[pptp-server] pptpd hanging

John Ferlito johnf at inodes.org
Sun Aug 11 18:07:14 CDT 2002


I'm having a problem on a couple of servers where after a while the
pptpd daemon hangs. I think I sort of know why this is happening but I'm
not sure.

Firstly I'm probably exacerbating the problem since I've got netsaint
basically telneting to port 1723 every 5 minutes to check that the
daemon is running. Anyway when it's in the non working state if you do
an strace you get

select(6, [], NULL NULL NULL)

ie it's sitting on an empty select with no timeout. Now this happens in
pptpmanager.c:167 when its waiting for a SIGCHLD. Now as far as I can
tell this happens when it thinks its run out of IPs to allocate. 

Now there aren't any children hanging around so there is no way its evr
going to get this signal. Hence its stuck there forever.

I can't see how it would get into this state however since I don't ever
see
MGR: No free connection slots or IPs
in my logs anywhere. Which would always get printed if the FD_CLR
instead of FD_SET happened.

I'm also seeing a few 
MGR: Reaped unknown child
which is making me think this maybe the cause too some degree ie it
really has free IPs left but the counter never decreases.

Has anyone come across this problem before?

The problem is its hard to reproduce the problem so hard to test. I'm
going to recompile pptpd with debugging symbols so I can gdb it and
check values next time it happens.


-- 
John
http://www.inodes.org/




More information about the pptp-server mailing list