Having a something_get() function call in a remove function is unusual.
A something_put() is more likely.

More over the remove() function does not match the error handling of the
probe().

Fix the remove() function to match the error handling path of the probe.

Fixes: f90142683f04 ("mmc: dw_mmc-rockchip: add runtime PM support")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
/!\   SPECULATIVE   /!\

I have a limited knowledge of the pm_ API.
However, as said, the error handling path of the probe looks more logical
to me.

Moreover, some more or less similar code can be found in
drivers/mmc/host/dw_mmc-exynos.c. This patch also align this rockchip
driver to the exynos's one.

So review with care.
---
 drivers/mmc/host/dw_mmc-rockchip.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/mmc/host/dw_mmc-rockchip.c b/drivers/mmc/host/dw_mmc-rockchip.c
index 2a99f15f527f..b5893c738b4a 100644
--- a/drivers/mmc/host/dw_mmc-rockchip.c
+++ b/drivers/mmc/host/dw_mmc-rockchip.c
@@ -373,8 +373,8 @@ static int dw_mci_rockchip_probe(struct platform_device *pdev)
 
 static int dw_mci_rockchip_remove(struct platform_device *pdev)
 {
-	pm_runtime_get_sync(&pdev->dev);
 	pm_runtime_disable(&pdev->dev);
+	pm_runtime_set_suspended(&pdev->dev);
 	pm_runtime_put_noidle(&pdev->dev);
 
 	dw_mci_pltfm_remove(pdev);
-- 
2.34.1

On Sun, 31 Jul 2022 at 19:35, Christophe JAILLET
<christophe.jaillet@wanadoo.fr> wrote:
>
> Having a something_get() function call in a remove function is unusual.
> A something_put() is more likely.
>
> More over the remove() function does not match the error handling of the
> probe().
>
> Fix the remove() function to match the error handling path of the probe.
>
> Fixes: f90142683f04 ("mmc: dw_mmc-rockchip: add runtime PM support")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> ---
> /!\   SPECULATIVE   /!\
>
> I have a limited knowledge of the pm_ API.
> However, as said, the error handling path of the probe looks more logical
> to me.
>
> Moreover, some more or less similar code can be found in
> drivers/mmc/host/dw_mmc-exynos.c. This patch also align this rockchip
> driver to the exynos's one.
>
> So review with care.
> ---
>  drivers/mmc/host/dw_mmc-rockchip.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/mmc/host/dw_mmc-rockchip.c b/drivers/mmc/host/dw_mmc-rockchip.c
> index 2a99f15f527f..b5893c738b4a 100644
> --- a/drivers/mmc/host/dw_mmc-rockchip.c
> +++ b/drivers/mmc/host/dw_mmc-rockchip.c
> @@ -373,8 +373,8 @@ static int dw_mci_rockchip_probe(struct platform_device *pdev)
>
>  static int dw_mci_rockchip_remove(struct platform_device *pdev)
>  {
> -       pm_runtime_get_sync(&pdev->dev);
>         pm_runtime_disable(&pdev->dev);
> +       pm_runtime_set_suspended(&pdev->dev);
>         pm_runtime_put_noidle(&pdev->dev);
>
>         dw_mci_pltfm_remove(pdev);

dw_mci_pltfm_remove() needs to be called with an active/powered host
device. That's why the call to pm_runtime_get_sync() is done, so I
don't think there is anything wrong with the existing code.

Kind regards
Uffe