Re: BPF sk_lookup v5 - TCP SYN and UDP 0-len flood benchmarks

From: Paolo Abeni <pabeni@redhat.com>
To: Alexei Starovoitov <alexei.starovoitov@gmail.com>,
	Jakub Sitnicki <jakub@cloudflare.com>
Cc: bpf <bpf@vger.kernel.org>,
	Network Development <netdev@vger.kernel.org>,
	kernel-team <kernel-team@cloudflare.com>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	"David S. Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>,
	Andrii Nakryiko <andriin@fb.com>,
	Lorenz Bauer <lmb@cloudflare.com>,
	Marek Majkowski <marek@cloudflare.com>,
	Martin KaFai Lau <kafai@fb.com>, Yonghong Song <yhs@fb.com>
Subject: Re: BPF sk_lookup v5 - TCP SYN and UDP 0-len flood benchmarks
Date: Mon, 24 Aug 2020 10:17:03 +0200	[thread overview]
Message-ID: <8ccf0b77c854a20f65026fdc68dcd64b93d07fc5.camel@redhat.com> (raw)
In-Reply-To: <CAADnVQKE6y9h2fwX6OS837v-Uf+aBXnT_JXiN_bbo2gitZQ3tA@mail.gmail.com>

On Tue, 2020-08-18 at 11:19 -0700, Alexei Starovoitov wrote:
> On Tue, Aug 18, 2020 at 8:49 AM Jakub Sitnicki <jakub@cloudflare.com> wrote:
> >          :                      rcu_read_lock();
> >          :                      run_array = rcu_dereference(net->bpf.run_array[NETNS_BPF_SK_LOOKUP]);
> >     0.01 :   ffffffff817f8624:       mov    0xd68(%r12),%rsi
> >          :                      if (run_array) {
> >     0.00 :   ffffffff817f862c:       test   %rsi,%rsi
> >     0.00 :   ffffffff817f862f:       je     ffffffff817f87a9 <__udp4_lib_lookup+0x2c9>
> >          :                      struct bpf_sk_lookup_kern ctx = {
> >     1.05 :   ffffffff817f8635:       xor    %eax,%eax
> >     0.00 :   ffffffff817f8637:       mov    $0x6,%ecx
> >     0.01 :   ffffffff817f863c:       movl   $0x110002,0x40(%rsp)
> >     0.00 :   ffffffff817f8644:       lea    0x48(%rsp),%rdi
> >    18.76 :   ffffffff817f8649:       rep stos %rax,%es:(%rdi)
> >     1.12 :   ffffffff817f864c:       mov    0xc(%rsp),%eax
> >     0.00 :   ffffffff817f8650:       mov    %ebp,0x48(%rsp)
> >     0.00 :   ffffffff817f8654:       mov    %eax,0x44(%rsp)
> >     0.00 :   ffffffff817f8658:       movzwl 0x10(%rsp),%eax
> >     1.21 :   ffffffff817f865d:       mov    %ax,0x60(%rsp)
> >     0.00 :   ffffffff817f8662:       movzwl 0x20(%rsp),%eax
> >     0.00 :   ffffffff817f8667:       mov    %ax,0x62(%rsp)
> >          :                      .sport          = sport,
> >          :                      .dport          = dport,
> >          :                      };
> 
> Such heavy hit to zero init 56-byte structure is surprising.
> There are two 4-byte holes in this struct. You can try to pack it and
> make sure that 'rep stoq' is used instead of 'rep stos' (8 byte at a time vs 4).

I think here rep stos is copying 8 bytes at a time (%rax operand, %ecx
initalized with '6').

I think that you can avoid the costly instruction explicitly
initializing each field individually:

	struct bpf_sk_lookup_kern ctx;

	ctx.family = AF_INET;
	ctx.protocol = protocol;
	// ...

note, you likely want to explicitly zero the v6 addresses, too.

Cheers,

Paolo