GPIO management via mem

Hi,
I have to monitor signals on many gpios but via proc fs it takes a lot of CPU. There is a faster way but I need some help. http://www.black-swift.com/wiki/index.php?title=Working_with_GPIOs_(C/C++)

We could have access to gpios directly via memory blocks but I dont know base address for gpios and mem block size. I have investigated topic and probably base addr is at 0x20000000 but I dont know mem block size ( probably 16K). Could you confirm the informations?

best regards,

 

Hi, I resolved a problem. Please check link https://community.freescale.com/thread/381277. Use example with mmap from userland and test on pin 40th. My previous thinking has been wrong.

Code:

example for pin 161 with no changed comments

#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <fcntl.h>
#include <sys/mman.h>
#include <errno.h>
#include <string.h>
#include <stdint.h>
#include <unistd.h>

#define IOMUXC                           0x020E0000
#define IOMUXC_SW_MUX_CTL_PAD_SD4_DATA0  0x020E031C
#define GPIO2_DR                         0x020B0000
//#define GPIO2_GDIR                       0x020A0004
#define GPIO2_GDIR      (GPIO2_DR + 0x04)
#define GPIO2_PSR                        1


static volatile uint32_t *gpio;

int main(int argc, char **argv)
{
    int fd ;

    //Obtain handle to physical memory
    if ((fd = open ("/dev/mem", O_RDWR | O_SYNC) ) < 0) {
        printf("Unable to open /dev/mem: %s\n", strerror(errno));
        return -1;
    }

    /**
      * Map some of the pin mux registers within the range of the
      * GPIO40 GPIO40 pin.
    **/
    gpio = (uint32_t *)mmap(0, (getpagesize() * 250), PROT_READ|PROT_WRITE, MAP_SHARED, fd, IOMUXC);

    // The first 3 bits are the only thing that matters here according to the
    // datasheet.
    *(gpio + ((IOMUXC_SW_MUX_CTL_PAD_SD4_DATA0 - IOMUXC) / 4)) = 0x5;

    //Be a good neighbor
    if((munmap((void *)gpio, (getpagesize() * 250))) < 0){
        printf("munmap failed: %s\n", strerror(errno));
        exit(1);
    }
    /** The i.MX 6 Dual and Quad Processor have 7 GPIO banks
      * Each bank amounts to 1mb of memory which is perfectly
      * aligned for a memory map. Mmap will fail if the offset
      * is not a multiple of 1mb. For this example I just want
      * the GPIO2 bank which starts at 0x020A0000. I want the
      * bank (1mb or 250 pages of memory). I could get away with
      * a single page, but I am greedy :-P.
    **/
    gpio = (uint32_t *)mmap(0, (getpagesize() * 250), PROT_READ|PROT_WRITE, MAP_SHARED, fd, GPIO2_DR);

    if ((int32_t)gpio < 0){
        printf("mmap failed: %s\n", strerror(errno));
        exit(1);
    }

    /**
      * Setup GPIO40 or GPIO2_IO08 as an output.
      * gpio[1] is the GPIO direction register (GPIO2_GDIR).
      * We want to only want to set bit 8 of this register
      * to a binary 1 while leaving the rest intact.
    **/
    *(gpio + ((GPIO2_GDIR - GPIO2_DR) / 4)) = *(gpio + ((GPIO2_GDIR - GPIO2_DR) / 4)) | 1<<GPIO2_PSR;

    /**
      * Blink GPIO40 every 500 milliseconds.
      * gpio[0] is the GPIO data register (GPIO2_DR).
      * We only want to toggle bit 8 of this register
      * while leaving the rest alone.
    **/
    while(1){
        *(gpio + ((GPIO2_DR - GPIO2_DR) / 4)) = *(gpio + ((GPIO2_DR - GPIO2_DR) / 4)) ^ 1<<GPIO2_PSR;
        usleep(250000);
        *(gpio + ((GPIO2_DR - GPIO2_DR) / 4)) = *(gpio + ((GPIO2_DR - GPIO2_DR) / 4)) ^ 1<<GPIO2_PSR;
        usleep(250000);
    }
}