[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Y36uhLebt0Kx26Nc@x130.lan>
Date: Wed, 23 Nov 2022 15:36:36 -0800
From: Saeed Mahameed <saeed@...nel.org>
To: Maciej Fijalkowski <maciej.fijalkowski@...el.com>
Cc: "David S. Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>,
Eric Dumazet <edumazet@...gle.com>,
Saeed Mahameed <saeedm@...dia.com>, netdev@...r.kernel.org,
Tariq Toukan <tariqt@...dia.com>,
Shay Drory <shayd@...dia.com>, Parav Pandit <parav@...dia.com>
Subject: Re: [net 03/14] net/mlx5: SF: Fix probing active SFs during driver
probe phase
On 23 Nov 15:57, Maciej Fijalkowski wrote:
>On Mon, Nov 21, 2022 at 06:25:48PM -0800, Saeed Mahameed wrote:
>> From: Shay Drory <shayd@...dia.com>
>>
>> When SF devices and SF port representors are located on different
>> functions, unloading and reloading of SF parent driver doesn't recreate
>> the existing SF present in the device.
>> Fix it by querying SFs and probe active SFs during driver probe phase.
>
>Maybe shed some light on how it's actually being done? Have a few words
>that you're adding a workqueue dedicated for that etc. There is also a
>new mutex, I was always expecting that such mechanisms get a bit of
>explanation/justification why there is a need for its introduction.
>
it's needed so we can synchronize the new sync operation on load with the
pre-existing SF adding mechanism "device events" ..
But i don't believe it's a requirement to explain the code behind a
commit, I think for this case it's self explanatory for someone who
understand how the basic mechanism of adding SFs to a PF funciton works in
mlx5. but yes I agree, some more information would've been useful, we will
be more verbose for future patches.
>Not sure if including some example reproducer in here is mandatory or not
>(and therefore splat, if any). General feeling is that commit message
>could be beefed up.
>
reproducer is really as simple as stated in the commit message,
a devlink reload when you have SFs loaded on a dual function system
(smart nic) where the switchdev SF admin is on a remote cpu function.
Powered by blists - more mailing lists