[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <8e9d830b-556b-b8e6-45df-0bf7971b4237@intel.com>
Date: Tue, 3 Oct 2023 15:09:39 +0200
From: Alexander Lobakin <aleksander.lobakin@...el.com>
To: Jakub Kicinski <kuba@...nel.org>, Larysa Zaremba
<larysa.zaremba@...el.com>
CC: <bpf@...r.kernel.org>, <ast@...nel.org>, <daniel@...earbox.net>,
<andrii@...nel.org>, <martin.lau@...ux.dev>, <song@...nel.org>, <yhs@...com>,
<john.fastabend@...il.com>, <kpsingh@...nel.org>, <sdf@...gle.com>,
<haoluo@...gle.com>, <jolsa@...nel.org>, David Ahern <dsahern@...il.com>,
Willem de Bruijn <willemb@...gle.com>, Jesper Dangaard Brouer
<brouer@...hat.com>, Anatoly Burakov <anatoly.burakov@...el.com>, "Alexander
Lobakin" <alexandr.lobakin@...el.com>, Magnus Karlsson
<magnus.karlsson@...il.com>, Maryam Tahhan <mtahhan@...hat.com>,
<xdp-hints@...-project.net>, <netdev@...r.kernel.org>, Willem de Bruijn
<willemdebruijn.kernel@...il.com>, Alexei Starovoitov
<alexei.starovoitov@...il.com>, Simon Horman <simon.horman@...igine.com>,
Tariq Toukan <tariqt@...lanox.com>, Saeed Mahameed <saeedm@...lanox.com>,
Maciej Fijalkowski <maciej.fijalkowski@...el.com>
Subject: Re: [RFC bpf-next v2 09/24] xdp: Add VLAN tag hint
From: Jakub Kicinski <kuba@...nel.org>
Date: Tue, 3 Oct 2023 05:35:19 -0700
> On Wed, 27 Sep 2023 09:51:09 +0200 Larysa Zaremba wrote:
>> Implement functionality that enables drivers to expose VLAN tag
>> to XDP code.
>>
>> VLAN tag is represented by 2 variables:
>> - protocol ID, which is passed to bpf code in BE
>> - VLAN TCI, in host byte order
>
> Sorry for a random chime-in but was there any discussion about
> the validity of VLAN stripping as an offload?
>
> I always thought this is a legacy "Windows" thing which allowed
> Windows drivers to operate on VLAN-tagged networks even before
> the OS itself understood VLANs... Do people actually care about
> having it enabled?
On MIPS routers, I actually have some perf gains from having it enabled.
So they do, I'd say. Mediatek even has DSA tag stripping. Both save you
some skb->data push-pulls, csum corrections when CHECKSUM_COMPLETE, skb
unsharing in some cases, reduce L3/L4 headers cacheline spanning etc.
Thanks,
Olek
Powered by blists - more mailing lists