Aim: This study adopted a three-wave random intercept cross-lagged panel model to explore the longitudinal reciprocal relationships between (a) the teacher–student relationship (TSR) quality and (b) the parent–child relationship (PCR) quality and positive affect among Chinese primary school students. Samples: Two primary school student samples, including 3505 and 2505 students, were tracked with their perceived relationship quality with math teachers and parents and their positive affect levels in mathematics learning over three academic years. Results: The results demonstrated that more closeness with parents could significantly predict students' subsequent higher level of positive affect in mathematics learning. However, more closeness with their math teachers did not show significant prediction. Meanwhile, more conflict with math teachers and parents could significantly predict their subsequent lower degree of perceived positive affect in mathematics learning. That is, a reciprocal association lines in the PCR quality and positive affect, whereas only a unidirectional association exists between the TSR and positive affect. The predictions of the experienced positive affect on their perceived interpersonal relationships with math teachers and parents were stronger than those in the reverse association. Conclusions: This study identifies that while the effects of closeness with math teachers and parents on positive affect in students' math learning differ, conflict with math teachers and parents indeed harms students' experienced positive affect in math learning. More attention should also be paid to fostering positive affect in math learning.