cvs commit: src/sys/i386/include atomic.h

[John Baldwin]

On Tuesday 28 March 2006 15:04, Daniel Eischen wrote:
> On Tue, 28 Mar 2006, Dag-Erling [iso-8859-1] Smørgrav wrote:
> 
> > John Baldwin <jhb at freebsd.org> writes:
> > > s/off/on/ I trust
> > >
> > > Not true.  The tinderbox would attest to that.  Please see code such as
> > > this:  [...]
> >
> > which uses uintptr_t, not actual pointers, to avoid warnings.  In
> > effect, that code is broken.
> >
> > Apply the attached patch, see how far a buildkernel gets...
> >
> > I think the proper thing to do, to cover all your bases, would be to
> > define a MD atomic_*_intptr family which operated on uintptr_t, and
> > define an MI atomic_*_ptr family which operates on void * based on
> > that.
> 
> I would like to see that.
> 
> > > Even userland uses casts when it uses void * rather than uintptr_t for
> > > the underlying type.  See src/lib/libpthread/sys/lock.c or
> > > src/lib/libthr/thr_umtx.h.
> 
> I didn't like having to do that in libpthread.  It was confusing
> to me.  I'd be happy if I never had to deal with uintptr_t.

Then please provide the C idiom for declaring a pointer to a volatile
void pointer.  volatile void ** declares a pointer to a pointer to a
volatile void.  This is a real issue that was causing problems with
refcounting on mbuf's on sparc64 last year due to us not properly
letting the compiler know what values were being invalidated.

cdecl doesn't seem up to the task.  And again, we already had this
discussion N months ago.

-- 
John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org