[PLUG] Re: debugging hyperthreading problems
Wil Cooley
wcooley at nakedape.cc
Thu Oct 12 21:25:26 UTC 2006
On Thu, 2006-10-12 at 10:53 -0700, drew wymore wrote:
> Bringing this thread back to life. Still trying to figure this one out. I
> have ruled out SSH as using Telnet causes the same behavior with the second
> CPU enabled. However with both CPU's going and logging in locally the issue
> doesn't happen. So its something with sockets/bash or some other weirdness
> but I'm at a loss now. Someone at work suggested maybe a glibc problem but
> I'm running a recent version that supports NPTL and bash is using those libs
> and when I force standard linux thread libs it still has the same problem.
>
> Anyone ideas?
Something funny is going on with interrupts, probably.
o Have you tried a different NIC, preferably not on-board?
o Have you tried pinging it, preferably more than the default 4 times
Windows ping does? You might find some odd variances in ping times
where there should be none.
o If you can do a flood ping or otherwise send other traffic to the
host while making the connection and see if that causes it to not
happen--it sounds like a buffer somewhere isn't getting flushed as
often as expected and pumping data through might cause it to be
flushed.
o Run Wireshark (nee Ethereal) while logging in and compare packet
timings, look for broken packets, etc.
o Check /etc/profile, /etc/profile.d, /etc/bashrc to see if something
gets run or set on login. You might temporarily enable tracing by
putting 'set -x' at the top of the /etc/profile and see if you can
pinpoint where it stops, if it actually stops.
Wil
--
Wil Cooley <wcooley at nakedape.cc>
http://nakedape.cc
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.pdxlinux.org/pipermail/plug/attachments/20061012/fae14155/attachment.asc>
More information about the PLUG
mailing list