Re: [net-next, v2, 3/7] net: dsa: free skb->cb usage in core driver

From: Vladimir Oltean <olteanv@gmail.com>
To: Yangbo Lu <yangbo.lu@nxp.com>
Cc: netdev@vger.kernel.org,
	Richard Cochran <richardcochran@gmail.com>,
	Vladimir Oltean <vladimir.oltean@nxp.com>,
	"David S . Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>,
	Jonathan Corbet <corbet@lwn.net>,
	Kurt Kanzenbach <kurt@linutronix.de>,
	Andrew Lunn <andrew@lunn.ch>,
	Vivien Didelot <vivien.didelot@gmail.com>,
	Florian Fainelli <f.fainelli@gmail.com>,
	Claudiu Manoil <claudiu.manoil@nxp.com>,
	Alexandre Belloni <alexandre.belloni@bootlin.com>,
	UNGLinuxDriver@microchip.com, linux-doc@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [net-next, v2, 3/7] net: dsa: free skb->cb usage in core driver
Date: Mon, 26 Apr 2021 21:16:37 +0300	[thread overview]
Message-ID: <20210426181637.2rneohfxkrvwctf2@skbuf> (raw)
In-Reply-To: <20210426093802.38652-4-yangbo.lu@nxp.com>

On Mon, Apr 26, 2021 at 05:37:58PM +0800, Yangbo Lu wrote:
> Free skb->cb usage in core driver and let device drivers decide to
> use or not. The reason having a DSA_SKB_CB(skb)->clone was because
> dsa_skb_tx_timestamp() which may set the clone pointer was called
> before p->xmit() which would use the clone if any, and the device
> driver has no way to initialize the clone pointer.
>
> Although for now putting memset(skb->cb, 0, 48) at beginning of
> dsa_slave_xmit() by this patch is not very good, there is still way
> to improve this. Otherwise, some other new features, like one-step
> timestamp which needs a flag of skb marked in dsa_skb_tx_timestamp(),
> and handles as one-step timestamp in p->xmit() will face same
> situation.
>
> Signed-off-by: Yangbo Lu <yangbo.lu@nxp.com>
> ---
> Changes for v2:
> 	- Added this patch.
> ---
>  drivers/net/dsa/ocelot/felix.c         |  1 +
>  drivers/net/dsa/sja1105/sja1105_main.c |  2 +-
>  drivers/net/dsa/sja1105/sja1105_ptp.c  |  4 +++-
>  drivers/net/ethernet/mscc/ocelot.c     |  6 +++---
>  drivers/net/ethernet/mscc/ocelot_net.c |  2 +-
>  include/linux/dsa/sja1105.h            |  3 ++-
>  include/net/dsa.h                      | 14 --------------
>  include/soc/mscc/ocelot.h              |  8 ++++++++
>  net/dsa/slave.c                        |  3 +--
>  net/dsa/tag_ocelot.c                   |  8 ++++----
>  net/dsa/tag_ocelot_8021q.c             |  8 ++++----
>  11 files changed, 28 insertions(+), 31 deletions(-)
>
> diff --git a/drivers/net/dsa/ocelot/felix.c b/drivers/net/dsa/ocelot/felix.c
> index d679f023dc00..8980d56ee793 100644
> --- a/drivers/net/dsa/ocelot/felix.c
> +++ b/drivers/net/dsa/ocelot/felix.c
> @@ -1403,6 +1403,7 @@ static bool felix_txtstamp(struct dsa_switch *ds, int port,
>
>  	if (ocelot->ptp && ocelot_port->ptp_cmd == IFH_REW_OP_TWO_STEP_PTP) {
>  		ocelot_port_add_txtstamp_skb(ocelot, port, clone);
> +		OCELOT_SKB_CB(skb)->clone = clone;
>  		return true;
>  	}
>

Uh-oh, this patch fails to build:

In file included from ./include/soc/mscc/ocelot_vcap.h:9:0,
                 from drivers/net/dsa/ocelot/felix.c:9:
drivers/net/dsa/ocelot/felix.c: In function ‘felix_txtstamp’:
drivers/net/dsa/ocelot/felix.c:1406:17: error: ‘skb’ undeclared (first use in this function)
   OCELOT_SKB_CB(skb)->clone = clone;
                 ^
./include/soc/mscc/ocelot.h:698:29: note: in definition of macro ‘OCELOT_SKB_CB’
  ((struct ocelot_skb_cb *)((skb)->cb))
                             ^~~
drivers/net/dsa/ocelot/felix.c:1406:17: note: each undeclared identifier is reported only once for each function it appears in
   OCELOT_SKB_CB(skb)->clone = clone;
                 ^
./include/soc/mscc/ocelot.h:698:29: note: in definition of macro ‘OCELOT_SKB_CB’
  ((struct ocelot_skb_cb *)((skb)->cb))
                             ^~~

It depends on changes made in patch 3.