With very long lenses the popular methods are difficult due to the distances involved. I do it outside, shooting a brick wall with a magazine page stuck to it with BluTack. Test at the closer end of the kind of distances you usually shoot at, as Bob suggests, though with a lens of that quality if there was any significant difference with distance I would send it back to Canon and get them to recalibrate it.
Always test at lowest f/number, centre AF point, rock steady tripod, mirror-lockup, IS off, square to wall. Bubble-level in the hot shoe helps, though within a couple of inches in set-up accuracy doesn't make much odds with a long lens as the change in angles is very slight. If you can light the target with flash at a lower power setting, for very brief durations, so much the better.
Shoot with AF micro adjust at say +10, +5, 0, -5, -10. Do several shots at each setting, defocusing each time. Check for sharpness, often the LCD on max magnification is good enough at this stage. Pick the sharpest and do the same again at finer increments and view on PC.
With any luck you should then be very close to optimum and when that's decided, do a final test shooting a series of six or more, defocusing each time, to check for consistency.
Then you might want to do the same at a much greater distance, like a house at the end of the road. So long as you are square to the target all will be well.
Edit: Forgot to add, a quick check comparing against live view AF, that should always be dead accurate, can save a lot of messing about
