netdev - Re: [PATCH] vxlan: using AF_PACKET socket to send multicast packet cause WARNING

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: <20140402075106.3eec7791@nehalam.linuxnetplumber.net>
Date:	Wed, 2 Apr 2014 07:51:06 -0700
From:	Stephen Hemminger <stephen@...workplumber.org>
To:	lucien xin <lucien.xin@...il.com>
Cc:	David Miller <davem@...emloft.net>,
	network dev <netdev@...r.kernel.org>
Subject: Re: [PATCH] vxlan: using AF_PACKET socket to send multicast packet
 cause WARNING

On Wed, 2 Apr 2014 13:31:59 +0800
lucien xin <lucien.xin@...il.com> wrote:

> On Wed, Apr 2, 2014 at 4:10 AM, David Miller <davem@...emloft.net> wrote:
> > This is not the correct fix.
> >
> > If vxlan is going to reinject dev_hard_start_xmit() bound packets into
> > the ipv4 stack, it must make sure that they are well formed.
> >
> > The SKB traversing the ipv4 stack must either be bound to an ipv4/ipv6
> > socket or be orphaned.
> 
> yeah,  it's right. then I want to bound skb to vxlan's sock in
> vxlan_xmit_skb(), like this:
> 
> diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
> index 1236812..8fd7e93 100644
> --- a/drivers/net/vxlan.c
> +++ b/drivers/net/vxlan.c
> @@ -1755,6 +1755,8 @@ int vxlan_xmit_skb(struct vxlan_sock *vs,
>         if (err)
>                 return err;
> 
> +       skb_set_owner_w(skb, vs->sock->sk);
> +
>         return iptunnel_xmit(rt, skb, src, dst, IPPROTO_UDP, tos, ttl, df,
>                              false);
>  }
> 
> which can not only solve the issue, but also make the logic more correct.
> David, how do you think?

That breaks local socket accounting for tunneled packets.
The idea is that a TCP socket sending over vxlan will keep it's
memory accounting all the way until the device sends it.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html