On Thu, Mar 06, 2008 at 09:06:09AM -0500, Mark M. Hoffman wrote: > Hi Darrick: > Notice how encode_temp() above is outside of the locked section. It is > instructive for potential reviewers of this patch series to see just how easy > it is to miss these. Yep... > REG_CFG5 (from which temp_twos_complement is updated) is never written. Do you > really need to read that during every update? If you move that one read into > the init routine, that will instantly kill all the race conditions associated > with temp_twos_complement (assuming the hardware does not modify this register > behind our backs). What do you think? Correct, the hardware doesn't modify that register at all. I see your patch, so I'll have a look through that. --D
