The carbonization of the metallic liner and limiters of TEXTOR led to plasmas which contained less than 10 −5 metal, and, in best cases, (0.6 ± 0.4)% low Z ions per D +. However disruptions damaged some of the carbon film and evaporated metal atoms. These were redeposited on the carbon layer and became a source of impurities. Graphite limiters have been used since mid-1985. The metal pollution is not a problem for more than 1000 discharges after one carbonization. However the low Z contamination was initially five times larger than earlier. Surprisingly, the oxygen concentration had considerably increased. Experiments were performed to identify the sources of C and O and to improve the conditioning methods in this “all carbon” surrounding. The main tools used were: RG (radiofrequency assisted DC glow) discharges in H 2, D 2, and their mixtures with CH 4 and CD 4, the study of the H/D ratio in water and hydrocarbons and of the CO and CO 2 released during the ramp heating and cooling of limiters, liner, vessel and main portholes, and the modulation of the deuterium fluxes to the liner and limiters during tokamak discharges. The quality of tokamak plasmas measures the progress of the wall conditioning. We describe the optimized conditioning procedure which was applied after the most probable oxygen source had been drastically reduced.