On Wed, Jun 05, 2019 at 12:43:44PM +0900, Lorenzo Colitti wrote: > On Wed, Jun 5, 2019 at 12:29 PM Hangbin Liu <liuhang...@gmail.com> wrote: > > > We rely on being able to add a rule and either have a dup be created > > > (in which case we'll remove it later) or have it fail with EEXIST (in > > > which case we won't remove it later). > > > > With Maciej said, how about add NLM_F_EXCL flag when you add a new rule. > > If it returned EEXIST, which means there is an dup rule, you just do not > > remove it later. > > > > Would that fix your issue? > > We can't do that without rewriting our code and making it more > complex. The way the code is structured is that an update is "add all > new rules; delete all old rules". To do what you suggest we would need > to either change that to "for rule in rules; add newrule; delete > oldrule" or we'd need to keep state on which rules already existed.
Hmm...Generally speaking we need to check the cmd's return value, or why it exists. > > The previous behaviour provided semantics that are useful to > userspace, and this commit broke those semantics. Please revert. Keep two exactally same rules in kernel looks strange and doesn't make sense to me. But let's see what does David Ahern think, he is more experienced in this part :) Thanks Hangbin
