ARM/serial: at91: switch atmel serial to use gpiolib
Subsystems (7)
PATCH 1/1 ARM/serial: at91: switch atmel serial to use gpiolib
This passes the errata fix using a GPIO to control the RTS pin on one of the AT91 chips to use gpiolib instead of the AT91-specific interfaces. Also remove the reliance on compile-time #defines and the cpu_* check and rely on the platform passing down the proper GPIO pin through platform data.
This is a prerequisite for getting rid of the local GPIO implementation in the AT91 platform and move toward multiplatform.
This also makes way for device tree conversion: the RTS GPIO pin can be passed by standard GPIO bindings.
---
This is an alternative to the patch entitled "ARM/serial: at91: move machine quirk into machine" which needs testing to confirm this approach. Seeking ACKs on this if the approach seems OK to all parties.
--- arch/arm/mach-at91/at91rm9200_devices.c | 1 + drivers/tty/serial/atmel_serial.c | 51 +++++++++++++++++++++------------ include/linux/platform_data/atmel.h | 1 + 3 files changed, 35 insertions(+), 18 deletions(-)
arch/arm/mach-at91/at91rm9200_devices.c +1 -0
@@ -961,6 +961,7 @@ static struct resource uart0_resources[] = {static struct atmel_uart_data uart0_data = { .use_dma_tx =1 , .use_dma_rx =1 , + .rts_gpio =AT91_PIN_PA21 , };static u64 uart0_dmamask =DMA_BIT_MASK (32 );
drivers/tty/serial/atmel_serial.c +33 -18
80 lines not on this page · raw message
include/linux/platform_data/atmel.h +1 -0
@@ -84,6 +84,7 @@ struct atmel_uart_data {short use_dma_rx;/* use receive DMA? */ void __iomem *regs;/* virt. base address, if any */ struct serial_rs485 rs485;/* rs485 settings */ +int rts_gpio;/* optional RTS GPIO */ };/* Touchscreen Controller */
Signature · 1 line
1.8.3.1
PATCH 1/1 ARM/serial: at91: switch atmel serial to use gpiolib
On 05/11/2013 13:35, Linus Walleij :
Quoted 80 lines
This passes the errata fix using a GPIO to control the RTS pin on one of the AT91 chips to use gpiolib instead of the AT91-specific interfaces. Also remove the reliance on compile-time #defines and the cpu_* check and rely on the platform passing down the proper GPIO pin through platform data.
This is a prerequisite for getting rid of the local GPIO implementation in the AT91 platform and move toward multiplatform.
This also makes way for device tree conversion: the RTS GPIO pin can be passed by standard GPIO bindings.
Signed-off-by Linus Walleij <[email protected]>---
This is an alternative to the patch entitled "ARM/serial: at91: move machine quirk into machine" which needs testing to confirm this approach. Seeking ACKs on this if the approach seems OK to all parties.--- arch/arm/mach-at91/at91rm9200_devices.c | 1 + drivers/tty/serial/atmel_serial.c | 51 +++++++++++++++++++++------------ include/linux/platform_data/atmel.h | 1 + 3 files changed, 35 insertions(+), 18 deletions(-)arch/arm/mach-at91/at91rm9200_devices.c +1
-0@@ -961,6 +961,7 @@ static struct resource uart0_resources[] = {static struct atmel_uart_data uart0_data = { .use_dma_tx =1 , .use_dma_rx =1 , + .rts_gpio =AT91_PIN_PA21 , };static u64 uart0_dmamask =DMA_BIT_MASK (32 );drivers/tty/serial/atmel_serial.c +2
-13@@ -41,15 +41,11 @@#include <linux/uaccess.h> #include <linux/platform_data/atmel.h> #include <linux/timer.h> +#include <linux/gpio.h> #include <asm/io.h> #include <asm/ioctls.h> -#ifdef CONFIG_ARM -#include <mach/cpu.h> -#include <asm/gpio.h> -#endif -#define PDC_BUFFER_SIZE 512/* Revisit: We should calculate this based on the actual port settings */ #define PDC_RX_TIMEOUT (3 * 10)/* 3 bytes */ @@ -167,6 +163,7 @@ struct atmel_uart_port {struct circ_buf rx_ring;struct serial_rs485 rs485;/* rs485 settings */ +int rts_gpio;/* optional RTS GPIO */ unsigned int tx_done_mask;bool is_usart;/* usart or uart */ struct timer_list uart_timer;/* uart timer */ @@ -300,20 +297,17 @@ static void atmel_set_mctrl(struct uart_port *port, u_int mctrl)unsigned int mode;struct atmel_uart_port *atmel_port =to_atmel_uart_port (port); -#ifdef CONFIG_ARCH_AT91RM9200 -if (cpu_is_at91rm9200 ()) { -/* -* AT91RM9200 Errata #39: RTS0 is not internally connected -* to PA21. We need to drive the pin manually. -*/ -if (port->mapbase ==AT91RM9200_BASE_US0 ) { -if (mctrl &TIOCM_RTS )
drivers/tty/serial/atmel_serial.c static void atmel_set_mctrl(struct uart_port *port, u_int mctrl) · line 305
Quoted 10 lines
drivers/tty/serial/atmel_serial.c +6
-4-at91_set_gpio_value (AT91_PIN_PA21 ,0 ); - else - at91_set_gpio_value(AT91_PIN_PA21, 1); - } +/* +* AT91RM9200 Errata #39: RTS0 is not internally connected +* to PA21. We need to drive the pin as a GPIO. +*/ +if (gpio_is_valid (atmel_port->rts_gpio) && + port->mapbase ==AT91RM9200_BASE_US0 ) {
Okay. Let's keep it like this for the moment. If I want to generalize this "feature" to other SoCs, I will remove the second test.
Quoted 10 lines
drivers/tty/serial/atmel_serial.c +4
-1+if (mctrl &TIOCM_RTS ) +gpio_set_value (atmel_port->rts_gpio,0 ); +else +gpio_set_value (atmel_port->rts_gpio,1 ); } -#endif if (mctrl &TIOCM_RTS ) control |=ATMEL_US_RTSEN ;
drivers/tty/serial/atmel_serial.c static int atmel_serial_probe(struct platform_device *pdev) · line 2362
Quoted 4 lines
drivers/tty/serial/atmel_serial.c +1
-0@@ -2365,6 +2359,27 @@ static int atmel_serial_probe(struct platform_device *pdev) port = &atmel_ports[ret]; port->backup_imr =0 ; port->uart.line = ret; + port->rts_gpio =-1 ;/* Invalid, zero could be valid */
What about -EINVAL?
drivers/tty/serial/atmel_serial.c static int atmel_serial_probe(struct platform_device *pdev) · line 2370
Quoted 8 lines
drivers/tty/serial/atmel_serial.c +8
-0+/* +* In theory the GPIO pin controlling RTS could be zero and +* this would be an improper check, but we know that the only +* existing case is != 0 and it's nice to use the zero-initialized +* structs to indicate "no RTS GPIO" instead of open-coding some +* invalid value everywhere. +*/ +if (pdata->rts_gpio >0 ) {
Okay, let's do the gpio validity checking like this. It is preventing us from big headache.
But, unfortunately, on DT-enabled machines we do not have a pdata structure. You would have to change the test to:
+ if (pdata && pdata->rts_gpio > 0) {
But anyway, I send an DT support for this feature as a follow-up of this patch. You can squash both patches if you want.
Quoted 28 lines
drivers/tty/serial/atmel_serial.c +12
-0+ ret =devm_gpio_request (&pdev->dev, pdata->rts_gpio,"RTS" ); +if (ret) { +dev_err (&pdev->dev,"error requesting RTS GPIO\n" ); + goto err; + } + port->rts_gpio = pdata->rts_gpio; + ret =gpio_direction_output (port->rts_gpio,0 ); +if (ret) { +dev_err (&pdev->dev,"error setting up RTS GPIO\n" ); + goto err; + } + } ret =atmel_init_port (port, pdev);if (ret)include/linux/platform_data/atmel.h +1
-0@@ -84,6 +84,7 @@ struct atmel_uart_data {short use_dma_rx;/* use receive DMA? */ void __iomem *regs;/* virt. base address, if any */ struct serial_rs485 rs485;/* rs485 settings */ +int rts_gpio;/* optional RTS GPIO */ };/* Touchscreen Controller */
Signature · 1 line
Nicolas Ferre
PATCH 1/1 ARM/serial: at91: specify RTS in DT using gpio
In reply to Linus Walleij--- Documentation/devicetree/bindings/serial/atmel-usart.txt | 3 +++ drivers/tty/serial/atmel_serial.c | 13 ++++++++++--- 2 files changed, 13 insertions(+), 3 deletions(-)
Documentation/devicetree/bindings/serial/atmel-usart.txt +3 -0
@@ -10,6 +10,8 @@ Required properties: Optional properties: - atmel,use-dma-rx: use of PDC or DMA for receiving data - atmel,use-dma-tx: use of PDC or DMA for transmitting data +- rts-gpios: specify a GPIO for RTS line. It will use specified PIO instead of the peripheral + function pin for the USART RTS feature. If unsure, don't specify this property. - add dma bindings for dma transfer: - dmas: DMA specifier, consisting of a phandle to DMA controller node, memory peripheral interface and USART DMA channel ID, FIFO configuration.
@@ -28,6 +30,7 @@ Example: interrupts = <7>; atmel,use-dma-rx; atmel,use-dma-tx; + rts-gpios = <&pioD 15 0>; }; - use DMA:
drivers/tty/serial/atmel_serial.c +10 -3
@@ -35,6 +35,7 @@#include <linux/platform_device.h> #include <linux/of.h> #include <linux/of_device.h> +#include <linux/of_gpio.h> #include <linux/dma-mapping.h> #include <linux/atmel_pdc.h> #include <linux/atmel_serial.h>
@@ -2327,6 +2328,7 @@ static int atmel_serial_probe(struct platform_device *pdev)struct device_node *np = pdev->dev.of_node;struct atmel_uart_data *pdata =dev_get_platdata (&pdev->dev);void *data; +int rts_pin = -EINVAL ;int ret = -ENODEV ;BUILD_BUG_ON (ATMEL_SERIAL_RINGSIZE & (ATMEL_SERIAL_RINGSIZE -1 ));
@@ -2364,13 +2366,18 @@ static int atmel_serial_probe(struct platform_device *pdev)* structs to indicate "no RTS GPIO" instead of open-coding some * invalid value everywhere. */ -if (pdata->rts_gpio >0 ) { - ret =devm_gpio_request (&pdev->dev, pdata->rts_gpio,"RTS" ); +if (pdata && pdata->rts_gpio >0 ) + rts_pin = pdata->rts_gpio; +else if (np) + rts_pin =of_get_named_gpio (np,"rts-gpios" ,0 ); + +if (gpio_is_valid (rts_pin)) { + ret =devm_gpio_request (&pdev->dev, rts_pin,"RTS" );if (ret) {dev_err (&pdev->dev,"error requesting RTS GPIO\n" ); goto err; } - port->rts_gpio = pdata->rts_gpio; + port->rts_gpio = rts_pin; ret =gpio_direction_output (port->rts_gpio,0 );if (ret) {dev_err (&pdev->dev,"error setting up RTS GPIO\n" );
Signature · 1 line
1.8.2.2
PATCH 1/1 ARM/serial: at91: specify RTS in DT using gpio
Quoted 20 lines · Nicolas Ferre
Signed-off-by Nicolas Ferre <[email protected]>---
Documentation/devicetree/bindings/serial/atmel-usart.txt| 3 +++drivers/tty/serial/atmel_serial.c| 13 ++++++++++---2 files changed, 13 insertions(+), 3 deletions(-)Documentation/devicetree/bindings/serial/atmel-usart.txt +0
-0@@ -10,6 +10,8 @@ Required properties:Optional properties:
- atmel,use-dma-rx: use of PDC or DMA for receiving data
- atmel,use-dma-tx: use of PDC or DMA for transmitting data+- rts-gpios: specify a GPIO for RTS line. It will use specified PIO instead of the peripheral + function pin for the USART RTS feature. If unsure, don't specify this property.- add dma bindings for dma transfer:
- dmas: DMA specifier, consisting of a phandle to DMA controller node, memory peripheral interface and USART DMA channel ID, FIFO configuration.
Documentation/devicetree/bindings/serial/atmel-usart.txt Example: · line 34
Quoted 5 lines
Documentation/devicetree/bindings/serial/atmel-usart.txt +1
-0@@ -28,6 +30,7 @@ Example: interrupts = <7>; atmel,use-dma-rx; atmel,use-dma-tx; + rts-gpios = <&pioD 15 0>; };
gpios means 2 gpio
I?ll rts-gpio as we can only use one anyway
Quoted 23 lines
Documentation/devicetree/bindings/serial/atmel-usart.txt +0
-1- use DMA:drivers/tty/serial/atmel_serial.c +2
-0@@ -35,6 +35,7 @@#include <linux/platform_device.h> #include <linux/of.h> #include <linux/of_device.h> +#include <linux/of_gpio.h> #include <linux/dma-mapping.h> #include <linux/atmel_pdc.h> #include <linux/atmel_serial.h> @@ -2327,6 +2328,7 @@ static int atmel_serial_probe(struct platform_device *pdev)struct device_node *np = pdev->dev.of_node;struct atmel_uart_data *pdata =dev_get_platdata (&pdev->dev);void *data; +int rts_pin = -EINVAL ;int ret = -ENODEV ;BUILD_BUG_ON (ATMEL_SERIAL_RINGSIZE & (ATMEL_SERIAL_RINGSIZE -1 ));
drivers/tty/serial/atmel_serial.c static int atmel_serial_probe(struct platform_device *pdev) · line 2373
Quoted 10 lines
drivers/tty/serial/atmel_serial.c +5
-2@@ -2364,13 +2366,18 @@ static int atmel_serial_probe(struct platform_device *pdev)* structs to indicate "no RTS GPIO" instead of open-coding some * invalid value everywhere. */ -if (pdata->rts_gpio >0 ) { - ret =devm_gpio_request (&pdev->dev, pdata->rts_gpio,"RTS" ); +if (pdata && pdata->rts_gpio >0 ) + rts_pin = pdata->rts_gpio; +else if (np) + rts_pin =of_get_named_gpio (np,"rts-gpios" ,0 ); +
use directly port->rts_gpio so we can drop the assign after if the gpio is invalid the port->rts_gpio need to be < 0 too
Best Regards,
J.
Quoted 14 lines
drivers/tty/serial/atmel_serial.c +3
-1+if (gpio_is_valid (rts_pin)) { + ret =devm_gpio_request (&pdev->dev, rts_pin,"RTS" );if (ret) {dev_err (&pdev->dev,"error requesting RTS GPIO\n" ); goto err; } - port->rts_gpio = pdata->rts_gpio; + port->rts_gpio = rts_pin; ret =gpio_direction_output (port->rts_gpio,0 );if (ret) {dev_err (&pdev->dev,"error setting up RTS GPIO\n" );Signature · 1 line
1.8.2.2
PATCH 1/1 ARM/serial: at91: specify RTS in DT using gpio
Quoted 31 lines · Jean-Christophe PLAGNIOL-VILLARD
Quoted 20 lines · Nicolas Ferre
Signed-off-by Nicolas Ferre <[email protected]>---
Documentation/devicetree/bindings/serial/atmel-usart.txt| 3 +++drivers/tty/serial/atmel_serial.c| 13 ++++++++++---2 files changed, 13 insertions(+), 3 deletions(-)Documentation/devicetree/bindings/serial/atmel-usart.txt +0
-0@@ -10,6 +10,8 @@ Required properties:Optional properties:
- atmel,use-dma-rx: use of PDC or DMA for receiving data
- atmel,use-dma-tx: use of PDC or DMA for transmitting data+- rts-gpios: specify a GPIO for RTS line. It will use specified PIO instead of the peripheral + function pin for the USART RTS feature. If unsure, don't specify this property.- add dma bindings for dma transfer:
- dmas: DMA specifier, consisting of a phandle to DMA controller node, memory peripheral interface and USART DMA channel ID, FIFO configuration.Documentation/devicetree/bindings/serial/atmel-usart.txt Example: · line 34
Quoted 5 lines
Documentation/devicetree/bindings/serial/atmel-usart.txt +1
-0@@ -28,6 +30,7 @@ Example: interrupts = <7>; atmel,use-dma-rx; atmel,use-dma-tx; + rts-gpios = <&pioD 15 0>; };gpios means 2 gpio
Quoted 1 line
I?ll rts-gpio as we can only use one anyway
Hi Jean
Have you read Documentation/devicetree/bindings/gpio/gpio.txt ?
It says:
GPIO properties should be named "[<name>-]gpios". Exact meaning of each gpios property must be documented in the device tree binding for each device.
There is also an example of a node with a single gpio using the property name gpios. So plural is correct, even if there is only one gpio.
Andrew
PATCH 1/1 ARM/serial: at91: specify RTS in DT using gpio
In reply to Jean-Christophe PLAGNIOL-VILLARDOn 05/11/2013 16:59, Jean-Christophe PLAGNIOL-VILLARD :
Quoted 31 lines · Nicolas Ferre
Quoted 21 lines
Signed-off-by Nicolas Ferre <[email protected]>---
Documentation/devicetree/bindings/serial/atmel-usart.txt| 3 +++drivers/tty/serial/atmel_serial.c| 13 ++++++++++---2 files changed, 13 insertions(+), 3 deletions(-)Documentation/devicetree/bindings/serial/atmel-usart.txt +0
-0@@ -10,6 +10,8 @@ Required properties:Optional properties:
- atmel,use-dma-rx: use of PDC or DMA for receiving data
- atmel,use-dma-tx: use of PDC or DMA for transmitting data+- rts-gpios: specify a GPIO for RTS line. It will use specified PIO instead of the peripheral + function pin for the USART RTS feature. If unsure, don't specify this property.- add dma bindings for dma transfer:
- dmas: DMA specifier, consisting of a phandle to DMA controller node, memory peripheral interface and USART DMA channel ID, FIFO configuration.Documentation/devicetree/bindings/serial/atmel-usart.txt Example: · line 34
Quoted 5 lines
Documentation/devicetree/bindings/serial/atmel-usart.txt +1
-0@@ -28,6 +30,7 @@ Example: interrupts = <7>; atmel,use-dma-rx; atmel,use-dma-tx; + rts-gpios = <&pioD 15 0>; };gpios means 2 gpio
Quoted 1 line
I?ll rts-gpio as we can only use one anyway
Cf. Andrew's comment.
Quoted 20 lines
Quoted 20 lines
Documentation/devicetree/bindings/serial/atmel-usart.txt +0
-1- use DMA:drivers/tty/serial/atmel_serial.c +2
-0@@ -35,6 +35,7 @@#include <linux/platform_device.h> #include <linux/of.h> #include <linux/of_device.h> +#include <linux/of_gpio.h> #include <linux/dma-mapping.h> #include <linux/atmel_pdc.h> #include <linux/atmel_serial.h> @@ -2327,6 +2328,7 @@ static int atmel_serial_probe(struct platform_device *pdev)struct device_node *np = pdev->dev.of_node;struct atmel_uart_data *pdata =dev_get_platdata (&pdev->dev);void *data; +int rts_pin = -EINVAL ;int ret = -ENODEV ;
Quoted 16 lines
drivers/tty/serial/atmel_serial.c static int atmel_serial_probe(struct platform_device *pdev) · line 2373
Quoted 12 lines
drivers/tty/serial/atmel_serial.c +5
-2BUILD_BUG_ON (ATMEL_SERIAL_RINGSIZE & (ATMEL_SERIAL_RINGSIZE -1 ));@@ -2364,13 +2366,18 @@ static int atmel_serial_probe(struct platform_device *pdev)* structs to indicate "no RTS GPIO" instead of open-coding some * invalid value everywhere. */ -if (pdata->rts_gpio >0 ) { - ret =devm_gpio_request (&pdev->dev, pdata->rts_gpio,"RTS" ); +if (pdata && pdata->rts_gpio >0 ) + rts_pin = pdata->rts_gpio; +else if (np) + rts_pin =of_get_named_gpio (np,"rts-gpios" ,0 ); +use directly port->rts_gpio so we can drop the assign after if the gpio is invalid the port->rts_gpio need to be < 0 too
Yes this is definitively an option. I cook a v2 right now.
Quoted 21 lines
Best Regards,
J.Quoted 14 lines
drivers/tty/serial/atmel_serial.c +3
-2+if (gpio_is_valid (rts_pin)) { + ret =devm_gpio_request (&pdev->dev, rts_pin,"RTS" );if (ret) {dev_err (&pdev->dev,"error requesting RTS GPIO\n" ); goto err; } - port->rts_gpio = pdata->rts_gpio; + port->rts_gpio = rts_pin; ret =gpio_direction_output (port->rts_gpio,0 );if (ret) {dev_err (&pdev->dev,"error setting up RTS GPIO\n" ); --1.8.2.2
Signature · 1 line
Nicolas Ferre
PATCH v2 1/1 ARM/serial: at91: specify RTS in DT using gpio
In reply to Nicolas Ferre--- Documentation/devicetree/bindings/serial/atmel-usart.txt | 3 +++ drivers/tty/serial/atmel_serial.c | 11 ++++++++--- 2 files changed, 11 insertions(+), 3 deletions(-)
Documentation/devicetree/bindings/serial/atmel-usart.txt +3 -0
@@ -10,6 +10,8 @@ Required properties: Optional properties: - atmel,use-dma-rx: use of PDC or DMA for receiving data - atmel,use-dma-tx: use of PDC or DMA for transmitting data +- rts-gpios: specify a GPIO for RTS line. It will use specified PIO instead of the peripheral + function pin for the USART RTS feature. If unsure, don't specify this property. - add dma bindings for dma transfer: - dmas: DMA specifier, consisting of a phandle to DMA controller node, memory peripheral interface and USART DMA channel ID, FIFO configuration.
@@ -28,6 +30,7 @@ Example: interrupts = <7>; atmel,use-dma-rx; atmel,use-dma-tx; + rts-gpios = <&pioD 15 0>; }; - use DMA:
drivers/tty/serial/atmel_serial.c +8 -3
@@ -35,6 +35,7 @@#include <linux/platform_device.h> #include <linux/of.h> #include <linux/of_device.h> +#include <linux/of_gpio.h> #include <linux/dma-mapping.h> #include <linux/atmel_pdc.h> #include <linux/atmel_serial.h>
@@ -2364,13 +2365,17 @@ static int atmel_serial_probe(struct platform_device *pdev)* structs to indicate "no RTS GPIO" instead of open-coding some * invalid value everywhere. */ -if (pdata->rts_gpio >0 ) { - ret =devm_gpio_request (&pdev->dev, pdata->rts_gpio,"RTS" ); +if (pdata && pdata->rts_gpio >0 ) + port->rts_gpio = pdata->rts_gpio; +else if (np) + port->rts_gpio =of_get_named_gpio (np,"rts-gpios" ,0 ); + +if (gpio_is_valid (port->rts_gpio)) { + ret =devm_gpio_request (&pdev->dev, port->rts_gpio,"RTS" );if (ret) {dev_err (&pdev->dev,"error requesting RTS GPIO\n" ); goto err; } - port->rts_gpio = pdata->rts_gpio; ret =gpio_direction_output (port->rts_gpio,0 );if (ret) {dev_err (&pdev->dev,"error setting up RTS GPIO\n" );
Signature · 1 line
1.8.2.2
PATCH 1/1 ARM/serial: at91: switch atmel serial to use gpiolib
In reply to Linus WalleijQuoted 80 lines · Linus Walleij
This passes the errata fix using a GPIO to control the RTS pin on one of the AT91 chips to use gpiolib instead of the AT91-specific interfaces. Also remove the reliance on compile-time #defines and the cpu_* check and rely on the platform passing down the proper GPIO pin through platform data.
This is a prerequisite for getting rid of the local GPIO implementation in the AT91 platform and move toward multiplatform.
This also makes way for device tree conversion: the RTS GPIO pin can be passed by standard GPIO bindings.
Signed-off-by Linus Walleij <[email protected]>---
This is an alternative to the patch entitled "ARM/serial: at91: move machine quirk into machine" which needs testing to confirm this approach. Seeking ACKs on this if the approach seems OK to all parties.--- arch/arm/mach-at91/at91rm9200_devices.c | 1 + drivers/tty/serial/atmel_serial.c | 51 +++++++++++++++++++++------------ include/linux/platform_data/atmel.h | 1 + 3 files changed, 35 insertions(+), 18 deletions(-)arch/arm/mach-at91/at91rm9200_devices.c +1
-0@@ -961,6 +961,7 @@ static struct resource uart0_resources[] = {static struct atmel_uart_data uart0_data = { .use_dma_tx =1 , .use_dma_rx =1 , + .rts_gpio =AT91_PIN_PA21 , };static u64 uart0_dmamask =DMA_BIT_MASK (32 );drivers/tty/serial/atmel_serial.c +2
-13@@ -41,15 +41,11 @@#include <linux/uaccess.h> #include <linux/platform_data/atmel.h> #include <linux/timer.h> +#include <linux/gpio.h> #include <asm/io.h> #include <asm/ioctls.h> -#ifdef CONFIG_ARM -#include <mach/cpu.h> -#include <asm/gpio.h> -#endif -#define PDC_BUFFER_SIZE 512/* Revisit: We should calculate this based on the actual port settings */ #define PDC_RX_TIMEOUT (3 * 10)/* 3 bytes */ @@ -167,6 +163,7 @@ struct atmel_uart_port {struct circ_buf rx_ring;struct serial_rs485 rs485;/* rs485 settings */ +int rts_gpio;/* optional RTS GPIO */ unsigned int tx_done_mask;bool is_usart;/* usart or uart */ struct timer_list uart_timer;/* uart timer */ @@ -300,20 +297,17 @@ static void atmel_set_mctrl(struct uart_port *port, u_int mctrl)unsigned int mode;struct atmel_uart_port *atmel_port =to_atmel_uart_port (port); -#ifdef CONFIG_ARCH_AT91RM9200 -if (cpu_is_at91rm9200 ()) { -/* -* AT91RM9200 Errata #39: RTS0 is not internally connected -* to PA21. We need to drive the pin manually. -*/ -if (port->mapbase ==AT91RM9200_BASE_US0 ) { -if (mctrl &TIOCM_RTS )
drivers/tty/serial/atmel_serial.c static void atmel_set_mctrl(struct uart_port *port, u_int mctrl) · line 305
Quoted 10 lines
drivers/tty/serial/atmel_serial.c +6
-4-at91_set_gpio_value (AT91_PIN_PA21 ,0 ); - else - at91_set_gpio_value(AT91_PIN_PA21, 1); - } +/* +* AT91RM9200 Errata #39: RTS0 is not internally connected +* to PA21. We need to drive the pin as a GPIO. +*/ +if (gpio_is_valid (atmel_port->rts_gpio) && + port->mapbase ==AT91RM9200_BASE_US0 ) {
not really a fon of this hack
if we use a uart for rs485 we need to use a gpio for rts too
so we may need to find a better way
Quoted 12 lines
drivers/tty/serial/atmel_serial.c +4
-1+if (mctrl &TIOCM_RTS ) +gpio_set_value (atmel_port->rts_gpio,0 ); +else +gpio_set_value (atmel_port->rts_gpio,1 ); } -#endif if (mctrl &TIOCM_RTS ) control |=ATMEL_US_RTSEN ;@@ -2365,6 +2359,27 @@ static int atmel_serial_probe(struct platform_device *pdev) port = &atmel_ports[ret]; port->backup_imr =0 ;
drivers/tty/serial/atmel_serial.c static int atmel_serial_probe(struct platform_device *pdev) · line 2370
Quoted 10 lines
drivers/tty/serial/atmel_serial.c +9
-0port->uart.line = ret; + port->rts_gpio =-1 ;/* Invalid, zero could be valid */ +/* +* In theory the GPIO pin controlling RTS could be zero and +* this would be an improper check, but we know that the only +* existing case is != 0 and it's nice to use the zero-initialized +* structs to indicate "no RTS GPIO" instead of open-coding some +* invalid value everywhere. +*/ +if (pdata->rts_gpio >0 ) {
this can not work 0 is a valid gpio if you do this you need to update ALL the bard filee too to set it as -EINVAL;
Quoted 30 lines
drivers/tty/serial/atmel_serial.c +12
-0+ ret =devm_gpio_request (&pdev->dev, pdata->rts_gpio,"RTS" ); +if (ret) { +dev_err (&pdev->dev,"error requesting RTS GPIO\n" ); + goto err; + } + port->rts_gpio = pdata->rts_gpio; + ret =gpio_direction_output (port->rts_gpio,0 ); +if (ret) { +dev_err (&pdev->dev,"error setting up RTS GPIO\n" ); + goto err; + } + } ret =atmel_init_port (port, pdev);if (ret)include/linux/platform_data/atmel.h +1
-0@@ -84,6 +84,7 @@ struct atmel_uart_data {short use_dma_rx;/* use receive DMA? */ void __iomem *regs;/* virt. base address, if any */ struct serial_rs485 rs485;/* rs485 settings */ +int rts_gpio;/* optional RTS GPIO */ };/* Touchscreen Controller */ Signature · 1 line
1.8.3.1
PATCH 1/1 ARM/serial: at91: switch atmel serial to use gpiolib
On 05/11/2013 17:28, Jean-Christophe PLAGNIOL-VILLARD :
Quoted 93 lines · Linus Walleij
Quoted 80 lines
This passes the errata fix using a GPIO to control the RTS pin on one of the AT91 chips to use gpiolib instead of the AT91-specific interfaces. Also remove the reliance on compile-time #defines and the cpu_* check and rely on the platform passing down the proper GPIO pin through platform data.
This is a prerequisite for getting rid of the local GPIO implementation in the AT91 platform and move toward multiplatform.
This also makes way for device tree conversion: the RTS GPIO pin can be passed by standard GPIO bindings.
Signed-off-by Linus Walleij <[email protected]>---
This is an alternative to the patch entitled "ARM/serial: at91: move machine quirk into machine" which needs testing to confirm this approach. Seeking ACKs on this if the approach seems OK to all parties.--- arch/arm/mach-at91/at91rm9200_devices.c | 1 + drivers/tty/serial/atmel_serial.c | 51 +++++++++++++++++++++------------ include/linux/platform_data/atmel.h | 1 + 3 files changed, 35 insertions(+), 18 deletions(-)arch/arm/mach-at91/at91rm9200_devices.c +1
-0@@ -961,6 +961,7 @@ static struct resource uart0_resources[] = {static struct atmel_uart_data uart0_data = { .use_dma_tx =1 , .use_dma_rx =1 , + .rts_gpio =AT91_PIN_PA21 , };static u64 uart0_dmamask =DMA_BIT_MASK (32 );drivers/tty/serial/atmel_serial.c +2
-13@@ -41,15 +41,11 @@#include <linux/uaccess.h> #include <linux/platform_data/atmel.h> #include <linux/timer.h> +#include <linux/gpio.h> #include <asm/io.h> #include <asm/ioctls.h> -#ifdef CONFIG_ARM -#include <mach/cpu.h> -#include <asm/gpio.h> -#endif -#define PDC_BUFFER_SIZE 512/* Revisit: We should calculate this based on the actual port settings */ #define PDC_RX_TIMEOUT (3 * 10)/* 3 bytes */ @@ -167,6 +163,7 @@ struct atmel_uart_port {struct circ_buf rx_ring;struct serial_rs485 rs485;/* rs485 settings */ +int rts_gpio;/* optional RTS GPIO */ unsigned int tx_done_mask;bool is_usart;/* usart or uart */ struct timer_list uart_timer;/* uart timer */ @@ -300,20 +297,17 @@ static void atmel_set_mctrl(struct uart_port *port, u_int mctrl)unsigned int mode;struct atmel_uart_port *atmel_port =to_atmel_uart_port (port); -#ifdef CONFIG_ARCH_AT91RM9200 -if (cpu_is_at91rm9200 ()) { -/* -* AT91RM9200 Errata #39: RTS0 is not internally connected -* to PA21. We need to drive the pin manually. -*/ -if (port->mapbase ==AT91RM9200_BASE_US0 ) { -if (mctrl &TIOCM_RTS )drivers/tty/serial/atmel_serial.c static void atmel_set_mctrl(struct uart_port *port, u_int mctrl) · line 305
Quoted 10 lines
drivers/tty/serial/atmel_serial.c +6
-4-at91_set_gpio_value (AT91_PIN_PA21 ,0 ); - else - at91_set_gpio_value(AT91_PIN_PA21, 1); - } +/* +* AT91RM9200 Errata #39: RTS0 is not internally connected +* to PA21. We need to drive the pin as a GPIO. +*/ +if (gpio_is_valid (atmel_port->rts_gpio) && + port->mapbase ==AT91RM9200_BASE_US0 ) {not really a fon of this hack
Quoted 3 lines
if we use a uart for rs485 we need to use a gpio for rts too
so we may need to find a better way
Yes... but the problem from now on is to remove another hack which prevents us from moving forward to the single zImage direction (with the help of Linus W.).
So, before that we can enhance the code for RS-485, I am in favor of a move with this patch which is equivalent to what we already have in the driver nowadays.
But is is true that we can add another patch on top of this one to remove the restriction on "AT91RM9200_BASE_US0". If you feel it is a good idea, go on and send a patch.
Bye,
Quoted 56 lines
Quoted 12 lines
drivers/tty/serial/atmel_serial.c +4
-1+if (mctrl &TIOCM_RTS ) +gpio_set_value (atmel_port->rts_gpio,0 ); +else +gpio_set_value (atmel_port->rts_gpio,1 ); } -#endif if (mctrl &TIOCM_RTS ) control |=ATMEL_US_RTSEN ;@@ -2365,6 +2359,27 @@ static int atmel_serial_probe(struct platform_device *pdev) port = &atmel_ports[ret]; port->backup_imr =0 ;drivers/tty/serial/atmel_serial.c static int atmel_serial_probe(struct platform_device *pdev) · line 2370
Quoted 10 lines
drivers/tty/serial/atmel_serial.c +9
-0port->uart.line = ret; + port->rts_gpio =-1 ;/* Invalid, zero could be valid */ +/* +* In theory the GPIO pin controlling RTS could be zero and +* this would be an improper check, but we know that the only +* existing case is != 0 and it's nice to use the zero-initialized +* structs to indicate "no RTS GPIO" instead of open-coding some +* invalid value everywhere. +*/ +if (pdata->rts_gpio >0 ) {this can not work 0 is a valid gpio if you do this you need to update ALL the bard filee too to set it as -EINVAL;Quoted 30 lines
drivers/tty/serial/atmel_serial.c +12
-0+ ret =devm_gpio_request (&pdev->dev, pdata->rts_gpio,"RTS" ); +if (ret) { +dev_err (&pdev->dev,"error requesting RTS GPIO\n" ); + goto err; + } + port->rts_gpio = pdata->rts_gpio; + ret =gpio_direction_output (port->rts_gpio,0 ); +if (ret) { +dev_err (&pdev->dev,"error setting up RTS GPIO\n" ); + goto err; + } + } ret =atmel_init_port (port, pdev);if (ret)include/linux/platform_data/atmel.h +1
-1@@ -84,6 +84,7 @@ struct atmel_uart_data {short use_dma_rx;/* use receive DMA? */ void __iomem *regs;/* virt. base address, if any */ struct serial_rs485 rs485;/* rs485 settings */ +int rts_gpio;/* optional RTS GPIO */ };/* Touchscreen Controller */ --1.8.3.1
Signature · 1 line
Nicolas Ferre
PATCH 1/1 ARM/serial: at91: switch atmel serial to use gpiolib
In reply to Jean-Christophe PLAGNIOL-VILLARDQuoted 9 lines · Jean-Christophe PLAGNIOL-VILLARD
drivers/tty/serial/atmel_serial.c static void atmel_set_mctrl(struct uart_port *port, u_int mctrl) · line 305
Quoted 6 lines · Linus Walleij
drivers/tty/serial/atmel_serial.c +6
-0+/* +* AT91RM9200 Errata #39: RTS0 is not internally connected +* to PA21. We need to drive the pin as a GPIO. +*/ +if (gpio_is_valid (atmel_port->rts_gpio) && + port->mapbase ==AT91RM9200_BASE_US0 ) {
Quoted 3 lines
not really a fon of this hack
if we use a uart for rs485 we need to use a gpio for rts too
Hm it was completely unnecessary anyway, as I now specify the RTS pin per UART instance, it will only be valid for those setting it right.
Quoted 11 lines
drivers/tty/serial/atmel_serial.c static int atmel_serial_probe(struct platform_device *pdev) · line 2370
Quoted 8 lines
drivers/tty/serial/atmel_serial.c +8
-0+/* +* In theory the GPIO pin controlling RTS could be zero and +* this would be an improper check, but we know that the only +* existing case is != 0 and it's nice to use the zero-initialized +* structs to indicate "no RTS GPIO" instead of open-coding some +* invalid value everywhere. +*/ +if (pdata->rts_gpio >0 ) {this can not work 0 is a valid gpio if you do this you need to update ALL the bard filee too to set it as
-EINVAL;
Isn't it the other way around? I did this exactly to avoid updating all the boardfiles, as the rts_gpio == 0 for all those not defining it. This is why I test for > 0.
Yours,
Linus Walleij