netdev - Re: [PATCH bpf-next v2 11/14] xsk: add shared umem support between devices

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAJ8uoz30f_jbtH4bM-YAxyPq2+zqC1CC3c+eQFg-ECwgkOfzSw@mail.gmail.com>
Date:   Thu, 16 Jul 2020 06:35:48 +0200
From:   Magnus Karlsson <magnus.karlsson@...il.com>
To:     Maxim Mikityanskiy <maximmi@...lanox.com>
Cc:     Magnus Karlsson <magnus.karlsson@...el.com>,
        Björn Töpel <bjorn.topel@...el.com>,
        Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Network Development <netdev@...r.kernel.org>,
        Jonathan Lemon <jonathan.lemon@...il.com>,
        bpf <bpf@...r.kernel.org>, jeffrey.t.kirsher@...el.com,
        "Fijalkowski, Maciej" <maciej.fijalkowski@...el.com>,
        Maciej Fijalkowski <maciejromanfijalkowski@...il.com>,
        cristian.dumitrescu@...el.com
Subject: Re: [PATCH bpf-next v2 11/14] xsk: add shared umem support between devices

On Tue, Jul 14, 2020 at 12:18 PM Maxim Mikityanskiy
<maximmi@...lanox.com> wrote:
>
> On 2020-07-10 17:16, Magnus Karlsson wrote:
> > Add support to share a umem between different devices. This mode
> > can be invoked with the XDP_SHARED_UMEM bind flag. Previously,
> > sharing was only supported within the same device. Note that when
> > sharing a umem between devices, just as in the case of sharing a
> > umem between queue ids, you need to create a fill ring and a
> > completion ring and tie them to the socket (with two setsockopts,
> > one for each ring) before you do the bind with the
> > XDP_SHARED_UMEM flag. This so that the single-producer
> > single-consumer semantics of the rings can be upheld.
>
> I'm not sure if you saw my comment under v1 asking about performance.
> Could you share what performance numbers (packet rate) you see when
> doing forwarding with xsk_fwd? I'm interested in:
>
> 1. Forwarding between two queues of the same netdev.
>
> 2. Forwarding between two netdevs.
>
> 3. xdpsock -l as the baseline.

Sorry for the delay Max, but it is all due to vacation. I will provide
you with the numbers once the weather turns sour and/or the family
gets tired of me ;-). From what I can remember, it did not scale
perfectly linearly, instead it hit some other bottleneck, though I did
not examine what at that time.

/Magnus

> Thanks,
> Max
>
> >
> > Signed-off-by: Magnus Karlsson <magnus.karlsson@...el.com>
> > ---
> >   net/xdp/xsk.c | 11 ++++-------
> >   1 file changed, 4 insertions(+), 7 deletions(-)
> >
> > diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
> > index 05fadd9..4bf47d3 100644
> > --- a/net/xdp/xsk.c
> > +++ b/net/xdp/xsk.c
> > @@ -695,14 +695,11 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
> >                       sockfd_put(sock);
> >                       goto out_unlock;
> >               }
> > -             if (umem_xs->dev != dev) {
> > -                     err = -EINVAL;
> > -                     sockfd_put(sock);
> > -                     goto out_unlock;
> > -             }
> >
> > -             if (umem_xs->queue_id != qid) {
> > -                     /* Share the umem with another socket on another qid */
> > +             if (umem_xs->queue_id != qid || umem_xs->dev != dev) {
> > +                     /* Share the umem with another socket on another qid
> > +                      * and/or device.
> > +                      */
> >                       xs->pool = xp_create_and_assign_umem(xs,
> >                                                            umem_xs->umem);
> >                       if (!xs->pool) {
> >
>