In this study, the problem of non-fragile passive filtering for a class of sampleded-data system with long time delay is addressed. The uncertain parameters are supposed to belong to norm-bounded uncertainties. A direct distribution processing methodology is developed to design a stable linear filter that assures asymptotic stability and a prescribed passive performance for the filtering error, in spite of the uncertainty and the long time-delay. The proposed algorithm is given in term of linear matrix inequality, whose feasibility and effectiveness has been shown by a numberical example.