On Wed, Dec 13, 2023 at 9:44 AM Eduard Zingerman <eddyz87@xxxxxxxxx> wrote: > > On Tue, 2023-12-12 at 15:25 -0800, Andrii Nakryiko wrote: > > Add a test validating that freplace'ing another main (entry) BPF program > > fails if the target BPF program doesn't have valid/expected func proto BTF. > > > > Signed-off-by: Andrii Nakryiko <andrii@xxxxxxxxxx> > > --- > > Acked-by: Eduard Zingerman <eddyz87@xxxxxxxxx> > > I have two nitpicks, fill free to ignore. > When test is run with -vvv, verifier log says: > > -- BEGIN PROG LOAD LOG -- > func#0 @0 > Cannot replace static functions > processed 0 insns ... > -- END PROG LOAD LOG -- > > Would it be possible to match the error message in this test? Yes, if we add a bunch of extra log grabbing and matching logic to fexit_bpf2bpf test. Which, honestly, I just didn't want to touch more than I absolutely needed to. So I'll use your permission to ignore this. > Also, maybe kernel should be tweaked to be a bit more informative, > as message about static function is confusing, wdyt? > Currently the verifier doesn't distinguish between reasons for "unreliable". Not sure if it's worth tracking more information just for this. Certainly that feels like an orthogonal to this series improvement.
