[PATCH] tty: serial: cpm_uart: Fix behaviour for non existing GPIOs

[PATCH] tty: serial: cpm_uart: Fix behaviour for non existing GPIOs

[Christophe Leroy]

devm_gpiod_get_index() doesn't return NULL but -ENOENT when the
requested GPIO doesn't exist,  leading to the following messages:

[    2.742468] gpiod_direction_input: invalid GPIO (errorpointer)
[    2.748147] can't set direction for gpio #2: -2
[    2.753081] gpiod_direction_input: invalid GPIO (errorpointer)
[    2.758724] can't set direction for gpio #3: -2
[    2.763666] gpiod_direction_output: invalid GPIO (errorpointer)
[    2.769394] can't set direction for gpio #4: -2
[    2.774341] gpiod_direction_input: invalid GPIO (errorpointer)
[    2.779981] can't set direction for gpio #5: -2
[    2.784545] ff000a20.serial: ttyCPM1 at MMIO 0xfff00a20 (irq = 39, base_baud = 8250000) is a CPM UART

Use IS_ERR_OR_NULL() to properly check gpiod validity.

Fixes: 97cbaf2c829b ("tty: serial: cpm_uart: Convert to use GPIO descriptors")
Cc: stable@vger.kernel.org
Cc: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
---
 drivers/tty/serial/cpm_uart/cpm_uart_core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/tty/serial/cpm_uart/cpm_uart_core.c b/drivers/tty/serial/cpm_uart/cpm_uart_core.c
index a04f74d2e854..3cbe24802296 100644
--- a/drivers/tty/serial/cpm_uart/cpm_uart_core.c
+++ b/drivers/tty/serial/cpm_uart/cpm_uart_core.c
@@ -1217,7 +1217,7 @@ static int cpm_uart_init_port(struct device_node *np,
 
 		gpiod = devm_gpiod_get_index(dev, NULL, i, GPIOD_ASIS);
 
-		if (gpiod) {
+		if (!IS_ERR_OR_NULL(gpiod)) {
 			if (i == GPIO_RTS || i == GPIO_DTR)
 				ret = gpiod_direction_output(gpiod, 0);
 			else
-- 
2.25.0

Re: [PATCH] tty: serial: cpm_uart: Fix behaviour for non existing GPIOs

[Johan Hovold]

On Sat, Jun 06, 2020 at 07:30:21AM +0000, Christophe Leroy wrote:
> devm_gpiod_get_index() doesn't return NULL but -ENOENT when the
> requested GPIO doesn't exist,  leading to the following messages:
> 
> [    2.742468] gpiod_direction_input: invalid GPIO (errorpointer)
> [    2.748147] can't set direction for gpio #2: -2
> [    2.753081] gpiod_direction_input: invalid GPIO (errorpointer)
> [    2.758724] can't set direction for gpio #3: -2
> [    2.763666] gpiod_direction_output: invalid GPIO (errorpointer)
> [    2.769394] can't set direction for gpio #4: -2
> [    2.774341] gpiod_direction_input: invalid GPIO (errorpointer)
> [    2.779981] can't set direction for gpio #5: -2
> [    2.784545] ff000a20.serial: ttyCPM1 at MMIO 0xfff00a20 (irq = 39, base_baud = 8250000) is a CPM UART
> 
> Use IS_ERR_OR_NULL() to properly check gpiod validity.

Why check for NULL at all?

> Fixes: 97cbaf2c829b ("tty: serial: cpm_uart: Convert to use GPIO descriptors")
> Cc: stable@vger.kernel.org
> Cc: Linus Walleij <linus.walleij@linaro.org>
> Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
> ---
>  drivers/tty/serial/cpm_uart/cpm_uart_core.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/tty/serial/cpm_uart/cpm_uart_core.c b/drivers/tty/serial/cpm_uart/cpm_uart_core.c
> index a04f74d2e854..3cbe24802296 100644
> --- a/drivers/tty/serial/cpm_uart/cpm_uart_core.c
> +++ b/drivers/tty/serial/cpm_uart/cpm_uart_core.c
> @@ -1217,7 +1217,7 @@ static int cpm_uart_init_port(struct device_node *np,
>  
>  		gpiod = devm_gpiod_get_index(dev, NULL, i, GPIOD_ASIS);
>  
> -		if (gpiod) {
> +		if (!IS_ERR_OR_NULL(gpiod)) {
>  			if (i == GPIO_RTS || i == GPIO_DTR)
>  				ret = gpiod_direction_output(gpiod, 0);
>  			else

Johan

Re: [PATCH] tty: serial: cpm_uart: Fix behaviour for non existing GPIOs

[Linus Walleij]

Hi Christophe!

On Sat, Jun 6, 2020 at 9:30 AM Christophe Leroy
<christophe.leroy@csgroup.eu> wrote:


>                 gpiod = devm_gpiod_get_index(dev, NULL, i, GPIOD_ASIS);
>
> -               if (gpiod) {
> +               if (!IS_ERR_OR_NULL(gpiod)) {
>                         if (i == GPIO_RTS || i == GPIO_DTR)
>                                 ret = gpiod_direction_output(gpiod, 0);
>                         else

This code, and the way descriptors are used in the driver leads
me to believe that the right solution is to use the optional
call with a hard error check:

gpiod = devm_gpiod_get_index_optional(...);

if (IS_ERR(gpiod))
    return PTR_ERR(gpiod);

if (gpiod) {
... followed by the old code ...

This makes sure that the array member is left NULL if there is no
GPIO on this line, and all other errors, such as -EPROBE_DEFER
which currently absolutely does not work, will lead to us properly
exiting with an error.

Yours,
Linus Walleij