It has been recognized long ago that there is a conceptual difficulty in thermodynamics of plasticity (Bridgman,1950). Clausius introduced the notion of entropy using the assumption that any two states of the body can be connected by a reversible process. This assumption works fine for gases and fluids but for metals it does not: in vicinity of every state of a metal there are states which can be reached only by irreversible processes. By a state one means here the physical state of material, i.e. crystal structure along with all its defects. This difficulty has practical consequences: it certainly contributed to failing of all attempts to construct the universal constitutive equations of metals that capture basic simultaneous interacting processes of plastic deformation, work hardening, recovery, and creep. In cases when the major deformation mechanism is pure dislocational, the resolution of the difficulty can be sought in averaging of equations of dislocation dynamics. This is the subject of the talk. I will address four questions: a. How do additional thermodynamic parameters, entropy and temperature of dislocation structure, arise in averaging of dislocation dynamics? b. What is the physical meaning of microstructure temperature? c. Why is stored energy of dislocation structure so small in apparent contradiction to the fact that nearly entire resistance to plastic deformation (in FCC metals) is due to dislocation interactions? d. Which constraints are imposed by dislocation dynamics on constitutive equations of plasticity?