On Wed, Aug 26, 2015 at 10:32:01AM +0200, Daniel Vetter wrote: > On Tue, Aug 18, 2015 at 04:06:17PM +0100, Chris Wilson wrote: > > On Tue, Aug 18, 2015 at 03:55:07PM +0100, Nick Hoath wrote: > > > >This is the wrong location. Just kill set_seqno, the experiment has run > > > >its course and we now have a n igt to exercise seqno wraparound. > > > It has to be here as the seqno has to be initialised before it is > > > used to create requests for the initialisation. > > > > It is the wrong location as init_hw() is called as part of resume and > > reset, both times where we don't want to actually touch seqno. > > Hm I thought it also makes sure to update the hws to avoid the driver > getting confused, which is probably something we should be doing on resume > (since the hws is in stolen) because hibernate. The hws page can't be in stolen since we need CPU access to it and the powers-that-be have said no. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
