The body of the function will consist in large part of a loop that will compute a root of f to some tolerance. The first question that arises when contemplating a loop implementation should be ``What condition will control the loop?'' And to answer that question, you must answer the question ``What must be true when the loop terminates?''
This is an important point. If we want the difference between pos and neg to be less than or equal to 1e-5 after the loop terminates, we must make sure that the loop continues to run as long as the difference is greater than 1e-5. Now what do we know about the implementation of bisection?
Note that fabs is the C function for finding the absolute value of a floating-point number.