On Thu, May 18, 2023 at 11:41:10AM +0200, Pablo Neira Ayuso wrote:
> On Wed, May 10, 2023 at 01:48:01PM +0200, Phil Sutter wrote:
> > Hi Pablo,
> >
> > On Wed, May 10, 2023 at 12:54:28PM +0200, Pablo Neira Ayuso wrote:
> > > On Wed, May 03, 2023 at 12:50:22PM +0200, Phil Sutter wrote:
> > > > This is a revert of commit 8d443adfcc8c1 ("evaluate: attempt to set_eval
> > > > flag if dynamic updates requested"), implementing the alternative
> > > > mentioned in the comment it added.
> > > >
> > > > Reason is the inconsistent behaviour when applying the same ruleset
> > > > twice: In the first call, the set lacking 'dynamic' flag does not exist
> > > > and is therefore added to the cache. Consequently, both the 'add set'
> > > > command and the set statement point at the same set object. In the
> > > > second call, a set with same name exists already, so the object created
> > > > for 'add set' command is not added to cache and consequently not updated
> > > > with the missing flag. The kernel thus rejects the NEWSET request as the
> > > > existing set differs from the new one.
> > >
> > > # cat test.nft
> > > flush ruleset
> >
> > Just remove this 'flush ruleset' call, then it should trigger.
>
> I cannot reproduce it yet :(
I can reproduce it. For the record, I have applied this alternate fix:
https://patchwork.ozlabs.org/project/netfilter-devel/patch/20230518125806.11100-1-pablo@xxxxxxxxxxxxx/
