On 12/16, Peter Zijlstra wrote:
quoted text
>
> So yea, I think we can simplify all this even further, but we do need a
> smp_rmb() in between those two checks, hmm, and RCU.

I am not sure we need rmb() or even read_barrier_depends(), I think
the simple barrier() should be enough. Although rcu_dereference()
may be more clean.

We don't really care about the "correct" result of *owner. All we need,
we should ensure it is safe to dereference this address. If CPU itself
does something like "readahead" before we verified "lock->owner == owner"
under RCU lock, this shouldn't lead to the fault.

Oleg.

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