Problems with creating a pointer to struct at a specific address

Hello, I have problems when I trying to create a pointer to a struct initialized at a specific address. I tied a lot of different combinations, but still I does not get it to work. Here is one example:

Code:

---


#include <stdio.h>
#define INIT_ADR ((pointerFOO) 0x00000000); 
 
typedef struct FOO_tag{
    int    PER;
    int    PDR;
    int    PSR;
} FOO, *pointerFOO;
 
 
 
int main(void)
{
 
  pointerFOO pFOO = INIT_ADR;
 
  printf("&(pFOO->PER)= %i\n",&(pFOO->PER));
  printf("&(pFOO->PDR)= %i\n",&(pFOO->PDR));
 
  pFOO->PER = 1; 
 
 
  return(0);
}

---

I get some error with pFOO->PER = 1 when I am trying to run the code. Is there some who knows how I can solve this problem?

/Dan

Quote:

---

Originally Posted by danba185

I get some error with pFOO->PER = 1 when I am trying to run the code. Is there some who knows how I can solve this problem?

---

yes try another address, preferable one that has been returned by a call to malloc().
Kurt

What do you mean by "a struct initialized at a specific address"? are you talking about that FOO global variable? What do you want to do exactly?

Maybe you should say what that "some error" is, then we might be able to help.

ZuK: I suspect this is for some kind of embedded system which has to use a specific address for some hardware registers, and malloc will likely not return the desired address.

Edit: I compiled the code myself and got warnings about format %i expecting an int but the argument is int *. After changing %i to %p it compiled without warnings but crashed with a segmentation fault at the line with "pFOO->PER = 1;" (which I fully expected to happen). Of course, I'm running this on Linux rather than whatever system you're writing this code for, so who knows what the desired behavior is for your system.

@ christop: the op specifically told that the error happens at the line "pFOO->PER = 1;
that's why I think that NULL is not a valid address to write to on his system ( embedded system or not ).
and yes I didn't spot the type missmatch in the printf() calls.
Kurt

Hi, thank you all for the replies. Yes I will run this on an embedded system and that is why the memory address is very important. I am also running Linux and it crashed with a segmentation fault for me as well. Do you have any suggestions of how I can specify the address of where the struct shall be located and still using pointers? I tried to use

Code:

---

pFOO = (struct FOO *) malloc(sizeof(struct FOO));

---

and it did work fine and I did not get any segmentation fault, but is it possible to define the address?

/Dan

You code is fine if you change the printf calls. Just make shure that the address you specify is valid. NULL seems not to be writable.
Kurt

Can you give me an example of how I shall do if I want to specify the address to 0x12345678?
/Dan

Code:

---

#include <stdio.h>
#define INIT_ADR ((pointerFOO)0x12345678);
 
typedef struct FOO_tag{
    int    PER;
    int    PDR;
    int    PSR;
} FOO, *pointerFOO;
 
int main(void)
{
  pointerFOO pFOO = INIT_ADR;
 
  pFOO->PER = 1;
  return(0);
}

---

Kurt

It does not work, this was my main problem and I still get a segmentation fault when I run the code?
/Daniel

You can't just pick an address randomly like that.
If the underlying O/S uses virtual memory, that address must be on a mapped page (static data, stack, heap, whatever), otherwise you get a seg fault.
If there's no virtual memory, then you can indicate the linker where objects are located (for instance), but this is beyond C programming.

If you run that on linux I don't know of a way to find a valid constant for that address.
If you run it on an embedded system check the spec.
Kurt

An address that is valid on an embedded system (or even from inside the Linux kernel, which sees "real" memory versus "virtual" memory) probably isn't going to be valid in a user-space program in Linux. For example, you might be able to write to address 0x00000000 or 0x12345678 in your embedded system (you will have to check the documentation for the embedded system to be sure), but it most likely won't be a valid address in your Linux program. Software for embedded systems generally isn't portable like that.

You are right, thank you all for the valuble information. It would be very powerful if I could debug the C-code before it is loaded into the embedded system. I am running eclipse, do you know if there is any plugin which makes this possible to run some virtual memory instead of accessing the real memory of the computer?

/Dan

Why not declare a static array at the beginning of your program to simulate the computer memory, e.g.:

Code:

---


char RAM[1024]; /* 1Kb of RAM */
...
int main(void) {
  ...
}

---

Then if you want to set an object at a specific address, just do:

Code:

---

struct foo *p = (struct foo*)(RAM+offset); /* where offset is your custom address */

---

As a last note, be careful about alignment, the address must be compatible with the pointed type.

A better solution would be to use a simulator for whatever type of embedded system you're working on, but maybe it's expensive.

Using a bit of conditional compilation to isolate the machine dependencies should enable you to get a lot of testing done on the host.
Eg

Code:

---

#include <stdio.h>

typedef struct FOO_tag{
    int    PER;
    int    PDR;
    int    PSR;
} FOO;  /* don't typedef pointer types *pointerFOO; */

#if defined TARGET
#define INIT_ADR ((FOO*)0x12345678)
FOO *getMyFoo ( void ) {
    FOO *result = INIT_ADR;
    return result;
}
#elif defined HOST
FOO *getMyFoo ( void ) {
    static FOO test = {
        1, 2, 3 /* useful data */
    };
    return &test;
}
#else
#error neither HOST or TARGET specified
#endif

int main ( ) {
    FOO *p = getMyFoo();
    printf("PER=%d\n", p->PER );
    return 0;
}

---

$ gcc -DHOST foo.c
$ ./a.out
PER=1

That is actually a very nice and useful solution, thank you Salem :)