Segmentation fault in inline application record for registration

I am running debian with kernel 3.14.14 on a Freescale iMX6 processor.

I have this code in an embedded application which is giving a segmentation fault.

volatile unsigned int& GPIO2IO26CTRL = *((volatile unsigned int*)0x20e0104);
GPIO2IO26CTRL = 0x5;

      

also tried:

volatile unsigned int* GPIO2IO26CTRL = (volatile unsigned int*)0x20e0104;
*GPIO2IO26CTRL = 0x5;

      

But I really want to write 0x5 at 0x20e0104. This tells the iMX6 chip that the io line should be GPIO, not spi;

How do I get around this?

+3


source to share


2 answers


This is possible from user space (although you may need to be root). The key reassigns memory to the process address space.



There is a program called devmem2

that can do this. The source code is here .

+2


source


Thanks for the help. If anyone has the same problem, I want to show my solution. Most of this code came from devmem2 from Jan-Derk Bakker.

Here are two routines I wrote to read and write from memory.

static const u_int32_t MAP_SIZE = 4096;
static const u_int32_t MAP_MASK = (MAP_SIZE - 1);

bool CGpio::writeMem(u_int32_t memAddr, u_int32_t &value)
{
    int fd;
    void *map_base, *virt_addr;
    off_t target = memAddr;

    // open the mem file
    if((fd = open("/dev/mem", O_RDWR | O_SYNC)) == -1)
        return false;

    // map one page
    map_base = mmap(0, MAP_SIZE, PROT_READ | PROT_WRITE, MAP_SHARED, fd, target & ~MAP_MASK);
    if (map_base == (void *)-1)
        return false;

    // map virutial space
    virt_addr = map_base + (target & MAP_MASK);

    // write the value
    *((u_int32_t *) virt_addr) = value;

    // clean up
    bool ret = (munmap(map_base, MAP_SIZE) != -1);
    close(fd);

    return ret;
}

bool CGpio::readMem(u_int32_t memAddr, u_int32_t &value)
{
    int fd;
    void *map_base, *virt_addr;
    off_t target = memAddr;

    // open the mem file
    if((fd = open("/dev/mem", O_RDWR | O_SYNC)) == -1)
        return false;

    // map one page
    map_base = mmap(0, MAP_SIZE, PROT_READ | PROT_WRITE, MAP_SHARED, fd, target & ~MAP_MASK);
    if (map_base == (void *)-1)
         return false;

    // map virutial space
    virt_addr = map_base + (target & MAP_MASK);

    // read the value
    value = *((u_int32_t *) virt_addr);

    // clean up
    bool ret = (munmap(map_base, MAP_SIZE) != -1);
    close(fd);

    return ret;
}

      



And here are the headers I used.

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <signal.h>
#include <fcntl.h>
#include <sys/mman.h>
#include <string.h>
#include <iostream>
#include <sys/types.h>

      

+1


source







All Articles