Re: [parisc-linux] coherent ops and mb() revisited

[James Bottomley <James.Bottomley@SteelEye.com>]

On Sat, 2004-09-04 at 21:38, Grant Grundler wrote:
> I haven't tested the following patch and just wanted to get
> feedback if it is "the right thing" or not.

It should work, but I don't think it's the right thing to do.

__ldcw already insists on operating on volatile data, which is the
correct thing to do.  Adding a memory barrier is probably harmless, but
it crimps any optimisations gcc might like to try with that function
(not that there really are any with such a simple loop).  Indeed, we
could even remove the __volatile__ from the asm.

I think the best implementation is probably

barrier();
a = __ldcw_align(x);
while (__ldcw(a) == 0)
	while(*a == 0)
		;
mb();

Which makes it totally clear what barriers we need where in the spinlock
(the first being simply a compiler ordering barrier and the second
actually a processor memory barrier).

James


_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux