netdev - Re: TX offloads for NVGRE (OVS GRE with inner protocol being TEB)

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: <CAJ3xEMggVJ6Lm7vmFhW2PSPv612WQ4-sNWPpefkMcy1RQv+7=g@mail.gmail.com>
Date:	Fri, 20 Feb 2015 09:54:21 +0200
From:	Or Gerlitz <gerlitz.or@...il.com>
To:	Jesse Gross <jesse@...ira.com>
Cc:	Or Gerlitz <ogerlitz@...lanox.com>,
	Joe Stringer <joestringer@...ira.com>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: TX offloads for NVGRE (OVS GRE with inner protocol being TEB)

On Thu, Feb 19, 2015 at 7:01 PM, Jesse Gross <jesse@...ira.com> wrote:
> On Thu, Feb 19, 2015 at 6:05 AM, Or Gerlitz <ogerlitz@...lanox.com> wrote:

>> It seems that the OVS GRE code lacks handling of offloads (e.g to come into
>> play with NICs that support NVGRE).
>>
>> I assume we need to place a call to iptunnel_handle_offloads before invoking
>> iptunnel_xmit, agree? so ~the quick patch below should do the work? I wasn't
>> sure how to set the value of the 2nd param for iptunnel_handle_offloads().
>
> I don't think that this is the issue. __build_header() calls
> gre_handle_offloads() which should do all of this work already.

yep... gre_handle_offloads() is called, so something else goes wrong here.

> I did notice that skb_set_inner_protocol() is not being called because
> it is in the wrong place in the GRE encapsulation code. It is
> currently in ip_gre.c():__gre_xmit(), if that line was moved to
> gre_demux.c:gre_build_header() then it would be used by all callers.

Oh, thanks, I'll add that and see if / how much does it help.

> One other thing that is potentially an issue for offloads is that all
> of the encapsulations call vlan_hwaccel_push_inside() after the
> respective handle_offloads code. This doesn't seem right as it could
> affect the layer pointers, although it only likely matters in a
> minority of cases where the inner MAC pointer is used.

mmm, I don't think we're failing on that now, but can look this up too.


Did you ever tested Linux OVS/NGRE with a NIC that can do offloads for
this type of traffic?
--
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