Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210919154337.9243-1-stachecki.tyler@gmail.com>
In-Reply-To: <20210919154337.9243-1-stachecki.tyler@gmail.com>
From:   Cong Wang <xiyou.wangcong@gmail.com>
Date:   Sun, 19 Sep 2021 16:33:47 -0700
Message-ID: <CAM_iQpUeH7sOWRDfaAsER4HdRJQUuODB-ht0NyEZgnCXEmm2RQ@mail.gmail.com>
Subject: Re: [PATCH] ovs: Only clear tstamp when changing namespaces
To:     "Tyler J. Stachecki" <stachecki.tyler@gmail.com>
Cc:     fankaixi.li@bytedance.com, xiexiaohui.xxh@bytedance.com,
        "Cong Wang ." <cong.wang@bytedance.com>,
        Pravin B Shelar <pshelar@ovn.org>,
        "David S. Miller" <davem@davemloft.net>,
        Jakub Kicinski <kuba@kernel.org>,
        Linux Kernel Network Developers <netdev@vger.kernel.org>,
        dev@openvswitch.org, LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Sun, Sep 19, 2021 at 10:59 AM Tyler J. Stachecki
<stachecki.tyler@gmail.com> wrote:
>
> As of "ovs: clear skb->tstamp in forwarding path", the
> tstamp is now being cleared unconditionally to fix fq qdisc
> operation with ovs vports.
>
> While this is mostly correct and fixes forwarding for that
> use case, a slight adjustment is necessary to ensure that
> the tstamp is cleared *only when the forwarding is across
> namespaces*.

Hmm? I am sure timestamp has already been cleared when
crossing netns:

void skb_scrub_packet(struct sk_buff *skb, bool xnet)
{
...
        if (!xnet)
                return;

        ipvs_reset(skb);
        skb->mark = 0;
        skb->tstamp = 0;
}

So, what are you trying to fix?

>
> Signed-off-by: Tyler J. Stachecki <stachecki.tyler@gmail.com>
> ---
>  net/openvswitch/vport.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/net/openvswitch/vport.c b/net/openvswitch/vport.c
> index cf2ce5812489..c2d32a5c3697 100644
> --- a/net/openvswitch/vport.c
> +++ b/net/openvswitch/vport.c
> @@ -507,7 +507,8 @@ void ovs_vport_send(struct vport *vport, struct sk_buff *skb, u8 mac_proto)
>         }
>
>         skb->dev = vport->dev;
> -       skb->tstamp = 0;
> +       if (dev_net(skb->dev))

Doesn't dev_net() always return a non-NULL pointer?

If you really want to check whether it is cross-netns, you should
use net_eq() to compare src netns with dst netns, something like:
if (!net_eq(dev_net(vport->dev), dev_net(skb->dev))).

Thanks.