On Mon, Aug 31, 2009 at 09:21, Thomas Bächler<thomas@xxxxxxxxxxxxx> wrote: > Theodore Tso schrieb: >> >> Hmm, well *that's* a change between the blkid that was in e2fsprogs >> and the version that was in util-linux-ng. In e2fsprogs' blkid, I >> stopped after I found the first match. As a result, the filesystems >> that had the most sophisticated (and thus less prone to false >> positives) were put *first*, and filesystems that had sloppy >> mkfs/mkswap/dmcrypt-setup programs that didn't zap the first and last >> 32k-64k of the partition to prevent false positives were placed later >> in the detection table. >> >> If the new blkid library in util-linux-ng searches the whole list and >> finds the *last* match, and it's using the same order of probes as in >> e2fsprogs' blkid, it means there will be a significant chance in >> behavior, and the danger is that there may be many more false >> positives. > > I think you read me wrong: It's not returning the last result. When it > finishes all the probes, it checks if there was exactly one result and > prints that. If there was more than one result, it discards them all and > returns no result at all. This will lead to no false positives at all, but > lots of false negatives with LUKS volumes created with old cryptsetup > versions. Right, and it's the expected behavior. We can not afford to return anything we are not sure about. It may result in serious data corruption and data loss to return the wrong type of filesystem. Therefore we don't return anything if we find conflicting signatures, and require the user to fix the metadada and remove invalid conflicting signatures from the device to make the auto-detection working. Or not to rely on auto-probing and use the specified values from fstab. Todays systems rely heavily on auto-detection and we can no longer play wild guessing games here like we did in the past, and risk to destroy data by returning the wrong guess. A logic which relies on any probing order for conflicting signatures is seriously broken. Thanks, Kay -- To unsubscribe from this list: send the line "unsubscribe util-linux-ng" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
