On Sat, Aug 06 2022 at 2:09P -0400, Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote: > On Fri, Aug 5, 2022 at 12:10 PM Mike Snitzer <snitzer@xxxxxxxxxx> wrote: > > > > All said: I think it worthwhile to merge these changes for 6.0, rather > > than hold until 6.1, now that we have confidence this _optional_ DM > > verity feature is working as expected. Please be aware there was a > > small linux-next merge fixup needed: > > https://lore.kernel.org/all/20220805125744.475531-1-broonie@xxxxxxxxxx/T/ > > Well, more importantly, the verity_target version numbers clash. > > I used the newer "{1, 9, 0}" version number, but if you want it to be > "{1, 9, 1}" to show that it's a superset of the previous one, you > should do that yourself. You did the right thing. > That said, the best option would be to remove version numbers > entirely. They are a completely broken concept as an ABI, and *never* > work. > > Feature bitmasks work. Version numbers don't. Version numbers > fundamentally break when something is backported or any other > non-linearity happens. > > Please don't use version numbers for ABI issues. Version numbers are > for human consumption, nothing more, and shouldn't be used for > anything that has semantics. Yes, I know you mentioned this before and I said I'd look to switch to feature bitmasks. Yet here we are. Sorry about that, but I will take a serious look at fixing this over the next development cycle(s). There is just quite a bit of innertia in these version numbers across all the disparate userspace tools that use DM. So the transition needs some design, planning and coordination but I'll get it done. Really ;) Thanks, Mike
