Re: [PATCH] kern: simple futex for gnumach (version 4)

[Richard Braun]

On Tue, Dec 24, 2013 at 12:52:23PM +0100, Marin Ramesa wrote:
> I need to start somewhere. I really want to learn how kernels are
> programmed.
> This task was listed in the small hacks entry, I really thought it
> would be a
> small hack.

That's probably a mistake on our part.

> There is a one call to vm_map lookup to retrive the offset and the
> object. Simple

How do you implement cross address space synchronization ?

> locks are used and there are calls to thread_resume() and
> thread_suspend(). I don't
> see anything so much difficult that I can't learn while doing it.

Well, using thread_resume/thread_suspend might work for non preemptible
kernels, but not for a preemptible one, or a multiprocessor one like
Mach was.

Look at your call to thread_suspend. If thread_resume is called after
the futex lock is released (it can't be called before since it's called
with that same lock reacquired), but before thread_suspend is called,
the thread will miss its wakeup. This is why interlocks (usually hidden
in wait queues) are normally used. See assert_wait().

> But I won't work on this if it's frustrating. I don't want to cause
> any trouble.

For now, I'm the only one reacting to this, not much of a trouble.
But I'm pretty sure we would all like contributors to pay great care
to what they're doing, whatever their technical level.

-- 
Richard Braun