On Sun, Mar 22, 2020 at 03:21:57AM +0100, Stefano Brivio wrote: > Phil reports that inserting an element, that includes a concatenated > range colliding with an existing one, fails silently. > > This is because so far set back-ends have no way to tell apart cases > of identical elements being inserted from clashing elements. On > insertion, the front-end would strip -EEXIST if NLM_F_EXCL is not > passed, so we return success to userspace while an error in fact > occurred. > > As suggested by Pablo, allow back-ends to return -ENOTEMPTY in case > of partial overlaps, with patch 1/4. Then, with patches 2/4 to 4/4, > update nft_set_pipapo and nft_set_rbtree to report partial overlaps > using the new error code. Applied, thanks.
