Re: [Qemu-devel] [PATCH 1/2] iothread: stash thread ID away

[Stefan Hajnoczi]

On Mon, Feb 24, 2014 at 05:48:13PM +0100, Paolo Bonzini wrote:
> Il 24/02/2014 16:53, Stefan Hajnoczi ha scritto:
> >>>> >+    qemu_cond_destroy(&init_info.init_done_cond);
> >>>> >+    qemu_mutex_destroy(&init_info.init_done_lock);
> >>>
> >>> Destroying the mutex here is racy.  You need to keep it until the
> >>> iothread is destroyed.
> >I don't think so:
> >
> >qemu_cond_signal() is called with the mutex held.  Therefore, our
> >qemu_cond_wait() followed by qemu_mutex_unlock() will only complete once
> >the thread has released the mutex.
> >
> >The thread will never touch the mutex again so it is safe to destroy it.
> >There is no race condition.
> 
> Could qemu_mutex_destroy run while the other thread has already
> released the main thread, but before it returns?  As far as I know,
> the only time when it is safe to destroy the "last" synchronization
> object (in this case the mutex is the last, the condvar is not) is
> after pthread_join.

I guess you're saying that while unlocking the mutex is atomic, that
doesn't guarantee pthread won't access the mutex internal state some
more after it has unlocked it.  Therefore it's not safe for another
thread to destroy the mutex even after it has acquired it.

POSIX does say that:

"It shall be safe to destroy an initialized mutex that is unlocked."

But maybe I am reading too much into that?

After poking around glibc a little I think you are right.  I can't say
for sure but it seems even after a futex call glibc might still mess
with internal state.  But if anyone knows for certain, please speak up.

Stefan