On Mon, May 04, 2020 at 11:59:16PM +0200, Richard Weinberger wrote: > ----- Ursprüngliche Mail ----- > > Von: "Johannes Thumshirn" <jth@xxxxxxxxxx> > > An: "David Sterba" <dsterba@xxxxxxx> > > CC: "linux-fsdevel" <linux-fsdevel@xxxxxxxxxxxxxxx>, "linux-btrfs" <linux-btrfs@xxxxxxxxxxxxxxx>, "Eric Biggers" > > <ebiggers@xxxxxxxxxx>, "richard" <richard@xxxxxx>, "Johannes Thumshirn" <johannes.thumshirn@xxxxxxx>, "Johannes > > Thumshirn" <jthumshirn@xxxxxxx> > > Gesendet: Dienstag, 28. April 2020 12:58:58 > > Betreff: [PATCH v2 1/2] btrfs: add authentication support > > > From: Johannes Thumshirn <johannes.thumshirn@xxxxxxx> > > > > Add authentication support for a BTRFS file-system. > > > > This works, because in BTRFS every meta-data block as well as every > > data-block has a own checksum. For meta-data the checksum is in the > > meta-data node itself. For data blocks, the checksums are stored in the > > checksum tree. > > Eric already raised doubts, let me ask more directly. > Does the checksum tree really cover all moving parts of BTRFS? > > I'm a little surprised how small your patch is. > Getting all this done for UBIFS was not easy and given that UBIFS is truly > copy-on-write it was still less work than it would be for other filesystems. The patch is small because the amount if cross-referencing between the structures and "noise" in the structures is assumed to be sufficient so just the calculation of the new checksum needs to be added. Using 'assumed' must naturally raise eyebrows, what we all want is a proof that it is so, and I believe this is the core of the work here but it's missing so we unfortunatelly have to take the rounds in this thread and actually dig out the details. The hmac support won't be merged without making things clear and documented.
