Re: [PATCH RFC V1 net-next 3/4] net: Let the active time stamping layer be selectable.

From: Vladimir Oltean <vladimir.oltean@nxp.com>
To: Richard Cochran <richardcochran@gmail.com>
Cc: Andrew Lunn <andrew@lunn.ch>,
	"netdev@vger.kernel.org" <netdev@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	David Miller <davem@davemloft.net>,
	Grygorii Strashko <grygorii.strashko@ti.com>,
	Jakub Kicinski <kuba@kernel.org>,
	Joakim Zhang <qiangqing.zhang@nxp.com>,
	Kurt Kanzenbach <kurt@linutronix.de>,
	Miroslav Lichvar <mlichvar@redhat.com>,
	Russell King <linux@arm.linux.org.uk>
Subject: Re: [PATCH RFC V1 net-next 3/4] net: Let the active time stamping layer be selectable.
Date: Fri, 21 Jan 2022 16:23:27 +0000	[thread overview]
Message-ID: <20220121162327.4p3iqbtt4qtnknhp@skbuf> (raw)
In-Reply-To: <20220121152820.GA15600@hoboy.vegasvil.org>

On Fri, Jan 21, 2022 at 07:28:20AM -0800, Richard Cochran wrote:
> On Fri, Jan 21, 2022 at 02:50:36PM +0000, Vladimir Oltean wrote:
> > So as I mentioned earlier, the use case would be hardware performance
> > testing and diagnosing. You may consider that as not that important, but
> > this is basically what I had to do for several months, and even wrote
> > a program for that, that collects packet timestamps at all possible points.
> 
> This is not possible without making a brand new CMSG to accommodate
> time stamps from all the various layers.
> 
> That is completely out of scope for this series.
> 
> The only practical use case of this series is to switch from PHY back to MAC.

I don't think my proposal is out of scope. It deals with the same thing
as what you propose: the kernel makes a selection by default, user space
can change it. The need for PHC identification in the cmsg arises as a
direct consequence of the fact that there are multiple PHCs in the path.
It isn't a requirement that I introduced artificially. Your solution has
a need to deal with that problem too, it's just that it omits to do so:

|               When changing the value, some packets in the kernel
|               networking stack may still be delivered with time
|               stamps from the previous provider.