latest CVS 2.0 beta hang problem

[Alan DeKok]

A.L.M.Buxey at lboro.ac.uk wrote:
> okay - got it to do the same with a straight radtest. 
> here is the radiusd -X  - several attempts that get proxied
> away..and never answered. remote site marked as zombie
> but then any valid requests just dont get serviced.

  OK.  There are a few issues:

> Rejecting request 0 due to lack of any response from home server 10.1.2.3 port 1812
> WARNING: Unresponsive child (id 0) for request 0, in module <server-core> component <server-core>
>   Found Post-Proxy-Type

  That's just a dumb message, as it's not running in threaded mode, AND
the request isn't being process in a child thread!  I've committed a fix
which prints that message only if there is really a child thread that's
unresponsive.

> Going to the next request
> WARNING: Marking home server 10.1.2.3 port 1812 as zombie (it looks like it is dead).
> Waking up in 0.2 seconds.
> Child is finally responsive for request 0
> 
> after this, its a kill -9 job too on the server process. not just kill/killall
> or server restart via 'service' init.d tool.

  I think what's happening is that the request is being free'd in a
place where it doesn't expect it to be free'd.

  I've re-arranged the code to note, and catch this race condition.  It
works in my tests, but my tests didn't see this race condition in the
first place...

  Alan DeKok.