On Wed, Jun 24, 2020 at 01:01:08PM +0000, Johannes Thumshirn wrote: > On 24/06/2020 14:21, Nikolay Borisov wrote: > > So the answer could go both ways, but this doesn't mean we shouldn't > > strive to build better/more robust interfaces. > > By changing the reserved16 field to a version, we could have both. I > just don't think there's much value add here. > > After Hans spoke up, I favor the flags field, as it's more flexible and > puts less assumptions on user-space. If the flag is set, field X is valid. Yes that's the idea and version does not make sense to me as it's too coarse. The whole problem here is that the default zeroing of the unused members does not help like for other members where zero usually does not make sense at all (like sectorsize, or the latest addition clone_alignment). So we need the bit to say 'this value is actually correct'.
