Re: [PATCH net-next v3 00/12] net: dsa: microchip: PTP support for KSZ956x

From: Christian Eggers <ceggers@arri.de>
To: <Tristram.Ha@microchip.com>
Cc: <olteanv@gmail.com>, <kuba@kernel.org>, <andrew@lunn.ch>,
	<richardcochran@gmail.com>, <robh+dt@kernel.org>,
	<vivien.didelot@gmail.com>, <davem@davemloft.net>,
	<kurt.kanzenbach@linutronix.de>, <george.mccollister@gmail.com>,
	<marex@denx.de>, <helmut.grohne@intenta.de>,
	<pbarker@konsulko.com>, <Codrin.Ciubotariu@microchip.com>,
	<Woojung.Huh@microchip.com>, <UNGLinuxDriver@microchip.com>,
	<netdev@vger.kernel.org>, <devicetree@vger.kernel.org>,
	<linux-kernel@vger.kernel.org>
Subject: Re: [PATCH net-next v3 00/12] net: dsa: microchip: PTP support for KSZ956x
Date: Wed, 25 Nov 2020 22:08:39 +0100	[thread overview]
Message-ID: <3569829.EPWo3g8d0Q@n95hx1g2> (raw)
In-Reply-To: <BYAPR11MB35582F880B533EB2EE0CDD1DECE00@BYAPR11MB3558.namprd11.prod.outlook.com>

I need some help from Microchip, please read below.

On Thursday, 19 November 2020, 19:51:15 CET, Tristram.Ha@microchip.com wrote:
> There is one more requirement that is a little difficult to do.  The calculated peer delay
> needs to be programmed in hardware register, but the regular PTP stack has no way to
> send that command.  I think the driver has to do its own calculation by snooping on the
> Pdelay_Req/Pdelay_Resp/Pdelay_Resp_Follow_Up messages.

In an (offline) discussion with Vladimir we discovered, that the KSZ switch
behaves different as ptp4l expects: 

The KSZ switch forwards PTP (e.g. SYNC) messages in hardware (with updating
the correction field). For this, the peer delays need be configured for each
port.

ptp4l in turn expects to do the forwarding in software (for the P2P_TC clock
configuration). For this, no hardware configuration of the peer delay is
necessary. But due to limitations of currently available hardware, this TC
forwarding is currently only supported for 2 step clocks, as a one-step clock
would probably fully replace the originTimestamp field (similar as a BC, but
not as a TC).

Vladimir suggested to configure an ACL in the KSZ switch to block forwarding
of PTP messages between the user ports and to run ptp4l as BC. My idea is to
simply block forwarding of UDP messages with destination ports 319+320 and
L2 messages with the PTP Ether-Type.

I installed the following ACL (for UDP) in the Port ACL Access registers 0-F:
|_0__1__2__3__4__5__6__7__8__9__A__B__C__D__E__F
| 00 39 01 40 01 3F 42 22 00 00 00 60 00 00 00 01
ACL index: 0

Match: 
- MD=11 (L4)
- ENB=10 (UDP ports)
- S/D=0 (dst)
- EQ=1 (equal)
- MAX_PORT=320
- MIN_PORT=319
- PC=01 (min or max)
- PRO=17 (UDP, don't care?)
- FME=0 (disabled)

Action:
- PM=0 (disabled)
- P=0 (don't care)
- RPE=0 (disabled)
- RP=0 (don't care)
- MM=11 (replace)
- PORT_FWD_MAP: all ports to 0

Processing entry:
- Ruleset=0x0001
- FRN=0

Unfortunately, with this configuration PTP messages are still forwarded from
port 1 to port 2. Although I was successful in blocking other communication
(e.g. by MAC address), the matching rules above seem not to work. Is there an
error in the ACL, or is forwarding of PTP traffic independent of configured
ACLs?

regards
Christian