public inbox for ecos-discuss@sourceware.org
 help / color / mirror / Atom feed
* [ECOS] high bandwidth network problems
@ 2000-09-06  2:58 Carl van Schaik
  2000-09-06  5:13 ` Gary Thomas
  0 siblings, 1 reply; 4+ messages in thread
From: Carl van Schaik @ 2000-09-06  2:58 UTC (permalink / raw)
  To: Ecos Mailing List

Hi

I have my board connected to our backbone which has a huge ammount of
broadcast traffic continously (nearly 2000 machines), after a while, ecos
starts spewing out messages: and stops responding to network traffic.

warning: eth_recv out of MBUFs
setsoftnet

Can anyone tell me what the setsoftnet is supposed to do? Do I have to 
implement somthing in the network driver? What can I do to
prevent this from occuring?

thanks

Carl van Schaik   <carl@leg.uct.ac.za>
-----------------------------------------------------
Dept. Electrical Engineering, UCT

621 Menzies Building
University of Cape Town
Rondebosch   
Western Cape
7700
South Africa

Tel: +27 21 650-3467
Fax: +27 21 650-3465

^ permalink raw reply	[flat|nested] 4+ messages in thread

* RE: [ECOS] high bandwidth network problems
  2000-09-06  2:58 [ECOS] high bandwidth network problems Carl van Schaik
@ 2000-09-06  5:13 ` Gary Thomas
  2000-09-06  5:19   ` Andrew Lunn
  0 siblings, 1 reply; 4+ messages in thread
From: Gary Thomas @ 2000-09-06  5:13 UTC (permalink / raw)
  To: Carl van Schaik; +Cc: Ecos Mailing List

On 06-Sep-2000 Carl van Schaik wrote:
> Hi
> 
> I have my board connected to our backbone which has a huge ammount of
> broadcast traffic continously (nearly 2000 machines), after a while, ecos
> starts spewing out messages: and stops responding to network traffic.
> 
> warning: eth_recv out of MBUFs
> setsoftnet
> 
> Can anyone tell me what the setsoftnet is supposed to do? Do I have to 
> implement somthing in the network driver? What can I do to
> prevent this from occuring?
> 

Try increasing the number of mbufs.  Set the configuration parameter
CYGPKG_NET_MEM_USAGE to something larger (the default is 256*1024).

Do you have the latest code from CVS?  I seem to recall something in there
about an mbuf memory leak which was fixed recently.  This might help.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [ECOS] high bandwidth network problems
  2000-09-06  5:13 ` Gary Thomas
@ 2000-09-06  5:19   ` Andrew Lunn
  2000-09-06  5:24     ` Gary Thomas
  0 siblings, 1 reply; 4+ messages in thread
From: Andrew Lunn @ 2000-09-06  5:19 UTC (permalink / raw)
  To: Gary Thomas; +Cc: carl, ecos-discuss

> Try increasing the number of mbufs.  Set the configuration parameter
> CYGPKG_NET_MEM_USAGE to something larger (the default is 256*1024).
> 
> Do you have the latest code from CVS?  I seem to recall something in there
> about an mbuf memory leak which was fixed recently.  This might help.

Increasing the memory is not the real solution. The code will cope
with running out of mbufs, its a normal occurance. Whats wrong is
printing these messages. Printing these messages is a classic denial
of service problem. What it should do is limit the number of times it
prints it, to a maximum of one per second, or even not at all.

        Andrew

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [ECOS] high bandwidth network problems
  2000-09-06  5:19   ` Andrew Lunn
@ 2000-09-06  5:24     ` Gary Thomas
  0 siblings, 0 replies; 4+ messages in thread
From: Gary Thomas @ 2000-09-06  5:24 UTC (permalink / raw)
  To: andrew.lunn; +Cc: ecos-discuss, carl

On 06-Sep-2000 Andrew Lunn wrote:
>> Try increasing the number of mbufs.  Set the configuration parameter
>> CYGPKG_NET_MEM_USAGE to something larger (the default is 256*1024).
>> 
>> Do you have the latest code from CVS?  I seem to recall something in there
>> about an mbuf memory leak which was fixed recently.  This might help.
> 
> Increasing the memory is not the real solution. The code will cope
> with running out of mbufs, its a normal occurance. Whats wrong is
> printing these messages. Printing these messages is a classic denial
> of service problem. What it should do is limit the number of times it
> prints it, to a maximum of one per second, or even not at all.
> 

A reasonable observation.  The messages are[were] there simply so we 
could see if the code was detecting such conditions and doing the right
thing.  If they messages themselves are getting in the way, just nuke 'em.

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2000-09-06  5:24 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2000-09-06  2:58 [ECOS] high bandwidth network problems Carl van Schaik
2000-09-06  5:13 ` Gary Thomas
2000-09-06  5:19   ` Andrew Lunn
2000-09-06  5:24     ` Gary Thomas

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).