On Sun, Sep 10, 2017 at 11:37:29PM -0700, Christoph Hellwig wrote: > On Sun, Sep 10, 2017 at 09:50:27PM +0200, Sebastian Andrzej Siewior wrote: > > On 2017-08-22 23:19:31 [-0700], Christoph Hellwig wrote: > > > as part of the reflink feature. These out of place writes are > > > a requirement to implement data checksums, as otherwise your data > > > and checksum can not be updated atomically. Everything else is > > > beyond that at this point is vapourware. We'd have to decided > > So using the journal for atomic updates doesn't work then. Which means > > fragmentation unavoidable. > > Well, in theory you could also do data journalling - but in the end > out of place writes are probably going to be more efficient, and at > least for XFS also area already implemented. >From a quick look at the dm-integrity code, it looks like it is just a journalling block device. i.e. it logs both the data and the calculated CRC in it's internal journal before writing the data and CRC to their separate destinations so it can avoid inconsistent data/CRCs if a crash occurs. Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx -- To unsubscribe from this list: send the line "unsubscribe linux-xfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
