On shallow-marine carbonate buildups in south China, Turkey, and Japan, uppermost Permian skeletal limestones are truncated by an erosional surface that exhibits as much as 10 cm of topography, including overhanging relief. Sedimentary facies, microfabrics, carbon isotopes, and cements together suggest that erosion occurred in a submarine setting. Moreover, biostratigraphic data from south China demonstrate that the surface postdates the uppermost Permian sequence boundary at the global stratotype section and truncates strata within the youngest known Permian conodont zone. The occurrences of similar truncation surfaces at the mass-extinction horizon on carbonate platforms across the global tropics, each overlain by microbial buildups, and their association with a large negative excursion in δ13C further suggest a causal link between erosion of shallow-marine carbonates and mass extinction. Previously proposed to account for marine extinctions, the hypothesis of rapid carbon release from sedimentary reservoirs or the deep ocean can also explain the petrographic observations. Rapid, unbuffered carbon release would cause submarine carbonate dissolution, accounting for erosion of uppermost Permian skeletal carbonates, and would be followed by a pulse of high carbonate saturation, explaining the precipitation of microbial limestones containing upward-growing carbonate crystal fans. Models for other carbon-release events suggest that at least 5 × 1018 g of carbon, released in <100 k.y., would be required. Of previously hypothesized Permian-Triassic boundary scenarios, thermogenic methane production from heating of coals during Siberian Traps emplacement best accounts for petrographic characteristics and depositional environment of the truncation surface and overlying microbial limestone, as well as an associated carbon isotope excursion and physiologically selective extinction in the marine realm.