Re: [PATCH v6 1/7] serial: Do not hold the port lock when setting rx-during-tx GPIO

[Lino Sanfilippo]

Hi,

On 25.12.23 at 13:31, Maarten Brock wrote:
> Lino Sanfilippo wrote on 2023-12-25 12:35:
>> diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
>> index f1348a509552..d155131f221d 100644
>> --- a/drivers/tty/serial/serial_core.c
>> +++ b/drivers/tty/serial/serial_core.c
>> @@ -1402,6 +1402,16 @@ static void uart_set_rs485_termination(struct
>> uart_port *port,
>>                   !!(rs485->flags & SER_RS485_TERMINATE_BUS));
>>  }
>>
>> +static void uart_set_rs485_rx_during_tx(struct uart_port *port,
>> +                    const struct serial_rs485 *rs485)
>> +{
>> +    if (!(rs485->flags & SER_RS485_ENABLED))
>> +        return;
>> +
>
> How about checking port->rs485_rx_during_tx_gpio here against NULL instead of
> before every call?
>

gpiod_set_value_cansleep() already checks for a NULL pointer, so doing this check
in the caller is not needed.

>> +    gpiod_set_value_cansleep(port->rs485_rx_during_tx_gpio,
>> +                 !!(rs485->flags & SER_RS485_RX_DURING_TX));
>> +}
>> +
>>  static int uart_rs485_config(struct uart_port *port)
>>  {
>>      struct serial_rs485 *rs485 = &port->rs485;
>> @@ -1413,12 +1423,17 @@ static int uart_rs485_config(struct uart_port *port)
>>
>>      uart_sanitize_serial_rs485(port, rs485);
>>      uart_set_rs485_termination(port, rs485);
>> +    uart_set_rs485_rx_during_tx(port, rs485);
>>
>>      uart_port_lock_irqsave(port, &flags);
>>      ret = port->rs485_config(port, NULL, rs485);
>>      uart_port_unlock_irqrestore(port, flags);
>> -    if (ret)
>> +    if (ret) {
>>          memset(rs485, 0, sizeof(*rs485));
>> +        /* unset GPIOs */
>> +        gpiod_set_value_cansleep(port->rs485_term_gpio, 0);
>> +        gpiod_set_value_cansleep(port->rs485_rx_during_tx_gpio, 0);
>> +    }
>>
>>      return ret;
>>  }
>> @@ -1457,6 +1472,7 @@ static int uart_set_rs485_config(struct
>> tty_struct *tty, struct uart_port *port,
>>          return ret;
>>      uart_sanitize_serial_rs485(port, &rs485);
>>      uart_set_rs485_termination(port, &rs485);
>> +    uart_set_rs485_rx_during_tx(port, &rs485);
>>
>>      uart_port_lock_irqsave(port, &flags);
>>      ret = port->rs485_config(port, &tty->termios, &rs485);
>> @@ -1468,8 +1484,14 @@ static int uart_set_rs485_config(struct
>> tty_struct *tty, struct uart_port *port,
>>              port->ops->set_mctrl(port, port->mctrl);
>>      }
>>      uart_port_unlock_irqrestore(port, flags);
>> -    if (ret)
>> +    if (ret) {
>> +        /* restore old GPIO settings */
>> +        gpiod_set_value_cansleep(port->rs485_term_gpio,
>> +            !!(port->rs485.flags & SER_RS485_TERMINATE_BUS));
>> +        gpiod_set_value_cansleep(port->rs485_rx_during_tx_gpio,
>> +            !!(port->rs485.flags & SER_RS485_RX_DURING_TX));
>
> This does not look like restoring.


Hmm. The rx-during-tx and terminate-bus GPIOs may have changed before the
drivers rs485_config() was called. If that function fails, the GPIOs
are set back to the values they had before (i.e what is still stored in
the ports serial_rs485 struct). So what is wrong with the term "restore"?

> Further this looks suspiciously like duplicated code

Since the added code consists of two one-liners I am not sure how to
decrease code duplication in this case. We could introduce wrapper functions (the only
ones we have so far to set the GPIOs are uart_set_rs485_termination() and
uart_set_rs485_rx_during_tx() which cannot be used here due to the initial
check for SER_RS485_ENABLED). But would that really help?


>
>>          return ret;
>> +    }
>>
>>      if (copy_to_user(rs485_user, &port->rs485, sizeof(port->rs485)))
>>          return -EFAULT;
>> diff --git a/drivers/tty/serial/stm32-usart.c b/drivers/tty/serial/stm32-usart.c
>> index 3048620315d6..ec9a72a5bea9 100644
>> --- a/drivers/tty/serial/stm32-usart.c
>> +++ b/drivers/tty/serial/stm32-usart.c
>> @@ -226,10 +226,7 @@ static int stm32_usart_config_rs485(struct
>> uart_port *port, struct ktermios *ter
>>
>>      stm32_usart_clr_bits(port, ofs->cr1, BIT(cfg->uart_enable_bit));
>>
>> -    if (port->rs485_rx_during_tx_gpio)
>> -        gpiod_set_value_cansleep(port->rs485_rx_during_tx_gpio,
>> -                     !!(rs485conf->flags & SER_RS485_RX_DURING_TX));
>> -    else
>> +    if (!port->rs485_rx_during_tx_gpio)
>
> Should the ! be there?
>

Thats a good point, the "else" seems indeed to be wrong. It has been introduced
with the code that added the GPIO support (c54d48543689 "serial: stm32: Add support for rs485 RX_DURING_TX output GPIO")

I will fix it in the next version of this patch, thanks.


>>          rs485conf->flags |= SER_RS485_RX_DURING_TX;
>>
>>      if (rs485conf->flags & SER_RS485_ENABLED) {
>
> Kind Regards
> Maarten Brock
>

Thanks a lot for the review.

BR,
Lino