On Sat, Jun 27, 2020 at 11:01:36PM +0200, Hans van Kranenburg wrote:
There's effectively no change in the implementation, other than documenting the 'in' semantics.
Although this is basically the same situation as in the LOGICAL_INO v1 and v2, the number of users of FS_INFO ioctl is presumably not high and the buffer has been write-only so far, there's no existing logic that would had to be tweaked.
Once the flags are there, all new implementations should take the semantics into account. Therefore I think this is a safe plan, but feel free to poke more holes to that.
In the V2 thread you mentioned generation, metadata_uuid, total_bytes as interesting missing ones. What about adding them just right now directly?
If you mean in the same patch, that would be mixing interface fix with interface extension, so no, unless there's a really good reason to do that in one patch.