Log whether a global variable has been read or written

Question

Requirement:

Given a C program I have to identify whether the functions accessing global variables are reading them or writing them.

Example code:

#include <stdio.h>

/* global variable declaration */
int g = 20;

int main()
{

    /* writing the global variable */
    g = 10;

    /* reading the global variable */
    printf ("value of g = %d\n",  g);

    return 0;
}

Executing the above code I want to generate a log file in the below format:

1- Global variable a written in function main() "TIME_STAMP"
2- Global variable a read in function main() "TIME_STAMP"

---

Research:

I am cetainly able to acheive this by doing a static analysis of source code as per below logic:

• Go through the c code and identify the statements where the global variable is read.
• Then analysis the c code statement to identify if it is a read or write statement.(Checking if ++ or -- operator is used with global variable or any assignemnt has been made to the global variable)
• Add a log statement above the identified statement which will execute along with this statement execution.

This is not a proper implementation.

Some studies:

I have gone through how debuggers are able to capture information.

Some links in the internet: How to catch a memory write and call function with address of write

Answer 1

Not completely answering your question, but to just log access you could do:

#include <stdio.h>

int g = 0;

#define g (*(fprintf(stderr, "accessing g from %s. g = %d\n", __FUNCTION__, g), &g))

void foo(void)
{
  g = 2;
  printf("g=%d\n", g);
}

void bar(void)
{
  g = 3;
  printf("g=%d\n", g);
}

int main(void)
{
  printf("g=%d\n", g);
  g = 1;
  foo();
  bar();
  printf("g=%d\n", g);
}

Which would print:

accessing g from main. g = 0
g=0
accessing g from main. g = 0
accessing g from foo. g = 1
accessing g from foo. g = 2
g=2
accessing g from bar. g = 2
accessing g from bar. g = 3
g=3
accessing g from main. g = 3
g=3

Answer 2

Below is the way i solved this problem:

• I created a utility(In java) which works as below(C program source file is the input to my utility):
  • Parse the file line by line identifying the variables and functions.
  • It stores global variables in a separate container and look for lines using them.
  • For every line which access the global variable i am analyzing them identifying whether it is a read operation or write operation(ex: ==, +=, -+ etc are write operation).
  • For every such operation i am instrumenting the code as suggested by @alk(https://stackoverflow.com/a/41158928/6160431) and that in turn will generate the log file when i execute the modified source file.

I am certainly able to achieve what i want but still looking for better implementation if anyone have.

For further discussion if anybody want we can have have a chat.

I refer the source code and algos from the below tools:

http://www.dyninst.org/

https://software.intel.com/en-us/articles/pin-a-dynamic-binary-instrumentation-tool