On 11/17/26 21:11, [email protected] wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential
> issue to consider:
> - [High] drm/meson: dw-hdmi: skipped HDMITX mem PD power-up in
>   meson_dw_hdmi_init()
> --
> 
> commit c8e782099f758bf635402705a20f57d7df9682a7
> Author: Ronald Claveau <[email protected]>
> 
> drm/meson: dw-hdmi: Drop manual HHI clock and memory PD setup
> 
> This commit removes the direct HHI_HDMI_CLK_CNTL and HHI_MEM_PD_REG0
> register writes in meson_dw_hdmi_init() and takes a runtime PM reference
> in meson_dw_hdmi_probe() to power on the VPU power domain.
> 
>> diff --git a/drivers/gpu/drm/meson/meson_dw_hdmi.c 
>> b/drivers/gpu/drm/meson/meson_dw_hdmi.c
>> index eb33c70f29c31..20a821936b3af 100644
>> --- a/drivers/gpu/drm/meson/meson_dw_hdmi.c
>> +++ b/drivers/gpu/drm/meson/meson_dw_hdmi.c
> [ ... ]
>> @@ -599,12 +598,6 @@ static void meson_dw_hdmi_init(struct meson_dw_hdmi 
>> *meson_dw_hdmi)
>>  {
>>      struct meson_drm *priv = meson_dw_hdmi->priv;
>>  
>> -    /* Enable clocks */
>> -    regmap_update_bits(priv->hhi, HHI_HDMI_CLK_CNTL, 0xffff, 0x100);
>> -
>> -    /* Bring HDMITX MEM output of power down */
>> -    regmap_update_bits(priv->hhi, HHI_MEM_PD_REG0, 0xff << 8, 0);
>> -
>>      /* Reset HDMITX APB & TX & PHY */
>>      reset_control_reset(meson_dw_hdmi->hdmitx_apb);
>>      reset_control_reset(meson_dw_hdmi->hdmitx_ctrl);
>> @@ -829,9 +822,29 @@ static int __maybe_unused 
>> meson_dw_hdmi_pm_resume(struct device *dev)
>>      return 0;
>>  }
>>  
>> +static void meson_dw_hdmi_pm_put(void *dev)
>> +{
>> +    pm_runtime_put_sync(dev);
>> +}
>> +
>>  static int meson_dw_hdmi_probe(struct platform_device *pdev)
>>  {
>> -    return component_add(&pdev->dev, &meson_dw_hdmi_ops);
>> +    struct device *dev = &pdev->dev;
>> +    int ret;
>> +
>> +    ret = devm_pm_runtime_enable(dev);
>> +    if (ret)
>> +            return ret;
>> +
>> +    ret = pm_runtime_resume_and_get(dev);
>> +    if (ret)
>> +            return dev_err_probe(dev, ret, "Failed to power on HDMI TX\n");
> 
> [Severity: High]
> Can dropping the HHI_MEM_PD_REG0 write leave the HDMITX memory powered
> down when the bootloader turns on the top-level VPU domain without
> enabling HDMI?
> 
> In the VPU power domain driver, pwrc_ee_is_powered_off() only checks the
> top-level AO sleep bit and does not check HHI_MEM_PD_REG0[15:8].  When
> the top-level VPU domain is already powered on at boot,
> meson_ee_pwrc_init_domain() registers the genpd as already on without
> calling meson_ee_pwrc_on():
> 
> drivers/pmdomain/amlogic/meson-ee-pwrc.c:meson_ee_pwrc_init_domain() {
>       ...
>       if (dom->num_clks && dom->desc.is_powered_off && 
> !dom->desc.is_powered_off(dom)) {
>               ret = clk_bulk_prepare_enable(dom->num_clks, dom->clks);
>               if (ret)
>                       return ret;
> 
>               dom->base.flags = GENPD_FLAG_ALWAYS_ON;
>               ret = pm_genpd_init(&dom->base, NULL, false);
>               if (ret)
>                       return ret;
>       }
>       ...
> }
> 
> Because the genpd is initialized in the ON state with
> GENPD_FLAG_ALWAYS_ON, pm_runtime_resume_and_get() is a no-op and
> meson_ee_pwrc_on() never runs to clear HHI_MEM_PD_REG0[15:8], leaving
> HDMITX memory powered down when meson_dw_hdmi_init() and dw_hdmi_probe()
> access the controller.
> 

I will add a new patch with a function to release the memories power off
bits in v2.

> In addition, power-domains and assigned-clocks are optional in the
> amlogic,meson-dw-hdmi binding and were only added to in-tree DTs in
> v6.11, so dropping the HHI_HDMI_CLK_CNTL and HHI_MEM_PD_REG0 writes also
> regresses older DTBs (where meson_vclk_set() only configures
> HHI_HDMI_CLK_CNTL later at atomic_enable time) and builds with CONFIG_PM
> disabled.
> 

I will add a depends on PM in v2.

-- 
Best regards,
Ronald

Reply via email to