Re: [PATCH net-next 0/8] nfp: offload LAG for tc flower egress

From: John Hurley <john.hurley@netronome.com>
To: Jakub Kicinski <jakub.kicinski@netronome.com>
Cc: Jiri Pirko <jiri@resnulli.us>, David Miller <davem@davemloft.net>,
	Linux Netdev List <netdev@vger.kernel.org>,
	oss-drivers@netronome.com, Jay Vosburgh <j.vosburgh@gmail.com>,
	Veaceslav Falico <vfalico@gmail.com>,
	Andy Gospodarek <andy@greyhouse.net>
Subject: Re: [PATCH net-next 0/8] nfp: offload LAG for tc flower egress
Date: Tue, 29 May 2018 15:08:48 +0100	[thread overview]
Message-ID: <CAK+XE=kAM=sv4vpSKuyaiHzdLCNZUwuEL0ggODCQnsM+sfCw8w@mail.gmail.com> (raw)
In-Reply-To: <20180525194728.7aa4a116@cakuba>

On Sat, May 26, 2018 at 3:47 AM, Jakub Kicinski
<jakub.kicinski@netronome.com> wrote:
> On Fri, 25 May 2018 08:48:09 +0200, Jiri Pirko wrote:
>> Thu, May 24, 2018 at 04:22:47AM CEST, jakub.kicinski@netronome.com wrote:
>> >Hi!
>> >
>> >This series from John adds bond offload to the nfp driver.  Patch 5
>> >exposes the hash type for NETDEV_LAG_TX_TYPE_HASH to make sure nfp
>> >hashing matches that of the software LAG.  This may be unnecessarily
>> >conservative, let's see what LAG maintainers think :)
>>
>> So you need to restrict offload to only certain hash algo? In mlxsw, we
>> just ignore the lag setting and do some hw default hashing. Would not be
>> enough? Note that there's a good reason for it, as you see, in team, the
>> hashing is done in a BPF function and could be totally arbitrary.
>> Your patchset effectively disables team offload for nfp.
>
> My understanding is that the project requirements only called for L3/L4
> hash algorithm offload, hence the temptation to err on the side of
> caution and not offload all the bond configurations.  John can provide
> more details.  Not being able to offload team is unfortunate indeed.

Hi Jiri,
Yes, as Jakub mentions, we restrict ourselves to L3/L4 hash algorithm
as this is currently what is supported in fw.
Hopefully this will change as fw features are expanded.
I understand the issue this presents with offloading team.
Perhaps resorting to a default hw hash for team is acceptable.
John