buy!   
(c) 1998 David Ford.  Redistribution via the Microsoft Network is prohibited.
 for linux-kernel: please read linux/Documentation/* before posting problems