From: John Fastabend <john.fastabend@gmail.com>
To: "Toke Høiland-Jørgensen" <toke@redhat.com>,
"Daniel Borkmann" <daniel@iogearbox.net>
Cc: Alexei Starovoitov <ast@kernel.org>,
Martin KaFai Lau <kafai@fb.com>, Song Liu <songliubraving@fb.com>,
Yonghong Song <yhs@fb.com>,
Marek Majkowski <marek@cloudflare.com>,
Lorenz Bauer <lmb@cloudflare.com>,
David Miller <davem@davemloft.net>,
Jesper Dangaard Brouer <brouer@redhat.com>,
netdev@vger.kernel.org, bpf@vger.kernel.org
Subject: RE: [PATCH bpf-next 0/9] xdp: Support multiple programs on a single interface through chain calls
Date: Wed, 02 Oct 2019 09:43:49 -0700 [thread overview]
Message-ID: <5d94d3c5a238f_22502b00ea21a5b4e9@john-XPS-13-9370.notmuch> (raw)
In-Reply-To: <157002302448.1302756.5727756706334050763.stgit@alrua-x1>
Toke Høiland-Jørgensen wrote:
> This series adds support for executing multiple XDP programs on a single
> interface in sequence, through the use of chain calls, as discussed at the Linux
> Plumbers Conference last month:
>
> https://linuxplumbersconf.org/event/4/contributions/460/
>
> # HIGH-LEVEL IDEA
>
> The basic idea is to express the chain call sequence through a special map type,
> which contains a mapping from a (program, return code) tuple to another program
> to run in next in the sequence. Userspace can populate this map to express
> arbitrary call sequences, and update the sequence by updating or replacing the
> map.
>
> The actual execution of the program sequence is done in bpf_prog_run_xdp(),
> which will lookup the chain sequence map, and if found, will loop through calls
> to BPF_PROG_RUN, looking up the next XDP program in the sequence based on the
> previous program ID and return code.
>
> An XDP chain call map can be installed on an interface by means of a new netlink
> attribute containing an fd pointing to a chain call map. This can be supplied
> along with the XDP prog fd, so that a chain map is always installed together
> with an XDP program.
>
> # PERFORMANCE
>
> I performed a simple performance test to get an initial feel for the overhead of
> the chain call mechanism. This test consists of running only two programs in
> sequence: One that returns XDP_PASS and another that returns XDP_DROP. I then
> measure the drop PPS performance and compare it to a baseline of just a single
> program that only returns XDP_DROP.
>
> For comparison, a test case that uses regular eBPF tail calls to sequence two
> programs together is also included. Finally, because 'perf' showed that the
> hashmap lookup was the largest single source of overhead, I also added a test
> case where I removed the jhash() call from the hashmap code, and just use the
> u32 key directly as an index into the hash bucket structure.
>
> The performance for these different cases is as follows (with retpolines disabled):
retpolines enabled would also be interesting.
>
> | Test case | Perf | Add. overhead | Total overhead |
> |---------------------------------+-----------+---------------+----------------|
> | Before patch (XDP DROP program) | 31.0 Mpps | | |
> | After patch (XDP DROP program) | 28.9 Mpps | 2.3 ns | 2.3 ns |
IMO even 1 Mpps overhead is too much for a feature that is primarily about
ease of use. Sacrificing performance to make userland a bit easier is hard
to justify for me when XDP _is_ singularly about performance. Also that is
nearly 10% overhead which is fairly large. So I think going forward the
performance gab needs to be removed.
> | XDP tail call | 26.6 Mpps | 3.0 ns | 5.3 ns |
> | XDP chain call (no jhash) | 19.6 Mpps | 13.4 ns | 18.7 ns |
> | XDP chain call (this series) | 17.0 Mpps | 7.9 ns | 26.6 ns |
>
> From this it is clear that while there is some overhead from this mechanism; but
> the jhash removal example indicates that it is probably possible to optimise the
> code to the point where the overhead becomes low enough that it is acceptable.
I'm missing why 'in theory' at least this can't be made as-fast as tail calls?
Again I can't see why someone would lose 30% of their performance when a userland
program could populate a tail call map for the same effect. Sure userland would
also have to enforce some program standards/conventions but it could be done and
at 30% overhead that pain is probably worth it IMO.
My thinking though is if we are a bit clever chaining and tail calls could be
performance-wise equivalent?
I'll go read the patches now ;)
.John
>
> # PATCH SET STRUCTURE
> This series is structured as follows:
>
> - Patch 1: Prerequisite
> - Patch 2: New map type
> - Patch 3: Netlink hooks to install the chain call map
> - Patch 4: Core chain call logic
> - Patch 5-7: Bookkeeping updates to tools
> - Patch 8: Libbpf support for installing chain call maps
> - Patch 9: Selftests with example user space code
>
> The whole series is also available in my git repo on kernel.org:
> https://git.kernel.org/pub/scm/linux/kernel/git/toke/linux.git/log/?h=xdp-multiprog-01
>
next prev parent reply other threads:[~2019-10-02 16:44 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-02 13:30 [PATCH bpf-next 0/9] xdp: Support multiple programs on a single interface through chain calls Toke Høiland-Jørgensen
2019-10-02 13:30 ` [PATCH bpf-next 1/9] hashtab: Add new bpf_map_fd_put_value op Toke Høiland-Jørgensen
2019-10-02 13:30 ` [PATCH bpf-next 2/9] xdp: Add new xdp_chain_map type for specifying XDP call sequences Toke Høiland-Jørgensen
2019-10-02 15:50 ` Lorenz Bauer
2019-10-02 18:25 ` Toke Høiland-Jørgensen
2019-10-02 13:30 ` [PATCH bpf-next 3/9] xdp: Support setting and getting device chain map Toke Høiland-Jørgensen
2019-10-02 15:50 ` Lorenz Bauer
2019-10-02 18:32 ` Toke Høiland-Jørgensen
2019-10-02 18:07 ` kbuild test robot
2019-10-02 18:29 ` kbuild test robot
2019-10-02 13:30 ` [PATCH bpf-next 4/9] xdp: Implement chain call logic to support multiple programs on one interface Toke Høiland-Jørgensen
2019-10-02 17:33 ` kbuild test robot
2019-10-02 17:53 ` kbuild test robot
2019-10-02 13:30 ` [PATCH bpf-next 5/9] tools/include/uapi: Add XDP chain map definitions Toke Høiland-Jørgensen
2019-10-02 13:30 ` [PATCH bpf-next 6/9] tools/libbpf_probes: Add support for xdp_chain map type Toke Høiland-Jørgensen
2019-10-02 13:30 ` [PATCH bpf-next 7/9] bpftool: Add definitions " Toke Høiland-Jørgensen
2019-10-02 13:30 ` [PATCH bpf-next 8/9] libbpf: Add support for setting and getting XDP chain maps Toke Høiland-Jørgensen
2019-10-02 13:30 ` [PATCH bpf-next 9/9] selftests: Add tests for XDP chain calls Toke Høiland-Jørgensen
2019-10-02 15:10 ` [PATCH bpf-next 0/9] xdp: Support multiple programs on a single interface through " Alan Maguire
2019-10-02 15:33 ` Toke Høiland-Jørgensen
2019-10-02 16:34 ` John Fastabend
2019-10-02 18:33 ` Toke Høiland-Jørgensen
2019-10-02 20:34 ` John Fastabend
2019-10-03 7:48 ` Toke Høiland-Jørgensen
2019-10-03 10:09 ` Jesper Dangaard Brouer
2019-10-03 19:45 ` John Fastabend
2019-10-02 16:35 ` Lorenz Bauer
2019-10-02 18:54 ` Toke Høiland-Jørgensen
2019-10-02 16:43 ` John Fastabend [this message]
2019-10-02 19:09 ` Toke Høiland-Jørgensen
2019-10-02 19:15 ` Daniel Borkmann
2019-10-02 19:29 ` Toke Høiland-Jørgensen
2019-10-02 19:46 ` Alexei Starovoitov
2019-10-03 7:58 ` Toke Høiland-Jørgensen
2019-10-02 18:38 ` Song Liu
2019-10-02 18:54 ` Song Liu
2019-10-02 19:25 ` Toke Høiland-Jørgensen
2019-10-03 8:53 ` Jesper Dangaard Brouer
2019-10-03 14:03 ` Alexei Starovoitov
2019-10-03 14:33 ` Toke Høiland-Jørgensen
2019-10-03 14:53 ` Edward Cree
2019-10-03 18:49 ` Jesper Dangaard Brouer
2019-10-03 19:35 ` John Fastabend
2019-10-04 8:09 ` Toke Høiland-Jørgensen
2019-10-04 10:34 ` Edward Cree
2019-10-04 15:58 ` Lorenz Bauer
2019-10-07 16:43 ` Edward Cree
2019-10-07 17:12 ` Lorenz Bauer
2019-10-07 19:21 ` Edward Cree
2019-10-07 21:01 ` Alexei Starovoitov
2019-10-02 19:23 ` Toke Høiland-Jørgensen
2019-10-02 19:49 ` Song Liu
2019-10-03 7:59 ` Toke Høiland-Jørgensen
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=5d94d3c5a238f_22502b00ea21a5b4e9@john-XPS-13-9370.notmuch \
--to=john.fastabend@gmail.com \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=brouer@redhat.com \
--cc=daniel@iogearbox.net \
--cc=davem@davemloft.net \
--cc=kafai@fb.com \
--cc=lmb@cloudflare.com \
--cc=marek@cloudflare.com \
--cc=netdev@vger.kernel.org \
--cc=songliubraving@fb.com \
--cc=toke@redhat.com \
--cc=yhs@fb.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).