On Thu, Nov 15, 2018 at 10:28:02AM +0100, Adam Rudziński wrote: > W dniu 2018-11-15 o 10:08, Uwe Kleine-König pisze: > > On Thu, Nov 15, 2018 at 09:30:40AM +0100, Adam Rudziński wrote: > > > W dniu 2018-11-15 o 08:36, Uwe Kleine-König pisze: > > > > Hello, > > > > > > > > On Wed, Nov 14, 2018 at 11:33:25PM +0100, Adam Rudziński wrote: > > > > > returning to our discussion on incorrect handling of CTSb used as "driver > > > > > enable" by SERIAL_IMX driver in RS485 mode, I've made the fix I was able to, > > > > > ie. the simple one with just keeping the receiver of the UART on. Eventually > > > > > this got tested in action and worked fine. > > > > > > > > > > The kernel I used is a version of ancient 4.14.39, possibly with some > > > > > vendor-specific patches. I tried to compare it with the most recent > > > > > long-term 4.14.81, but there are changes which I can't interpret without > > > > > deeper studies (however, it seems to me that the original problem would > > > > > still be present). Newer kernels have even more differences. > > > > > > > > > > If such fix could be useful, and somebody feels like undertaking the effort > > > > > to implement it in a proper way, the fix can be found under this address > > > > > (along with a short description): > > > > > http://arf.net.pl/iMX6/RS485fix.html > > > > How does the relevant part of your dts look like? Did you try to use a > > > > gpio instead of the hardware function "CTS" as rts? > > > > > > > > Best regards > > > > Uwe > > > It goes like this, first in dtsi: > > > > > > &uart5 { > > > pinctrl-names = "default"; > > > pinctrl-0 = <&pinctrl_uart5>; > > > uart-has-rtscts; > > > status = "disabled"; > > > }; > > > > > > and then in dts for the board: > > > > > > &uart5 { > > > rs485-rts-active-high; > > > linux,rs485-enabled-at-boot-time; > > > status = "okay"; > > > }; > > Then try adding > > > > rts-gpios = <&gpio1 6 GPIO_ACTIVE_LOW>; > > > > to your board's dts and adapt the pinctrol accordingly to configure the > > cts pin to it's GPIO function instead of CTS. There should be no need to > > adapt the driver for that. > > The driver controls the CTS pin by setting a field in one of UART-related > registers. To me it looks like it can't be just simply "reconfigured" to use > GPIO. Do I underestimate it? No, but the driver does both, poking the UART register and drive a gpio. See https://git.kernel.org/linus/58362d5be35216f196b4a4d16aa2c6ef938087f0 Best regards Uwe -- Pengutronix e.K. | Uwe Kleine-König | Industrial Linux Solutions | http://www.pengutronix.de/ |
