Re: [RFC PATCH bpf-next 3/3] samples/bpf: Add a simple bridge example accelerated with XDP

From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
To: Yoshiki Komachi <komachi.yoshiki@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Jesper Dangaard Brouer <hawk@kernel.org>,
	John Fastabend <john.fastabend@gmail.com>,
	Jakub Kicinski <kuba@kernel.org>, Martin KaFai Lau <kafai@fb.com>,
	Song Liu <songliubraving@fb.com>, Yonghong Song <yhs@fb.com>,
	Andrii Nakryiko <andriin@fb.com>, KP Singh <kpsingh@chromium.org>,
	Roopa Prabhu <roopa@cumulusnetworks.com>,
	Nikolay Aleksandrov <nikolay@cumulusnetworks.com>,
	David Ahern <dsahern@kernel.org>,
	Networking <netdev@vger.kernel.org>,
	bridge@lists.linux-foundation.org, bpf <bpf@vger.kernel.org>
Subject: Re: [RFC PATCH bpf-next 3/3] samples/bpf: Add a simple bridge example accelerated with XDP
Date: Fri, 31 Jul 2020 10:48:18 -0700	[thread overview]
Message-ID: <CAEf4BzaRKhJqFmXJEQy5LOjKx9nkPgAKHa3cesvywy2qqg93YA@mail.gmail.com> (raw)
In-Reply-To: <1596170660-5582-4-git-send-email-komachi.yoshiki@gmail.com>

On Thu, Jul 30, 2020 at 9:45 PM Yoshiki Komachi
<komachi.yoshiki@gmail.com> wrote:
>
> This patch adds a simple example of XDP-based bridge with the new
> bpf_fdb_lookup helper. This program simply forwards packets based
> on the destination port given by FDB in the kernel. Note that both
> vlan filtering and learning features are currently unsupported in
> this example.
>
> There is another plan to recreate a userspace application
> (xdp_bridge_user.c) as a daemon process, which helps to automate
> not only detection of status changes in bridge port but also
> handling vlan protocol updates.
>
> Note: David Ahern suggested a new bpf helper [1] to get master
> vlan/bonding devices in XDP programs attached to their slaves
> when the master vlan/bonding devices are bridge ports. If this
> idea is accepted and the helper is introduced in the future, we
> can handle interfaces slaved to vlan/bonding devices in this
> sample by calling the suggested bpf helper (I guess it can get
> vlan/bonding ifindex from their slave ifindex). Notice that we
> don't need to change bpf_fdb_lookup() API to use such a feature,
> but we just need to modify bpf programs like this sample.
>
> [1]: http://vger.kernel.org/lpc-networking2018.html#session-1
>
> Signed-off-by: Yoshiki Komachi <komachi.yoshiki@gmail.com>
> ---

Have you tried using a BPF skeleton for this? It could have saved a
bunch of mechanical code for your example. Also libbpf supports map
pinning out of the box now, I wonder if it would just work in your
case. Also it would be nice if you tried using BPF link-based approach
for this example, to show how it can be used. Thanks!

>  samples/bpf/Makefile          |   3 +
>  samples/bpf/xdp_bridge_kern.c | 129 ++++++++++++++++++
>  samples/bpf/xdp_bridge_user.c | 239 ++++++++++++++++++++++++++++++++++
>  3 files changed, 371 insertions(+)
>  create mode 100644 samples/bpf/xdp_bridge_kern.c
>  create mode 100644 samples/bpf/xdp_bridge_user.c
>

[...]