serial: imx: half-duplex RS485 operation with RTS active low

[Stefan Agner]

On 16.04.2018 15:42, Uwe Kleine-König wrote:
> On Mon, Apr 16, 2018 at 03:01:44PM +0200, Stefan Agner wrote:
>> On 16.04.2018 12:29, Stefan Agner wrote:
>> > On 16.04.2018 11:22, Uwe Kleine-König wrote:
>> >> Hi Stefan,
>> >>
>> >> On Mon, Apr 16, 2018 at 11:14:32AM +0200, Stefan Agner wrote:
>> >>> Using upstream I noticed that RS-485 does not work in the default
>> >>> configuration for our platforms (Toradex Apalis/Colibri). Closer
>> >>
>> >> This is an i.MX6?
>> >>
>> >
>> > Yes, sorry about that.
>> >
>> > But I think it is i.MX UART specific, I noticed the same behavior on
>> > i.MX 7 too.
>> >
>> >>> debugging shows that it is related to "serial: imx: default to half
>> >>> duplex rs485".
>> >>>
>> >>> We use the i.MX UART in DTE mode and control the RS-485 transceiver
>> >>> using the RTS signal in low-active mode.
>> >>>
>> >>> 	uart-has-rtscts;
>> >>> 	fsl,dte-mode;
>> >>> 	linux,rs485-enabled-at-boot-time;
>> >>> 	rs485-rts-active-low;
>> >>
>> >> That means you're not using a GPIO for RTS signaling, right?
>> >>
>> >
>> > I use native RTS capability (which is in DTE mode the CTS signal...)
>> >
>> >>> Using this setting leads to the RTS signal not getting asserted (the
>> >>> oscilloscope only shows a very short fluke before the start bit is
>> >>> sent).
>> >
>> > Just FYI, the fluke looks like a proper assert, but it is really only
>> > 50ns wide.
>> >
>>
>> Also tried DCE mode, the same behavior.
>>
>> Two screenshots showing TX/RTS (CTS_B) in different settings:
>> https://imgur.com/a/PbUex
>>
>> The little fluke disappears when disabling RX before setting CTSC:
>>
>> --- a/drivers/tty/serial/imx.c
>> +++ b/drivers/tty/serial/imx.c
>> @@ -657,6 +657,9 @@ static void imx_uart_start_tx(struct uart_port
>> *port)
>>         if (port->rs485.flags & SER_RS485_ENABLED) {
>>                 u32 ucr2;
>>
>> +               if (!(port->rs485.flags & SER_RS485_RX_DURING_TX))
>> +                       imx_uart_stop_rx(port);
>> +
>>                 ucr2 = imx_uart_readl(sport, UCR2);
>>                 if (port->rs485.flags & SER_RS485_RTS_ON_SEND)
>>                         imx_uart_rts_active(sport, &ucr2);
>> @@ -664,9 +667,6 @@ static void imx_uart_start_tx(struct uart_port
>> *port)
>>                         imx_uart_rts_inactive(sport, &ucr2);
>>                 imx_uart_writel(sport, ucr2, UCR2);
>>
>> -               if (!(port->rs485.flags & SER_RS485_RX_DURING_TX))
>> -                       imx_uart_stop_rx(port);
>> -
>>                 /*
>>                  * Enable transmitter and shifter empty irq only if DMA
>> is off.
>>                  * In the DMA case this is done in the tx-callback.
>>
>> It seems that if the RX path is disabled, CTS_B is no longer
>> controllable. It just stays high. That is not a problem in the high
>> active RTS case... However, it breaks low active half-duplex...
>>
>> It seems that this thread is describing this situation:
>> https://community.nxp.com/thread/385047
> 
> So better use a gpio instead of the hardware-function?

I guess so, did not test that though.

Or use full duplex mode. That is what we used anyways in the past, even
when using a single RS485 bus (half duplex) on the physical layer
anyway.

> 
> Adding a respective check in the driver would be nice though.

Hm, yes agreed. Where would that check go? In probe? Then I guess it
only would warn when using device tree properties...?

--
Stefan