On Tue, May 01, 2012 at 01:44:58AM +0200, Hans J. Koch wrote:

On Mon, Apr 30, 2012 at 10:43:33PM +0100, Mark Brown wrote:

...

I'm not entirely sure you've looked at the code here...

Oh yes, sorry. I got distracted by your mentioning of the suppressed writes. Actually, we have suppressed reads. ac97_read() does a "return cache[reg];" for most registers, assuming the hardware register cannot change after a ac97_write(). That might or might not be true, to me it looks dangerous.

The driver knows exactly which registers are volatile and takes advantage of this - this is entirely idiomatic for devices like this and not really that hard to get right.