> > seems wrong. Does it cause an actual problem though? Back when that code was written and submitted, ISTR doing a leak analysis and finding none on start_io() paths or their cleanup siblings. Did something change since then? I don't recall problems being reported in this area over the past several years, either... - Dave