[plug] GDM hanging, leaving zombies

[Craig Ringer]

>>Ah, sorry, I though that was clear from my eariler message. Process 
>>24242 is the parent of all those processes.
> 
> To clarify: you mean that it is the direct parent of each process --
> there is no nesting of processes (hence the `ps -F` idea)?

Direct parent, yes. I couldn't send ppid info if I wanted to then, 
because I'd already killed it off. However, pid 24242 was the direct 
parent of all other gdm processes on the system, and was listed as the 
ppid on those processes.

>>I guess it could be a kernel bug - I'm using the stock RH kernel with 
>>their threading stuff - but it's unlikely to be a hardware issue.
> 
> Did we migrate toward "sounds like a kernel bug" last time, too?
> (Hey,...that's right...were those zombie processes still holding
> onto sockets today...? It is surely a kernel bug if they were.)

Nope. That was me being an idiot, and misreading the netstat output last 
time. pid 24242, the parent, does hold a listening socket on udp/177 and 
is not responding to requests, however it is a real process and not a 
zombie.

Personally, though, I think Bernard's idea makes sense - though it 
doesn't explain why I couldn't trace the process or attach gdb.

Craig Ringer


_______________________________________________
plug mailing list
plug at plug.linux.org.au
http://mail.plug.linux.org.au/cgi-bin/mailman/listinfo/plug