Grant Coady wrote: >You ever focus on wrong issues and build wrong model of >what is important? That's where I'm at trying to build >understanding of a system I don't even have end-user >experience with yet. > > Certainly, wouldn't be an experiecned programmer if I hadn't. >I get to the stage of cherry-picking the best ideas I've seen >in the drivers, not one driver has struck me as a model driver. > > > >>In addition to the way I know my drivers worked, I believe that all >>other drivers have an update_lock semaphore that could be used to >>protect read-modify-write's in the same way. >> >> > >But they don't, and I wonder if they should? > > Ahh... That's where I haven't really done a good job explaining myself. I fully agree with you that they should. It's obvious to me that the functions that *set* the values need to be protected to prevent an SMP machine from borking things up. That's why I added the down/up calls in all of the write paths of the drivers I wrote. Perhaps I should have made a bigger deal about this when I first realized it in the drivers I created and submitted patches for the other drivers... Would have saved you the trouble... :v)
