[Intel-gfx] [PATCH 5/5] drm/i915/bxt: add DDI port HW readout support

[Ville Syrjälä]

On Mon, Jun 22, 2015 at 05:22:33PM +0300, Imre Deak wrote:
> On ma, 2015-06-22 at 16:44 +0300, Ville Syrjälä wrote:
> > On Thu, Jun 18, 2015 at 05:25:57PM +0300, Imre Deak wrote:
> > > Add support for reading out the HW state for DDI ports. Since the actual
> > > programming is very similar to the CHV/VLV DPIO PLL programming we can
> > > reuse much of the logic from there.
> > > 
> > > This fixes the state checker failures I saw on my BXT with HDMI output.
> > > 
> > > Signed-off-by: Imre Deak <imre.deak at intel.com>
> > > ---
> > >  drivers/gpu/drm/i915/i915_reg.h  | 15 +++++++++------
> > >  drivers/gpu/drm/i915/intel_ddi.c | 22 ++++++++++++++++++++--
> > >  2 files changed, 29 insertions(+), 8 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> > > index bba0691..fcf6ad5 100644
> > > --- a/drivers/gpu/drm/i915/i915_reg.h
> > > +++ b/drivers/gpu/drm/i915/i915_reg.h
> > > @@ -1169,10 +1169,12 @@ enum skl_disp_power_wells {
> > >  #define _PORT_PLL_EBB_0_A		0x162034
> > >  #define _PORT_PLL_EBB_0_B		0x6C034
> > >  #define _PORT_PLL_EBB_0_C		0x6C340
> > > -#define   PORT_PLL_P1_MASK		(0x07 << 13)
> > > -#define   PORT_PLL_P1(x)		((x)  << 13)
> > > -#define   PORT_PLL_P2_MASK		(0x1f << 8)
> > > -#define   PORT_PLL_P2(x)		((x)  << 8)
> > > +#define   PORT_PLL_P1_SHIFT		13
> > > +#define   PORT_PLL_P1_MASK		(0x07 << PORT_PLL_P1_SHIFT)
> > > +#define   PORT_PLL_P1(x)		((x)  << PORT_PLL_P1_SHIFT)
> > > +#define   PORT_PLL_P2_SHIFT		8
> > > +#define   PORT_PLL_P2_MASK		(0x1f << PORT_PLL_P2_SHIFT)
> > > +#define   PORT_PLL_P2(x)		((x)  << PORT_PLL_P2_SHIFT)
> > >  #define BXT_PORT_PLL_EBB_0(port)	_PORT3(port, _PORT_PLL_EBB_0_A, \
> > >  						_PORT_PLL_EBB_0_B,	\
> > >  						_PORT_PLL_EBB_0_C)
> > > @@ -1192,8 +1194,9 @@ enum skl_disp_power_wells {
> > >  /* PORT_PLL_0_A */
> > >  #define   PORT_PLL_M2_MASK		0xFF
> > >  /* PORT_PLL_1_A */
> > > -#define   PORT_PLL_N_MASK		(0x0F << 8)
> > > -#define   PORT_PLL_N(x)			((x) << 8)
> > > +#define   PORT_PLL_N_SHIFT		8
> > > +#define   PORT_PLL_N_MASK		(0x0F << PORT_PLL_N_SHIFT)
> > > +#define   PORT_PLL_N(x)			((x) << PORT_PLL_N_SHIFT)
> > >  /* PORT_PLL_2_A */
> > >  #define   PORT_PLL_M2_FRAC_MASK		0x3FFFFF
> > >  /* PORT_PLL_3_A */
> > > diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c
> > > index ca970ba..6859068 100644
> > > --- a/drivers/gpu/drm/i915/intel_ddi.c
> > > +++ b/drivers/gpu/drm/i915/intel_ddi.c
> > > @@ -971,8 +971,26 @@ static void hsw_ddi_clock_get(struct intel_encoder *encoder,
> > >  static int bxt_calc_pll_link(struct drm_i915_private *dev_priv,
> > >  				enum intel_dpll_id dpll)
> > >  {
> > > -	/* FIXME formula not available in bspec */
> > > -	return 0;
> > > +	struct intel_shared_dpll *pll;
> > > +	struct intel_dpll_hw_state *state;
> > > +	intel_clock_t clock;
> > > +
> > > +	/* For DDI ports we always use a shared PLL. */
> > > +	if (WARN_ON(dpll == DPLL_ID_PRIVATE))
> > > +		return 0;
> > > +
> > > +	pll = &dev_priv->shared_dplls[dpll];
> > > +	state = &pll->config.hw_state;
> > > +
> > > +	clock.m1 = 2;
> > 
> > For chv I opted to read out the m1 divider from PLL_DW1 too even though
> > 2 is the only supported value. Doing so might catch some cases where the
> > register gets misprogrammed.
> 
> I haven't found the BXT counterpart for this at least when looking
> through BSpec. PORT_PLL_1[2:0] is "i_fbpredivratio" and it reads 0 for
> me.

That's the one.

> 
> > But that's a minor issue, so even without this is
> > Reviewed-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > 
> > > +	clock.m2 = (state->pll0 & PORT_PLL_M2_MASK) << 22;
> > > +	if (state->pll3 & PORT_PLL_M2_FRAC_ENABLE)
> > > +		clock.m2 |= state->pll2 & PORT_PLL_M2_FRAC_MASK;
> > 
> > I just realized we're missing the fracnen check from chv_crtc_clock_get().
> > Care to send a patch to add the check there as well?
> 
> Ok.
> 
> > > +	clock.n = (state->pll1 & PORT_PLL_N_MASK) >> PORT_PLL_N_SHIFT;
> 
> Btw, now that I looked through BSpec and the code again, this looks like
> also be tied to 1, so we'd need to check for this in
> bxt_ddi_pll_select(). chv_find_best_dpll() does the right thing already.

You mean WARN if n != 1? I suppose it can't hurt.

> 
> > > +	clock.p1 = (state->ebb0 & PORT_PLL_P1_MASK) >> PORT_PLL_P1_SHIFT;
> > > +	clock.p2 = (state->ebb0 & PORT_PLL_P2_MASK) >> PORT_PLL_P2_SHIFT;
> > > +
> > > +	return vlv_calc_port_clock(100000, &clock);
> > >  }
> > >  
> > >  static void bxt_ddi_clock_get(struct intel_encoder *encoder,
> > > -- 
> > > 2.1.4
> > > 
> > > _______________________________________________
> > > Intel-gfx mailing list
> > > Intel-gfx at lists.freedesktop.org
> > > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
> > 
> 

-- 
Ville Syrjälä
Intel OTC