netdev - Re: [PATCH bpf-next v3 5/8] bpftool: support dumping metadata

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAEf4BzYi+DjDHZTWWOPyJaU13civjeFRO-bGOzvEPjsimk906A@mail.gmail.com>
Date:   Tue, 8 Sep 2020 15:35:00 -0700
From:   Andrii Nakryiko <andrii.nakryiko@...il.com>
To:     Stanislav Fomichev <sdf@...gle.com>
Cc:     Networking <netdev@...r.kernel.org>, bpf <bpf@...r.kernel.org>,
        "David S. Miller" <davem@...emloft.net>,
        Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        YiFei Zhu <zhuyifei@...gle.com>,
        YiFei Zhu <zhuyifei1999@...il.com>
Subject: Re: [PATCH bpf-next v3 5/8] bpftool: support dumping metadata

On Tue, Sep 8, 2020 at 1:53 PM Stanislav Fomichev <sdf@...gle.com> wrote:
>
> On Wed, Sep 2, 2020 at 10:00 PM Andrii Nakryiko
> <andrii.nakryiko@...il.com> wrote:
> >
> > On Fri, Aug 28, 2020 at 12:37 PM Stanislav Fomichev <sdf@...gle.com> wrote:
> > >
> > > From: YiFei Zhu <zhuyifei@...gle.com>
> > >
> > > Added a flag "--metadata" to `bpftool prog list` to dump the metadata
> > > contents. For some formatting some BTF code is put directly in the
> > > metadata dumping. Sanity checks on the map and the kind of the btf_type
> > > to make sure we are actually dumping what we are expecting.
> > >
> > > A helper jsonw_reset is added to json writer so we can reuse the same
> > > json writer without having extraneous commas.
> > >
> > > Sample output:
> > >
> > >   $ bpftool prog --metadata
> > >   6: cgroup_skb  name prog  tag bcf7977d3b93787c  gpl
> > >   [...]
> > >         btf_id 4
> > >         metadata:
> > >                 metadata_a = "foo"
> > >                 metadata_b = 1
> > >
> > >   $ bpftool prog --metadata --json --pretty
> > >   [{
> > >           "id": 6,
> > >   [...]
> > >           "btf_id": 4,
> > >           "metadata": {
> > >               "metadata_a": "foo",
> > >               "metadata_b": 1
> > >           }
> > >       }
> > >   ]
> > >
> > > Cc: YiFei Zhu <zhuyifei1999@...il.com>
> > > Signed-off-by: YiFei Zhu <zhuyifei@...gle.com>
> > > Signed-off-by: Stanislav Fomichev <sdf@...gle.com>
> > > ---
> > >  tools/bpf/bpftool/json_writer.c |   6 ++
> > >  tools/bpf/bpftool/json_writer.h |   3 +
> > >  tools/bpf/bpftool/main.c        |  10 +++
> > >  tools/bpf/bpftool/main.h        |   1 +
> > >  tools/bpf/bpftool/prog.c        | 130 ++++++++++++++++++++++++++++++++
> > >  5 files changed, 150 insertions(+)
> > >
> >
> > [...]
> >
> > > +
> > > +       if (bpf_map_lookup_elem(map_fd, &key, value)) {
> > > +               p_err("metadata map lookup failed: %s", strerror(errno));
> > > +               goto out_free;
> > > +       }
> > > +
> > > +       err = btf__get_from_id(map_info.btf_id, &btf);
> >
> > what if the map has no btf_id associated (e.g., because of an old
> > kernel?); why fail in this case?
> Thank you for the review, coming back at it a bit late :-(
>
> This functionality is guarded by --metadata bpftool flag (off by default).
> In case of no btf_id, it might be helpful to show why we don't have
> the metadata rather than just quietly failing.
> WDYT?

we might do it similarly to PID info I added with bpf_iter: if it's
supported -- emit it, if not -- skip and still succeed. So maybe we
don't really need extra --metadata flag and should do all this always?

>
> > > +       if (err || !btf) {
> > > +               p_err("metadata BTF get failed: %s", strerror(-err));
> > > +               goto out_free;
> > > +       }
> > > +
> > > +       t_datasec = btf__type_by_id(btf, map_info.btf_value_type_id);
> > > +       if (BTF_INFO_KIND(t_datasec->info) != BTF_KIND_DATASEC) {
> >
> > btf_is_datasec(t_datasec)
> >
> > > +               p_err("bad metadata BTF");
> > > +               goto out_free;
> > > +       }
> > > +
> > > +       vlen = BTF_INFO_VLEN(t_datasec->info);
> >
> > btf_vlen(t_datasec)
> >
> > > +       vsi = (struct btf_var_secinfo *)(t_datasec + 1);
> >
> > btf_var_secinfos(t_datasec)
> >
> > > +
> > > +       /* We don't proceed to check the kinds of the elements of the DATASEC.
> > > +        * The verifier enforce then to be BTF_KIND_VAR.
> >
> > typo: then -> them
> >
> > > +        */
> > > +
> > > +       if (json_output) {
> > > +               struct btf_dumper d = {
> > > +                       .btf = btf,
> > > +                       .jw = json_wtr,
> > > +                       .is_plain_text = false,
> > > +               };
> > > +
> > > +               jsonw_name(json_wtr, "metadata");
> > > +
> > > +               jsonw_start_object(json_wtr);
> > > +               for (i = 0; i < vlen; i++) {
> >
> > nit: doing ++vsi here
> Agreed with all the above, except this one.
> It feels like it's safer to do [i] in case somebody adds a 'continue'
> clause later and we miss that '++vsi'.
> Let me know if you feel strongly about it.

I meant to add vsi++ inside the for clause, no way to miss it:

for (i = 0; i < vlen, i++, vsi++) {
  continue/break/whatever you want, except extra i++ or vsi++
}

it's the safest way, imo

>
> > > +                       t_var = btf__type_by_id(btf, vsi[i].type);
> >
> > and vsi->type here and below would look a bit cleaner
> >
> > > +
> > > +                       jsonw_name(json_wtr, btf__name_by_offset(btf, t_var->name_off));
> > > +                       err = btf_dumper_type(&d, t_var->type, value + vsi[i].offset);
> > > +                       if (err) {
> > > +                               p_err("btf dump failed");
> > > +                               break;
> > > +                       }
> > > +               }
> > > +               jsonw_end_object(json_wtr);
> >
> > [...]