On Wed, Apr 13, 2016 at 11:59:06AM +0200, Daniel Vetter wrote: > On Tue, Apr 12, 2016 at 09:03:08PM +0100, Chris Wilson wrote: > > After mi_set_context() succeeds, we need to update the state of the > > engine's last_context. This ensures that we hold a pin on the context > > whilst the hardware may write to it. However, since we didn't complete > > the post-switch setup of the context, we need to force the subsequent > > use of the same context to complete the setup (which means updating > > should_skip_switch()). > > > > Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> > > Cc: Daniel Vetter <daniel.vetter@xxxxxxxx> > > Cc: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxxxxxxxx> > > Specializing do_rcs_switch makes sense, but shouldn't be intermingled with > the bugfix. Assuming I'm reading things correctly the real bugfix is to > add the check for hw_ctx.initialized to should_skip_switch()? If so, > please split into 2 patches - first to add just that check, 2nd to do the > RCS specialization. Not quite. That is a critical step, but we also have to set the engine->last_context earlier. That gets quite muddled with threading the !rcs path through it. I felt that separating them and making it clear what was happening to each was much easier to understand. The proviso being as always we duplicate some code (later patches will ease some of that, but nothing can eliminate that RCS is special). -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
