netdev - RE: curent state of GSO over IPv6 tunnel

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]

Message-ID: <B5657A6538887040AD3A81F1008BEC63B37694@avmb3.qlogic.org>
Date:	Tue, 16 Sep 2014 06:30:51 +0000
From:	Yuval Mintz <Yuval.Mintz@...gic.com>
To:	Eric Dumazet <eric.dumazet@...il.com>
CC:	netdev <netdev@...r.kernel.org>
Subject: RE: curent state of GSO over IPv6 tunnel

 > > > Hi,
> > > >
> > > > Tried running simple scenarios using ipv6/ipv6 via gre-tunnel [ip6_gre].
> > > >
> > > > When I've tried sniffing Tx traffic [tcpdump] I've noticed there
> > > > were no aggregations being set, i.e., only MTU-sized packets reached NIC.
> > > >
> > > > Looking at output of `ethtool -k', I can see `tx_gre_segmentation' is on.
> > > >
> > > > What's the current state of the feature? Is it feasible and I'm
> > > > missing some configuration, or is it yet an unsupported feature?
> > >
> > > As far as I know, only bnx2x driver currently supports
> > > NETIF_F_GSO_GRE
> > >
> >
> > Hi Eric,
> >
> > Might be true, but should this flag hold any sway over IPv6?
> > [Both at the moment and in general, i.e., if it does not would we need
> > an additional feature or simply extend the functionality so that it
> > would effect IPv6 as well?]
> 
> Both ipv4 and ipv6 are supported under this single flag.
> 
> Really, in 2014 all NIC supporting TSO for GRE should support IPv6 as well.
> 

O.k. - I can see that natively ip6_gre device doesn't show any features.
If I would like to modify its code in order for it to support GSO/TSO,
what would that entail?

[I assume it would require a deep-dive otherwise someone would have done
it already; But still, I'd like to get a pointer if possible].

Thanks,
Yuval