Re: [PATCH 6/8] i386: bitops: Don't mark memory as clobberedunnecessarily

[Trond Myklebust]

On Tue, 2007-07-24 at 11:13 -0700, Linus Torvalds wrote:
> 
> On Tue, 24 Jul 2007, Trond Myklebust wrote:
> > 
> > That's not what the Documentation/memory-barriers.txt states:
> 
> Hmm.. You're right. We only actually need it for the unconditional bitops, 
> like the *unlock* side.
> 
> IOW, if you do a spinlock with the bitops, the locking side should be able 
> to use a "test_and_set_bit()" on its own, but the unlocking side should be
> 
> 	smp_mb__before_clear_bit();
> 	clear_bit();
> 
> because the ones that don't return a value also don't imply a memory 
> barrier.

Yup, and this is exactly what we currently do in bit_spin_unlock().

Trond

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/