[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <Yrqogz8U+BxtFWiu@nanopsycho>
Date: Tue, 28 Jun 2022 09:06:43 +0200
From: Jiri Pirko <jiri@...nulli.us>
To: Jakub Kicinski <kuba@...nel.org>
Cc: Shannon Nelson <snelson@...sando.io>,
Ido Schimmel <idosch@...dia.com>, netdev@...r.kernel.org,
davem@...emloft.net, petrm@...dia.com, pabeni@...hat.com,
edumazet@...gle.com, mlxsw@...dia.com
Subject: Re: [patch net-next 00/11] mlxsw: Implement dev info and dev flash
for line cards
Mon, Jun 27, 2022 at 08:52:09PM CEST, kuba@...nel.org wrote:
>On Mon, 27 Jun 2022 11:38:50 -0700 Shannon Nelson wrote:
>> >> Can you encode the base device's PCI info into the auxiliary device's id
>> >
>> > Would look odd to he PCI BDF in auxdev addsess, wouldn't it?
>>
>> Sure, it looks a little odd to see something like mycore.app.1281, but
>> it does afford the auxiliary driver, and any other observer, a way to
>> figure out which device it is representing. This also works nicely when
>> trying to associate an auxiliary driver instance for a VF with the
>> matching VF PCI driver instance.
>
>I'd personally not mind divorcing devlink from bus devices a little
>more. On one hand we have cases like this where there's naturally no
How exactly do you envision to do this? There is a good reason to have
the handle based on bus/name, as it is constant and predictable.
>bus device, on the other we have multi-link PCI devices which want to
>straddle NUMA nodes but otherwise are just a logical unit.
Powered by blists - more mailing lists