Add CTS/RTS gpio support to STM32 UART
Subsystems (3)
v2 cover Add CTS/RTS gpio support to STM32 UART
From: Manivannan Sadhasivam <[email protected]>
Hello,
This patchset adds CTS/RTS gpio support to STM32 UART controller. Eventhough the UART controller supports using dedicated CTS/RTS gpios, sometimes we need to use different set of gpios for flow control.
This is necessary for the upcoming STM32MP1 based board called Stinger96 IoT-Box. On that board, a bluetooth chip is connected to one of the UART controller but the CTS/RTS lines got swapped mistakenly. So this patchset serves as a workaround for that hardware bug and also supports the usecase of using any gpio for CTS/RTS functionality. As per the sugggestion provided by Andy for v1, I've now switched to mctrl_gpio driver.
This patchset has been validated with Stinger96 IoT-Box connected to Murata WiFi-BT combo chip.
Thanks,
Mani
Changes in v2:
As per the review by Andy:
* Switched to mctrl_gpio driver instead of using custom CTS/RTS implementation
* Removed the use of software flow control terminology.
Manivannan Sadhasivam (2):
tty: serial: Add modem control gpio support for STM32 UART
dt-bindings: serial: Document CTS/RTS gpios in STM32 UART
.../bindings/serial/st,stm32-uart.yaml | 14 ++++++ drivers/tty/serial/Kconfig | 1 + drivers/tty/serial/stm32-usart.c | 43 ++++++++++++++++++- drivers/tty/serial/stm32-usart.h | 1 + 4 files changed, 58 insertions(+), 1 deletion(-)
Signature · 1 line
2.17.1
PATCH v2 1/2 tty: serial: Add modem control gpio support for STM32 UART
From: Manivannan Sadhasivam <[email protected]>
STM32 UART controllers have the built in modem control support using dedicated gpios which can be enabled using 'st,hw-flow-ctrl' flag in DT. But there might be cases where the board design need to use different gpios for modem control.
For supporting such cases, this commit adds modem control gpio support to STM32 UART controller using mctrl_gpio driver.
--- drivers/tty/serial/Kconfig | 1 + drivers/tty/serial/stm32-usart.c | 43 +++++++++++++++++++++++++++++++- drivers/tty/serial/stm32-usart.h | 1 + 3 files changed, 44 insertions(+), 1 deletion(-)
drivers/tty/serial/Kconfig +1 -0
@@ -1462,6 +1462,7 @@ config SERIAL_STM32 tristate "STMicroelectronics STM32 serial port support" select SERIAL_CORE depends on ARCH_STM32 || COMPILE_TEST + select SERIAL_MCTRL_GPIO if GPIOLIB help This driver is for the on-chip Serial Controller on STMicroelectronics STM32 MCUs.
drivers/tty/serial/stm32-usart.c +42 -1
84 lines not on this page · raw message
drivers/tty/serial/stm32-usart.h +1 -0
@@ -274,6 +274,7 @@ struct stm32_port {bool fifoen;int wakeirq;int rdr_mask;/* receive data register mask */ +struct mctrl_gpios *gpios;/* modem control gpios */ };static struct stm32_port stm32_ports[STM32_MAX_PORTS ];
Signature · 1 line
2.17.1
PATCH v2 2/2 dt-bindings: serial: Document CTS/RTS gpios in STM32 UART
In reply to [email protected]From: Manivannan Sadhasivam <[email protected]>
Document the use of CTS/RTS gpios for flow control in STM32 UART controller. These properties can be used instead of 'st,hw-flow-ctrl' for making use of any gpio pins for flow control instead of dedicated pins. It should be noted that both CTS/RTS and 'st,hw-flow-ctrl' properties cannot co-exist in a design.
--- .../devicetree/bindings/serial/st,stm32-uart.yaml | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
Documentation/devicetree/bindings/serial/st,stm32-uart.yaml +14 -0
@@ -48,6 +48,12 @@ properties: minItems: 1 maxItems: 2 + cts-gpios: + maxItems: 1 + + rts-gpios: + maxItems: 1 + wakeup-source: true rs485-rts-delay: true
@@ -55,6 +61,14 @@ properties: linux,rs485-enabled-at-boot-time: true rs485-rx-during-tx: true +if: + required: + - st,hw-flow-ctrl +then: + properties: + cts-gpios: false + rts-gpios: false + required: - compatible - reg
Signature · 1 line
2.17.1
Quoted 16 lines · On Thu, Apr 16, 2020 at 8:58 PM <[email protected]> wrote:
From: Manivannan Sadhasivam <[email protected]>
Hello,
This patchset adds CTS/RTS gpio support to STM32 UART controller. Eventhough the UART controller supports using dedicated CTS/RTS gpios, sometimes we need to use different set of gpios for flow control.
This is necessary for the upcoming STM32MP1 based board called Stinger96 IoT-Box. On that board, a bluetooth chip is connected to one of the UART controller but the CTS/RTS lines got swapped mistakenly. So this patchset serves as a workaround for that hardware bug and also supports the usecase of using any gpio for CTS/RTS functionality. As per the sugggestion provided by Andy for v1, I've now switched to
mctrl_gpiodriver.
Quoted 3 lines
This patchset has been validated with Stinger96 IoT-Box connected to Murata WiFi-BT combo chip.
Looks good to me,
Quoted 24 lines
Thanks,
ManiChanges in v2:
As per the review by Andy:
* Switched to
mctrl_gpiodriver instead of using custom CTS/RTS implementation
* Removed the use of software flow control terminology.Manivannan Sadhasivam (2):
tty: serial: Add modem control gpio support for STM32 UART
dt-bindings: serial: Document CTS/RTS gpios in STM32 UART.../bindings/serial/st,stm32-uart.yaml | 14 ++++++ drivers/tty/serial/Kconfig | 1 + drivers/tty/serial/stm32-usart.c | 43 ++++++++++++++++++- drivers/tty/serial/stm32-usart.h | 1 + 4 files changed, 58 insertions(+), 1 deletion(-)--
2.17.1
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 1/2] tty: serial: Add modem control gpio support for STM32 UART
In reply to [email protected]Quoted 43 lines · On 4/16/20 7:57 PM, [email protected] wrote:
From: Manivannan Sadhasivam <[email protected]>
STM32 UART controllers have the built in modem control support using dedicated gpios which can be enabled using 'st,hw-flow-ctrl' flag in DT. But there might be cases where the board design need to use different gpios for modem control.
For supporting such cases, this commit adds modem control gpio support to STM32 UART controller using
mctrl_gpiodriver.Signed-off-by Manivannan Sadhasivam <[email protected]>--- drivers/tty/serial/Kconfig | 1 + drivers/tty/serial/stm32-usart.c | 43 +++++++++++++++++++++++++++++++- drivers/tty/serial/stm32-usart.h | 1 + 3 files changed, 44 insertions(+), 1 deletion(-)drivers/tty/serial/Kconfig +1
-0@@ -1462,6 +1462,7 @@ config SERIAL_STM32 tristate "STMicroelectronics STM32 serial port support" select SERIAL_CORE depends on ARCH_STM32 || COMPILE_TEST + select SERIAL_MCTRL_GPIO if GPIOLIB help This driver is for the on-chip Serial Controller on STMicroelectronics STM32 MCUs.drivers/tty/serial/stm32-usart.c +1
-0@@ -31,6 +31,7 @@#include <linux/tty_flip.h> #include <linux/tty.h> +#include "serial_mctrl_gpio.h" #include "stm32-usart.h" static void stm32_stop_tx (struct uart_port *port);@@ -510,12 +511,29 @@ static void stm32_set_mctrl(struct uart_port *port, unsigned int mctrl)stm32_set_bits (port, ofs->cr3,USART_CR3_RTSE );
drivers/tty/serial/stm32-usart.c static void stm32_set_mctrl(struct uart_port *port, unsigned int mctrl) · line 521
Quoted 10 lines
drivers/tty/serial/stm32-usart.c +4
-0else stm32_clr_bits (port, ofs->cr3,USART_CR3_RTSE ); + +mctrl_gpio_set (stm32_port->gpios, mctrl); }static unsigned int stm32_get_mctrl (struct uart_port *port) { +struct stm32_port *stm32_port =to_stm32_port (port); +int ret;
Hi Mani,
Please find few minor remarks and a question from my side.
'ret' could be an unsigned int
drivers/tty/serial/stm32-usart.c static void stm32_set_mctrl(struct uart_port *port, unsigned int mctrl) · line 530
Quoted 10 lines
drivers/tty/serial/stm32-usart.c +8
-1+/* This routine is used to get signals of: DCD, DSR, RI, and CTS */ -return TIOCM_CAR |TIOCM_DSR |TIOCM_CTS ; + ret =TIOCM_CAR |TIOCM_DSR |TIOCM_CTS ; + +return mctrl_gpio_get (stm32_port->gpios, &ret); +} + +static void stm32_enable_ms (struct uart_port *port) +{
Just a question here: purpose of your patch is to handle the gpio case. So you may get modem control interrupts from gpios with this patch.
In other drivers, I can see the implementation checks gpio usage (like in atmel_serial). When there's no gpio, the corresponding interrupt at the serial controller level is enabled, e.g. :
if (!mctrl_gpio_to_gpiod(atmel_port->gpios, UART_GPIO_CTS)) ier |= ...
Do you need modem control interrupts in your case ?
In case the Stinger96 board signals gets fixed in a future revision, would it be needed to enable modem control interrupts in the USART controller ?
Quoted 41 lines
drivers/tty/serial/stm32-usart.c +16
-0+mctrl_gpio_enable_ms (to_stm32_port (port)->gpios); +} + +static void stm32_disable_ms (struct uart_port *port) +{ +mctrl_gpio_disable_ms (to_stm32_port (port)->gpios); }/* Transmit stop */ @@ -626,6 +644,9 @@ static void stm32_shutdown(struct uart_port *port)u32 val, isr;int ret; +/* Disable modem control interrupts */ +stm32_disable_ms (port); + val =USART_CR1_TXEIE |USART_CR1_TE ; val |= stm32_port->cr1_irq |USART_CR1_RE ; val |=BIT (cfg->uart_enable_bit);@@ -764,6 +785,12 @@ static void stm32_set_termios(struct uart_port *port, struct ktermios *termios, cr3 |=USART_CR3_CTSE |USART_CR3_RTSE ; } +/* Handle modem control interrupts */ +if (UART_ENABLE_MS (port, termios->c_cflag)) +stm32_enable_ms (port); +else +stm32_disable_ms (port); + usartdiv =DIV_ROUND_CLOSEST (port->uartclk, baud);/* @@ -898,6 +925,7 @@ static const struct uart_ops stm32_uart_ops = { .throttle = stm32_throttle, .unthrottle = stm32_unthrottle, .stop_rx = stm32_stop_rx, + .enable_ms = stm32_enable_ms, .break_ctl = stm32_break_ctl, .startup = stm32_startup, .shutdown = stm32_shutdown,
drivers/tty/serial/stm32-usart.c static int stm32_init_port(struct stm32_port *stm32port, · line 992
Quoted 1 line
drivers/tty/serial/stm32-usart.c +0
-0@@ -964,6 +992,19 @@ static int stm32_init_port(struct stm32_port *stm32port, ret = -EINVAL ;
return -EINVAL;
drivers/tty/serial/stm32-usart.c static int stm32_init_port(struct stm32_port *stm32port, · line 996
Quoted 4 lines
drivers/tty/serial/stm32-usart.c +2
-0} + stm32port->gpios =mctrl_gpio_init (&stm32port->port,0 ); +if (IS_ERR (stm32port->gpios))
Please add error path: add a clk_disable_unprepare() here, before the return.
drivers/tty/serial/stm32-usart.c static int stm32_init_port(struct stm32_port *stm32port, · line 1003
Quoted 7 lines
drivers/tty/serial/stm32-usart.c +7
-0+return PTR_ERR (stm32port->gpios); + +/* Both CTS/RTS gpios and "st,hw-flow-ctrl" should not be specified */ +if (stm32port->hw_flow_control) { +if (mctrl_gpio_to_gpiod (stm32port->gpios,UART_GPIO_CTS ) || +mctrl_gpio_to_gpiod (stm32port->gpios,UART_GPIO_RTS )) { +dev_err (&pdev->dev,"Conflicting RTS/CTS config\n" );
same here
Best Regards,
Thanks,
Fabrice
Quoted 20 lines
drivers/tty/serial/stm32-usart.c +4
-0+return -EINVAL ; + } + } +return ret; }drivers/tty/serial/stm32-usart.h +1
-0@@ -274,6 +274,7 @@ struct stm32_port {bool fifoen;int wakeirq;int rdr_mask;/* receive data register mask */ +struct mctrl_gpios *gpios;/* modem control gpios */ };static struct stm32_port stm32_ports[STM32_MAX_PORTS ];
Re: [PATCH v2 1/2] tty: serial: Add modem control gpio support for STM32 UART
Hi,
Quoted 57 lines · Fabrice Gasnier
Quoted 43 lines · On 4/16/20 7:57 PM, [email protected] wrote:
From: Manivannan Sadhasivam <[email protected]>
STM32 UART controllers have the built in modem control support using dedicated gpios which can be enabled using 'st,hw-flow-ctrl' flag in DT. But there might be cases where the board design need to use different gpios for modem control.
For supporting such cases, this commit adds modem control gpio support to STM32 UART controller using
mctrl_gpiodriver.Signed-off-by Manivannan Sadhasivam <[email protected]>--- drivers/tty/serial/Kconfig | 1 + drivers/tty/serial/stm32-usart.c | 43 +++++++++++++++++++++++++++++++- drivers/tty/serial/stm32-usart.h | 1 + 3 files changed, 44 insertions(+), 1 deletion(-)drivers/tty/serial/Kconfig +1
-0@@ -1462,6 +1462,7 @@ config SERIAL_STM32 tristate "STMicroelectronics STM32 serial port support" select SERIAL_CORE depends on ARCH_STM32 || COMPILE_TEST + select SERIAL_MCTRL_GPIO if GPIOLIB help This driver is for the on-chip Serial Controller on STMicroelectronics STM32 MCUs.drivers/tty/serial/stm32-usart.c +1
-0@@ -31,6 +31,7 @@#include <linux/tty_flip.h> #include <linux/tty.h> +#include "serial_mctrl_gpio.h" #include "stm32-usart.h" static void stm32_stop_tx (struct uart_port *port);@@ -510,12 +511,29 @@ static void stm32_set_mctrl(struct uart_port *port, unsigned int mctrl)stm32_set_bits (port, ofs->cr3,USART_CR3_RTSE );drivers/tty/serial/stm32-usart.c static void stm32_set_mctrl(struct uart_port *port, unsigned int mctrl) · line 521
Quoted 10 lines
drivers/tty/serial/stm32-usart.c +4
-0else stm32_clr_bits (port, ofs->cr3,USART_CR3_RTSE ); + +mctrl_gpio_set (stm32_port->gpios, mctrl); }static unsigned int stm32_get_mctrl (struct uart_port *port) { +struct stm32_port *stm32_port =to_stm32_port (port); +int ret;Hi Mani,
Quoted 4 lines
Please find few minor remarks and a question from my side.
'ret' could be an unsigned int
Ok
Quoted 5 lines
Quoted 5 lines
drivers/tty/serial/stm32-usart.c +3
-1+/* This routine is used to get signals of: DCD, DSR, RI, and CTS */ -return TIOCM_CAR |TIOCM_DSR |TIOCM_CTS ; + ret =TIOCM_CAR |TIOCM_DSR |TIOCM_CTS ; +
Quoted 9 lines
drivers/tty/serial/stm32-usart.c static void stm32_set_mctrl(struct uart_port *port, unsigned int mctrl) · line 530
Quoted 5 lines
drivers/tty/serial/stm32-usart.c +5
-0+return mctrl_gpio_get (stm32_port->gpios, &ret); +} + +static void stm32_enable_ms (struct uart_port *port) +{Just a question here: purpose of your patch is to handle the gpio case. So you may get modem control interrupts from gpios with this patch.
We will only get CTS gpio interrupt which will get handled by mctrl_gpio driver.
Quoted 7 lines
In other drivers, I can see the implementation checks gpio usage (like in
atmel_serial). When there's no gpio, the corresponding interrupt at the serial controller level is enabled, e.g. :if (!mctrl_gpio_to_gpiod(atmel_port->gpios, UART_GPIO_CTS)) ier |= ...
Those drivers support both gpio and hardware modem control interrupts, so they are checking to make sure either one of them is enabled at a time. But on STM32 UART, modem interrupts are not enabled (USART_CR3_CTSIE etc...)
Quoted 2 lines
Do you need modem control interrupts in your case ?
Nope. Since modem control interrupts in UART block are not enabled currently, I don't bother about it.
Quoted 4 lines
In case the Stinger96 board signals gets fixed in a future revision, would it be needed to enable modem control interrupts in the USART controller ?
I don't think so. The current driver supports hardware flow control without using interrupts, so we are fine.
Quoted 45 lines
Quoted 41 lines
drivers/tty/serial/stm32-usart.c +16
-0+mctrl_gpio_enable_ms (to_stm32_port (port)->gpios); +} + +static void stm32_disable_ms (struct uart_port *port) +{ +mctrl_gpio_disable_ms (to_stm32_port (port)->gpios); }/* Transmit stop */ @@ -626,6 +644,9 @@ static void stm32_shutdown(struct uart_port *port)u32 val, isr;int ret; +/* Disable modem control interrupts */ +stm32_disable_ms (port); + val =USART_CR1_TXEIE |USART_CR1_TE ; val |= stm32_port->cr1_irq |USART_CR1_RE ; val |=BIT (cfg->uart_enable_bit);@@ -764,6 +785,12 @@ static void stm32_set_termios(struct uart_port *port, struct ktermios *termios, cr3 |=USART_CR3_CTSE |USART_CR3_RTSE ; } +/* Handle modem control interrupts */ +if (UART_ENABLE_MS (port, termios->c_cflag)) +stm32_enable_ms (port); +else +stm32_disable_ms (port); + usartdiv =DIV_ROUND_CLOSEST (port->uartclk, baud);/* @@ -898,6 +925,7 @@ static const struct uart_ops stm32_uart_ops = { .throttle = stm32_throttle, .unthrottle = stm32_unthrottle, .stop_rx = stm32_stop_rx, + .enable_ms = stm32_enable_ms, .break_ctl = stm32_break_ctl, .startup = stm32_startup, .shutdown = stm32_shutdown,drivers/tty/serial/stm32-usart.c static int stm32_init_port(struct stm32_port *stm32port, · line 992
Quoted 1 line
drivers/tty/serial/stm32-usart.c +0
-0@@ -964,6 +992,19 @@ static int stm32_init_port(struct stm32_port *stm32port, ret = -EINVAL ;return
-EINVAL;
Quoted 8 lines
drivers/tty/serial/stm32-usart.c static int stm32_init_port(struct stm32_port *stm32port, · line 996
Quoted 4 lines
drivers/tty/serial/stm32-usart.c +2
-0} + stm32port->gpios =mctrl_gpio_init (&stm32port->port,0 ); +if (IS_ERR (stm32port->gpios))Please add error path: add a
clk_disable_unprepare()here, before the return.
Ok.
Thanks,
Mani
Quoted 34 lines
drivers/tty/serial/stm32-usart.c static int stm32_init_port(struct stm32_port *stm32port, · line 1003
Quoted 7 lines
drivers/tty/serial/stm32-usart.c +7
-0+return PTR_ERR (stm32port->gpios); + +/* Both CTS/RTS gpios and "st,hw-flow-ctrl" should not be specified */ +if (stm32port->hw_flow_control) { +if (mctrl_gpio_to_gpiod (stm32port->gpios,UART_GPIO_CTS ) || +mctrl_gpio_to_gpiod (stm32port->gpios,UART_GPIO_RTS )) { +dev_err (&pdev->dev,"Conflicting RTS/CTS config\n" );same here
Best Regards,
Thanks,
FabriceQuoted 20 lines
drivers/tty/serial/stm32-usart.c +4
-0+return -EINVAL ; + } + } +return ret; }drivers/tty/serial/stm32-usart.h +1
-0@@ -274,6 +274,7 @@ struct stm32_port {bool fifoen;int wakeirq;int rdr_mask;/* receive data register mask */ +struct mctrl_gpios *gpios;/* modem control gpios */ };static struct stm32_port stm32_ports[STM32_MAX_PORTS ];