netdev - Re: [PATCH bpf-next 3/3] bpf/bpftool: handle libbpf_probe_prog

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: <CACdoK4JbhtOpQeGo+NUh5t3nQG8No8Di6ce-9gwgNw3az2Fu=A@mail.gmail.com>
Date:   Fri, 1 Apr 2022 22:33:16 +0100
From:   Quentin Monnet <quentin@...valent.com>
To:     Andrii Nakryiko <andrii.nakryiko@...il.com>
Cc:     Milan Landaverde <milan@...verde.com>, bpf <bpf@...r.kernel.org>,
        Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Andrii Nakryiko <andrii@...nel.org>, Martin Lau <kafai@...com>,
        Song Liu <songliubraving@...com>, Yonghong Song <yhs@...com>,
        john fastabend <john.fastabend@...il.com>,
        KP Singh <kpsingh@...nel.org>,
        Dave Marchevsky <davemarchevsky@...com>,
        Stanislav Fomichev <sdf@...gle.com>,
        Networking <netdev@...r.kernel.org>,
        open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH bpf-next 3/3] bpf/bpftool: handle libbpf_probe_prog_type errors

On Fri, 1 Apr 2022 at 19:42, Andrii Nakryiko <andrii.nakryiko@...il.com> wrote:
>
> On Fri, Apr 1, 2022 at 9:05 AM Quentin Monnet <quentin@...valent.com> wrote:
> >
> > 2022-03-31 11:45 UTC-0400 ~ Milan Landaverde <milan@...verde.com>
> > > Previously [1], we were using bpf_probe_prog_type which returned a
> > > bool, but the new libbpf_probe_bpf_prog_type can return a negative
> > > error code on failure. This change decides for bpftool to declare
> > > a program type is not available on probe failure.
> > >
> > > [1] https://lore.kernel.org/bpf/20220202225916.3313522-3-andrii@kernel.org/
> > >
> > > Signed-off-by: Milan Landaverde <milan@...verde.com>
> > > ---
> > >  tools/bpf/bpftool/feature.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/tools/bpf/bpftool/feature.c b/tools/bpf/bpftool/feature.c
> > > index c2f43a5d38e0..b2fbaa7a6b15 100644
> > > --- a/tools/bpf/bpftool/feature.c
> > > +++ b/tools/bpf/bpftool/feature.c
> > > @@ -564,7 +564,7 @@ probe_prog_type(enum bpf_prog_type prog_type, bool *supported_types,
> > >
> > >               res = probe_prog_type_ifindex(prog_type, ifindex);
> > >       } else {
> > > -             res = libbpf_probe_bpf_prog_type(prog_type, NULL);
> > > +             res = libbpf_probe_bpf_prog_type(prog_type, NULL) > 0;
> > >       }
> > >
> > >  #ifdef USE_LIBCAP
> >
>
> A completely unrelated question to you, Quentin. How hard is bpftool's
> dependency on libcap? We've recently removed libcap from selftests, I
> wonder if it would be possible to do that for bpftool as well to
> reduce amount of shared libraries bpftool depends on.

There's not a super-strong dependency on it. It's used in feature
probing, for two things.

First one is to be accurate when we check that the user has the right
capabilities for probing efficiently the system. A workaround consists
in checking that we run with uid=0 (root), although it's less
accurate.

Second thing is probing as an unprivileged user: if bpftool is run to
probe as root but with the "unprivileged" keyword, libcap is used to
drop the CAP_SYS_ADMIN and run the probes without it. I don't know if
there's an easy alternative to libcap for that. Also I don't know how
many people use this feature, but I remember that this was added
because there was some demand at the time, so presumably there are
users relying on this.

This being said, libcap is optional for compiling bpftool, so you
should be able to have it work just as well if the library is not
available on the system? Basically you'd just lose the ability to
probe as an unprivileged user. Do you need to remove the optional
dependency completely?

Quentin

PS: Not directly related but since we're talking of libcap, we
recently discovered that the lib is apparently changing errno when it
maybe shouldn't and plays badly with batch mode:
https://stackoverflow.com/questions/71608181/bpf-xdp-bpftool-batch-file-returns-error-reading-batch-file-failed-opera