ARM/serial: at91: move machine quirk into machine
Subsystems (6)
PATCH 1/4 ARM/serial: at91: move machine quirk into machine
This removes the dependency on <mach/gpio.h> from the AT91 serial driver by adding an mctrl callback quirk.
Long term it is better if the driver calls the generic GPIO interface (gpio_set_value()), but this gets a hairy cross-depency into the machine-local headers out of the way for now.
---
Greg, if you're OK with this approach please give me an ACK so I can take this through the ARM SoC or GPIO tree in the end.
--- arch/arm/mach-at91/at91rm9200_devices.c | 14 ++++++++++++++ drivers/tty/serial/atmel_serial.c | 19 ++++--------------- include/linux/platform_data/atmel.h | 1 + 3 files changed, 19 insertions(+), 15 deletions(-)
arch/arm/mach-at91/at91rm9200_devices.c +14 -0
@@ -12,6 +12,7 @@*/ #include <asm/mach/arch.h> #include <asm/mach/map.h> +#include <asm/termios.h> #include <linux/dma-mapping.h> #include <linux/gpio.h>
@@ -957,9 +958,22 @@ static struct resource uart0_resources[] = { }, }; +static void uart0_set_mctrl (unsigned int mctrl) +{ +/* +* AT91RM9200 Errata #39: RTS0 is not internally connected +* to PA21. We need to drive the pin manually. +*/ +if (mctrl &TIOCM_RTS ) +at91_set_gpio_value (AT91_PIN_PA21 ,0 ); +else +at91_set_gpio_value (AT91_PIN_PA21 ,1 ); +} +static struct atmel_uart_data uart0_data = { .use_dma_tx =1 , .use_dma_rx =1 , + .set_mctrl = uart0_set_mctrl, };static u64 uart0_dmamask =DMA_BIT_MASK (32 );
drivers/tty/serial/atmel_serial.c +4 -15
@@ -47,7 +47,6 @@#ifdef CONFIG_ARM #include <mach/cpu.h> -#include <asm/gpio.h> #endif #define PDC_BUFFER_SIZE 512
@@ -176,6 +175,7 @@ struct atmel_uart_port {void (*schedule_tx)(struct uart_port *port);void (*release_rx)(struct uart_port *port);void (*release_tx)(struct uart_port *port); +void (*set_mctrl)(unsigned int mctrl); };static struct atmel_uart_port atmel_ports[ATMEL_MAX_UART ];
@@ -300,20 +300,8 @@ 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 ) -at91_set_gpio_value (AT91_PIN_PA21 ,0 ); -else -at91_set_gpio_value (AT91_PIN_PA21 ,1 ); - } - } -#endif +if (atmel_port->set_mctrl) + atmel_port->set_mctrl (mctrl);if (mctrl &TIOCM_RTS ) control |=ATMEL_US_RTSEN ;
@@ -2365,6 +2353,7 @@ static int atmel_serial_probe(struct platform_device *pdev) port = &atmel_ports[ret]; port->backup_imr =0 ; port->uart.line = ret; + port->set_mctrl = pdata->set_mctrl; 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 */ +void (*set_mctrl)(unsigned int mctrl);/* mctrl callback */ };/* Touchscreen Controller */
Signature · 1 line
1.8.3.1
PATCH 1/4 ARM/serial: at91: move machine quirk into machine
On 05/11/2013 10:28, Linus Walleij :
Quoted 9 lines
This removes the dependency on <
mach/gpio.h> from the AT91 serial driver by adding an mctrl callback quirk.Long term it is better if the driver calls the generic GPIO interface (
gpio_set_value()), but this gets a hairy cross-depency into the machine-local headers out of the way for now.Cc Greg Kroah-Hartman <[email protected]>Signed-off-by Linus Walleij <[email protected]>
Linus,
I do think that we have to remove this dependency but the only thing that embarrasses me is that your change only allows non-DT at91rm9200 to work well with its RTS line.
If we use Device Tree with this product (which is actually possible) we don't have this feature available anymore...
A possible solution can be to give an optional RTS pin specification to the DT binding and to the pdata structure, if defined it can be used in the atmel_set_mctrl() function. This way we do not have a special behavior for at91rm9200 anymore. For the use of specialized AT91 gpio functions, well, we can implement the move to gpiolib.
What do you think Linus? Is there a trend to remove the calls to GPIO functions in drivers?
Best regards,
Quoted 65 lines
---
Greg, if you're OK with this approach please give me an ACK so I can take this through the ARM SoC or GPIO tree in the end.--- arch/arm/mach-at91/at91rm9200_devices.c | 14 ++++++++++++++ drivers/tty/serial/atmel_serial.c | 19 ++++--------------- include/linux/platform_data/atmel.h | 1 + 3 files changed, 19 insertions(+), 15 deletions(-)arch/arm/mach-at91/at91rm9200_devices.c +14
-0@@ -12,6 +12,7 @@*/ #include <asm/mach/arch.h> #include <asm/mach/map.h> +#include <asm/termios.h> #include <linux/dma-mapping.h> #include <linux/gpio.h> @@ -957,9 +958,22 @@ static struct resource uart0_resources[] = { }, }; +static void uart0_set_mctrl (unsigned int mctrl) +{ +/* +* AT91RM9200 Errata #39: RTS0 is not internally connected +* to PA21. We need to drive the pin manually. +*/ +if (mctrl &TIOCM_RTS ) +at91_set_gpio_value (AT91_PIN_PA21 ,0 ); +else +at91_set_gpio_value (AT91_PIN_PA21 ,1 ); +} +static struct atmel_uart_data uart0_data = { .use_dma_tx =1 , .use_dma_rx =1 , + .set_mctrl = uart0_set_mctrl, };static u64 uart0_dmamask =DMA_BIT_MASK (32 );drivers/tty/serial/atmel_serial.c +1
-1@@ -47,7 +47,6 @@#ifdef CONFIG_ARM #include <mach/cpu.h> -#include <asm/gpio.h> #endif #define PDC_BUFFER_SIZE 512@@ -176,6 +175,7 @@ struct atmel_uart_port {void (*schedule_tx)(struct uart_port *port);void (*release_rx)(struct uart_port *port);void (*release_tx)(struct uart_port *port); +void (*set_mctrl)(unsigned int mctrl); };static struct atmel_uart_port atmel_ports[ATMEL_MAX_UART ];
drivers/tty/serial/atmel_serial.c static void atmel_set_mctrl(struct uart_port *port, u_int mctrl) · line 304
Quoted 5 lines
drivers/tty/serial/atmel_serial.c +0
-2@@ -300,20 +300,8 @@ 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 ()) {
By removing this call, you can also remove
#include <mach/cpu.h>
as well (maybe in another patch).
Quoted 38 lines
drivers/tty/serial/atmel_serial.c +3
-12-/* -* 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 ) -at91_set_gpio_value (AT91_PIN_PA21 ,0 ); -else -at91_set_gpio_value (AT91_PIN_PA21 ,1 ); - } - } -#endif +if (atmel_port->set_mctrl) + atmel_port->set_mctrl (mctrl);if (mctrl &TIOCM_RTS ) control |=ATMEL_US_RTSEN ;@@ -2365,6 +2353,7 @@ static int atmel_serial_probe(struct platform_device *pdev) port = &atmel_ports[ret]; port->backup_imr =0 ; port->uart.line = ret; + port->set_mctrl = pdata->set_mctrl; 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 */ +void (*set_mctrl)(unsigned int mctrl);/* mctrl callback */ };/* Touchscreen Controller */
Signature · 1 line
Nicolas Ferre
PATCH 1/4 ARM/serial: at91: move machine quirk into machine
Quoted 7 lines · Nicolas Ferre
On 05/11/2013 10:28, Linus Walleij :
I do think that we have to remove this dependency but the only thing that embarrasses me is that your change only allows non-DT at91rm9200 to work well with its RTS line.
If we use Device Tree with this product (which is actually possible) we don't have this feature available anymore...
You can still supply the callback into the machine using AUXDATA if you wish.
But I can do the thing mentioned in the commit message and switch the driver to use gpiolib instead, I'll send another patch, can you please test it...
Yours,
Linus Walleij
Greg Kroah-Hartman
maintainerPATCH 1/4 ARM/serial: at91: move machine quirk into machine
In reply to Linus WalleijQuoted 7 lines · Linus Walleij
This removes the dependency on <
mach/gpio.h> from the AT91 serial driver by adding an mctrl callback quirk.Long term it is better if the driver calls the generic GPIO interface (
gpio_set_value()), but this gets a hairy cross-depency into the machine-local headers out of the way for now.
Quoted 5 lines
Cc Greg Kroah-Hartman <[email protected]>Signed-off-by Linus Walleij <[email protected]>---
Greg, if you're OK with this approach please give me an ACK so I can take this through the ARM SoC or GPIO tree in the end.