From: Takashi Iwai Subject: Re: [PATCH] drm/i915: Re-enable mt forcewake check for IVB Date: Fri, 16 Nov 2012 14:59:17 +0100 Message-ID: References: Mime-Version: 1.0 (generated by SEMI 1.14.6 - "Maruoka") Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Received: from mx2.suse.de (cantor2.suse.de [195.135.220.15]) by gabe.freedesktop.org (Postfix) with ESMTP id 23309436BF for ; Fri, 16 Nov 2012 05:59:19 -0800 (PST) In-Reply-To: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org Errors-To: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org To: Daniel Vetter Cc: intel-gfx@lists.freedesktop.org List-Id: intel-gfx@lists.freedesktop.org At Thu, 15 Nov 2012 21:22:41 +0100, Takashi Iwai wrote: > > At Thu, 15 Nov 2012 21:07:33 +0100, > Daniel Vetter wrote: > > > > On Thu, Nov 15, 2012 at 6:03 PM, Takashi Iwai wrote: > > > Some IVB machines seem to need non-MT forcewake. Using MT forcewake > > > on such machines result in broken or blank screens. > > > > > > Signed-off-by: Takashi Iwai > > > --- > > > The commit 36ec8f877481449bdfa072e6adf2060869e2b970 triggered a > > > regression on my test IVY laptops. Instead of reverting the commit, > > > I cooked a bit to make fitting only for IS_IVYBRIDGE(). > > > > Hm, that's rather bad since proper usage of DERRMR requires that we > > enable rc6 on ivb, which is only really possible when we have > > multi-threaded forcewake. > > Hm, so just disabling RC6 doesn't suffice? > > > Now I've thought that all production > > machines have that enabled (it's something the bios should do), so the > > big question: Is this a pre-production hw? If so, a bios upgrade > > should fix it since historically we've killed pre-production > > workarounds once the real hw ships, there are simply too many ... > > Indeed a machine showing the issue is a prototype machine, but it's a > product-ready unit, so I don't think CPU/GPU is different from the > same model in the market. It seems that I've looked at a lspci output on a wrong machine while working remotely. The machines showing the problem have CPU revisions less than 8. So, you are right, it must be specific to a pre-production CPU. Assuming it being not in the market, I can live without that. Time to throw away a crap :) Sorry for the noise. Takashi