Deadlocked cygrunsrv processes with strace message "_cygtls::remove: wait 0x0"

[Max Bowsher]

Max Bowsher wrote:
> Something about Apache 2 is causing its controlling cygrunsrv service
> process to deadlock within the Cygwin DLL when it attempts to exit.
> Attaching with strace to the hung process shows "_cygtls::remove: wait 
> 0x0".
> Signals sent to the hung process are not delivered (as shown by strace), 
> but
> a kill -9 will succeed.
> Attaching with gdb does show anything that looks promising.
>
> Does anyone have any suggestions that could save me some time, compared to
> an in-depth study of the Cygwin DLL code in this area?

Ummmmm. Oh. Dear.

Since gdb backtrace wasn't helpful, I decided to try single stepping in gdb 
onwards from the last strace output.

I found out that _cygtls.remove is being called successfully from 
dll_entry(DLL_THREAD_DETACH), which returns, and the hang finally occurs 
inside ZwTerminateThread() !!!
What now? Do I scream and give up?

Max.


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/