Re: [Intel-gfx] [PATCH 02/15] drm/i915: Don't emit mbox updates without semaphores
On Tue, Dec 17, 2013 at 02:02:23PM -0800, Ben Widawsky wrote: > On Tue, Dec 17, 2013 at 07:24:41PM +, Chris Wilson wrote: > > On Mon, Dec 16, 2013 at 08:50:38PM -0800, Ben Widawsky wrote: > > > Aside from the fact that it leaves confusing dumps on error capture, it > > > is entirely unnecessary, and potentially harmful in cases like BDW, > > > where the instruction has changed. > > > > > > In reality (seemingly), this will have no behavioral impact. > > > > > > Signed-off-by: Ben Widawsky > > > > The reason why we currently do is because i915.semaphores can change at > > runtime. So we emit the instructions whilst i915.semaphores=0 just in > > case, it is enabled later. This restriction can be lifted with a little > > more work in handling the missed semaphores, I think, or it may just > > require a proof that everything is safe as is. > > -Chris > > > > > It should still check the module parameter - I guess it would be nice to > guard changing the module parameter with struct_mutex (generally, not > just here), as that also breaks the emit path. > > So in short, I think it's broken for two reasons. > > My (and Daniel's) vote is to just make the module param static. Dynamic i915.semaphores is something I can live happily without. If we ever do need such a thing, it needs to be internal to the kernel. -Chris -- Chris Wilson, Intel Open Source Technology Centre ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH 02/15] drm/i915: Don't emit mbox updates without semaphores
On Tue, Dec 17, 2013 at 07:24:41PM +, Chris Wilson wrote: > On Mon, Dec 16, 2013 at 08:50:38PM -0800, Ben Widawsky wrote: > > Aside from the fact that it leaves confusing dumps on error capture, it > > is entirely unnecessary, and potentially harmful in cases like BDW, > > where the instruction has changed. > > > > In reality (seemingly), this will have no behavioral impact. > > > > Signed-off-by: Ben Widawsky > > The reason why we currently do is because i915.semaphores can change at > runtime. So we emit the instructions whilst i915.semaphores=0 just in > case, it is enabled later. This restriction can be lifted with a little > more work in handling the missed semaphores, I think, or it may just > require a proof that everything is safe as is. > -Chris > It should still check the module parameter - I guess it would be nice to guard changing the module parameter with struct_mutex (generally, not just here), as that also breaks the emit path. So in short, I think it's broken for two reasons. My (and Daniel's) vote is to just make the module param static. -- Ben Widawsky, Intel Open Source Technology Center ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH 02/15] drm/i915: Don't emit mbox updates without semaphores
On Mon, Dec 16, 2013 at 08:50:38PM -0800, Ben Widawsky wrote: > Aside from the fact that it leaves confusing dumps on error capture, it > is entirely unnecessary, and potentially harmful in cases like BDW, > where the instruction has changed. > > In reality (seemingly), this will have no behavioral impact. > > Signed-off-by: Ben Widawsky The reason why we currently do is because i915.semaphores can change at runtime. So we emit the instructions whilst i915.semaphores=0 just in case, it is enabled later. This restriction can be lifted with a little more work in handling the missed semaphores, I think, or it may just require a proof that everything is safe as is. -Chris -- Chris Wilson, Intel Open Source Technology Centre ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx