Some MediaTek SoCs got a gated UART baud clock, which currently gets
disabled as the clk subsystem believes it would be unused. This results in
the uart freezing right after "clk: Disabling unused clocks" on those
platforms.

To fix this request the "baud" clock as enabled to prevent disabling it
among with the unused clocks.
Runtime power management can probably also be restored to the state before
commit e32a83c70cf9 ("serial: 8250-mtk: modify mtk uart power and clock
management"), but that isn't strictly needed to fix the regression
introduced by that commit.

Fixes: e32a83c70cf9 ("serial: 8250-mtk: modify mtk uart power and clock management")
Suggested-by: Sam Shih <sam.shih@mediatek.com>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
---
 drivers/tty/serial/8250/8250_mtk.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_mtk.c b/drivers/tty/serial/8250/8250_mtk.c
index b44de2ed7413..9329ed1f759d 100644
--- a/drivers/tty/serial/8250/8250_mtk.c
+++ b/drivers/tty/serial/8250/8250_mtk.c
@@ -475,13 +475,13 @@ static int mtk8250_probe_of(struct platform_device *pdev, struct uart_port *p,
 	int dmacnt;
 #endif
 
-	data->uart_clk = devm_clk_get(&pdev->dev, "baud");
+	data->uart_clk = devm_clk_get_enabled(&pdev->dev, "baud");
 	if (IS_ERR(data->uart_clk)) {
 		/*
 		 * For compatibility with older device trees try unnamed
 		 * clk when no baud clk can be found.
 		 */
-		data->uart_clk = devm_clk_get(&pdev->dev, NULL);
+		data->uart_clk = devm_clk_get_enabled(&pdev->dev, NULL);
 		if (IS_ERR(data->uart_clk)) {
 			dev_warn(&pdev->dev, "Can't get uart clock\n");
 			return PTR_ERR(data->uart_clk);
-- 
2.51.0

Il 14/09/25 01:59, Daniel Golle ha scritto:
> Some MediaTek SoCs got a gated UART baud clock, which currently gets
> disabled as the clk subsystem believes it would be unused. This results in
> the uart freezing right after "clk: Disabling unused clocks" on those
> platforms.
> 
> To fix this request the "baud" clock as enabled to prevent disabling it
> among with the unused clocks.
> Runtime power management can probably also be restored to the state before
> commit e32a83c70cf9 ("serial: 8250-mtk: modify mtk uart power and clock
> management"), but that isn't strictly needed to fix the regression
> introduced by that commit.
> 
> Fixes: e32a83c70cf9 ("serial: 8250-mtk: modify mtk uart power and clock management")
> Suggested-by: Sam Shih <sam.shih@mediatek.com>
> Signed-off-by: Daniel Golle <daniel@makrotopia.org>

Is there any particular reason why you didn't en/disable the uart_clk in
runtime PM ops?

Cheers,
Angelo

> ---
>   drivers/tty/serial/8250/8250_mtk.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/tty/serial/8250/8250_mtk.c b/drivers/tty/serial/8250/8250_mtk.c
> index b44de2ed7413..9329ed1f759d 100644
> --- a/drivers/tty/serial/8250/8250_mtk.c
> +++ b/drivers/tty/serial/8250/8250_mtk.c
> @@ -475,13 +475,13 @@ static int mtk8250_probe_of(struct platform_device *pdev, struct uart_port *p,
>   	int dmacnt;
>   #endif
>   
> -	data->uart_clk = devm_clk_get(&pdev->dev, "baud");
> +	data->uart_clk = devm_clk_get_enabled(&pdev->dev, "baud");
>   	if (IS_ERR(data->uart_clk)) {
>   		/*
>   		 * For compatibility with older device trees try unnamed
>   		 * clk when no baud clk can be found.
>   		 */
> -		data->uart_clk = devm_clk_get(&pdev->dev, NULL);
> +		data->uart_clk = devm_clk_get_enabled(&pdev->dev, NULL);
>   		if (IS_ERR(data->uart_clk)) {
>   			dev_warn(&pdev->dev, "Can't get uart clock\n");
>   			return PTR_ERR(data->uart_clk);