Hi All,

          I am using Samsung S3C2440 board to which linux kernel is
ported and now i have an application board i need to make the system bus
interface for my application board and S3C2440 board(i.e., memory
mapping of Address and Data bus of S3C2440 to Address and Data bus of
application board). How to do this.. any suggestions please.

Your question is somewhat short of essential detail.  The term
"application board" is rather vague.

Either way you probably need to familiarise your self with Chapter 5 -
Memory Controller of the S3C2440 User Manual.  There are 8 banks each of
which select a 128Mb address space, all banks support conventional ROM,
SRAM type address/data bus, and two additionally support SDRAM. All
banks support 8/16/23 bit data bus, except Bank 0 which has only 16/32
bit bus support. One of these banks will already be used by your board
for SDRAM, and another for Flash memory.  Possibly there are additional
memory mapped devices on your board, which may use additional bank slect
lines.  Your board's user guide or schematic will tell you which ones
are available.  Basically you need to hook up your board to the
necessary data and address lines, and the bank select to the data bus
tri-state latch (chip select), and hook up the R/W select line.

You will also need to enable the selected memory region in the MMU
configuration.

Hi,

     Thanks for your reply i have attached the application board
schematic here i mapped the address bus (A1-A3) and Data bus (D0-D7)to
IDE connector of address bus (LADDR1-LADDR3) and data bus
(LDATA0-LDATA7). Also i have made a code the if i select the particular
memory location say 0x08000000 then the chip select(nGCS1) has to enable
and if i continuously read the content of the memory location then the
chip select(nGCS1) has to enable continuously till i kill the execution
of the program. If i check the nGCS1 pin of the IDE connector in the CRO
chip select will happens only once but not continuously. Please find the
attachment and below code.

/******************test.c****************/

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <fcntl.h>
#include <sys/mman.h>
#include <sys/ioctl.h>

#define MAP_SIZE 4096UL    
#define MAP_MASK (MAP_SIZE - 1)  

#define DATA      0x08000000 

int main(int argc, char *argv[]) {
    int fd;
    void *map_base, *virt_addr; 
    unsigned long read_result, writeval;

    if((fd = open("/dev/mem", O_RDWR | O_SYNC)) == -1)
  return -1;
   while(1){
    map_base = mmap(0, MAP_SIZE, PROT_READ | PROT_WRITE, MAP_SHARED, fd, DATA & ~MAP_MASK);
    if(map_base == (void *) -1)
  return -1;
   
    virt_addr = map_base + (DATA & MAP_MASK);
    read_result = *((unsigned long *) virt_addr);

    // Write into the memory location 0x08000000
    writeval = read_result;
    writeval = 0x0AA; //WRITE THE DATA TO MEMORY
    *((unsigned long *) virt_addr) = writeval;
    printf("writeval = 0x%X\n", writeval);

    //Read the content of the memory location 0x08000000
    read_result = *((unsigned long *) virt_addr);
    printf("read_result = 0x%X\n",read_result);
    }

    if(munmap(map_base, MAP_SIZE) == -1)
  return -1;

    close(fd);
    return 0;

}

I am no expert on Linux, but you will at least need to disable the cache
for the application board's memory mapped region, otherwise you will
simply read what is in the cache rather than what is on the board -
which might explain the apparent single access.

Similarly you should declare pointers that map to the region as
volatile; otherwise the optimiser may potentially remove a read and use
the value previously read into a register.