On Wed, Feb 15, 2023 at 1:55 PM Ilya Leoshkevich <iii@xxxxxxxxxxxxx> wrote: > > In any case, I don't have a really strong opinion here, just explaining > my thought process. If the consensus is to drop it, I would not mind > dropping it at all. I think doing: + desc = find_kfunc_desc(prog, func_id, offset); + if (!desc) + return -EFAULT; would be enough to catch that verifier issue. We have similar code in several places. When we can print verbose() into the verifier log, we do. In other cases we just EFAULT which is a sign that something wrong with the verifier. We don't use WARN in all those cases.
