On Mon, Jul 18, 2011 at 05:41, Ajay Bhargav wrote:
--- /dev/null +++ b/arch/arm/include/asm/arch-armada100/gpio.h
+#ifndef __ASSEMBLY__ +#include <asm/types.h> +#endif ... +struct armdgpio_registers {
u32 gplr0; /* Pin Level Register - 0x0000 */...
considering you dont protect actual C code in this file with __ASSEMBLY__, it doesnt make much sense to protect the include
--- a/drivers/gpio/Makefile +++ b/drivers/gpio/Makefile
COBJS-$(CONFIG_AT91_GPIO) += at91_gpio.o +COBJS-$(CONFIG_ARMADA100_GPIO) += armada100_gpio.o
"R" comes before "T", so your new entry should be above the at91 one
--- /dev/null +++ b/drivers/gpio/armada100_gpio.c
+int gpio_request(int gp, const char *label) +{
- /*
- * Assumes corresponding MFP is configured peoperly
- * for use as GPIO
- */
- return 0;
+}
i think you mean "properly" in the comment, but this assumption is bogus. error checking for valid pins has to be in the gpio_request() so that all the other funcs can assume the pin is valid. not the other way around.
+void gpio_free(int gp) +{
- /* default direction for GPIO is input */
- gpio_direction_input(gp);
+}
usually people dont do this. usually the free step leaves the pin alone. i know in the linux kernel, they assume (and thus require) that behavior. -mike