The linearity of most operators you cite is easy to check because they "operate" on the result of the function, so the answer to "is the operator linear" is pretty much the same as "is the operation applied after the function f linear".
That is:
v -> a*v+b is linear, and so is
f -> a*f+b
The exception in your examples is
f -> df/dx
which is not a transformation applied on individual output values of f. Though, you can check what happens if you scale f or if you add g to it.
Operators are much more general and a vast set than the above examples suggest. In the quest of distribution theories, Radon (a person I guess, not the element) measures were introduced to add local infinite values (=Dirac pulses) to functions, but they would not include an operator like
f(x, y) -> f(y, x) which swaps the arguments
this operator being perfectly legitimate in distribution theories
http://en.wikipedia.org/wiki/Distribution_(mathematics)