We study relation between stellar mass and halo mass for high-mass halos using a sample of galaxy clusters with accurate measurements of stellar masses from optical and IR data and total masses from X-ray observations. We find that stellar mass of the brightest cluster galaxies (BCGs) scales as M*BCG\propto M500^a_BCG with the best fit slope of a_BCG~0.35+-0.1 and scatter of M*BCG at a fixed M500 of ~0.2 dex. We show that M*-M relations from abundance matching or halo modelling reported in recent studies underestimate stellar masses of BCGs by a factor of ~2-4, because these studies used stellar mass functions (SMF) based on photometry that severely underestimates the outer surface brightness profiles of massive galaxies. We show that M*-M relation derived using abundance matching with the recent SMF calibration by Bernardi et al. (2013) based on improved photometry is in a much better agreement with the relation we derive. The total stellar mass of galaxies correlates with total mass M500 with the slope of \approx 0.6+-0.1 and scatter of 0.1 dex. This indicates that efficiency with which baryons are converted into stars decreases with increasing cluster mass. We show that for a fixed choice of the initial mass function (IMF) total stellar fraction in clusters is only a factor of ~3-5 lower than the peak stellar fraction reached in M\approx 10^12 Msun halos, and only a factor of ~1.5-3 if the IMF becomes progressively more bottom heavy with increasing mass in early type galaxies, as indicated by recent observations. The larger normalization and slope of the M*-M relation derived in this study shows that the overall efficiency of star formation in massive halos is suppressed much less than was thought before and that feedback and associated suppression of star formation in massive halos should be weaker than assumed in most of the current semi-analytic models and simulations.