Restorer line is one of the key factors in hybrid rice production, and the knowledge about its genetic diversity may facilitate rice heterosis application. The restorer resources of three-line hybrid rice in China are rich in quantity, but the overall knowledge of these resources are less, although some results regarding the genetic diversity of restorers with a small quantity of materials have been reported since 1990s. To understand the genetic diversity of the restorers of three-line hybrid rice at a whole resource level, we collected 128 accessions of restorer line from the 6 rice eco-regions and 16 provinces in China in the present study, among which 101 were indica and 27 japonica. The materials could be divided into three types, WA (92), HL (8), and BT (28) types respectively, based on sterile-and-restoring relation. Thirty six pairs of microsatellite (SSR) markers covering the 12 rice chromosomes evenly and 9 phenotypic traits including days of heading, growth duration, plant height, spike weight, spike length, grains per spike, 1 000-grain weight, grain length, and grain length/width ratio were used to reveal the genetic diversity of the materials. The neighbor-jointing cluster diagram based on SSR results showed that the 128 accessions were distributed in 6 clusters, including 44, 42, 28, 3, 7, and 5 accessions in 1 to 6 clusters, respectively. Most of the indica (83.2%) and WA type (82.6%) were distributed in clusters 1 and 2, the japonica (81.5%) and BT type (78.6%) in cluster 3, and HL type (75%) in cluster 1. Materials of IR 24 progeny distributed in 4 clusters, while those of Minghui 63 progeny were all in cluster 1. In the 128 accessions, a total of 281 alleles were found with an average of 8.0 per locus by SSR markers. The alleles were 265 and 186 in indica and japonica subspecies with the average of 7.4 and 5.2 per locus, respectively. The average genetic diversity index (He) of total materials was 0.6190, in which indica (0.5770)>japonica (0.5656) with no significant difference (P>0.05), BT (0.5816)>WA (0.5705)>HL (0.4989) types with no significance (P>0.05) between BT and WA types. The He values lined as eco-region Ⅱ>eco-region Ⅲ>eco-region Ⅵ>eco-region Ⅰ>eco-region Ⅴ> eco-region Ⅳ, in which both eco-regions Ⅱ and Ⅳ were significant (P<0.01) with others. Two Hunan (0.5615) and Sichuan (0.5236), which had comparative sample size, showed significant (P<0.01) difference on He value. Detected by the 9 phenotypic traits, the Shannon-weiner indices (I) were 1.3746 (indica) and 1.3789 (japonica), BT (1.4026)>WA (1.3567)>HL (1.1732) types, rice region Ⅱ (1.3921)>region Ⅲ (1.1209)>region Ⅴ (1.0570)>region Ⅰ (0.9954)>region Ⅵ (0.7818)>region Ⅳ (0.7411). Phenotypic and SSR data showed similar results with the correlation coefficient of 0.9981* (by subspecies), 0.9418 (by sterile-restoring relation), and 0.8835** (by rice eco-regions), respectively, indicating the accordant evolution between genotype and phenotype.