On 2019/1/4 上午11:43, Chris Murphy wrote: > On Thu, Jan 3, 2019 at 6:32 PM Qu Wenruo <[email protected]> wrote: >> >> >> >> On 2019/1/4 上午9:15, Chris Murphy wrote: >>> If you use btrfs-image -ss option, there won't be any sensitive >>> information included. Files are hashed. Some short name files or dirs >>> can't be hashed (you'll see a warning) and those are replaced with >>> random garbage instead. There is only metadata with the image, no user >>> data. >> >> Please don't advice -ss for btrfs-image. It's super slow and will easily >> cause tons of noise when running btrfs check on it. > > I thought in recent btrfs-progs that -ss was quite a bit faster. The > last time I used it I don't think it was much slower than -s. Anyway I > don't have a preference between hash or garbage, but I don't think > it's realistic for users to give out filenames in images. So is -s OK?
For -s/-ss the problem is, btrfs check will report tons of hash mismatch, and recent kernel will even refuse to mount due to enhanced dir item hash check. The extra problem of -ss is, it can't ensure all filenames to have a conflict hash, so it's not that much different from -s. For User don't want to give any filenames, then don't give any files names. We could find a way to get needed info just using btrfs ins dump-tree, especially for this case where only one error is reported. Thanks, Qu > > >
signature.asc
Description: OpenPGP digital signature
