Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.35 as permitted sender) client-ip=23.128.96.35;
MIME-Version: 1.0
References: <20230914083716.57443-1-larysa.zaremba@intel.com>
 <ZQM1BUzcZQtXusA3@google.com> <ZQM5kt8qHKUH0Iob@lincoln>
In-Reply-To: <ZQM5kt8qHKUH0Iob@lincoln>
From:   Stanislav Fomichev <sdf@google.com>
Date:   Thu, 14 Sep 2023 10:05:47 -0700
Message-ID: <CAKH8qBuw68AixQabgP5wNfAQBcc0RuVNEyV9rf9vgVi__c4Y9A@mail.gmail.com>
Subject: Re: [PATCH bpf-next] bpf: Allow to use kfunc XDP hints and frags together
To:     Larysa Zaremba <larysa.zaremba@intel.com>
Cc:     bpf@vger.kernel.org, Alexei Starovoitov <ast@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Andrii Nakryiko <andrii@kernel.org>,
        Martin KaFai Lau <martin.lau@linux.dev>,
        Song Liu <song@kernel.org>,
        Yonghong Song <yonghong.song@linux.dev>,
        John Fastabend <john.fastabend@gmail.com>,
        KP Singh <kpsingh@kernel.org>, Hao Luo <haoluo@google.com>,
        Jiri Olsa <jolsa@kernel.org>, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Sep 14, 2023 at 9:55=E2=80=AFAM Larysa Zaremba <larysa.zaremba@inte=
l.com> wrote:
>
> On Thu, Sep 14, 2023 at 09:29:57AM -0700, Stanislav Fomichev wrote:
> > On 09/14, Larysa Zaremba wrote:
> > > There is no fundamental reason, why multi-buffer XDP and XDP kfunc RX=
 hints
> > > cannot coexist in a single program.
> > >
> > > Allow those features to be used together by modifying the flags condi=
tions.
> > >
> > > Suggested-by: Stanislav Fomichev <sdf@google.com>
> > > Link: https://lore.kernel.org/bpf/CAKH8qBuzgtJj=3DOKMdsxEkyML36VsAuZp=
crsXcyqjdKXSJCBq=3DQ@mail.gmail.com/
> > > Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
> > > ---
> > >  kernel/bpf/offload.c | 6 +++++-
> > >  1 file changed, 5 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/kernel/bpf/offload.c b/kernel/bpf/offload.c
> > > index ee35f33a96d1..43aded96c79b 100644
> > > --- a/kernel/bpf/offload.c
> > > +++ b/kernel/bpf/offload.c
> > > @@ -232,7 +232,11 @@ int bpf_prog_dev_bound_init(struct bpf_prog *pro=
g, union bpf_attr *attr)
> > >         attr->prog_type !=3D BPF_PROG_TYPE_XDP)
> > >             return -EINVAL;
> > >
> > > -   if (attr->prog_flags & ~BPF_F_XDP_DEV_BOUND_ONLY)
> > > +   if (attr->prog_flags & ~(BPF_F_XDP_DEV_BOUND_ONLY | BPF_F_XDP_HAS=
_FRAGS))
> > > +           return -EINVAL;
> > > +
> >
> > [..]
> >
> > > +   if (attr->prog_flags & BPF_F_XDP_HAS_FRAGS &&
> > > +       !(attr->prog_flags & BPF_F_XDP_DEV_BOUND_ONLY))
> > >             return -EINVAL;
> >
> > Any reason we have 'attr->prog_flags & BPF_F_XDP_HAS_FRAGS' part here?
> > Seems like doing '!(attr->prog_flags & BPF_F_XDP_DEV_BOUND_ONLY)' shoul=
d
> > be enough, right? We only want to bail out here when BPF_F_XDP_DEV_BOUN=
D_ONLY
> > is not set and we don't really care whether BPF_F_XDP_HAS_FRAGS is set
> > or not at this point.
>
> If !(attr->prog_flags & BPF_F_XDP_DEV_BOUND_ONLY) at this point, program =
could
> be requesting offload.
>
> Now I have thought about those conditions once more and they could be red=
uced to
> this:
>
> if (attr->prog_flags & ~(BPF_F_XDP_DEV_BOUND_ONLY) &&
>     attr->prog_flags !=3D (BPF_F_XDP_DEV_BOUND_ONLY | BPF_F_XDP_HAS_FRAGS=
))
>         return -EINVAL;
>
> What do you think?

Ah, so this check is here to protect against the mbuf+offloaded
combination? (looking at that other thread with Maciej)
Let's keep your current way with two separate checks, but let's add
your "/* Frags are allowed only if program is dev-bound-only, but not
if it is requesting
bpf offload. */" as a comment to the second check?