To configure audio registers, the clock of the video port in use must be
enabled.
As those clocks are managed by the VOP driver, they can't be enabled here
to write the registers even when the HDMI cable is disconnected.

Furthermore, the registers values are computed from the TMDS char rate,
which is not available when disconnected.

Returning -ENODEV seemed reasonable at first, but ASoC will log an error
multiple times if dw_hdmi_qp_audio_prepare() return an error.
Userspace might also retry multiple times, filling the kernel log with:

hdmi-audio-codec hdmi-audio-codec.0.auto: ASoC error (-19): at snd_soc_dai_prepare() on i2s-hifi

This has become even worse with the support of the second HDMI TX port.

Activating the clocks to write fake data (fake because the TMDS char
rate is unavailable) would require API changes to communicate between
VOP and HDMI, which doesn't really make sense.

Using a cached regmap to be dumped when a cable is connected won't work
because writing order is important and some data needs to be retrieved
from registers to write others.

Returning 0 to silently fail sounds like the best and simplest solution.

Fixes: fd0141d1a8a2 ("drm/bridge: synopsys: Add audio support for dw-hdmi-qp")
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
---
 drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c b/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c
index 5e5f8c2f95be1..9b9d43c02e3a5 100644
--- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c
+++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c
@@ -458,8 +458,16 @@ static int dw_hdmi_qp_audio_prepare(struct drm_connector *connector,
 	struct dw_hdmi_qp *hdmi = dw_hdmi_qp_from_bridge(bridge);
 	bool ref2stream = false;
 
+	/*
+	 * Silently return if tmds_char_rate is not set.
+	 *
+	 * Writing audio registers requires that the clock of the Video Port currently in
+	 * use by the VOP (dclk_vp<id>) is enabled.
+	 * That clock is guaranteed to be enabled when hdmi->tmds_char_rate is set, so we
+	 * only configure audio when it is set.
+	 */
 	if (!hdmi->tmds_char_rate)
-		return -ENODEV;
+		return 0;
 
 	if (fmt->bit_clk_provider | fmt->frame_clk_provider) {
 		dev_err(hdmi->dev, "unsupported clock settings\n");
-- 
2.50.1

On Tue, Jul 22, 2025 at 03:54:35PM -0400, Detlev Casanova wrote:
> To configure audio registers, the clock of the video port in use must be
> enabled.
> As those clocks are managed by the VOP driver, they can't be enabled here
> to write the registers even when the HDMI cable is disconnected.
> 
> Furthermore, the registers values are computed from the TMDS char rate,
> which is not available when disconnected.
> 
> Returning -ENODEV seemed reasonable at first, but ASoC will log an error
> multiple times if dw_hdmi_qp_audio_prepare() return an error.
> Userspace might also retry multiple times, filling the kernel log with:
> 
> hdmi-audio-codec hdmi-audio-codec.0.auto: ASoC error (-19): at snd_soc_dai_prepare() on i2s-hifi
> 
> This has become even worse with the support of the second HDMI TX port.
> 
> Activating the clocks to write fake data (fake because the TMDS char
> rate is unavailable) would require API changes to communicate between
> VOP and HDMI, which doesn't really make sense.
> 
> Using a cached regmap to be dumped when a cable is connected won't work
> because writing order is important and some data needs to be retrieved
> from registers to write others.
> 
> Returning 0 to silently fail sounds like the best and simplest solution.
> 
> Fixes: fd0141d1a8a2 ("drm/bridge: synopsys: Add audio support for dw-hdmi-qp")
> Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
> ---
>  drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c b/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c
> index 5e5f8c2f95be1..9b9d43c02e3a5 100644
> --- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c
> +++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c
> @@ -458,8 +458,16 @@ static int dw_hdmi_qp_audio_prepare(struct drm_connector *connector,
>  	struct dw_hdmi_qp *hdmi = dw_hdmi_qp_from_bridge(bridge);
>  	bool ref2stream = false;
>  
> +	/*
> +	 * Silently return if tmds_char_rate is not set.
> +	 *
> +	 * Writing audio registers requires that the clock of the Video Port currently in
> +	 * use by the VOP (dclk_vp<id>) is enabled.
> +	 * That clock is guaranteed to be enabled when hdmi->tmds_char_rate is set, so we
> +	 * only configure audio when it is set.
> +	 */
>  	if (!hdmi->tmds_char_rate)
> -		return -ENODEV;
> +		return 0;

What if the cable gets diconnected _while_ this function is running?

>  
>  	if (fmt->bit_clk_provider | fmt->frame_clk_provider) {
>  		dev_err(hdmi->dev, "unsupported clock settings\n");
> -- 
> 2.50.1
> 

-- 
With best wishes
Dmitry

Hi Dmitry
On Wednesday, 23 July 2025 07:41:43 EDT Dmitry Baryshkov wrote:
> On Tue, Jul 22, 2025 at 03:54:35PM -0400, Detlev Casanova wrote:
> > To configure audio registers, the clock of the video port in use must be
> > enabled.
> > As those clocks are managed by the VOP driver, they can't be enabled here
> > to write the registers even when the HDMI cable is disconnected.
> > 
> > Furthermore, the registers values are computed from the TMDS char rate,
> > which is not available when disconnected.
> > 
> > Returning -ENODEV seemed reasonable at first, but ASoC will log an error
> > multiple times if dw_hdmi_qp_audio_prepare() return an error.
> > Userspace might also retry multiple times, filling the kernel log with:
> > 
> > hdmi-audio-codec hdmi-audio-codec.0.auto: ASoC error (-19): at
> > snd_soc_dai_prepare() on i2s-hifi
> > 
> > This has become even worse with the support of the second HDMI TX port.
> > 
> > Activating the clocks to write fake data (fake because the TMDS char
> > rate is unavailable) would require API changes to communicate between
> > VOP and HDMI, which doesn't really make sense.
> > 
> > Using a cached regmap to be dumped when a cable is connected won't work
> > because writing order is important and some data needs to be retrieved
> > from registers to write others.
> > 
> > Returning 0 to silently fail sounds like the best and simplest solution.
> > 
> > Fixes: fd0141d1a8a2 ("drm/bridge: synopsys: Add audio support for
> > dw-hdmi-qp") Signed-off-by: Detlev Casanova
> > <detlev.casanova@collabora.com>
> > ---
> > 
> >  drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c | 10 +++++++++-
> >  1 file changed, 9 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c
> > b/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c index
> > 5e5f8c2f95be1..9b9d43c02e3a5 100644
> > --- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c
> > +++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c
> > @@ -458,8 +458,16 @@ static int dw_hdmi_qp_audio_prepare(struct
> > drm_connector *connector,> 
> >  	struct dw_hdmi_qp *hdmi = dw_hdmi_qp_from_bridge(bridge);
> >  	bool ref2stream = false;
> > 
> > +	/*
> > +	 * Silently return if tmds_char_rate is not set.
> > +	 *
> > +	 * Writing audio registers requires that the clock of the Video Port
> > currently in +	 * use by the VOP (dclk_vp<id>) is enabled.
> > +	 * That clock is guaranteed to be enabled when hdmi->tmds_char_rate 
is
> > set, so we +	 * only configure audio when it is set.
> > +	 */
> > 
> >  	if (!hdmi->tmds_char_rate)
> > 
> > -		return -ENODEV;
> > +		return 0;
> 
> What if the cable gets diconnected _while_ this function is running?

Unplugging the cable will not deactivate the clocks, that's the job of the 
VOP's vop2_crtc_atomic_disable() function. My understanding is that that 
function will not be called until this one has returned, but I could be wrong.

> >  	if (fmt->bit_clk_provider | fmt->frame_clk_provider) {
> >  	
> >  		dev_err(hdmi->dev, "unsupported clock settings\n");

Regards,
Detlev.