Hi. Worrying about stale data reads, i.e. no internal corruption, but storage reading back data which was confirmed as rewritten. Here's a relatively simple patch to annotate the iolog with checksums for verification. I patched this into fio-2.0.11, and for now it's more of an rfc than calling for inclusion. Some things I've been wondering about: * Metadata mode is nice for debugging. Currently, metadata writes disable checksumming, which weakens integrity checks. So I'm wondering if the two shouldn't rather have nested. * This (s/w obviously) won't work with async verification: collisions between verify reads and generated re-writes are not tracked. I guess one could consider letting async mode consume an rbtree'd block map, maybe add generation ids to the block headers, and detect collisions. * Requires additional host memory, according to backlog size. Didn't add an option turn it on/off yet. More thoughts? Thanks, Daniel -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
