netdev - Re: Is bug 200755 in anyone's queue??

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] [day] [month] [year] [list]

Date:   Tue, 10 Sep 2019 13:10:14 -0400
From:   Willem de Bruijn <willemdebruijn.kernel@...il.com>
To:     Paolo Abeni <pabeni@...hat.com>
Cc:     Willem de Bruijn <willemdebruijn.kernel@...il.com>,
        Steve Zabele <zabele@...cast.net>,
        Eric Dumazet <eric.dumazet@...il.com>,
        Mark KEATON <mark.keaton@...theon.com>,
        Network Development <netdev@...r.kernel.org>,
        "shum@...ndrew.org" <shum@...ndrew.org>,
        "vladimir116@...il.com" <vladimir116@...il.com>,
        "saifi.khan@...ikr.in" <saifi.khan@...ikr.in>,
        Daniel Borkmann <daniel@...earbox.net>,
        Stephen Hemminger <stephen@...workplumber.org>,
        Craig Gallek <kraig@...gle.com>
Subject: Re: Is bug 200755 in anyone's queue??

On Tue, Sep 10, 2019 at 12:56 PM Paolo Abeni <pabeni@...hat.com> wrote:
>
> Hi all,
>
> On Tue, 2019-09-10 at 11:52 -0400, Willem de Bruijn wrote:
> > This clearly has some loose ends and is no shorter or simpler. So
> > unless anyone has comments or a different solution, I'll finish
> > up the first variant.
>
> I'm sorry for the late feedback.
>
> I was wondering if we could use a new UDP-specific setsockopt to remove
> the connected socket from the reuseport group at connect() time?
>
> That would not have any behavioral change for existing application
> leveraging the current reuseport implementation and requires possibly a
> simpler implementation, but would need application changes for UDP
> servers doing reuse/connect().
>
> WDYT?

Thanks for taking a look, too, Paolo.

I looked into detaching the sockets from the group at connect time. It
could be done without setsockopt, even. Unfortunately, it brings other
problems.

The reuseport group is still there, so may still match sockets
before the connection. If the connected socket no longer has
sk_reuseport set, binding new sockets will fail on conflict. And so a
few more.