[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <2E243F8C-76F8-4792-B8C4-201E65F124F6@net-swift.com>
Date: Wed, 26 Jul 2023 11:12:41 +0800
From: "mengyuanlou@...-swift.com" <mengyuanlou@...-swift.com>
To: Jakub Kicinski <kuba@...nel.org>
Cc: netdev@...r.kernel.org
Subject: Re: [PATCH net-next 1/2] net: ngbe: add ncsi_enable flag for wangxun
nics
> 2023年7月26日 10:44,Jakub Kicinski <kuba@...nel.org> 写道:
>
> On Wed, 26 Jul 2023 09:59:15 +0800 mengyuanlou@...-swift.com wrote:
>>> 2023年7月26日 07:22,Jakub Kicinski <kuba@...nel.org> 写道:
>>> On Mon, 24 Jul 2023 17:24:58 +0800 Mengyuan Lou wrote:
>>>> + netdev->ncsi_enabled = wx->ncsi_hw_supported;
>>>
>>> I don't think that enabled and supported are the same thing.
>>> If server has multiple NICs or a NIC with multiple ports and
>>> BMC only uses one, or even none, we shouldn't keep the PHY up.
>>> By that logic 99% of server NICs should report NCSI as enabled.
>>
>> For a NIC with multiple ports, BMC switch connection for port0 to port1 online,
>> Drivers can not know port1 should keep up, if do not set ncsi_enabled before.
>
> I'm not crystal clear on what you're saying. But BMC sends a enable
> command to the NIC to enable a channel (or some such). This is all
> handled by FW. The FW can tell the host that the NCSI is now active
> on port1 and not port0.
>
>
Ok, I think I understand.
Thanks.
Another question.
Then, after drivers know that portx is using for BMC, it is necessary to
let phy to know this port should not be suspended?
I mean this patch 2/2 is useful.
Powered by blists - more mailing lists