SMC card working?

[Robert G. Brown]

On Tue, 9 May 2000, Thomas Gagne wrote:

> I noticed only one strange phenomona today: when computers send my laptop a
> lot of data quickly, the Xserver seems to get confused and the program
> *appears* to hang.  I was watching the network via tcpdump and it looks like
> the systems are still running, they just stop displaying.  Affected programs
> include telnet, rlogin, and X.

Hmmm.  This sounds like it could be a problem in the TCP stack that
Josip Loncaric on the kernel list has been working on for some time.  It
appears that under certain heavy TCP usage patterns a hideous latency is
introduced (I want to say on the order of ms but I truly cannot
remember).  While the driver is "hung" waiting to get back to the stream
packets build up and the system can get very unhappy or even crash.  He
wrote a tuning patch for the kernel.

Being the sort who squirrels this sort of thing away against the day it
bites, I even have the URL handy:

 Explanation/use:   http://www.icase.edu/coral/LinuxTCP2.html
 Patch for RH6.2:  http://www.icase.edu/~josip/tcp-patch-for-2.2.14-5.0

I'm certainly not certain that this is your problem -- there are also
some memory map problems and other sorts of things that could produce
this sort of result on rare systems (I have one, so I know:-) or just
plain not having enough memory can of course do it -- a high speed data
stream can eat memory buffering it.

If your laptop has only 64 MB and is running X, you might watch "free"
(or procstat, or xosview or MGM or even top) to see what happens to your
memory when this happens.  You might just be running out and swapping.
At this moment I'm doing a make -j3 kernel build on a dual with 128 MB
and X "hangs" just like you describe when I change views because the
parallel make sucks up memory like crazy.

HTH,

   rgb

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb@phy.duke.edu



-------------------------------------------------------------------
To unsubscribe send a message body containing "unsubscribe"
to linux-tulip-request@beowulf.org