Context. As a ubiquitous phenomenon, large-amplitude longitudinal filament oscillations usually decay in 1-4 periods. Recently, we observed a decayless case of such oscillations in the corona. Aims. We try to understand the physical process that maintains the decayless oscillation of the filament. Methods. Multi-wavelength imaging observations and magnetograms are collected to study the dynamics of the filament oscillation and its associated phenomena. To explain the decayless oscillations, we also perform one-dimensional hydrodynamic numerical simulations using the MPI-AMRVAC code. Results. In observations, the filament oscillates decaylessly with a period of 36.4 ± 0.3 min for almost 4 hours before eruption. During oscillations, four quasi-periodic jets emanate from a magnetic cancellation site near the filament. The time interval between neighboring jets is ∼ 68.9 ± 1.0 min. Numerical simulations constrained by the observations reproduced the decayless longitudinal oscillations. However, it is surprising to find that the period of the decayless oscillations is not consistent with the pendulum model. Conclusions. We propose that the decayless longitudinal oscillations of the filament are maintained by quasi-periodic jets, which is verified by the hydrodynamic simulations. More importantly, it is found that, when driven by quasi-periodic jets, the period of the filament longitudinal oscillations depends also on the driving period of the jets, not simply the pendulum period. With a parameter survey in simulations, we derived a formula, by which one can derive the pendulum oscillation period using the observed period of decayless filament oscillations and the driving periods of jets.