kern/139072: [zfs] zfs marked as production ready but it used a deprecated checksum algorithm

[James R. Van Artsdalen]

Olivier Smedts wrote:
> Hello,
>
> Now that this PR is closed, is there something to change on *existing*
> zfs filesystems to make them use fletcher4 (for new data) when they
> have the default property "checksum=on"?

# zfs set checksum=fletcher4 pool

>  Is there something to do
> (other than dumping and restoring) to change checksums to fletcher4
> for existing data and metadata ?

No.

Even "fletcher4" has the undesirable property that the checksum of every
group of zeros, of any length, is the same as the initial value of the
accumulator.  This means that fletcher4 is insensitive to the number of
leading zeros in the checksummed data.

The ZFS team needs to revisit the checksum issue and add another
algorithm but they have other things to worry about at the moment.  Some
SHA-3 contestants claim to be very fast though it's not clear they're
fast enough to replace a true Fletcher sum in the real world, at least
not yet.