On Fri, Feb 07, 2025 at 05:15:42AM +0100, Christoph Hellwig wrote: > On Thu, Feb 06, 2025 at 12:13:51PM -0800, Darrick J. Wong wrote: > > Hmm. So this is to support doing COW on non-reflink zoned filesystems? > > Yes. > > > How then do we protect the refcount intent log items from being replayed > > on an oler kernel? > > There are no refcount intent log items for this case. Er... right, there aren't any refcount log items because there's no refcount btree. Looking through the rest of the code, though, the same question about protection against recovery by old kernels applies to the bmap intent item logged in xfs_zoned_map_extent. And I think it has the same answer -- zoned storage is an incompat feature bit that's newer than the bmap intent items, so we're protected. I got there clumsily, but here I am :P Reviewed-by: "Darrick J. Wong" <djwong@xxxxxxxxxx> --D
