On Wed, Oct 12, 2022 at 01:49:36PM +0300, Ville Syrjala wrote: > From: Ville Syrjälä <[email protected]> > > On HSW/BDW the hardware ELD buffer does not work if the controller > is suspended. I'm not 100% which thing in there is needed to make it > work (at least just forcing the controller into D0 with setpci is > not enough). But a full runtime resume seems to do the trick here > at least, and so far it looks like this doesn't even deadlock or > anything. > > Cc: Chaitanya Kumar Borah <[email protected]> > Cc: Kai Vehmanen <[email protected]> > Cc: Takashi Iwai <[email protected]> > Signed-off-by: Ville Syrjälä <[email protected]> > --- > drivers/gpu/drm/i915/display/intel_audio.c | 38 ++++++++++++++++++++++ > 1 file changed, 38 insertions(+) > > diff --git a/drivers/gpu/drm/i915/display/intel_audio.c > b/drivers/gpu/drm/i915/display/intel_audio.c > index 328c47719fd8..467f3f260969 100644 > --- a/drivers/gpu/drm/i915/display/intel_audio.c > +++ b/drivers/gpu/drm/i915/display/intel_audio.c > @@ -23,6 +23,7 @@ > > #include <linux/component.h> > #include <linux/kernel.h> > +#include <linux/pm_runtime.h> > > #include <drm/drm_edid.h> > #include <drm/i915_component.h> > @@ -480,6 +481,16 @@ hsw_audio_config_update(struct intel_encoder *encoder, > hsw_hdmi_audio_config_update(encoder, crtc_state); > } > > +static struct pci_dev *hsw_hda_controller(struct drm_i915_private *i915) > +{ > + int domain = pci_domain_nr(to_pci_dev(i915->drm.dev)->bus); > + > + if (!IS_HASWELL(i915) && !IS_BROADWELL(i915)) > + return NULL; > + > + return pci_get_domain_bus_and_slot(domain, 0, PCI_DEVFN(3, 0)); > +} > + > static int hsw_eld_buffer_size(struct drm_i915_private *i915, > enum transcoder cpu_transcoder) > { > @@ -497,8 +508,14 @@ static void hsw_audio_codec_get_config(struct > intel_encoder *encoder, > enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; > u32 *eld = (u32 *)crtc_state->eld; > int eld_buffer_size, len, i; > + struct pci_dev *hsw_hdac; > u32 tmp; > > + /* on HSW/BDW ELD access doesn't work if the HDA controller is supended > */ > + hsw_hdac = hsw_hda_controller(i915); > + if (hsw_hdac) > + pm_runtime_get_sync(&hsw_hdac->dev); > + > tmp = intel_de_read(i915, HSW_AUD_PIN_ELD_CP_VLD); > if ((tmp & AUDIO_ELD_VALID(cpu_transcoder)) == 0) > return; > @@ -511,6 +528,9 @@ static void hsw_audio_codec_get_config(struct > intel_encoder *encoder, > > for (i = 0; i < len; i++) > eld[i] = intel_de_read(i915, HSW_AUD_EDID_DATA(cpu_transcoder)); > + > + if (hsw_hdac) > + pm_runtime_put(&hsw_hdac->dev);
Maybe these should be put_autosuspend()? Shrug. The runtime pm api remains as impenetrable as ever. -- Ville Syrjälä Intel
