From: Chris Wilson Subject: Re: [Intel-gfx] [PATCH] drm/i915: correctly restore fences with objects attached Date: Wed, 17 Jul 2013 16:47:32 +0100 Message-ID: <20130717154732.GF5249@cantiga.alporthouse.com> References: <1374065488-4485-1-git-send-email-daniel.vetter@ffwll.ch> <20130717083448.30b8fd5f@jbarnes-desktop> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20130717083448.30b8fd5f@jbarnes-desktop> Sender: stable-owner@vger.kernel.org To: Jesse Barnes Cc: Daniel Vetter , Intel Graphics Development , "for 3.10 only" List-Id: intel-gfx@lists.freedesktop.org On Wed, Jul 17, 2013 at 08:34:48AM -0700, Jesse Barnes wrote: > On Wed, 17 Jul 2013 14:51:28 +0200 > Daniel Vetter wrote: > > > + if (reg->obj) { > > + i915_gem_object_update_fence(reg->obj, reg, > > + reg->obj->tiling_mode); > > + } else { > > + i915_gem_write_fence(dev, i, NULL); > > + } > > Why do we look at reg->obj here? Can it be NULL? Or would > reg->obj->tiling_mode != I915_TILING_NONE do the same thing? Yes. This is the array of fence registers, irrespective of whether an object is using one. -Chris -- Chris Wilson, Intel Open Source Technology Centre