[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <be5f26b8-fbe5-841d-5469-3adcfc178ab3@amd.com>
Date: Mon, 5 Jun 2023 09:26:25 -0700
From: Shannon Nelson <shannon.nelson@....com>
To: Jakub Kicinski <kuba@...nel.org>
Cc: netdev@...r.kernel.org, davem@...emloft.net, brett.creeley@....com,
drivers@...sando.io, Nitya Sunkad <nitya.sunkad@....com>
Subject: Re: [PATCH net-next] ionic: add support for ethtool extended stat
link_down_count
On 6/2/23 11:47 PM, Jakub Kicinski wrote:
> On Fri, 2 Jun 2023 10:32:52 -0700 Shannon Nelson wrote:
>> Following the example of 9a0f830f8026 ("ethtool: linkstate: add a statistic
>> for PHY down events"), added support for link down events.
>>
>> Added callback ionic_get_link_ext_stats to ionic_ethtool.c to support
>> link_down_count, a property of netdev that gets incremented every time
>> the device link goes down.
>
> Hm, could you say more about motivation? The ethtool stat is supposed to
> come from HW and represent PHY-level flap count. It's used primarily to
> find bad cables in a datacenter. Is this also the use case for ionic?
> It's unclear to me whether ionic interfaces are seeing an actual PHY or
> just some virtual link state of the IPU and in the latter case it's not
> really a match.
This is a fair question - yes, it is possible that the FW could fake it,
but normally this is a signal from the HW. However, I suppose it would
be best to only have the PF reporting this counter, and not the VFs,
which is currently the case here. I'll have Nitya modify this for PF only.
Thanks,
sln
Powered by blists - more mailing lists