> Well, now that I know what to look for, tracking down more of these 
> problems should be significantly faster and easier. Are you generally 
> going to accept similar patches for other unprotected syscalls?

Until now the rule of thumb was to consider only time-consuming syscalls
(as I said, the primary goal is optimizing multi-threaded I/O). But I
guess we can accept such patches if they have no downsides. Martin, what
do you think?

> Still, I'd be extremly grateful if someone could tell me the trick how 
> to create a backtrace in such a deadlock situation... Or am I the only 
> one for which a simple "gdb -a" does not work?

Perhaps you can try to crash the process (using `kill -ABRT`) and then
load the core dump using gdb.