Re: [PATCH] kcm: use BPF_PROG_RUN

[Daniel Borkmann]

On Mon, Sep 23, 2019 at 02:31:04PM -0700, Eric Dumazet wrote:
> On 9/6/19 10:06 AM, Alexei Starovoitov wrote:
> > On Fri, Sep 6, 2019 at 3:03 AM Yonghong Song <yhs@xxxxxx> wrote:
> >> On 9/5/19 2:15 PM, Sami Tolvanen wrote:
> >>> Instead of invoking struct bpf_prog::bpf_func directly, use the
> >>> BPF_PROG_RUN macro.
> >>>
> >>> Signed-off-by: Sami Tolvanen <samitolvanen@xxxxxxxxxx>
> >>
> >> Acked-by: Yonghong Song <yhs@xxxxxx>
> > 
> > Applied. Thanks
> 
> Then we probably need this as well, what do you think ?

Yep, it's broken. 6cab5e90ab2b ("bpf: run bpf programs with preemption
disabled") probably forgot about it since it wasn't using BPF_PROG_RUN()
in the first place. If you get a chance, please send a proper patch,
thanks!

> diff --git a/net/kcm/kcmsock.c b/net/kcm/kcmsock.c
> index 8f12f5c6ab875ebaa6c59c6268c337919fb43bb9..6508e88efdaf57f206b84307f5ad5915a2ed21f7 100644
> --- a/net/kcm/kcmsock.c
> +++ b/net/kcm/kcmsock.c
> @@ -378,8 +378,13 @@ static int kcm_parse_func_strparser(struct strparser *strp, struct sk_buff *skb)
>  {
>         struct kcm_psock *psock = container_of(strp, struct kcm_psock, strp);
>         struct bpf_prog *prog = psock->bpf_prog;
> +       int res;
>  
> -       return BPF_PROG_RUN(prog, skb);
> +       preempt_disable();
> +       res = BPF_PROG_RUN(prog, skb);
> +       preempt_enable();
> +
> +       return res;
>  }
>  
>  static int kcm_read_sock_done(struct strparser *strp, int err)