[Intel-gfx] [PATCH v2] drm/i915: Always program CSR if CSR is uninitialized

[Patrik Jakobsson]

On Tue, Oct 27, 2015 at 08:41:31PM +0200, Imre Deak wrote:
> On pe, 2015-10-23 at 11:41 +0200, Patrik Jakobsson wrote:
> > The current CSR loading code depends on the CSR program memory to be
> > cleared after boot. This is unfortunately not true on all hardware.
> > Instead make use of the FW_UNINITIALIZED state in init and check for
> > FW_LOADED to prevent init path from skipping the actual programming.
> > 
> > v2: Move initialization of state to after HAS_CSR() check
> > 
> > Signed-off-by: Patrik Jakobsson <patrik.jakobsson at linux.intel.com>
> > Tested-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
> > Reviewed-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
> > ---
> >  drivers/gpu/drm/i915/intel_csr.c | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/intel_csr.c b/drivers/gpu/drm/i915/intel_csr.c
> > index 9e530a7..e74c09e 100644
> > --- a/drivers/gpu/drm/i915/intel_csr.c
> > +++ b/drivers/gpu/drm/i915/intel_csr.c
> > @@ -271,7 +271,7 @@ void intel_csr_load_program(struct drm_device *dev)
> >  	 * Unfortunately the ACPI subsystem doesn't yet give us a way to
> >  	 * differentiate this, hence figure it out with this hack.
> >  	 */
> > -	if (I915_READ(CSR_PROGRAM(0)))
> > +	if (I915_READ(CSR_PROGRAM(0)) && dev_priv->csr.state == FW_LOADED)
> >  		return;
> 
> As Animesh said csr->state is being removed, so I'd prefer if we didn't
> add new users for it.
> 
> I think the proper solution for this would be to reprogram the firmware
> only when we know we have to and not have here either of the above
> checks. The points we need to reprogram the firmware:
> 
> S3/S4 both on SKL.
> S3/S4/DC9 (runtime suspend) on BXT.

Isn't this what csr->state was intended to solve in the first place? It's very
easy to get lost in the discussions around DMC so I've might have missed
something.

> 
> To fix this particular bug you could also just add an
> intel_csr_reprogram() that would check for CSR_PROGRAM(0) as now, while
> during driver loading you would program the firmware unconditionally.

Did some more digging and it seems the CSR program is not cleared on warm boot.
So checking CSR_PROGRAM(0) == payload[0] will not work. Don't think we can
figure out the CSR status by probing hardware so we must keep a flag for this
somewhere. The CSR program loading is async so the state need to be stored
somewhere and not just passed along as a function argument.

-Patrik

> 
> --Imre
> 
> 
> >  
> >  	mutex_lock(&dev_priv->csr_lock);
> > @@ -428,6 +428,8 @@ void intel_csr_ucode_init(struct drm_device *dev)
> >  	if (!HAS_CSR(dev))
> >  		return;
> >  
> > +	intel_csr_load_status_set(dev_priv, FW_UNINITIALIZED);
> > +
> >  	if (IS_SKYLAKE(dev))
> >  		csr->fw_path = I915_CSR_SKL;
> >  	else if (IS_BROXTON(dev_priv))
> 
> 
>