Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210824101238.21105-1-harini.katakam@xilinx.com> <20210824140542.GA17195@hoboy.vegasvil.org>
In-Reply-To: <20210824140542.GA17195@hoboy.vegasvil.org>
From:   Harini Katakam <harinik@xilinx.com>
Date:   Tue, 24 Aug 2021 20:59:20 +0530
Message-ID: <CAFcVECKXOVwpsR=bEUmTXZpSQTjez1fjf1X9bXV_sFCe_U3_qA@mail.gmail.com>
Subject: Re: [RFC PATCH] net: macb: Process tx timestamp only on ptp packets
To:     Richard Cochran <richardcochran@gmail.com>
Cc:     Harini Katakam <harini.katakam@xilinx.com>,
        Nicolas Ferre <nicolas.ferre@microchip.com>,
        David Miller <davem@davemloft.net>,
        Claudiu Beznea <claudiu.beznea@microchip.com>,
        Andrei Pistirica <andrei.pistirica@microchip.com>,
        kuba@kernel.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org,
        Michal Simek <michal.simek@xilinx.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi Richard,

On Tue, Aug 24, 2021 at 7:35 PM Richard Cochran
<richardcochran@gmail.com> wrote:
>
> On Tue, Aug 24, 2021 at 03:42:38PM +0530, Harini Katakam wrote:
> > The current implementation timestamps all packets and also processes
> > the BD timestamp for the same. While it is true that HWTSTAMP_TX_ON
> > enables timestamps for outgoing packets, the sender of the packet
> > i.e. linuxptp enables timestamp for PTP or PTP event packets. Cadence
> > GEM IP has a provision to enable this in HW only for PTP packets.
> > Enable this option in DMA BD settings register to decrease overhead.
>
> NAK, because the HWTSTAMP_TX_ON means to time stamp any frame marked
> by user space, not just PTP frames.
>
> This patch does not "decrease overhead" because the code tests whether
> time stamping was request per packet:
>

Thanks for the review.
Yes, there is no SW overhead because the  skb check ensures timestamp
post processing is done only on requested packets. But the IP
timestamps all packets
because this is a register level setting, not per packet. That's the
overhead I was referring to.
But based on your explanation, it looks like we have no option but to enable
TSTAMP_ALL_FRAMES. Thanks.

Regards,
Harini