Re: [PATCH/RFC net-next 1/2] devlink: expose port function commands to assign VFs to multiple netdevs

From: Jiri Pirko <jiri@resnulli.us>
To: Simon Horman <simon.horman@corigine.com>
Cc: David Miller <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
	Michael Chan <michael.chan@broadcom.com>,
	Andy Gospodarek <andy@greyhouse.net>,
	Gal Pressman <gal@nvidia.com>, Saeed Mahameed <saeed@kernel.org>,
	Jesse Brandeburg <jesse.brandeburg@intel.com>,
	Tony Nguyen <anthony.l.nguyen@intel.com>,
	Edward Cree <ecree.xilinx@gmail.com>,
	Vladimir Oltean <vladimir.oltean@nxp.com>,
	Andrew Lunn <andrew@lunn.ch>, Fei Qin <fei.qin@corigine.com>,
	netdev@vger.kernel.org, oss-drivers@corigine.com
Subject: Re: [PATCH/RFC net-next 1/2] devlink: expose port function commands to assign VFs to multiple netdevs
Date: Wed, 8 Feb 2023 13:34:19 +0100	[thread overview]
Message-ID: <Y+OWy0prxf5pNWpv@nanopsycho> (raw)
In-Reply-To: <Y+OQmjJFeQeF2kJx@corigine.com>

Wed, Feb 08, 2023 at 01:07:54PM CET, simon.horman@corigine.com wrote:
>On Wed, Feb 08, 2023 at 12:40:45PM +0100, Jiri Pirko wrote:
>> Mon, Feb 06, 2023 at 04:36:02PM CET, simon.horman@corigine.com wrote:
>> >From: Fei Qin <fei.qin@corigine.com>
>> >
>> >Multiple physical ports of the same NIC may share the single
>> >PCI address. In some cases, assigning VFs to different physical
>> >ports can be demanded, especially under high-traffic scenario.
>> >Load balancing can be realized in virtualised use¬cases through
>> >distributing packets between different physical ports with LAGs
>> >of VFs which are assigned to those physical ports.
>> >
>> >This patch adds new attribute "vf_count" to 'devlink port function'
>> >API which only can be shown and configured under devlink ports
>> >with flavor "DEVLINK_PORT_FLAVOUR_PHYSICAL".
>> 
>> I have to be missing something. That is the meaning of "assigning VF"
>> to a physical port? Why there should be any relationship between
>> physical port and VF other than configured forwarding (using TC for
>> example)?
>> 
>> This seems very wrong. Preliminary NAK.
>
>Of course if TC is involved, then we have flexibility.
>
>What we are talking about here is primarily legacy mode.

I don't see any reason to add knobs for purpose of supporting the legacy
mode, sorry.

If you need this functionality, use TC.

>And the behaviour described would, when enabled allow NFP based NICs
>to behave more like most other multi-port NICs.
>
>That is, we can envisage a VEB with some VFs and one physical port.
>And anther with other VFs and another physical port.
>
>This is as opposed to a single VEB with all VFs, as is currently
>the case on NFP based NICs (but not most other multi-port NICs).
>