1=========================
2NXP SJA1105 switch driver
3=========================
4
5Overview
6========
7
8The NXP SJA1105 is a family of 6 devices:
9
10- SJA1105E: First generation, no TTEthernet
11- SJA1105T: First generation, TTEthernet
12- SJA1105P: Second generation, no TTEthernet, no SGMII
13- SJA1105Q: Second generation, TTEthernet, no SGMII
14- SJA1105R: Second generation, no TTEthernet, SGMII
15- SJA1105S: Second generation, TTEthernet, SGMII
16
17These are SPI-managed automotive switches, with all ports being gigabit
18capable, and supporting MII/RMII/RGMII and optionally SGMII on one port.
19
20Being automotive parts, their configuration interface is geared towards
21set-and-forget use, with minimal dynamic interaction at runtime. They
22require a static configuration to be composed by software and packed
23with CRC and table headers, and sent over SPI.
24
25The static configuration is composed of several configuration tables. Each
26table takes a number of entries. Some configuration tables can be (partially)
27reconfigured at runtime, some not. Some tables are mandatory, some not:
28
29============================= ================== =============================
30Table                          Mandatory          Reconfigurable
31============================= ================== =============================
32Schedule                       no                 no
33Schedule entry points          if Scheduling      no
34VL Lookup                      no                 no
35VL Policing                    if VL Lookup       no
36VL Forwarding                  if VL Lookup       no
37L2 Lookup                      no                 no
38L2 Policing                    yes                no
39VLAN Lookup                    yes                yes
40L2 Forwarding                  yes                partially (fully on P/Q/R/S)
41MAC Config                     yes                partially (fully on P/Q/R/S)
42Schedule Params                if Scheduling      no
43Schedule Entry Points Params   if Scheduling      no
44VL Forwarding Params           if VL Forwarding   no
45L2 Lookup Params               no                 partially (fully on P/Q/R/S)
46L2 Forwarding Params           yes                no
47Clock Sync Params              no                 no
48AVB Params                     no                 no
49General Params                 yes                partially
50Retagging                      no                 yes
51xMII Params                    yes                no
52SGMII                          no                 yes
53============================= ================== =============================
54
55
56Also the configuration is write-only (software cannot read it back from the
57switch except for very few exceptions).
58
59The driver creates a static configuration at probe time, and keeps it at
60all times in memory, as a shadow for the hardware state. When required to
61change a hardware setting, the static configuration is also updated.
62If that changed setting can be transmitted to the switch through the dynamic
63reconfiguration interface, it is; otherwise the switch is reset and
64reprogrammed with the updated static configuration.
65
66Traffic support
67===============
68
69The switches do not support switch tagging in hardware. But they do support
70customizing the TPID by which VLAN traffic is identified as such. The switch
71driver is leveraging ``CONFIG_NET_DSA_TAG_8021Q`` by requesting that special
72VLANs (with a custom TPID of ``ETH_P_EDSA`` instead of ``ETH_P_8021Q``) are
73installed on its ports when not in ``vlan_filtering`` mode. This does not
74interfere with the reception and transmission of real 802.1Q-tagged traffic,
75because the switch does no longer parse those packets as VLAN after the TPID
76change.
77The TPID is restored when ``vlan_filtering`` is requested by the user through
78the bridge layer, and general IP termination becomes no longer possible through
79the switch netdevices in this mode.
80
81The switches have two programmable filters for link-local destination MACs.
82These are used to trap BPDUs and PTP traffic to the master netdevice, and are
83further used to support STP and 1588 ordinary clock/boundary clock
84functionality.
85
86The following traffic modes are supported over the switch netdevices:
87
88+--------------------+------------+------------------+------------------+
89|                    | Standalone |   Bridged with   |   Bridged with   |
90|                    |    ports   | vlan_filtering 0 | vlan_filtering 1 |
91+====================+============+==================+==================+
92| Regular traffic    |     Yes    |       Yes        |  No (use master) |
93+--------------------+------------+------------------+------------------+
94| Management traffic |     Yes    |       Yes        |       Yes        |
95|    (BPDU, PTP)     |            |                  |                  |
96+--------------------+------------+------------------+------------------+
97
98Switching features
99==================
100
101The driver supports the configuration of L2 forwarding rules in hardware for
102port bridging. The forwarding, broadcast and flooding domain between ports can
103be restricted through two methods: either at the L2 forwarding level (isolate
104one bridge's ports from another's) or at the VLAN port membership level
105(isolate ports within the same bridge). The final forwarding decision taken by
106the hardware is a logical AND of these two sets of rules.
107
108The hardware tags all traffic internally with a port-based VLAN (pvid), or it
109decodes the VLAN information from the 802.1Q tag. Advanced VLAN classification
110is not possible. Once attributed a VLAN tag, frames are checked against the
111port's membership rules and dropped at ingress if they don't match any VLAN.
112This behavior is available when switch ports are enslaved to a bridge with
113``vlan_filtering 1``.
114
115Normally the hardware is not configurable with respect to VLAN awareness, but
116by changing what TPID the switch searches 802.1Q tags for, the semantics of a
117bridge with ``vlan_filtering 0`` can be kept (accept all traffic, tagged or
118untagged), and therefore this mode is also supported.
119
120Segregating the switch ports in multiple bridges is supported (e.g. 2 + 2), but
121all bridges should have the same level of VLAN awareness (either both have
122``vlan_filtering`` 0, or both 1). Also an inevitable limitation of the fact
123that VLAN awareness is global at the switch level is that once a bridge with
124``vlan_filtering`` enslaves at least one switch port, the other un-bridged
125ports are no longer available for standalone traffic termination.
126
127Topology and loop detection through STP is supported.
128
129L2 FDB manipulation (add/delete/dump) is currently possible for the first
130generation devices. Aging time of FDB entries, as well as enabling fully static
131management (no address learning and no flooding of unknown traffic) is not yet
132configurable in the driver.
133
134A special comment about bridging with other netdevices (illustrated with an
135example):
136
137A board has eth0, eth1, swp0@eth1, swp1@eth1, swp2@eth1, swp3@eth1.
138The switch ports (swp0-3) are under br0.
139It is desired that eth0 is turned into another switched port that communicates
140with swp0-3.
141
142If br0 has vlan_filtering 0, then eth0 can simply be added to br0 with the
143intended results.
144If br0 has vlan_filtering 1, then a new br1 interface needs to be created that
145enslaves eth0 and eth1 (the DSA master of the switch ports). This is because in
146this mode, the switch ports beneath br0 are not capable of regular traffic, and
147are only used as a conduit for switchdev operations.
148
149Device Tree bindings and board design
150=====================================
151
152This section references ``Documentation/devicetree/bindings/net/dsa/sja1105.txt``
153and aims to showcase some potential switch caveats.
154
155RMII PHY role and out-of-band signaling
156---------------------------------------
157
158In the RMII spec, the 50 MHz clock signals are either driven by the MAC or by
159an external oscillator (but not by the PHY).
160But the spec is rather loose and devices go outside it in several ways.
161Some PHYs go against the spec and may provide an output pin where they source
162the 50 MHz clock themselves, in an attempt to be helpful.
163On the other hand, the SJA1105 is only binary configurable - when in the RMII
164MAC role it will also attempt to drive the clock signal. To prevent this from
165happening it must be put in RMII PHY role.
166But doing so has some unintended consequences.
167In the RMII spec, the PHY can transmit extra out-of-band signals via RXD[1:0].
168These are practically some extra code words (/J/ and /K/) sent prior to the
169preamble of each frame. The MAC does not have this out-of-band signaling
170mechanism defined by the RMII spec.
171So when the SJA1105 port is put in PHY role to avoid having 2 drivers on the
172clock signal, inevitably an RMII PHY-to-PHY connection is created. The SJA1105
173emulates a PHY interface fully and generates the /J/ and /K/ symbols prior to
174frame preambles, which the real PHY is not expected to understand. So the PHY
175simply encodes the extra symbols received from the SJA1105-as-PHY onto the
176100Base-Tx wire.
177On the other side of the wire, some link partners might discard these extra
178symbols, while others might choke on them and discard the entire Ethernet
179frames that follow along. This looks like packet loss with some link partners
180but not with others.
181The take-away is that in RMII mode, the SJA1105 must be let to drive the
182reference clock if connected to a PHY.
183
184RGMII fixed-link and internal delays
185------------------------------------
186
187As mentioned in the bindings document, the second generation of devices has
188tunable delay lines as part of the MAC, which can be used to establish the
189correct RGMII timing budget.
190When powered up, these can shift the Rx and Tx clocks with a phase difference
191between 73.8 and 101.7 degrees.
192The catch is that the delay lines need to lock onto a clock signal with a
193stable frequency. This means that there must be at least 2 microseconds of
194silence between the clock at the old vs at the new frequency. Otherwise the
195lock is lost and the delay lines must be reset (powered down and back up).
196In RGMII the clock frequency changes with link speed (125 MHz at 1000 Mbps, 25
197MHz at 100 Mbps and 2.5 MHz at 10 Mbps), and link speed might change during the
198AN process.
199In the situation where the switch port is connected through an RGMII fixed-link
200to a link partner whose link state life cycle is outside the control of Linux
201(such as a different SoC), then the delay lines would remain unlocked (and
202inactive) until there is manual intervention (ifdown/ifup on the switch port).
203The take-away is that in RGMII mode, the switch's internal delays are only
204reliable if the link partner never changes link speeds, or if it does, it does
205so in a way that is coordinated with the switch port (practically, both ends of
206the fixed-link are under control of the same Linux system).
207As to why would a fixed-link interface ever change link speeds: there are
208Ethernet controllers out there which come out of reset in 100 Mbps mode, and
209their driver inevitably needs to change the speed and clock frequency if it's
210required to work at gigabit.
211
212MDIO bus and PHY management
213---------------------------
214
215The SJA1105 does not have an MDIO bus and does not perform in-band AN either.
216Therefore there is no link state notification coming from the switch device.
217A board would need to hook up the PHYs connected to the switch to any other
218MDIO bus available to Linux within the system (e.g. to the DSA master's MDIO
219bus). Link state management then works by the driver manually keeping in sync
220(over SPI commands) the MAC link speed with the settings negotiated by the PHY.
221