Get struct address from struct "method"

**gna** · 01-14-2010

Hi,

if I have some struct thing like

Code:

struct Foo {
  void (*method)();
}

void foo_method() {}

// and later, in to-be-runned-code
struct Foo foo;
foo.method = &foo_method;
foo.method();

My question: Is there any chance for figuring out the memory address of `foo` in `foo_method` (which has been called from/as member of `foo`)?

I've highlighted "any" to emphasize that I'm not scared about "ugly" methods to do this, not about code execution right before the call, not by Assembler code execution and so on.

Thanks! :-)

P.S.: I'm not interested in using such hacks in productive code... it's just academic interest ;-)

**bithub** · 01-14-2010

The memory address of "foo" doesn't exist until foo_method is called and the local variables are allocated memory on the stack.

**gna** · 01-14-2010

Sorry, but I don't get

why the memory address of foo "doesn't exist until foo_method is called" (foo exists before, so its address exists, too?)
what I should do with the second information about the local variables on the stack.

**bithub** · 01-14-2010

Oh I see, you are calling foo_method after the variable is created. In that case, can't you just pass the address of "foo" to foo_method as a parameter? Or store the address in a local variable?

**quzah** · 01-14-2010

You don't need the & operator, since the name of a function, like the name of an array, can be used as a pointer to that object; at least in this case. Also, () in function definitions is different than () in function declaration. A () in a prototype isn't the same as () in the actual function.

Quzah.

**gna** · 01-14-2010

bithub, no, that would be too simple (and not the stuff I requested in my first question).

quzah, I know the & thing but I like to write that operator down explicitly, just for readability, you know. The other thing, you're right, it should have been `void (*method)(void);`.

**bithub** · 01-14-2010

Originally Posted by gna

bithub, no, that would be too simple (and not the stuff I requested in my first question).

Of course, simple is overrated.

If you really want, you can get the address of the stack pointer (using embedded assembly) and add or subtract (depending if your implementation's stack grows up or down) values to get the address of "foo" on the stack. Of course the offset from the stack pointer depends on a lot of factors which will change depending on what compiler you use, what architecture you are using, and what compiler optimizations are set. I ran a little test on my machine, and found that the magic number (in my case) was 56 bytes with optimizations turned off, and 4 bytes with them turned on.

**gna** · 01-15-2010

Originally Posted by bithub

Of course, simple is overrated.

Like said before, this is for academic/testing interest/purposes only.

If you really want, you can get the address of the stack pointer (using embedded assembly) and add or subtract (depending if your implementation's stack grows up or down) values to get the address of "foo" on the stack. Of course the offset from the stack pointer depends on a lot of factors which will change depending on what compiler you use, what architecture you are using, and what compiler optimizations are set. I ran a little test on my machine, and found that the magic number (in my case) was 56 bytes with optimizations turned off, and 4 bytes with them turned on.

So I would crawl through the call stack... but how do I recognize the memory range that belongs to the `foo` struct in the stack if there are other objects/variables stored?

**pointer_to_HELP** · 01-15-2010

hi gna, ur code is quite interesting but can u please expllain it a bit more so that i (am a beginner ;-) ) can understand it in a better way

**bithub** · 01-15-2010

Originally Posted by gna

but how do I recognize the memory range that belongs to the `foo` struct in the stack if there are other objects/variables stored?

Make the first member of "foo" to be an integer, and assign it a unique value. Then just scan the stack memory looking for that value in your function. Of course it is impossible to choose a value that is 100% unique on the stack, but that's why this is an exercise in futility

**Elysia** · 01-15-2010

Originally Posted by quzah

You don't need the & operator, since the name of a function, like the name of an array, can be used as a pointer to that object; at least in this case. Also, () in function definitions is different than () in function declaration. A () in a prototype isn't the same as () in the actual function.

Quzah.

I, too, consider explicitly using & good practice. I do believe the reason it isn't required on functions is due to backwards compatibility.

But gna, is there a reason you are using C for this? Is there no chance of using a language that supports member functions natively such as C++?

**gna** · 01-15-2010

bithub, hm, thanks, I'll try.

Elysia, yes I could, but like said before, that would be too simple.

**Elysia** · 01-16-2010

Oh, I must have missed that.

Thread: Get struct address from struct "method"

Thread Tools

Search Thread

Display

Get struct address from struct "method"

Similar Threads

help with structs and malloc!

Writing array, to file

I thought pointers were pointers...

Function validation.

Binary Search Tree