Aims: To investigate the association of C-reactive protein (CRP) gene single nucleotide polymorphisms (SNPs), additional gene–gene, and gene–smoking interaction with ischemic stroke (IS) risk.Methods: Logistic regression is performed to investigate association between SNPs within CRP gene and IS risk. Generalized multifactor dimensionality reduction (GMDR) was used to analyze the gene–gene and gene–smoking interaction, cross-validation consistency, the testing balanced accuracy and the sign test were calculated.Results: Logistic analysis showed that three SNPs were all associated with decreased IS risk in additive and dominant models. The IS risks were lower in carriers of homozygous mutant of rs2794521 polymorphism and heterozygous of rs3093059 and rs1205 than those with wild-type homozygotes genotype, OR (95%CI) were 0.62 (0.40–0.90), 0.68 (0.50–0.96) and 0.65 (0.46–0.97), respectively. GMDR analysis suggested a significant two-locus model (P = 0.0010) involving rs2794521 and rs3093059. We also found a significant two-locus model (P = 0.0010) involving rs2794521 and smoking. Participants with rs2794521-AG or GG and rs3093059-AG or GG genotype have the lowest IS risk, compared to participants with rs2794521-AA and rs3093059-AA genotype, OR (95%CI) was 0.4 2 (0.233–0.61). In addition, non-smokers with rs2794521-AG or GG genotype have the lowest IS risk, compared to smokers with rs2794521-AA genotype, OR (95%CI) was 0.47 (0.23–0.76).Conclusions: We found that rs2794521, rs3093059, and rs1205 were associated with decreased IS risk; we also found that gene–gene interaction between rs2794521 and rs3093059, and gene–environment interaction between rs2794521 and smoking were associated with decreased IS risk.