Chris Murphy wrote:
> dmverity is a valid option geared explicitly for this and can include
> Reed-Solomon error correction;
Ouch, please no! Reed-Solomon is going to blow up the size of the live
images a lot, when all that's really needed is a simple checksum.
> and Btrfs. The Btrfs option is kinda need because a.) it's on-the-fly
> instead of all at once b.) it's every time a block is read, not one time,
> c.) can't be bypassed by the user, and d.) Btrfs supports overlays with
> the seed device function.
That's not quite the same thing as verification of the entire media. It
would also only be possible to replace the inner ext4 image with Btrfs, not
the outer SquashFS one, because Btrfs natively only supports zlib or LZO
compression, not the xz compression we are currently using. Switching (back)
to zlib would blow up the size of the live images a lot, and LZO is even
worse. The documentation says that LZMA/xz support is not currently being
considered in Btrfs. Btrfs also only compresses file contents, not metadata,
which is another reason an inner image would still be needed (to turn the
metadata into compressible contents). And whatever you do, the size and
performance impact of switching to Btrfs would have to be measured, it could
be significant in either direction.
Kevin Kofler
_______________________________________________
devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx
