Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 5 Dec 2011 11:24:16 -0800
From:      Jack Vogel <jfvogel@gmail.com>
To:        Andre Oppermann <andre@freebsd.org>
Cc:        FreeBSD Net <freebsd-net@freebsd.org>
Subject:   Re: VLAN HWTSO question/problem
Message-ID:  <CAFOYbcng-8=vCxitx%2BGVtuOJx983kCVs3mCZGfhLaTvND8Sdqg@mail.gmail.com>
In-Reply-To: <4EDCD23C.8050706@freebsd.org>
References:  <CAFOYbc=JbAUbqJmsNUoRo_anSomvSLz_O-pR%2B49_f-i5Oy-tVQ@mail.gmail.com> <4EDCD23C.8050706@freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
I have solved the problem. Was my bug :)  Turns out the xmit path has been
wrong for quite some time, it was only putting the VLAN tag and command bit
into the last descriptor of a frame, it should have been doing it in ALL of
them.

Its fixed in my internal code base, and will be coming down the pipe soon.
I've now turned HWTSO back on as default as well. Thanks to everyone
that replied.

Jack


On Mon, Dec 5, 2011 at 6:16 AM, Andre Oppermann <andre@freebsd.org> wrote:

> On 02.12.2011 21:06, Jack Vogel wrote:
>
>> I have just discovered during some validation testing on the em driver,
>> that if HWTSO is enabled
>> at the same time as VLAN_HWTAGGING and/or VLAN_HWFILTER that its broken.
>>
>> Looking at the vlan source its not obvious why this would be the case, but
>> netperf just fails completely.
>> I can turn TAGGING off and I can get TSO, or turn off TSO and get
>> TAGGING/FILTERING, but not
>> both :(
>>
>> My tester only has 8.2 to try this on, not sure if its a bug or a design
>> limitation, can someone
>> enlighten me?  At the moment I'm thinking of just turning HWTSO off by
>> default?
>>
>
> Maybe the HW or drive chokes on the overall ethernet frame being longer
> than 64K due to the VLAN header even though the IP packet part stays
> within the limits?
>
> --
> Andre
>



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAFOYbcng-8=vCxitx%2BGVtuOJx983kCVs3mCZGfhLaTvND8Sdqg>