On 02/05/2014 10:27 PM, Thomas Gleixner wrote:

On Wed, 5 Feb 2014, Ivan Khoronzhuk wrote:

quoted

+	/* here we have to be sure the timer has been disabled */

Sigh. This is not a proper explanation for a barrier, really. You want
to explain what it serializes against what. i.e. you want to explain
why you are using the relaxed functions and avoid a separate non
relaxed variant in favour of an explicit barrier.

quoted

+	__iowmb();

The proper thing is to have an inline function key_stone_barrier() and
a full explanation of the issue in exactly that place instead of
handwaving comments here and there.

Thanks,

	tglx

I can add new inline function like:

/**
  * keystone_timer_barrier: write memory barrier
  * use explicit barrier to avoid using readl/writel non relaxed function
  * variants, because in our case relaxed variants hide the true places
  * where barrier is needed.
  */
static inline void keystone_timer_barrier(void)
{
     __iowmb();
}

and use it where it is needed.
Are you OK with it?

And I propose to leave comments under the barriers in order to be
able to understand why they are used.

-- 
Regards,
Ivan Khoronzhuk

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html