Re: [PATCH] x86_64/__iowrite32_copy: don't use string move for PIOwrites

[H. Peter Anvin]

Pavel Machek wrote:
>> - the current code is fine as long as it doesn't cross from cached to
>>   uncached memory, which it never should in any kind of sensible driver;
> 
> Should we add a comment somewhere saying that is not permitted (and why)?

It sounds like what we should do is simply remove
arch/x86/lib/iomap_copy_64.S, and probably add a comment to
lib/iomap_copy.c stating that rep string instructions have been reported
to cause anomalies when accessing WC memory, and that we rely on gcc not
generating them due thanks to the volatiles in the implementation of
__raw_writel().  H.J. Lu tells me that gcc will not generate string
instructions for volatiles, and that the x86 gcc team considers that
part of the definition for volatile.  (HJL: please correct if I got that
incorrect.)

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.

--
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/