>
> 在 2021/4/21 19:43, Ananyev, Konstantin 写道:
> >>
> >> 'rd->u.max = rd->u.max' is meaningless which should be deleted.
> >>
> >> This patch fixed it.
> >>
> >> Fixes: 8021917293d0 ("bpf: add extra validation for input BPF program")
> >> Cc: sta...@dpdk.org
> >
> > As I remember, I did it on purpose.
> > Some old (but still supported) version of clang complained
> > about unitialized variable.
>
> Well, how about like this ?
> uint64_t tmp = rd->u.max;
> if (op == BPF_MOD)
> tmp = RTE_MIN(tmp, tmp - 1);
> rd->u.max = tmp;
Then, I think it should be:
tmp = RTE_MIN(tmp, rs->u.max - 1);
Or just leave things as they are right now.
Nothing is broken here.
> >
> >> Signed-off-by: Min Hu (Connor) <humi...@huawei.com>
> >> ---
> >> lib/librte_bpf/bpf_validate.c | 2 --
> >> 1 file changed, 2 deletions(-)
> >>
> >> diff --git a/lib/librte_bpf/bpf_validate.c b/lib/librte_bpf/bpf_validate.c
> >> index 9214f15..c5ad951 100644
> >> --- a/lib/librte_bpf/bpf_validate.c
> >> +++ b/lib/librte_bpf/bpf_validate.c
> >> @@ -517,8 +517,6 @@ eval_divmod(uint32_t op, struct bpf_reg_val *rd,
> >> struct bpf_reg_val *rs,
> >> } else {
> >> if (op == BPF_MOD)
> >> rd->u.max = RTE_MIN(rd->u.max, rs->u.max - 1);
> >> - else
> >> - rd->u.max = rd->u.max;
> >> rd->u.min = 0;
> >> }
> >>
> >> --
> >> 2.7.4
> >
> > .
> >
