All of lore.kernel.org
 help / color / mirror / Atom feed
From: Alexei Starovoitov <alexei.starovoitov@gmail.com>
To: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Brenden Blanco <bblanco@plumgrid.com>,
	davem@davemloft.net, netdev@vger.kernel.org,
	Martin KaFai Lau <kafai@fb.com>,
	Jesper Dangaard Brouer <brouer@redhat.com>,
	Ari Saha <as754m@att.com>, Or Gerlitz <gerlitz.or@gmail.com>,
	john.fastabend@gmail.com, hannes@stressinduktion.org,
	Thomas Graf <tgraf@suug.ch>, Tom Herbert <tom@herbertland.com>,
	Daniel Borkmann <daniel@iogearbox.net>
Subject: Re: [PATCH v6 12/12] net/mlx4_en: add prefetch in xdp rx path
Date: Thu, 7 Jul 2016 21:16:14 -0700	[thread overview]
Message-ID: <20160708041612.GA15452@ast-mbp.thefacebook.com> (raw)
In-Reply-To: <1467950191.17638.3.camel@edumazet-glaptop3.roam.corp.google.com>

On Fri, Jul 08, 2016 at 05:56:31AM +0200, Eric Dumazet wrote:
> On Thu, 2016-07-07 at 19:15 -0700, Brenden Blanco wrote:
> > XDP programs read and/or write packet data very early, and cache miss is
> > seen to be a bottleneck.
> > 
> > Add prefetch logic in the xdp case 3 packets in the future. Throughput
> > improved from 10Mpps to 12.5Mpps.  LLC misses as reported by perf stat
> > reduced from ~14% to ~7%.  Prefetch values of 0 through 5 were compared
> > with >3 showing dimishing returns.
> 
> This is what I feared with XDP.
> 
> Instead of making generic changes in the driver(s), we now have 'patches
> that improve XDP numbers'
> 
> Careful prefetches make sense in NIC drivers, regardless of XDP being
> used or not.
> 
> On mlx4, prefetching next cqe could probably help as well.

I've tried this style of prefetching in the past for normal stack
and it didn't help at all.
It helps XDP because inner processing loop is short with small number
of memory accesses, so prefetching Nth packet in advance helps.
Prefetching next packet doesn't help as much, since bpf prog is
too short and hw prefetch logic doesn't have time to actually pull
the data in.
The effectiveness of this patch depends on size of the bpf program
and ammount of work it does. For small and medium sizes it works well.
For large prorgrams probably not so much, but we didn't get to
this point yet. I think eventually the prefetch distance should
be calculated dynamically based on size of prog and amount of work
it does or configured via knob (which would be unfortunate).
The performance gain is sizable, so I think it makes sense to
keep it... even to demonstrate the prefetch logic.
Also note this is ddio capable cpu. On desktop class cpus
the prefetch is mandatory for all bpf programs to have good performance.

Another alternative we considered is to allow bpf programs to
indicate to xdp infra how much in advance to prefetch, so
xdp side will prefetch only when program gives a hint.
But that would be the job of future patches.

  reply	other threads:[~2016-07-08  4:16 UTC|newest]

Thread overview: 59+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-07-08  2:15 [PATCH v6 00/12] Add driver bpf hook for early packet drop and forwarding Brenden Blanco
2016-07-08  2:15 ` [PATCH v6 01/12] bpf: add XDP prog type for early driver filter Brenden Blanco
2016-07-09  8:14   ` Jesper Dangaard Brouer
2016-07-09 13:47     ` Tom Herbert
2016-07-10 13:37       ` Jesper Dangaard Brouer
2016-07-10 17:09         ` Brenden Blanco
2016-07-10 20:30           ` Tom Herbert
2016-07-11 10:15             ` Daniel Borkmann
2016-07-11 12:58               ` Jesper Dangaard Brouer
2016-07-10 20:27         ` Tom Herbert
2016-07-11 11:36           ` Jesper Dangaard Brouer
2016-07-10 20:56   ` Tom Herbert
2016-07-11 16:51     ` Brenden Blanco
2016-07-11 21:21       ` Daniel Borkmann
2016-07-10 21:04   ` Tom Herbert
2016-07-11 13:53     ` Jesper Dangaard Brouer
2016-07-08  2:15 ` [PATCH v6 02/12] net: add ndo to set xdp prog in adapter rx Brenden Blanco
2016-07-10 20:59   ` Tom Herbert
2016-07-11 10:35     ` Daniel Borkmann
2016-07-08  2:15 ` [PATCH v6 03/12] rtnl: add option for setting link xdp prog Brenden Blanco
2016-07-08  2:15 ` [PATCH v6 04/12] net/mlx4_en: add support for fast rx drop bpf program Brenden Blanco
2016-07-09 14:07   ` Or Gerlitz
2016-07-10 15:40     ` Brenden Blanco
2016-07-10 16:38       ` Tariq Toukan
2016-07-09 19:58   ` Saeed Mahameed
2016-07-09 21:37     ` Or Gerlitz
2016-07-10 15:25     ` Tariq Toukan
2016-07-10 16:05       ` Brenden Blanco
2016-07-11 11:48         ` Saeed Mahameed
2016-07-11 21:49           ` Brenden Blanco
2016-07-08  2:15 ` [PATCH v6 05/12] Add sample for adding simple drop program to link Brenden Blanco
2016-07-09 20:21   ` Saeed Mahameed
2016-07-11 11:09   ` Jamal Hadi Salim
2016-07-11 13:37     ` Jesper Dangaard Brouer
2016-07-16 14:55       ` Jamal Hadi Salim
2016-07-08  2:15 ` [PATCH v6 06/12] net/mlx4_en: add page recycle to prepare rx ring for tx support Brenden Blanco
2016-07-08  2:15 ` [PATCH v6 07/12] bpf: add XDP_TX xdp_action for direct forwarding Brenden Blanco
2016-07-08  2:15 ` [PATCH v6 08/12] net/mlx4_en: break out tx_desc write into separate function Brenden Blanco
2016-07-08  2:15 ` [PATCH v6 09/12] net/mlx4_en: add xdp forwarding and data write support Brenden Blanco
2016-07-08  2:15 ` [PATCH v6 10/12] bpf: enable direct packet data write for xdp progs Brenden Blanco
2016-07-08  2:15 ` [PATCH v6 11/12] bpf: add sample for xdp forwarding and rewrite Brenden Blanco
2016-07-08  2:15 ` [PATCH v6 12/12] net/mlx4_en: add prefetch in xdp rx path Brenden Blanco
2016-07-08  3:56   ` Eric Dumazet
2016-07-08  4:16     ` Alexei Starovoitov [this message]
2016-07-08  6:56       ` Eric Dumazet
2016-07-08 16:49         ` Brenden Blanco
2016-07-10 20:48           ` Tom Herbert
2016-07-10 20:50           ` Tom Herbert
2016-07-11 14:54             ` Jesper Dangaard Brouer
2016-07-08 15:20     ` Jesper Dangaard Brouer
2016-07-08 16:02       ` [net-next PATCH RFC] mlx4: RX prefetch loop Jesper Dangaard Brouer
2016-07-11 11:09         ` Jesper Dangaard Brouer
2016-07-11 16:00           ` Brenden Blanco
2016-07-11 23:05           ` Alexei Starovoitov
2016-07-12 12:45             ` Jesper Dangaard Brouer
2016-07-12 16:46               ` Alexander Duyck
2016-07-12 19:52                 ` Jesper Dangaard Brouer
2016-07-13  1:37                   ` Alexei Starovoitov
2016-07-10 16:14 ` [PATCH v6 00/12] Add driver bpf hook for early packet drop and forwarding Tariq Toukan

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=20160708041612.GA15452@ast-mbp.thefacebook.com \
    --to=alexei.starovoitov@gmail.com \
    --cc=as754m@att.com \
    --cc=bblanco@plumgrid.com \
    --cc=brouer@redhat.com \
    --cc=daniel@iogearbox.net \
    --cc=davem@davemloft.net \
    --cc=eric.dumazet@gmail.com \
    --cc=gerlitz.or@gmail.com \
    --cc=hannes@stressinduktion.org \
    --cc=john.fastabend@gmail.com \
    --cc=kafai@fb.com \
    --cc=netdev@vger.kernel.org \
    --cc=tgraf@suug.ch \
    --cc=tom@herbertland.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.