Cosmic far-infrared background (CFIRB) is a powerful probe of the history of star formation rate (SFR) and the connection between baryons and dark matter across cosmic time. In this work, we explore to which extent the CFIRB anisotropies can be reproduced by a simple physical framework for galaxy evolution, the gas regulator (bathtub) model. This model is based on continuity equations for gas, stars, and metals, taking into account cosmic gas accretion, star formation, and gas ejection. We model the large-scale galaxy bias and small-scale shot noise self-consistently, and we constrain our model using the CFIRB power spectra measured by Planck. Because of the simplicity of the physical model, the goodness of fit is limited. We compare our model predictions with the observed correlation between CFIRB and gravitational lensing, bolometric infrared luminosity functions, and submillimetre source counts. The strong clustering of CFIRB indicates a large galaxy bias, which corresponds to haloes of mass $10^{12.5} M_\odot$ at z=2, higher than the mass associated with the peak of the star formation efficiency. We also find that the far-infrared luminosities of haloes above $10^{12} M_\odot$ are higher than the expectation from the SFR observed in ultraviolet and optical surveys.