[Intel-gfx] 4.3.3 / skylake / dual graphics: multiple CSR errors at boot from i915 driver

[Daniel Vetter]

On Mon, Feb 15, 2016 at 06:59:53PM -0800, Marc MERLIN wrote:
> On Mon, Feb 15, 2016 at 11:07:23PM +0100, Daniel Vetter wrote:
> > On Mon, Feb 15, 2016 at 02:11:24PM +0800, Zhi Wang wrote:
> > > From SKL, i915 will try to load DMC firmware when system is starting up. You
> > > can find it from 01.org. Personally, it looks without the firmware, the
> > > system is also able to work, except a lot warnings. :)
> > 
> > We pretty much require dmc, since supporting a non-dmc path would double
> > the amount of testing we'd have to do. And right now we can do barely
> > enough to keep that one configuration working. This is the same with all
> > the various pm features that are disabled by default - if you touch any of
> > those we taint the kernel.
> 
> Right.
> I think the problem was that because I was loading from initrd and i915 was
> built in the kernel, it couldn't load the firmware.
> Apparently the solution is to build with 
> CONFIG_EXTRA_FIRMWARE="i915/skl_dmc_ver1.bin i915/skl_guc_ver4.bin"
> CONFIG_EXTRA_FIRMWARE_DIR="../linux-firmware/"
> 
> This allowed things to work, but it's not exactly an obvious step if you
> have i915 built in the kernel.
> On the plus side, my system would hang 1mn while i915 was trying to do
> things without its firmware, but the boot continued and X was usable without
> the firmware loaded.
> 
> Actually outside of the 1mn hang at boot, I'm not sure if my system is
> running X better with the firmware than without :)

dmc is required for some power saving stuff, not for rendering or basic
display functionality. And the problem with not having it really is in
testing all possible combinations, like you've just found out.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch