Impact ejecta from the surfaces of some small planetary satellites (such as martian moon Phobos) should give rise to dust belts surrounding the satellites' orbits. The mechanism of dust belt formation is, principally, as follows. A satellite is subjected to impacts of meteoroids with encounter velocities of about 10 km sec.^-1. The total mass of the ejecta is 3 to 4 orders of magnitude greater than the mass of a projectile. A great number of excavated particles is thrown to planetocentric trajectories close to the satellite orbit. This results in the generation of a dust toroid with an axial line coinciding with the orbit of the parent body. Simultaneously the grains are removed from the torus by the satellite. For not very small grains, their removal by the satellite is the major loss mechanism, so that the matter injection into the torus and the reaccretion come to a dynamical equilibrium at the increased density level. We study this mechanism in detail, construct the general model of a dust torus, and apply it to the Phobos case as the most typical and important in view of the martian space missions planned for the near future. Postulating a stationary meteoroidal flux to the Phobos surface and assuming the most probable regolith properties and impact parameters, we specify the model and, performing semianalitical evaluations, come to the following conclusions. The total mass of the Phobos dust belt is several times 10⁴ kg. The lifetime of a particle inside the torus should be about 1 year. The matter density reaches a maximal value at the Phobos orbit, being 10⁶ to 10⁷ times the background density of the interplanetary medium. The geometry of the equidensity curves in the meridional section of the torus is investigated. All estimations are probably within one order of magnitude accuracy due to the physical data uncertainties and errors of the simulation. Our results are valid for macroscopic particles, greater than about 100 μm in size. The fine-dust component of the belt, affected strongly by radiative forces and charging, should be the subject of a separate study. Ways of refining the model and directions of further investigation are discussed.