On Thu, 05 Apr 2018, Abhay Kumar <abhay.kumar@xxxxxxxxx> wrote: > In glk when device boots with 1366x768 panel, HDA codec doesn't comeup. > This result in no audio forever as cdclk is < 96Mhz. > This chagne will ensure CD clock to be twice of BCLK. In short, we can't poke around CDCLK like this. It needs a full modeset, and it's non-trivial from the path you're calling this from. I'm considering pushing the original patch [1], because we haven't come up with working alternatives. Please confirm that the patch reliably fixes the issue. (Though I think if you boot with *all* outputs disabled, we'll choose the lowest CDCLK possible regardless of the patch, reproducing the same issue.) What's the CDCLK frequency set by the BIOS/GOP at boot? There are no logs with drm.debug=14 attached to the referenced bug. I see that bspec says, "158.4 MHz CD (Display Audio enumeration and playback OK)" but that's *not* twice the BCLK. I'm inclined to lean towards 192 MHz min leading to max CDCLK on GLK. BR, Jani. [1] https://patchwork.freedesktop.org/patch/184778/ > > v2: > - Address comment (Jani) > - New design approach > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102937 > Signed-off-by: Abhay Kumar <abhay.kumar@xxxxxxxxx> > --- > drivers/gpu/drm/i915/intel_audio.c | 33 ++++++++++++++++++++++----------- > drivers/gpu/drm/i915/intel_cdclk.c | 21 +++++++++++++++++++++ > drivers/gpu/drm/i915/intel_drv.h | 1 + > 3 files changed, 44 insertions(+), 11 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c > index 709d6ca68074..f7dd3d532e93 100644 > --- a/drivers/gpu/drm/i915/intel_audio.c > +++ b/drivers/gpu/drm/i915/intel_audio.c > @@ -723,15 +723,37 @@ static void i915_audio_component_put_power(struct device *kdev) > intel_display_power_put(kdev_to_i915(kdev), POWER_DOMAIN_AUDIO); > } > > +/* Get CDCLK in kHz */ > +static int i915_audio_component_get_cdclk_freq(struct device *kdev) > +{ > + struct drm_i915_private *dev_priv = kdev_to_i915(kdev); > + > + if (WARN_ON_ONCE(!HAS_DDI(dev_priv))) > + return -ENODEV; > + > + return dev_priv->cdclk.hw.cdclk; > +} > + > static void i915_audio_component_codec_wake_override(struct device *kdev, > bool enable) > { > struct drm_i915_private *dev_priv = kdev_to_i915(kdev); > u32 tmp; > + int current_cdclk; > > if (!IS_GEN9_BC(dev_priv)) > return; > > + current_cdclk = i915_audio_component_get_cdclk_freq(kdev); > + > + /* > + * Before probing for HDA Codec we need to make sure > + * "The CD clock frequency must be at least twice > + * the frequency of the Azalia BCLK." > + */ > + if (INTEL_GEN(dev_priv) >= 9 && current_cdclk <= 192000) > + intel_cdclk_bump(dev_priv); > + > i915_audio_component_get_power(kdev); > > /* > @@ -753,17 +775,6 @@ static void i915_audio_component_codec_wake_override(struct device *kdev, > i915_audio_component_put_power(kdev); > } > > -/* Get CDCLK in kHz */ > -static int i915_audio_component_get_cdclk_freq(struct device *kdev) > -{ > - struct drm_i915_private *dev_priv = kdev_to_i915(kdev); > - > - if (WARN_ON_ONCE(!HAS_DDI(dev_priv))) > - return -ENODEV; > - > - return dev_priv->cdclk.hw.cdclk; > -} > - > /* > * get the intel_encoder according to the parameter port and pipe > * intel_encoder is saved by the index of pipe > diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c > index dc7db8a2caf8..9426e1b7badc 100644 > --- a/drivers/gpu/drm/i915/intel_cdclk.c > +++ b/drivers/gpu/drm/i915/intel_cdclk.c > @@ -1516,6 +1516,27 @@ void bxt_init_cdclk(struct drm_i915_private *dev_priv) > } > > /** > + * intel_cdclk_bump - Increase cdclk to 2* BCLK > + * @dev_priv: i915 device > + * > + * Increase CDCLK for GKL and CNL. This is done only > + * during HDA codec probe. > + */ > +void intel_cdclk_bump(struct drm_i915_private *dev_priv) > +{ > + struct intel_cdclk_state cdclk_state; > + > + cdclk_state = dev_priv->cdclk.hw; > + > + if (IS_GEMINILAKE(dev_priv)) { > + cdclk_state.cdclk = glk_calc_cdclk((2*96000)); > + cdclk_state.vco = glk_de_pll_vco(dev_priv, cdclk_state.cdclk); > + cdclk_state.voltage_level = bxt_calc_voltage_level(cdclk_state.cdclk); > + bxt_set_cdclk(dev_priv, &cdclk_state); > + } > +} > + > +/** > * bxt_uninit_cdclk - Uninitialize CDCLK on BXT > * @dev_priv: i915 device > * > diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h > index d1452fd2a58d..5192753df3dc 100644 > --- a/drivers/gpu/drm/i915/intel_drv.h > +++ b/drivers/gpu/drm/i915/intel_drv.h > @@ -1417,6 +1417,7 @@ void skl_uninit_cdclk(struct drm_i915_private *dev_priv); > void cnl_init_cdclk(struct drm_i915_private *dev_priv); > void cnl_uninit_cdclk(struct drm_i915_private *dev_priv); > void bxt_init_cdclk(struct drm_i915_private *dev_priv); > +void intel_cdclk_bump(struct drm_i915_private *dev_priv); > void bxt_uninit_cdclk(struct drm_i915_private *dev_priv); > void icl_init_cdclk(struct drm_i915_private *dev_priv); > void icl_uninit_cdclk(struct drm_i915_private *dev_priv); -- Jani Nikula, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
