Re: [RFC PATCH bpf-next 2/8] bpf: Support ->fill_link_info for kprobe_multi

[Yafang Shao <laoar.shao@gmail.com>]

On Mon, May 29, 2023 at 8:49 PM Jiri Olsa <olsajiri@gmail.com> wrote:
>
> On Sun, May 28, 2023 at 02:20:21PM +0000, Yafang Shao wrote:
>
> SNIP
>
> > diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> > index 0d84a7a..00a0009 100644
> > --- a/kernel/trace/bpf_trace.c
> > +++ b/kernel/trace/bpf_trace.c
> > @@ -2564,10 +2564,41 @@ static void bpf_kprobe_multi_link_show_fdinfo(const struct bpf_link *link,
> >       }
> >  }
> >
> > +static int bpf_kprobe_multi_link_fill_link_info(const struct bpf_link *link,
> > +                                             struct bpf_link_info *info)
> > +{
> > +     struct bpf_kprobe_multi_link *kmulti_link;
> > +     u64 *uaddrs = u64_to_user_ptr(info->kprobe_multi.addrs);
> > +     u32 ucount = info->kprobe_multi.count;
> > +     int i;
> > +
> > +     if (!uaddrs ^ !ucount)
> > +             return -EINVAL;
> > +
> > +     kmulti_link = container_of(link, struct bpf_kprobe_multi_link, link);
> > +     if (!uaddrs) {
> > +             info->kprobe_multi.count = kmulti_link->cnt;
> > +             return 0;
> > +     }
> > +
> > +     if (!ucount)
> > +             return 0;
> > +
> > +     if (ucount != kmulti_link->cnt)
> > +             return -EINVAL;
> > +
> > +     for (i = 0; i < ucount; i++)
> > +             if (copy_to_user(uaddrs + i, kmulti_link->addrs + i,
> > +                              sizeof(u64)))
> > +                     return -EFAULT;
>
> let's use put_user instead copy_to_user? or even better why not
> copy that with single copy_to_user from kmulti_link->addrs
>

Good point. I will utilize a single copy_to_user instead.

-- 
Regards
Yafang