netdev - Re: [PATCH v1 1/1] veth: Do not drop packets larger then the mtu set on the receiving side

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: <CAJieiUj8AbiH-ZzCO75kx3OxX=xxOgZUA=CJkZybWtn7whaH9w@mail.gmail.com>
Date:   Thu, 10 Jan 2019 07:39:21 -0800
From:   Roopa Prabhu <roopa@...ulusnetworks.com>
To:     Andrew Lunn <andrew@...n.ch>
Cc:     Fredrik Gustavsson <gustfred@...il.com>,
        netdev <netdev@...r.kernel.org>,
        David Miller <davem@...emloft.net>,
        Toshiaki Makita <makita.toshiaki@....ntt.co.jp>,
        Daniel Borkmann <daniel@...earbox.net>
Subject: Re: [PATCH v1 1/1] veth: Do not drop packets larger then the mtu set
 on the receiving side

On Thu, Jan 10, 2019 at 6:21 AM Andrew Lunn <andrew@...n.ch> wrote:
>
> On Thu, Jan 10, 2019 at 02:26:55PM +0100, Fredrik Gustavsson wrote:
> > commit affede4a779420bd8510ab937251a3796d3228df
> > Author: Fredrik Gustavsson <gustfred@...il.com>
> > Date:   Tue Jan 8 11:21:39 2019 +0100
> >
> > veth: Do not drop packets larger then the mtu set on the receiving side
> >
> > Currently veth drops all packets larger then the mtu set on the receiving
> > end of the pair. This is inconsistent with most hardware ethernet drivers
> > that happily receives packets up the the ethernet MTU independent of the
> > configured MTU.
>
> I agree with your argument, but i wonder if there could be a better
> implementation.
>
> ____dev_forward_skb() is on the hot path, so your additional check is
> going to slow down packet forwarding for everybody, not just veth.
> is_skb_forwardable() also does some additional checks which you are
> now skipping. Is that O.K?
>
> Since we are talking about a veth pair here, you have access to both
> ends of the link. Maybe consider if the mtu is changed on one end, you
> also change it on the other?
>
> Lets see what others think of that.
>

adding a IFF_VETH just for this seems like an overkill. especially
when there are other ways to indicate a virtual device type like
rtnetlink kind.
Also, its best to keep such checks local to veth, maybe with something
along the lines of what Andrew suggests above.