Re: [PATCH] drm/imagination: Clear runtime PM errors while resetting the GPU

[Matt Coster]

On 24/06/2025 16:01, Alessio Belle wrote:
> The runtime PM might be left in error state if one of the callbacks
> returned an error, e.g. if the (auto)suspend callback failed following
> a firmware crash.
> 
> When that happens, any further attempt to acquire or release a power
> reference will then also fail, making it impossible to do anything else
> with the GPU. The driver logic will eventually reach the reset code.
> 
> In pvr_power_reset(), replace pvr_power_get() with a new API
> pvr_power_get_clear() which also attempts to clear any runtime PM error
> state if acquiring a power reference is not possible.
> 
> Signed-off-by: Alessio Belle <alessio.belle@xxxxxxxxxx>

Reviewed-by: Matt Coster <matt.coster@xxxxxxxxxx>

I'll apply this to drm-misc-next on Friday if nobody has any objections.

Cheers,
Matt

> ---
>  drivers/gpu/drm/imagination/pvr_power.c | 59 ++++++++++++++++++++++++++++++++-
>  1 file changed, 58 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/imagination/pvr_power.c b/drivers/gpu/drm/imagination/pvr_power.c
> index 41f5d89e78b854cf6993838868a4416a220b490a..65642ded051db83e82e32e3c3e9f82508ad8d4cc 100644
> --- a/drivers/gpu/drm/imagination/pvr_power.c
> +++ b/drivers/gpu/drm/imagination/pvr_power.c
> @@ -340,6 +340,63 @@ pvr_power_device_idle(struct device *dev)
>         return pvr_power_is_idle(pvr_dev) ? 0 : -EBUSY;
>  }
> 
> +static int
> +pvr_power_clear_error(struct pvr_device *pvr_dev)
> +{
> +       struct device *dev = from_pvr_device(pvr_dev)->dev;
> +       int err;
> +
> +       /* Ensure the device state is known and nothing is happening past this point */
> +       pm_runtime_disable(dev);
> +
> +       /* Attempt to clear the runtime PM error by setting the current state again */
> +       if (pm_runtime_status_suspended(dev))
> +               err = pm_runtime_set_suspended(dev);
> +       else
> +               err = pm_runtime_set_active(dev);
> +
> +       if (err) {
> +               drm_err(from_pvr_device(pvr_dev),
> +                       "%s: Failed to clear runtime PM error (new error %d)\n",
> +                       __func__, err);
> +       }
> +
> +       pm_runtime_enable(dev);
> +
> +       return err;
> +}
> +
> +/**
> + * pvr_power_get_clear() - Acquire a power reference, correcting any errors
> + * @pvr_dev: Device pointer
> + *
> + * Attempt to acquire a power reference on the device. If the runtime PM
> + * is in error state, attempt to clear the error and retry.
> + *
> + * Returns:
> + *  * 0 on success, or
> + *  * Any error code returned by pvr_power_get() or the runtime PM API.
> + */
> +static int
> +pvr_power_get_clear(struct pvr_device *pvr_dev)
> +{
> +       int err;
> +
> +       err = pvr_power_get(pvr_dev);
> +       if (err == 0)
> +               return err;
> +
> +       drm_warn(from_pvr_device(pvr_dev),
> +                "%s: pvr_power_get returned error %d, attempting recovery\n",
> +                __func__, err);
> +
> +       err = pvr_power_clear_error(pvr_dev);
> +       if (err)
> +               return err;
> +
> +       return pvr_power_get(pvr_dev);
> +}
> +
>  /**
>   * pvr_power_reset() - Reset the GPU
>   * @pvr_dev: Device pointer
> @@ -364,7 +421,7 @@ pvr_power_reset(struct pvr_device *pvr_dev, bool hard_reset)
>          * Take a power reference during the reset. This should prevent any interference with the
>          * power state during reset.
>          */
> -       WARN_ON(pvr_power_get(pvr_dev));
> +       WARN_ON(pvr_power_get_clear(pvr_dev));
> 
>         down_write(&pvr_dev->reset_sem);
> 
> 
> ---
> base-commit: 1a45ef022f0364186d4fb2f4e5255dcae1ff638a
> change-id: 20250619-clear-rpm-errors-gpu-reset-359ecbc85689
> 
> Best regards,
> --
> Alessio Belle <alessio.belle@xxxxxxxxxx>
> 


-- 
Matt Coster
E: matt.coster@xxxxxxxxxx

Attachment: OpenPGP_signature.asc
Description: OpenPGP digital signature