On Thu, Sep 18, 2014 at 04:36:47PM +0100, Chris Wilson wrote: > On Thu, Sep 18, 2014 at 05:58:30PM +0300, Mika Kuoppala wrote: > > We don't know in what shape the default context was before reset. > > The reset also dropped our changes that were done in > > ring->init_context. > > > > Mark our default context as uninitialized for it to be properly > > setup up on reset recovery . > > The fundamental problem here is that context_init is not correctly split > into init/init_hw (or init/enable) pairs. That is we simply do not > attempt to reset the default state after reset via i915_gem_init_hw(). I > would rather we brought rationality to the gem init/init_hw sequence > first. Yeah I agree with Chris that context setup doesn't follow our established split with _init doing the software setup (and doesn't touch the hw at all) and _init_hw doing the hw setup. _init is run at driver load, while _init_hw is run at driver load, resume, after gpu reset and maybe even from runtime pm code. But I don't mind if we do this after or before this series, whatever is less fuzz is fine with me. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
