On Wed, Jun 12, 2013 at 11:42:56PM +0100, Chris Wilson wrote: > On Wed, Jun 12, 2013 at 02:15:48PM -0700, Ben Widawsky wrote: > > On Mon, Jun 10, 2013 at 01:54:50PM +0100, Chris Wilson wrote: > > > In particular, we were reseting our GEM tracking before even attempting > > > (and failing) to reset the chip. Do it with the other bookkeeping after > > > successfully reseting. Note that we can simplify the reset to always do > > > the GPU reinitialisation as for UMS it will simply be redundant in the > > > case where X was switched away. And that we can safely re-establish all > > > the features that were setup during initialisation - any that were not > > > enabled for UMS/DRI1 will not be restored and should still not be enabled. > > > > > > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk> > > > > > > > FWIW: I have some patches in my PPGTT series which I think accomplish the same > > thing with a nicer end result. > > But the alliteration! If you already have something better, please > ignore these. > > Thanks, > -Chris If only it didn't cause me rebase conflicts, I'd be totally on board! Let me double check that I don't completely fix the problem. In general I have some use cases which could benefit from using the ring during reset which I don't have a clear vision on how to fix. For now I am bending over backwards to not use the ring. -- Ben Widawsky, Intel Open Source Technology Center
