xref: /linux/Documentation/networking/dsa/sja1105.rst (revision bdd1a21b52557ea8f61d0a5dc2f77151b576eb70)
1=========================
2NXP SJA1105 switch driver
3=========================
4
5Overview
6========
7
8The NXP SJA1105 is a family of 10 SPI-managed automotive switches:
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- SJA1110A: Third generation, TTEthernet, SGMII, integrated 100base-T1 and
17  100base-TX PHYs
18- SJA1110B: Third generation, TTEthernet, SGMII, 100base-T1, 100base-TX
19- SJA1110C: Third generation, TTEthernet, SGMII, 100base-T1, 100base-TX
20- SJA1110D: Third generation, TTEthernet, SGMII, 100base-T1
21
22Being automotive parts, their configuration interface is geared towards
23set-and-forget use, with minimal dynamic interaction at runtime. They
24require a static configuration to be composed by software and packed
25with CRC and table headers, and sent over SPI.
26
27The static configuration is composed of several configuration tables. Each
28table takes a number of entries. Some configuration tables can be (partially)
29reconfigured at runtime, some not. Some tables are mandatory, some not:
30
31============================= ================== =============================
32Table                          Mandatory          Reconfigurable
33============================= ================== =============================
34Schedule                       no                 no
35Schedule entry points          if Scheduling      no
36VL Lookup                      no                 no
37VL Policing                    if VL Lookup       no
38VL Forwarding                  if VL Lookup       no
39L2 Lookup                      no                 no
40L2 Policing                    yes                no
41VLAN Lookup                    yes                yes
42L2 Forwarding                  yes                partially (fully on P/Q/R/S)
43MAC Config                     yes                partially (fully on P/Q/R/S)
44Schedule Params                if Scheduling      no
45Schedule Entry Points Params   if Scheduling      no
46VL Forwarding Params           if VL Forwarding   no
47L2 Lookup Params               no                 partially (fully on P/Q/R/S)
48L2 Forwarding Params           yes                no
49Clock Sync Params              no                 no
50AVB Params                     no                 no
51General Params                 yes                partially
52Retagging                      no                 yes
53xMII Params                    yes                no
54SGMII                          no                 yes
55============================= ================== =============================
56
57
58Also the configuration is write-only (software cannot read it back from the
59switch except for very few exceptions).
60
61The driver creates a static configuration at probe time, and keeps it at
62all times in memory, as a shadow for the hardware state. When required to
63change a hardware setting, the static configuration is also updated.
64If that changed setting can be transmitted to the switch through the dynamic
65reconfiguration interface, it is; otherwise the switch is reset and
66reprogrammed with the updated static configuration.
67
68Traffic support
69===============
70
71The switches do not have hardware support for DSA tags, except for "slow
72protocols" for switch control as STP and PTP. For these, the switches have two
73programmable filters for link-local destination MACs.
74These are used to trap BPDUs and PTP traffic to the master netdevice, and are
75further used to support STP and 1588 ordinary clock/boundary clock
76functionality. For frames trapped to the CPU, source port and switch ID
77information is encoded by the hardware into the frames.
78
79But by leveraging ``CONFIG_NET_DSA_TAG_8021Q`` (a software-defined DSA tagging
80format based on VLANs), general-purpose traffic termination through the network
81stack can be supported under certain circumstances.
82
83Depending on VLAN awareness state, the following operating modes are possible
84with the switch:
85
86- Mode 1 (VLAN-unaware): a port is in this mode when it is used as a standalone
87  net device, or when it is enslaved to a bridge with ``vlan_filtering=0``.
88- Mode 2 (fully VLAN-aware): a port is in this mode when it is enslaved to a
89  bridge with ``vlan_filtering=1``. Access to the entire VLAN range is given to
90  the user through ``bridge vlan`` commands, but general-purpose (anything
91  other than STP, PTP etc) traffic termination is not possible through the
92  switch net devices. The other packets can be still by user space processed
93  through the DSA master interface (similar to ``DSA_TAG_PROTO_NONE``).
94- Mode 3 (best-effort VLAN-aware): a port is in this mode when enslaved to a
95  bridge with ``vlan_filtering=1``, and the devlink property of its parent
96  switch named ``best_effort_vlan_filtering`` is set to ``true``. When
97  configured like this, the range of usable VIDs is reduced (0 to 1023 and 3072
98  to 4094), so is the number of usable VIDs (maximum of 7 non-pvid VLANs per
99  port*), and shared VLAN learning is performed (FDB lookup is done only by
100  DMAC, not also by VID).
101
102To summarize, in each mode, the following types of traffic are supported over
103the switch net devices:
104
105+-------------+-----------+--------------+------------+
106|             |   Mode 1  |    Mode 2    |   Mode 3   |
107+=============+===========+==============+============+
108|   Regular   |    Yes    | No           |     Yes    |
109|   traffic   |           | (use master) |            |
110+-------------+-----------+--------------+------------+
111| Management  |    Yes    |     Yes      |     Yes    |
112| traffic     |           |              |            |
113| (BPDU, PTP) |           |              |            |
114+-------------+-----------+--------------+------------+
115
116To configure the switch to operate in Mode 3, the following steps can be
117followed::
118
119  ip link add dev br0 type bridge
120  # swp2 operates in Mode 1 now
121  ip link set dev swp2 master br0
122  # swp2 temporarily moves to Mode 2
123  ip link set dev br0 type bridge vlan_filtering 1
124  [   61.204770] sja1105 spi0.1: Reset switch and programmed static config. Reason: VLAN filtering
125  [   61.239944] sja1105 spi0.1: Disabled switch tagging
126  # swp3 now operates in Mode 3
127  devlink dev param set spi/spi0.1 name best_effort_vlan_filtering value true cmode runtime
128  [   64.682927] sja1105 spi0.1: Reset switch and programmed static config. Reason: VLAN filtering
129  [   64.711925] sja1105 spi0.1: Enabled switch tagging
130  # Cannot use VLANs in range 1024-3071 while in Mode 3.
131  bridge vlan add dev swp2 vid 1025 untagged pvid
132  RTNETLINK answers: Operation not permitted
133  bridge vlan add dev swp2 vid 100
134  bridge vlan add dev swp2 vid 101 untagged
135  bridge vlan
136  port    vlan ids
137  swp5     1 PVID Egress Untagged
138
139  swp2     1 PVID Egress Untagged
140           100
141           101 Egress Untagged
142
143  swp3     1 PVID Egress Untagged
144
145  swp4     1 PVID Egress Untagged
146
147  br0      1 PVID Egress Untagged
148  bridge vlan add dev swp2 vid 102
149  bridge vlan add dev swp2 vid 103
150  bridge vlan add dev swp2 vid 104
151  bridge vlan add dev swp2 vid 105
152  bridge vlan add dev swp2 vid 106
153  bridge vlan add dev swp2 vid 107
154  # Cannot use mode than 7 VLANs per port while in Mode 3.
155  [ 3885.216832] sja1105 spi0.1: No more free subvlans
156
157\* "maximum of 7 non-pvid VLANs per port": Decoding VLAN-tagged packets on the
158CPU in mode 3 is possible through VLAN retagging of packets that go from the
159switch to the CPU. In cross-chip topologies, the port that goes to the CPU
160might also go to other switches. In that case, those other switches will see
161only a retagged packet (which only has meaning for the CPU). So if they are
162interested in this VLAN, they need to apply retagging in the reverse direction,
163to recover the original value from it. This consumes extra hardware resources
164for this switch. There is a maximum of 32 entries in the Retagging Table of
165each switch device.
166
167As an example, consider this cross-chip topology::
168
169  +-------------------------------------------------+
170  | Host SoC                                        |
171  |           +-------------------------+           |
172  |           | DSA master for embedded |           |
173  |           |   switch (non-sja1105)  |           |
174  |  +--------+-------------------------+--------+  |
175  |  |   embedded L2 switch                      |  |
176  |  |                                           |  |
177  |  |   +--------------+     +--------------+   |  |
178  |  |   |DSA master for|     |DSA master for|   |  |
179  |  |   |  SJA1105 1   |     |  SJA1105 2   |   |  |
180  +--+---+--------------+-----+--------------+---+--+
181
182  +-----------------------+ +-----------------------+
183  |   SJA1105 switch 1    | |   SJA1105 switch 2    |
184  +-----+-----+-----+-----+ +-----+-----+-----+-----+
185  |sw1p0|sw1p1|sw1p2|sw1p3| |sw2p0|sw2p1|sw2p2|sw2p3|
186  +-----+-----+-----+-----+ +-----+-----+-----+-----+
187
188To reach the CPU, SJA1105 switch 1 (spi/spi2.1) uses the same port as is uses
189to reach SJA1105 switch 2 (spi/spi2.2), which would be port 4 (not drawn).
190Similarly for SJA1105 switch 2.
191
192Also consider the following commands, that add VLAN 100 to every sja1105 user
193port::
194
195  devlink dev param set spi/spi2.1 name best_effort_vlan_filtering value true cmode runtime
196  devlink dev param set spi/spi2.2 name best_effort_vlan_filtering value true cmode runtime
197  ip link add dev br0 type bridge
198  for port in sw1p0 sw1p1 sw1p2 sw1p3 \
199              sw2p0 sw2p1 sw2p2 sw2p3; do
200      ip link set dev $port master br0
201  done
202  ip link set dev br0 type bridge vlan_filtering 1
203  for port in sw1p0 sw1p1 sw1p2 sw1p3 \
204              sw2p0 sw2p1 sw2p2; do
205      bridge vlan add dev $port vid 100
206  done
207  ip link add link br0 name br0.100 type vlan id 100 && ip link set dev br0.100 up
208  ip addr add 192.168.100.3/24 dev br0.100
209  bridge vlan add dev br0 vid 100 self
210
211  bridge vlan
212  port    vlan ids
213  sw1p0    1 PVID Egress Untagged
214           100
215
216  sw1p1    1 PVID Egress Untagged
217           100
218
219  sw1p2    1 PVID Egress Untagged
220           100
221
222  sw1p3    1 PVID Egress Untagged
223           100
224
225  sw2p0    1 PVID Egress Untagged
226           100
227
228  sw2p1    1 PVID Egress Untagged
229           100
230
231  sw2p2    1 PVID Egress Untagged
232           100
233
234  sw2p3    1 PVID Egress Untagged
235
236  br0      1 PVID Egress Untagged
237           100
238
239SJA1105 switch 1 consumes 1 retagging entry for each VLAN on each user port
240towards the CPU. It also consumes 1 retagging entry for each non-pvid VLAN that
241it is also interested in, which is configured on any port of any neighbor
242switch.
243
244In this case, SJA1105 switch 1 consumes a total of 11 retagging entries, as
245follows:
246
247- 8 retagging entries for VLANs 1 and 100 installed on its user ports
248  (``sw1p0`` - ``sw1p3``)
249- 3 retagging entries for VLAN 100 installed on the user ports of SJA1105
250  switch 2 (``sw2p0`` - ``sw2p2``), because it also has ports that are
251  interested in it. The VLAN 1 is a pvid on SJA1105 switch 2 and does not need
252  reverse retagging.
253
254SJA1105 switch 2 also consumes 11 retagging entries, but organized as follows:
255
256- 7 retagging entries for the bridge VLANs on its user ports (``sw2p0`` -
257  ``sw2p3``).
258- 4 retagging entries for VLAN 100 installed on the user ports of SJA1105
259  switch 1 (``sw1p0`` - ``sw1p3``).
260
261Switching features
262==================
263
264The driver supports the configuration of L2 forwarding rules in hardware for
265port bridging. The forwarding, broadcast and flooding domain between ports can
266be restricted through two methods: either at the L2 forwarding level (isolate
267one bridge's ports from another's) or at the VLAN port membership level
268(isolate ports within the same bridge). The final forwarding decision taken by
269the hardware is a logical AND of these two sets of rules.
270
271The hardware tags all traffic internally with a port-based VLAN (pvid), or it
272decodes the VLAN information from the 802.1Q tag. Advanced VLAN classification
273is not possible. Once attributed a VLAN tag, frames are checked against the
274port's membership rules and dropped at ingress if they don't match any VLAN.
275This behavior is available when switch ports are enslaved to a bridge with
276``vlan_filtering 1``.
277
278Normally the hardware is not configurable with respect to VLAN awareness, but
279by changing what TPID the switch searches 802.1Q tags for, the semantics of a
280bridge with ``vlan_filtering 0`` can be kept (accept all traffic, tagged or
281untagged), and therefore this mode is also supported.
282
283Segregating the switch ports in multiple bridges is supported (e.g. 2 + 2), but
284all bridges should have the same level of VLAN awareness (either both have
285``vlan_filtering`` 0, or both 1). Also an inevitable limitation of the fact
286that VLAN awareness is global at the switch level is that once a bridge with
287``vlan_filtering`` enslaves at least one switch port, the other un-bridged
288ports are no longer available for standalone traffic termination.
289
290Topology and loop detection through STP is supported.
291
292L2 FDB manipulation (add/delete/dump) is currently possible for the first
293generation devices. Aging time of FDB entries, as well as enabling fully static
294management (no address learning and no flooding of unknown traffic) is not yet
295configurable in the driver.
296
297A special comment about bridging with other netdevices (illustrated with an
298example):
299
300A board has eth0, eth1, swp0@eth1, swp1@eth1, swp2@eth1, swp3@eth1.
301The switch ports (swp0-3) are under br0.
302It is desired that eth0 is turned into another switched port that communicates
303with swp0-3.
304
305If br0 has vlan_filtering 0, then eth0 can simply be added to br0 with the
306intended results.
307If br0 has vlan_filtering 1, then a new br1 interface needs to be created that
308enslaves eth0 and eth1 (the DSA master of the switch ports). This is because in
309this mode, the switch ports beneath br0 are not capable of regular traffic, and
310are only used as a conduit for switchdev operations.
311
312Offloads
313========
314
315Time-aware scheduling
316---------------------
317
318The switch supports a variation of the enhancements for scheduled traffic
319specified in IEEE 802.1Q-2018 (formerly 802.1Qbv). This means it can be used to
320ensure deterministic latency for priority traffic that is sent in-band with its
321gate-open event in the network schedule.
322
323This capability can be managed through the tc-taprio offload ('flags 2'). The
324difference compared to the software implementation of taprio is that the latter
325would only be able to shape traffic originated from the CPU, but not
326autonomously forwarded flows.
327
328The device has 8 traffic classes, and maps incoming frames to one of them based
329on the VLAN PCP bits (if no VLAN is present, the port-based default is used).
330As described in the previous sections, depending on the value of
331``vlan_filtering``, the EtherType recognized by the switch as being VLAN can
332either be the typical 0x8100 or a custom value used internally by the driver
333for tagging. Therefore, the switch ignores the VLAN PCP if used in standalone
334or bridge mode with ``vlan_filtering=0``, as it will not recognize the 0x8100
335EtherType. In these modes, injecting into a particular TX queue can only be
336done by the DSA net devices, which populate the PCP field of the tagging header
337on egress. Using ``vlan_filtering=1``, the behavior is the other way around:
338offloaded flows can be steered to TX queues based on the VLAN PCP, but the DSA
339net devices are no longer able to do that. To inject frames into a hardware TX
340queue with VLAN awareness active, it is necessary to create a VLAN
341sub-interface on the DSA master port, and send normal (0x8100) VLAN-tagged
342towards the switch, with the VLAN PCP bits set appropriately.
343
344Management traffic (having DMAC 01-80-C2-xx-xx-xx or 01-19-1B-xx-xx-xx) is the
345notable exception: the switch always treats it with a fixed priority and
346disregards any VLAN PCP bits even if present. The traffic class for management
347traffic has a value of 7 (highest priority) at the moment, which is not
348configurable in the driver.
349
350Below is an example of configuring a 500 us cyclic schedule on egress port
351``swp5``. The traffic class gate for management traffic (7) is open for 100 us,
352and the gates for all other traffic classes are open for 400 us::
353
354  #!/bin/bash
355
356  set -e -u -o pipefail
357
358  NSEC_PER_SEC="1000000000"
359
360  gatemask() {
361          local tc_list="$1"
362          local mask=0
363
364          for tc in ${tc_list}; do
365                  mask=$((${mask} | (1 << ${tc})))
366          done
367
368          printf "%02x" ${mask}
369  }
370
371  if ! systemctl is-active --quiet ptp4l; then
372          echo "Please start the ptp4l service"
373          exit
374  fi
375
376  now=$(phc_ctl /dev/ptp1 get | gawk '/clock time is/ { print $5; }')
377  # Phase-align the base time to the start of the next second.
378  sec=$(echo "${now}" | gawk -F. '{ print $1; }')
379  base_time="$(((${sec} + 1) * ${NSEC_PER_SEC}))"
380
381  tc qdisc add dev swp5 parent root handle 100 taprio \
382          num_tc 8 \
383          map 0 1 2 3 5 6 7 \
384          queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
385          base-time ${base_time} \
386          sched-entry S $(gatemask 7) 100000 \
387          sched-entry S $(gatemask "0 1 2 3 4 5 6") 400000 \
388          flags 2
389
390It is possible to apply the tc-taprio offload on multiple egress ports. There
391are hardware restrictions related to the fact that no gate event may trigger
392simultaneously on two ports. The driver checks the consistency of the schedules
393against this restriction and errors out when appropriate. Schedule analysis is
394needed to avoid this, which is outside the scope of the document.
395
396Routing actions (redirect, trap, drop)
397--------------------------------------
398
399The switch is able to offload flow-based redirection of packets to a set of
400destination ports specified by the user. Internally, this is implemented by
401making use of Virtual Links, a TTEthernet concept.
402
403The driver supports 2 types of keys for Virtual Links:
404
405- VLAN-aware virtual links: these match on destination MAC address, VLAN ID and
406  VLAN PCP.
407- VLAN-unaware virtual links: these match on destination MAC address only.
408
409The VLAN awareness state of the bridge (vlan_filtering) cannot be changed while
410there are virtual link rules installed.
411
412Composing multiple actions inside the same rule is supported. When only routing
413actions are requested, the driver creates a "non-critical" virtual link. When
414the action list also contains tc-gate (more details below), the virtual link
415becomes "time-critical" (draws frame buffers from a reserved memory partition,
416etc).
417
418The 3 routing actions that are supported are "trap", "drop" and "redirect".
419
420Example 1: send frames received on swp2 with a DA of 42:be:24:9b:76:20 to the
421CPU and to swp3. This type of key (DA only) when the port's VLAN awareness
422state is off::
423
424  tc qdisc add dev swp2 clsact
425  tc filter add dev swp2 ingress flower skip_sw dst_mac 42:be:24:9b:76:20 \
426          action mirred egress redirect dev swp3 \
427          action trap
428
429Example 2: drop frames received on swp2 with a DA of 42:be:24:9b:76:20, a VID
430of 100 and a PCP of 0::
431
432  tc filter add dev swp2 ingress protocol 802.1Q flower skip_sw \
433          dst_mac 42:be:24:9b:76:20 vlan_id 100 vlan_prio 0 action drop
434
435Time-based ingress policing
436---------------------------
437
438The TTEthernet hardware abilities of the switch can be constrained to act
439similarly to the Per-Stream Filtering and Policing (PSFP) clause specified in
440IEEE 802.1Q-2018 (formerly 802.1Qci). This means it can be used to perform
441tight timing-based admission control for up to 1024 flows (identified by a
442tuple composed of destination MAC address, VLAN ID and VLAN PCP). Packets which
443are received outside their expected reception window are dropped.
444
445This capability can be managed through the offload of the tc-gate action. As
446routing actions are intrinsic to virtual links in TTEthernet (which performs
447explicit routing of time-critical traffic and does not leave that in the hands
448of the FDB, flooding etc), the tc-gate action may never appear alone when
449asking sja1105 to offload it. One (or more) redirect or trap actions must also
450follow along.
451
452Example: create a tc-taprio schedule that is phase-aligned with a tc-gate
453schedule (the clocks must be synchronized by a 1588 application stack, which is
454outside the scope of this document). No packet delivered by the sender will be
455dropped. Note that the reception window is larger than the transmission window
456(and much more so, in this example) to compensate for the packet propagation
457delay of the link (which can be determined by the 1588 application stack).
458
459Receiver (sja1105)::
460
461  tc qdisc add dev swp2 clsact
462  now=$(phc_ctl /dev/ptp1 get | awk '/clock time is/ {print $5}') && \
463          sec=$(echo $now | awk -F. '{print $1}') && \
464          base_time="$(((sec + 2) * 1000000000))" && \
465          echo "base time ${base_time}"
466  tc filter add dev swp2 ingress flower skip_sw \
467          dst_mac 42:be:24:9b:76:20 \
468          action gate base-time ${base_time} \
469          sched-entry OPEN  60000 -1 -1 \
470          sched-entry CLOSE 40000 -1 -1 \
471          action trap
472
473Sender::
474
475  now=$(phc_ctl /dev/ptp0 get | awk '/clock time is/ {print $5}') && \
476          sec=$(echo $now | awk -F. '{print $1}') && \
477          base_time="$(((sec + 2) * 1000000000))" && \
478          echo "base time ${base_time}"
479  tc qdisc add dev eno0 parent root taprio \
480          num_tc 8 \
481          map 0 1 2 3 4 5 6 7 \
482          queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
483          base-time ${base_time} \
484          sched-entry S 01  50000 \
485          sched-entry S 00  50000 \
486          flags 2
487
488The engine used to schedule the ingress gate operations is the same that the
489one used for the tc-taprio offload. Therefore, the restrictions regarding the
490fact that no two gate actions (either tc-gate or tc-taprio gates) may fire at
491the same time (during the same 200 ns slot) still apply.
492
493To come in handy, it is possible to share time-triggered virtual links across
494more than 1 ingress port, via flow blocks. In this case, the restriction of
495firing at the same time does not apply because there is a single schedule in
496the system, that of the shared virtual link::
497
498  tc qdisc add dev swp2 ingress_block 1 clsact
499  tc qdisc add dev swp3 ingress_block 1 clsact
500  tc filter add block 1 flower skip_sw dst_mac 42:be:24:9b:76:20 \
501          action gate index 2 \
502          base-time 0 \
503          sched-entry OPEN 50000000 -1 -1 \
504          sched-entry CLOSE 50000000 -1 -1 \
505          action trap
506
507Hardware statistics for each flow are also available ("pkts" counts the number
508of dropped frames, which is a sum of frames dropped due to timing violations,
509lack of destination ports and MTU enforcement checks). Byte-level counters are
510not available.
511
512Device Tree bindings and board design
513=====================================
514
515This section references ``Documentation/devicetree/bindings/net/dsa/sja1105.txt``
516and aims to showcase some potential switch caveats.
517
518RMII PHY role and out-of-band signaling
519---------------------------------------
520
521In the RMII spec, the 50 MHz clock signals are either driven by the MAC or by
522an external oscillator (but not by the PHY).
523But the spec is rather loose and devices go outside it in several ways.
524Some PHYs go against the spec and may provide an output pin where they source
525the 50 MHz clock themselves, in an attempt to be helpful.
526On the other hand, the SJA1105 is only binary configurable - when in the RMII
527MAC role it will also attempt to drive the clock signal. To prevent this from
528happening it must be put in RMII PHY role.
529But doing so has some unintended consequences.
530In the RMII spec, the PHY can transmit extra out-of-band signals via RXD[1:0].
531These are practically some extra code words (/J/ and /K/) sent prior to the
532preamble of each frame. The MAC does not have this out-of-band signaling
533mechanism defined by the RMII spec.
534So when the SJA1105 port is put in PHY role to avoid having 2 drivers on the
535clock signal, inevitably an RMII PHY-to-PHY connection is created. The SJA1105
536emulates a PHY interface fully and generates the /J/ and /K/ symbols prior to
537frame preambles, which the real PHY is not expected to understand. So the PHY
538simply encodes the extra symbols received from the SJA1105-as-PHY onto the
539100Base-Tx wire.
540On the other side of the wire, some link partners might discard these extra
541symbols, while others might choke on them and discard the entire Ethernet
542frames that follow along. This looks like packet loss with some link partners
543but not with others.
544The take-away is that in RMII mode, the SJA1105 must be let to drive the
545reference clock if connected to a PHY.
546
547RGMII fixed-link and internal delays
548------------------------------------
549
550As mentioned in the bindings document, the second generation of devices has
551tunable delay lines as part of the MAC, which can be used to establish the
552correct RGMII timing budget.
553When powered up, these can shift the Rx and Tx clocks with a phase difference
554between 73.8 and 101.7 degrees.
555The catch is that the delay lines need to lock onto a clock signal with a
556stable frequency. This means that there must be at least 2 microseconds of
557silence between the clock at the old vs at the new frequency. Otherwise the
558lock is lost and the delay lines must be reset (powered down and back up).
559In RGMII the clock frequency changes with link speed (125 MHz at 1000 Mbps, 25
560MHz at 100 Mbps and 2.5 MHz at 10 Mbps), and link speed might change during the
561AN process.
562In the situation where the switch port is connected through an RGMII fixed-link
563to a link partner whose link state life cycle is outside the control of Linux
564(such as a different SoC), then the delay lines would remain unlocked (and
565inactive) until there is manual intervention (ifdown/ifup on the switch port).
566The take-away is that in RGMII mode, the switch's internal delays are only
567reliable if the link partner never changes link speeds, or if it does, it does
568so in a way that is coordinated with the switch port (practically, both ends of
569the fixed-link are under control of the same Linux system).
570As to why would a fixed-link interface ever change link speeds: there are
571Ethernet controllers out there which come out of reset in 100 Mbps mode, and
572their driver inevitably needs to change the speed and clock frequency if it's
573required to work at gigabit.
574
575MDIO bus and PHY management
576---------------------------
577
578The SJA1105 does not have an MDIO bus and does not perform in-band AN either.
579Therefore there is no link state notification coming from the switch device.
580A board would need to hook up the PHYs connected to the switch to any other
581MDIO bus available to Linux within the system (e.g. to the DSA master's MDIO
582bus). Link state management then works by the driver manually keeping in sync
583(over SPI commands) the MAC link speed with the settings negotiated by the PHY.
584
585By comparison, the SJA1110 supports an MDIO slave access point over which its
586internal 100base-T1 PHYs can be accessed from the host. This is, however, not
587used by the driver, instead the internal 100base-T1 and 100base-TX PHYs are
588accessed through SPI commands, modeled in Linux as virtual MDIO buses.
589
590The microcontroller attached to the SJA1110 port 0 also has an MDIO controller
591operating in master mode, however the driver does not support this either,
592since the microcontroller gets disabled when the Linux driver operates.
593Discrete PHYs connected to the switch ports should have their MDIO interface
594attached to an MDIO controller from the host system and not to the switch,
595similar to SJA1105.
596
597Port compatibility matrix
598-------------------------
599
600The SJA1105 port compatibility matrix is:
601
602===== ============== ============== ==============
603Port   SJA1105E/T     SJA1105P/Q     SJA1105R/S
604===== ============== ============== ==============
6050      xMII           xMII           xMII
6061      xMII           xMII           xMII
6072      xMII           xMII           xMII
6083      xMII           xMII           xMII
6094      xMII           xMII           SGMII
610===== ============== ============== ==============
611
612
613The SJA1110 port compatibility matrix is:
614
615===== ============== ============== ============== ==============
616Port   SJA1110A       SJA1110B       SJA1110C       SJA1110D
617===== ============== ============== ============== ==============
6180      RevMII (uC)    RevMII (uC)    RevMII (uC)    RevMII (uC)
6191      100base-TX     100base-TX     100base-TX
620       or SGMII                                     SGMII
6212      xMII           xMII           xMII           xMII
622       or SGMII                                     or SGMII
6233      xMII           xMII           xMII
624       or SGMII       or SGMII                      SGMII
625       or 2500base-X  or 2500base-X                 or 2500base-X
6264      SGMII          SGMII          SGMII          SGMII
627       or 2500base-X  or 2500base-X  or 2500base-X  or 2500base-X
6285      100base-T1     100base-T1     100base-T1     100base-T1
6296      100base-T1     100base-T1     100base-T1     100base-T1
6307      100base-T1     100base-T1     100base-T1     100base-T1
6318      100base-T1     100base-T1     n/a            n/a
6329      100base-T1     100base-T1     n/a            n/a
63310     100base-T1     n/a            n/a            n/a
634===== ============== ============== ============== ==============
635