For the Hoek Brown failure criterion, are the values of mi, s, and a for the intact rock or for the rock mass?

This is a very good question.  If you measure the GSI for the entire rock mass, and use the corresponding Generalized Hoek-Brown parameters in a numerical analysis which also includes the discrete joints, then you are basically double counting the effect of the joints.  When using discrete joints in RS2, the Generalized Hoek-Brown properties should be of the rock between the joints.  If the rock between the joints is intact rock, then use the intact properties for it.  If you are modeling only major joints or faults in RS2 then the material between those interfaces may not be intact rock.  In that case, determine a GSI value for that material between the joints but not including the major joints that are modeled discretely.

For intact rock:  mb=mi; s=1.0; a=0.5

In RocData, the Use Lab Data option computes the intact strength.  Setting GSI=100 sets the rock mass strength to be equal to the intact strength.