Skip to content

J873g: USB4 on the rear USB-C ports through tipd and atc.c (T8152 routers, PCIe-C, M6 Mac mini) - #2178

Open
sithglan wants to merge 148 commits into
aurora-silicon:integration/aurorafrom
sithglan:thomas/j873-usb4
Open

sithglan wants to merge 148 commits into
aurora-silicon:integration/aurorafrom
sithglan:thomas/j873-usb4

Conversation

@sithglan

@sithglan sithglan commented Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

USB4 on the three rear USB-C ports of the M6 Mac mini (J873g, T8152), through the Type-C stack. This is the last step of the order agreed above. tipd owns the PD controllers, atc.c runs the ports' ATC PHYs in CIO mode, and the routers start from the Type-C state, as on the J813 (#2142). There is no private port monitor any more.

Stacking. Stacked on #2486, which is on #1800, #1743 and #1797. It also carries these commits unchanged:

Review the top 16 commits.

Carried from the earlier version, unchanged

  • Mailbox: each router's mailbox is "apple,t8152-cio-mailbox", "apple,t8152-smc-mailbox", with the two interrupts of its ADT acio-cpuN node.
  • CIO reset: T8152 (request 0x68, acknowledge 0x6c).
  • PCIe-C: T8152, with the MSI slice taken from the DT.
  • Routers: thunderbolt/apple T8152 support. The CIO power domains go through pmdomain: apple: T8152 PSD services as power domains #2459 and return to cold when unused.

Reworked

New

  • phy: apple: atc: run the T8152 Type-C ports in CIO mode
    • A T8152 PHY without apple,fixed-usb3 is a CIO-only Type-C port. It uses J873g: USB3 on the front ports (T8152 in the ATC PHY driver, M6 Mac mini) #1799's T8152 layout: power/reset block at 0x60000, sync reset at 0x60030, crossbar at 0x4c.
    • Both lanes are powered up. A reversed plug also sets the lane-swap bit.
    • Shutdown clamps AUX and drops the LFPS receivers, then releases the four power domains, so the next cable starts from a cold common block.
    • Thunderbolt partners are refused for now. Only USB4 has run on these routers.
  • DisplayPort and USB4 on one PHY
    • drm/apple's interim DP code powers the port's PHY through its DP PHY (phy_power_on/phy_power_off) instead of writing the PMGR/PSD registers itself, so genpd is the only power path.
    • While DP holds the PHY, a USB4 mux set returns -ETIMEDOUT, which tipd retries. While CIO runs, the DP power-on returns -EBUSY.
    • Both sides take the PHY only when it reads idle: 0x60000 = 0x4, sleep status 0 and sync reset 0.
    • This arbitration goes away once DP moves into atc.c.
  • thunderbolt: apple: stop the T8152 routers before a reboot: a running router is stopped before PCIe and DART shut down. The earlier port monitor showed this is needed.
  • DT:
    • the rear ports' ATC PHY nodes, with the ATC0/1/3 power states and PSD 55/56/58;
    • connector port@1 to the PHY's mux and switch, and port@4 to the router;
    • the router subtrees, identical to the hardware-tested ones apart from the graph links;
    • each router's NHI DART takes the router's first PSD service (59, 60, 62), for the DART power-domain lint. The router populates the DART only once it is powered and removes it before it powers down.
  • apple-t8152-port.c and its config line are removed.

Testing (M6, this tree)

The USB4 host-to-host and display rows, from one boot of this tree:

  • USB4: the back-middle port's partner is the upstream debug host (a Linux machine with USB4).
    • tipd entered USB4, atc.c brought the PHY up in CIO mode, and router ACIO0 started.
    • The router's firmware came up on its mailbox interrupts: 401178080.mailbox-recv counted 25 and -send 0.
    • The link trained (USB4, Gen 2, RS-FEC, both lanes in CL0), an Apple retimer was found, and the host-to-host link came up at 20 Gb/s on 2 lanes: "new host found … Linux mini".
  • Displays: all three come up, with the USB-C displays now powered through atc.c. 6 of 6 DPMS off/on cycles pass.
  • Other ports: the front hub enumerates at 10000M, and the two unused routers stay off.
  • Reboot: a reboot with the USB4 link up completes, through the new notifier.

Checks:

USB4 NVMe enclosure (Samsung 980 PRO in a Ugreen enclosure, back-right, this head):

  • Rows run: plugged at boot, 3 software replugs, a physical replug with the plug flipped (both orientations seen), DP and then USB4 again on the same port in one boot, and a reboot with the enclosure attached.
  • Every row: USB4 at 20 Gb/s on 2 lanes, PCIe at 16 GT/s x4, matching read checksums, 2911–3406 MiB/s sequential and 125k–141k IOPS random reads. The only errors were two reads in flight when the enclosure was pulled during a test.
  • Details are in the comment below.

Not covered: system sleep (the J873g has none), DP tunnelling over USB4 and two enclosures at once.

cc @Acelogic @RyanTheTide @DjDeveloperr @LiamRay10

Assisted-by: LLM (Claude:Opus 5.5)

sithglan and others added 30 commits November 6, 2026 11:23
The T8152 (M6) Always-On Processor differs from the T8140 one: its setup
port is not reachable from the AP, so it has one mailbox, and its audio
service uses six serializer power domains and the jack, speaker and
speaker-sense ADMAC channels.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
Add the Mac mini M6 (J873) sound card.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The J873 (Mac mini M6) AOP runs the same firmware generation as T8140
(EPIC v4, iBoot-loaded, started by the AP), but the T8140 "setup port"
is not reachable from the AP: mailbox 1 of the AOP's mailbox bank and the
ADT aop-exclave-mailbox both bus-error. The firmware boots without it:
RTKit protocol 12, then its "audio" EPIC service. Add apple,t8152-aop
with the T8140 configuration minus the setup port, starting every
advertised endpoint.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The Mac mini (M6, J873) speaker runs a 250-BCLK TDM frame, which the
SN012776's ratio auto-detection rejects. Describe the property that fixes
the ratio instead.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The J873 (Mac mini M6) AOP runs the speaker frame at 250 BCLKs of 12 MHz.
The SN012776 cannot auto-detect that ratio and mutes with "fault: bad
TDM clock". With TAS2764_CLK_CFG (register 0x60) set to 0x30 and
TDM_CFG0 bit 6 set it accepts the frame and plays. Do that when the new
ti,bclk-ratio property names the ratio; only this one ratio is supported.

Source of the value: the register write was first seen in the macOS
SN012776 driver; it is a single register value, and its effect (no fault,
correct playback) was verified on J873 hardware. No table or code was
taken from macOS.

The field is initialised in probe, so non-OF matches keep auto-detection.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The SDOUT1 mux offers the ASP receive channels as a source, but no DAPM
route connected them, so selecting "ASP RX CH1" or "ASP RX CH2" left the
capture path without an input. The J873 and J813 AOP audio services use
it to loop the playback stream back.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
On T8152 (J873, Mac mini M6) the AOP firmware owns the audio fabric, as
on T8140. Its "audio" EPIC service offers cout/cin (the CS42L84 jack,
base-ns ADMAC TX5/RX5, MCA4) and spkr (the SN012776 speaker, TX1, MCA0).

Each service is an ASoC back-end DAI: startup attaches it, requests pw0
(which raises the audio fabric) and powers the MCA leaves, all of which
are required. pwrd starts the serializer clocks at prepare, before ASoC
unmutes the codecs: the jack codec locks its PLL to the bit clock, and
the speaker amplifier reports "bad TDM clock" whenever it runs without
one. The jack restarts its clocks on every prepare so that a stream
restarted after an underrun begins on a frame boundary; the mono speaker
(one slot per frame) keeps them running. The front-ends are dmaengine
PCMs on the base-ns channels: the jack in stereo, the speaker in mono
(its serializer runs one 32-bit slot; stereo frames play at half pitch).

The firmware's serializer configuration (property 701) is byte-identical
to the one macOS publishes, so the driver leaves it alone.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The J873 (Mac mini M6) uses the T8152 AOP audio driver as its CPU
component: the CS42L84 jack on "cout" (DSP_A | IB_NF, one-BCLK sync
pulse) and one SN012776 on "spkr" (I2S, NB_IF: data sampled on the rising
edge, as macOS programs the amp).

Each AOP front-end owns the DMA channel of one serializer, so the speaker
can only be fed from the secondary front-end; the new spk_secondary_only
sets the speaker mux accordingly, and spk_dai_fmt gives the speaker link
its own format.

There is no qualified speaker protection for J873 yet (no verified sense
path, thermal model or power budget), so the speaker fails closed: it is
only enabled with please_blow_up_my_speakers. Opted in, it runs as macOS
runs it: amp gain 15 (ADT speaker-config) and no volume lock (safe_vol 0),
since there is no sense stream for a protection daemon. The jack is not
affected.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The T8142 AOP keeps its crash buffer in its own slot of its DART stream
and asks for it on endpoint 1 with that IOVA. A client without
shmem_setup refused the request, so the firmware never got its crash
buffer acknowledged.

APPLE_RTKIT_FIRMWARE_OWNED_CRASHLOG takes such a buffer as the firmware's,
without a CPU mapping, and acknowledges the request unchanged; a later
crash is reported without a log. RtKit::new_with_firmware_crashlog()
selects it at runtime, so a driver can use it for one SoC only.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The T8142 AOP (MacBook Air J813) waits for the host on a setup port, as
the T8140 one does: the ADT's aop-exclave-mailbox at 0x3d067c000 (AOP
ASC + 0x4007c000, not behind the arm-io translation). Its EPMAP announces
other service endpoints than T8140's fixed set (0x4, 0x20, 0x23 and 0x32
on J813), so take the endpoints that must get buffers from the EPMAP when
the match data says so.

The firmware places its own crash log buffer in its firmware-mapped DART
slot; accept it as the firmware's, as for G17P.

After the setup boot all AFK endpoints start, and with the audio fabric
left to the firmware (see the J813 device tree) the firmware announces
its "audio" EPIC service.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
Behind an AOP the serializer feeding a capture channel is the firmware's,
and it stops in the middle of a frame whenever it likes. A capture that
stays open across such stops (the speaker V/I sense, which a protection
daemon reads for as long as it runs) then carries the partial frame
forward: every later frame is off by its words, and the samples of one
amplifier are read as another's. On the MacBook Air J813 one restart in
three slipped by one 16-bit word.

Let a client pass the frame size in bytes as a u32 in
dma_slave_config.peripheral_config. A paused cyclic capture on such a
channel resumes at the next frame boundary of its current period: the
reports completed before the pause are accounted for, the rings are
reset and the rest of the period is submitted from the aligned address,
then whole periods as before. Without a frame size nothing changes.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The J813 speaker stream (ADT iop-audio-controller/audio-speaker) differs
from J873's: four SN012776 in 32-bit playback slots 0-3, and their voltage
and current sense back on the same stream's RX side (base-ns RX0) as eight
16-bit slots. Make the speaker and sense shapes per-SoC match data:
J873 keeps one playback channel; T8142 gets four, and an 8-channel S16_LE
"Speaker Sense" front-end.

The sense capture is the speaker serializer's RX, which the firmware
stops anywhere in a frame. Before every `pwrd` of the speaker, restart a
running sense capture at a frame boundary (apple-admac realignment, the
partial frame reads as zeros), so that a protection daemon that keeps it
open attributes each amplifier's voltage and current to the right one.

The sense back-end also attaches the speaker service and requests `pw0 `
at startup: its front-end requests the ADMAC channel next, and the ADMAC
only answers while the firmware holds the audio fabric up. Opening the
sense before any other stream since the AOP started (speakersafetyd at
boot) otherwise faults the bus.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
MacBook Air 13" M5 (T8142): as on J873 the AOP firmware owns the audio
fabric and the CPU component is the T8152-style AOP audio driver. The
jack is a CS42L84 on "cout", the speakers four SN012776 (one woofer and
one tweeter per side) on "spkr", with their V/I sense on the same stream.
As on the other SN012776 laptops the speaker volume stays locked until a
protection daemon (speakersafetyd) holds it.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The MacBook Air J813's three microphones are the AOP's "hpai" service
(ADT iop-audio-controller/audio-hp-mic): the firmware runs the PDM part
and delivers 48 kHz float32 frames of three channels (the ADT's
audio-hp-mic-proxy description: 12 bytes per frame) on leap-s ADMAC RX0.
There is no codec, so it gets a card of its own with one capture PCM,
named like the other AOP microphone cards (AppleJ813HPAI, "MacBook Air
J813 High-Power Audio Interface", driver aop_audio) so that the existing
UCM and DSP configurations for those apply.

As on T8140: open attaches the service and requests pw0 (which raises
the LEAP fabric), powers the LEAP leaves (leap-s ADMAC, RX0, LEAP MCA),
and only then requests the channel, whose registers fault without them.
The trigger starts the DMA and queues pwrd behind it; started earlier,
the LEAP produces into a channel nobody services. Close releases the
channel, then pw0 and the leaves.

The card exists only where the DT gives the LEAP power domains (all
three); J873 is unchanged. The dmaengine substream setup moves into a
helper shared with the ASoC front-ends.

Tested on J813: a tone from the speakers reads -44 to -49 dBFS on all
three channels, 70 dB over the quiet floor, exactly while it plays;
PipeWire picks up the card with asahi-audio's J613 beamformer graph.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
integration/aurora added the nCal boot argument and word-wise boot-argument
access to AopHwConfig. Both machine configurations keep the firmware's own
boot arguments unchanged, so set ncal to zero and leave word access off.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
Every boot on the J616s logged "spi-nor spi0.0: supply vcc not found,
using dummy regulator". The M1, M2, M3, J713 and J700 trees describe the
flash's always-on PMU rail as a fixed regulator; the M4 Pro/Max tree
came later and has none. Add the same regulator and point the flash at it.

Signed-off-by: Ryan Murray <81541646+RyanTheTide@users.noreply.github.com>
Assisted-by: LLM
Preserve current RTKit crashlog and GPU handling while retaining the expanded macaudio bindings and adding J813/J873 compatibles. Original audio commits and authorship are preserved.
…ation

Test whether COUT standby supplies the missing cold-boot fabric setup. A muted zero-data jack stream permitted HPAI capture and subsequent capture-only opens on J813; standby without any playback is the next hardware qualification step.

Preserve Thomas Glanzmann’s driver and adapt the T8140 DMA sync-stop pattern from Ryan Murray, also flushing queued HPAI work before DMA teardown. No LPAI buffer binding or codec unmute is performed.
…queue

Describe the T8142 microphone DMA and LEAP power domains and card names in the shared audio binding. Queue sleeping firmware power requests on system_dfl instead of the deprecated system workqueue, as directed by the runtime warning observed during hardware capture tests.
…-initialization-20261106

ASoC: apple: integrate J813 audio with first-open microphone initialization
…616-nor-vcc

arm64: dts: apple: t604x-j614-j616: Describe the boot NOR flash supply
Resolves sound/soc/apple/macaudio.c: aurora-wip added spk_dai_fmt and
spk_secondary_only (J813/J873) where integration added tweeter_hpf_granted
(speaker protection ownership); the struct keeps all three.

Signed-off-by: Ryan Murray <81541646+RyanTheTide@users.noreply.github.com>
Assisted-by: LLM
Signed-off-by: Ryan Murray <81541646+RyanTheTide@users.noreply.github.com>
Assisted-by: LLM
…nc-20261107

Merge integration/aurora into aurora-wip (11-07)
Brings aurora-wip up to integration/aurora 068dfce: pcie-apple endpoint
reset retry (aurora-silicon#1069) and the
tools/aurora-sep PKGBUILD checksum fix
(aurora-silicon#1068).

Signed-off-by: Ryan Murray <81541646+RyanTheTide@users.noreply.github.com>
Assisted-by: LLM
…n/wip-sync-20261107b

aurora-wip <- integration/aurora 068dfce (aurora-silicon#1068, aurora-silicon#1069). Clean merge, W=1 pcie-apple.o 0 warnings on the merged tree, CI host-tests/rust-build/source pass. Merge train (MTR).
Brings aurora-wip up to the tested release branch: CTXR execmem
(aurora-silicon#985, aurora-silicon#1079), KFIX speaker sleep and resume fixes (aurora-silicon#1077, aurora-silicon#1080),
JPEG v2 (aurora-silicon#1072), Jonathan's G17 and USB work (aurora-silicon#1073-aurora-silicon#1076), brcmfmac
flowring count (aurora-silicon#1081), ANE H16 startup restart (aurora-silicon#1066) and the
Touch ID per-prompt match (aurora-silicon#904).

Signed-off-by: Ryan Murray <81541646+RyanTheTide@users.noreply.github.com>
Assisted-by: LLM
Adds the 11-07 evening batch: t8140 disabled SPI3/FPWM0/SPMI nodes
(aurora-silicon#1034), J700 CPU zones (aurora-silicon#1043), J616C fans (aurora-silicon#1092) and the thunderbolt
DP PM workqueue fix (aurora-silicon#1098).

Signed-off-by: Ryan Murray <81541646+RyanTheTide@users.noreply.github.com>
Assisted-by: LLM
…n/wip-sync-20261107c

Gate on the exact head tree 6c684e8 (buildbox): fleet22-config Image+modules+dtbs build rc=0, 0 errors, 0 warnings, 5890 modules (= fleet); run-host-tests 193/193; apple-dcp host selftests 86/86 (fleet baseline 86/86). CI host-tests/source/rust-build green. Brings aurora-wip up to integration/aurora 810b92c.
Brings aurora-wip up to fleet25: DT descriptions (aurora-silicon#1029, aurora-silicon#1030, aurora-silicon#1112),
the J715 board (aurora-silicon#1129), Stockholm NFC (aurora-silicon#1042, aurora-silicon#1136), the AV1 tile check
(aurora-silicon#1100), the M4 Max ANE profile and AVD quirk (aurora-silicon#1088, aurora-silicon#1091), boot-log
fixes (aurora-silicon#1103, aurora-silicon#1104) and the Neo ANE memory floor (aurora-silicon#1137).

Conflicts: macaudio.c and apple,macaudio.yaml, where aurora-wip's J813
and J873 entries and integration's J715 entries were added at the same
spot. Keep all of them.

Signed-off-by: Ryan Murray <81541646+RyanTheTide@users.noreply.github.com>
Assisted-by: LLM
The USB-C PD controllers behind the J873g's (T8152) SPMI controller at
0x308908000 answer as the T8132 ones do: no register-selection
interrupt, so a selection is polled, and the payload length of the
selected register is read from register 0x1f. Their event, sleep and
wake sources are interrupts 11/17/19, 37/43/45 and 63/69/71 of that
controller.

Add apple,t8152-sn201202x with apple,t8132-sn201202x as fallback, so
that the T8132 interrupt layout (irq, sleep, wake) applies.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
… as Type-C mode switches

A T8152 DCP drives the rear USB-C port whose PD controller reports a
DisplayPort sink. So far the DCP read the controllers itself, through
apple,usb-pd. The controllers belong to the USB Type-C port driver
instead, which passes the DisplayPort alt mode state of a port to the
Type-C mode switches its connector's graph leads to.

Describe one display switch per port, shared by the DCPs: the port's
usb-c node in the shared ports node (the one apple,usb-c-ports refers
to) is the mode switch, with an endpoint at the display port (port@3)
of the port's usb-c-connector. The DCPs already share these nodes for
the port's PHY registers and tunables, and each port has one owner at
a time. One switch per DCP instead would fill the three mode switches
a connector's port driver looks at (fwnode_typec_mux_get() collects
the matches into a three-entry array and stops at the third endpoint
without an error), and leave out the port's ATC PHY, which needs one
for USB4. Drop apple,usb-pd, which nothing reads any more.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The T8152 DCPs found a USB-C display by reading the rear ports' PD
controllers over SPMI every 250 ms (t8152_pd.c): plug and orientation
from STATUS (0x1a), DisplayPort connection, pin assignment and HPD
from DATA_STATUS (0x5f), through the controllers' register window
(select at 0x00, length at 0x1f, data from 0x20). With tipd bound to
the same controllers, tipd failed to read its events while the
display path polled. The controllers belong to tipd now.

Take the state from the Type-C class instead: one Type-C mode switch
per rear port, the port's usb-c node in the shared ports node, with an
endpoint at the display port of the port's connector. tipd sets the
port's mode on it. The switch records whether the port is in
DisplayPort alt mode with HPD high, the lanes of the pin assignment
(two for B, D and F, which share the cable with USB 3) and the plug
orientation of the Type-C port, once, next to the port claims of
t8152_atc.c, and tells every DCP that drives USB-C to look at the
ports. The port claim and the PHY, crossbar and DPTX handling of
t8152_atc.c and iboot_drm.c stay as they were.

One switch per port, not one per DCP: a connector's port driver takes
three mode switches at most (TYPEC_MUX_MAX_DEVS;
fwnode_connection_find_matches() stops at the third endpoint and
returns success). A switch per DCP would fill all three on every J873g
rear connector, and the port's ATC PHY, which needs one for USB4,
would never be called.

A HPD pulse is not a new sink. Without a video stream, an EV2780 drops
HPD for about 160 ms every 4.7 s; the PD controller reports it after
the fact, and tipd passes it on as HPD low and then high. Connecting
the sink anew on each pulse kept its link from coming up (link faults,
one lane, then a standby link trained at RBR whose mode list lacked
the enabled mode, so the next DPMS on failed atomic check). The 250 ms
poll never saw these pulses. A switch event is looked at 100 ms later,
which takes such a low and high together, and only a changed plug
(orientation, lanes) or a sink that leaves DP alt mode with its
partner is a new connection.

dcp-fabric.c's route layer itself does not fit these DCPs: it switches
a DP PHY and the display crossbar through their drivers and connects
the sink through the DCP's IOMFB DPTX service, while the iBoot
interface DCPs program the ATC PHY and the crossbar themselves and
connect through their own DPTX port service. Their switches only
record the state, in t8152_typec.c.

The 250 ms work now only re-arms while a DCP can take the shared HDMI
port, whose HPD line has no interrupt.

The PD controller reset that made a sink which stopped answering AUX
enter DP alt mode again (a cold reset, "Gaid", then system power state
S0, "SSPS") is gone: the controller is tipd's, and tipd puts it into
S0 whenever it sets the controller up. What remains is the DisplayPort
side, at most once a minute: release the link and connect the port's
sink again. HPD changes now reach the display path as they happen.

The first DCP that probes registers the switches, ahead of everything
else in its probe, and they stay until the driver is unloaded: tipd
keeps the switch it found, and must not wait for a DCP that cannot
start. A switch that cannot be registered is disabled in the live tree,
so that tipd goes on without it.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
… tipd

The PD controllers of the three rear USB-C ports were described for the
display path alone, which read them over SPMI itself. Describe them
for tipd, which owns them from now on, and the display path takes their
DisplayPort alt mode state from the Type-C class:

- the controllers as apple,t8152-sn201202x, with the T8132 fallback:
  tipd's SPMI transport then polls the register selection and reads
  the payload length from register 0x1f, as these controllers need;
- the SPMI controller (AIC 682) as the interrupt controller of their
  event, sleep and wake sources: 11/17/19 (back-middle), 37/43/45
  (back-left) and 63/69/71 (back-right), rising edge, as in the ADT.
  In a test with the compatibles and interrupts alone, tipd bound all
  three, took its events by interrupt, and both display partners showed
  DisplayPort alt mode active;
- a usb-c-connector per controller, whose display port (port@3) leads
  to the port's usb-c node in the main DCP's shared ports node, the
  port's one display switch for all three DCPs. The USB port (port@0)
  stays unlinked: the USB data path is not changed here.

The DCP's usb-c nodes lose apple,usb-pd, which nothing reads any more.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
On T8152 the video decoder, the USB4 routers and their PHYs and the USB
controllers are powered through PMGR2 PSD services instead of PMGR
power-state registers. Each service has a 0x100-byte block with a target
state, an acknowledge bit and the current state.

Describe one service per node as a power domain, with an optional
apple,always-on for a service whose consumers cannot restore what
another agent set up in its blocks yet.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
Provide each T8152 PMGR2 PSD service as a power domain. Power-on
requests state 0 and power-off the cold state 6. The driver waits for the
acknowledge before it reads the state, because reading the state
consumes the acknowledge. A service marked apple,always-on is powered at
probe and kept on. Otherwise the domain starts in the state the
hardware reports, so genpd powers off a service that nothing uses.

The consumers include IRQ-safe devices, the DARTs, so the domains are
IRQ-safe and a transition is polled atomically, as the PMGR power states
are. On the J873g the AVD service (50) reaches state 0 in 390-500 us and
the cold state in about 240 us; the timeout is 10 ms.

The AVD and the USB4 routers' CIO services, which so far each had their
own variant of this, and the USB controllers (services 175 and 176) use
it.

It has its own Kconfig symbol, APPLE_PMGR_PSD, on by default with
ARCH_APPLE and built in by the shared asahi.config profile.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The two front USB-C ports of the Mac mini (2026, J873g) sit behind an
on-board USB3 hub. The hub's SuperSpeed lanes are wired to the Type-C
PHY of the SoC's ATC2 directly, without a Type-C mux or port controller
in between, so that PHY only ever runs one fixed USB3 configuration.

Describe the T8152 variant in that role, with a boolean
apple,fixed-usb3 that marks such a port: the PHY runs USB3 on lane 0
from probe on, ignores mux and orientation changes and provides only
the USB3 PHY of its controller. The controller's USB2 side and reset
stay with the bootloader, so the T8152 PHY has no #reset-cells and no
Type-C switch properties.

The PHY is described by its core and pipehandler banks. On T8152 the
core bank also holds the USB controller's DARTs, and the PHY's power
and reset block sits behind them. It needs four power domains, listed
in the order they are powered up: the ATC's USB_AON power state, its
common PSD service and its COMMON_DP_0 and COMMON_DP_1 power states.
The other variants keep their single power domain and #reset-cells.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The J873g front ports reach their on-board USB3 hub through ATC2's
Type-C PHY, wired to the hub without a Type-C mux. So far the hub ran at
USB2 only: the bootloader leaves the PHY powered down and the PIPE on
the controller's dummy PHY. Support the T8152 PHY in that role in the
shared ATC driver.

T8152 shares the T8142 generation's power and reset block, tunables and
PIPE handler, with a different layout: the USB controller's DARTs sit
at 0x20000 and 0x40000 of the ATC window, so the power and reset block
follows at 0x60000, its lane swap is bit 1 there, and the reset the
power-up releases first is a separate always-on sync reset at 0x60030.
The crossbar is at 0x4c and the USB3 mode's existing setting, USB3+DP
with single PMA 0x008, gives the value 0x110; there is no lane mode
register to program. Describe this in the match data, with zero for the
layout of the other SoCs, and leave the power-up and power-down, the
tunables, the crossbar and the PIPE switch to the existing code.

What remains T8152 code:
 - Lane 0 is powered up on its own after the common-a tunables (LFPS
   receivers, sleep, clamp, AHB reset) and down again with the PHY.
   Without it the lane registers stay in AHB reset, drop the lane
   tunables and the PHY never reports PIPE PhyStatus. The crossbar is
   set right after, before the remaining tunables, as in the T8152
   sequence.
 - While the PIPE moves to the PHY, the BIST overrides hold lane 0's
   PIPE reset and power-down, stepping from P2 to P3, and the
   non-selected PHY override is released under them.

The T8152 PIPE handler takes the lock request but does not acknowledge
it: with the controller out of reset, no acknowledge came within 150 ms
in either direction, while the PIPE moved to USB3 and back to the dummy
PHY under the request. So T8152 changes the PIPE under the request
without waiting for the acknowledge; otherwise every USB3 start fails.

A fixed-USB3 port, apple,fixed-usb3, is a host by construction. Probe
maps only the core and PIPE windows, the core without claiming it as
it holds the DARTs, powers the PHY's power domains in the order they
are listed and brings the PHY up in USB3. It registers only the USB3
PHY: no reset controller, Type-C mux or switch, as the controller's
reset and USB2 side stay with the bootloader. The PIPE is routed when
the controller powers the USB3 PHY on, after its core reset, which also
sets the PHY up again after a power-off. If that fails the host starts
with its PIPE parked, so the hub keeps USB2.

The T8152 register sequence is the one of the earlier standalone
driver, with which the hub enumerated at 10 Gb/s. The other SoCs keep
their register accesses: the new layout fields are zero for them, and
the new code runs for T8152 or a fixed-USB3 port only. KUnit cases
check the T8152 crossbar, lane and power block accesses.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The two front USB-C ports sit behind an on-board VL822 hub on usb-drd2.
The hub's USB2 side is on the controller's eUSB2 PHY, which the loader
sets up, and its USB3 side on ATC2's Type-C PHY, wired directly without
a Type-C mux. So far the controller was limited to high speed and the
hub ran at USB2 only.

Describe ATC2's PHY as an apple,t8152-atcphy with apple,fixed-usb3,
hand it to usb-drd2 as its USB3 PHY and drop the controller's high-speed
limit. Nothing changes for usb-drd1 (the back-left port).

The PHY's power domains are ATC2's USB_AON, COMMON_DP_0 and COMMON_DP_1
PMGR power states, in a new node for the ATCs' block of PMGR states at
0x308290000, and its common domain, PSD service 57 (ADT clock-gates
57). The T8152 USB3 bring-up runs the power states without automatic
power management, so they are marked apple,no-auto-pm.

The tunables are the groups of the ADT's atc-phy2 node (26A434) that
apply to USB3 on lane 0, in the order they are applied: the common and
PLL defaults as common-a, the fabric, PLL and USBC0 groups as common-b
and the lane 0 groups as lane0-usb. Some of them differ between the
four ATCs of the same machine, so they likely carry per-device
calibration. There is no atcphy2 alias, as the loader's ATC tunable
copy would replace them with values for another PHY layout.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
Allow the five ACIO register windows, LSTX GPIO and router firmware
tunables used by T8142. Describe its native PCIe-C compatible and
require the shared MSI vector base. Include the T8142 pinctrl fallback
chain used by the AOP LSTX pin.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
T8142 uses the Intr2AXI controller layout but its three-cell AIC
interrupt specifier and MSI window differ from T6050. Match it natively.
This allows a tunnelled NVMe controller to allocate working MSI-X
interrupts.

For T8132 and T8142 the slice is checked against the device tree alone.
msi-ranges names the AIC input of global vector apple,msi-vector-base in
a linear window, so the window's AIC base is msi-ranges minus
apple,msi-vector-base, and the slice is checked by its shape: an
edge-rising SPI, 32 vectors, 32-aligned, above the internal port's
vectors and inside the 8-bit map target. This replaces T8132's table of
ECAM apertures and its AIC base 1506 behind a machine check. The T8132
trees give 1602 - 0x60 = 1634 - 0x80 = 1506, and every J813 PCIe node
gives 1741.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
Each T8152 USB4 router's coprocessor has a 0x80-byte mailbox window with
the counted FIFOs of the SMC mailbox, at offset 0x80 of its register
block. Describe it with apple,t8152-smc-mailbox as the fallback, so the
SMC mailbox's match data serves it, and with its send-empty and
recv-not-empty interrupts.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The T8152 CIO reset block resets the rear USB4 routers of the J873g.
It has a request and an acknowledge register instead of T8142's control
register, so it gets its own compatible.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The T8152 CIO reset block requests a router reset in register 0x68 and
acknowledges it in 0x6c. Wait for the router's pair of busy bits to
clear, request, wait for the acknowledgment and clear it.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The DARTs of the J873g rear USB4 routers (NHI streams) and of their
PCIe-C tunnels are Gen3 DARTs like the display ones, but start unlocked
with inactive streams. Each has its own slot in the shared TLB
invalidation aperture, and its ADT tunables and DMA window.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The DARTs of the T8152 USB4 routers and their PCIe-C tunnels are Gen3
DARTs that, unlike the display ones, the bootloader leaves unlocked with
inactive streams. Admit cold Gen3 setup for the instances their
compatibles select, with version 0x300, their tunables and inactive
streams, and initialize them before enabling translation: clear the
streams, apply the tunables and invalidate through the Gen3 TLBI
aperture, which is now mapped before the reset. Domains use two-level
16K tables inside a single 64 GiB DMA window, as iBoot's display
streams do.

- ACIO DART (apple,t8152-acio-dart): 32 streams, NHI streams 0-11 and
  16-27.
- PCIe-C DART (apple,t8152-pciec-dart): translated streams 0-252, with
  253 and 255 in bypass for host transactions and MSI writes to the AIC
  (the ADT reserves them), firmware's stream 254 preserved, 16K pages,
  42-bit addresses; it must be tunnelled and have its power domain.
  Without the bypass on 255, NVMe completions were polled and the DART
  faulted at the AIC.

Each DART invalidates through the TLBI slot its "tlbi" resource names.
The slot has to be the port's own: with port 3's slot on port 1's DART,
reads returned wrong data with DART faults. The locked display DARTs
and all other Gen3 instances stay as they were.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The PCIe-C ports tunnelled through the J873g rear USB4 routers are laid
out like T8142's: Intr2AXI MSI slices, the T6050 tunnel controls, and no
port reset of their own.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The T8152 tunnelled PCIe-C ports use the T6050 tunnel controls with the
Intr2AXI MSI path, like T8142, and their MSI slices are checked the same
way: from msi-ranges and apple,msi-vector-base alone. The J873g's three
bridges use 0x60 (CIO0, ECAM 0x1c50000000), 0x80 (CIO1, 0x1c60000000)
and 0xc0 (CIO3, 0x1c80000000) of the window at AIC 1879.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
Add the T8152 ACIO and NHI. The T8152 router CPU is started by a
reconfiguration request through its mailbox, so it has no cpu window.
A second range maps the shared Gen3 TLB invalidation aperture for the
router's DART.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The T8152 router is a Type 7 MXWrap-v2p3 that a reconfiguration request
starts through its counted mailbox. Its DP IN path is not qualified.
Release the CIO domains in reverse order and suspend each one
synchronously, because deleting the device link only queues an idle
request: the PHY reset and the next router start need them cold.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The J873g's three rear USB-C ports have T8152 ATC PHYs behind a Type-C
mux, with a USB4 router behind each. Linux runs them in CIO (USB4)
mode only for now: their DisplayPort alt mode is still driven by the
display controllers, and their USB2 side by the bootloader.

Allow a T8152 PHY without apple,fixed-usb3 for such a port: its core
bank alone (the power and reset block at 0x60000 is part of it, and
the display controllers map the same bank), the orientation and mode
switches with port@0 to the connector's SuperSpeed port, the ATC's
four power domains in power-up order, and the common and lane CIO
tunables. The display controller takes the PHY through its
DisplayPort PHY.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The J873g's three rear USB-C ports have T8152 ATC PHYs with a USB4
router behind each. Their DisplayPort alt mode is still programmed by
the display driver (drm/apple's T8152 code), and their USB2 side is the
bootloader's. Run such a PHY in CIO mode for a USB4 partner and leave it
to the display driver otherwise, as a Type-C port of its own kind
(cio_only, selected by the T8152 match data for a PHY without
apple,fixed-usb3). The Type-C port driver sets the mux before it starts
the router and after it stops it, as on T8142.

CIO mode uses the T8152 layout of the fixed-USB3 support: the power
and reset block at 0x60000 with the sync reset at 0x60030, the crossbar
at 0x4c (USB4 0, USB4 swapped 1), no lane mode register. It differs
from USB3 in that
both lanes are powered up, one after the other, between the common-a
tunables and the others (LFPS receivers 0x1e24 bits 12/16 and 14/18,
then the lane controls at 0x1e18 and 0x1e1c out of sleep, unclamped and
out of AHB reset), and in that a reversed plug sets the lane swap (bit 1
of 0x60008) next to the swapped crossbar, where T8142 leaves it clear.
On the way down AUX is clamped (0x1e24 bit 10, sleeps 6 and 8 cleared)
and both lanes' LFPS receivers are dropped before the lanes, as the
T8152 CIO shutdown does: these always-on controls keep their state
while the common block is off. The lanes are only touched while the
register bus (0x60000 bit 3) is out of reset. This is the sequence the
earlier port monitor ran, with which a rear port trained USB4 at
2 x 20 Gb/s and carried an NVMe enclosure at 16 GT/s x4.

The PHY's four power domains (USB_AON, the common PSD service, DP_0,
DP_1) are attached without device links. They are held while the PHY
runs CIO mode, and dropped synchronously after the CIO shutdown, so
that the common block goes cold before the next partner: a training
from a powered common block asserted in the router's firmware.

The PHY and the display driver take turns, explicitly:

- The display driver powers the PHY's DP PHY on before it touches the
  PHY and off when it has let go of it. The DP PHY holds the domains for
  it and keeps CIO mode out: a USB4 mode change fails with -ETIMEDOUT,
  and the mux declares its timeouts transient, so the port driver
  retries until the display is gone. A DP PHY power-on in CIO mode
  fails with -EBUSY.
- Either takes the PHY only when its power and reset block is as its
  power-up leaves it (0x60000 = 0x4, no sleep acknowledged, sync reset
  0); CIO mode is refused otherwise.
- A display the bootloader started has all four domains on and a busy
  block at probe. Its domains are then held for the display driver,
  which takes the PHY over, until its DP PHY goes off.

Once DisplayPort moves into this driver, the DP PHY programs it itself,
both modes switch under the PHY's lock and this arbitration goes away.

Thunderbolt alt mode is refused on these ports for now: the T8152
routers have only been run with USB4 partners. The other SoCs are
unchanged: the new paths are taken for a cio_only PHY or the T8152
generation only, and T8152 now sets has_usb4, which a fixed-USB3 port's
tunable list does not consult. KUnit cases check the CIO lanes,
crossbar and swap in both orientations, the power-down, the idle test
and which Type-C states ask for CIO. The DCP host-test fixtures that
compile the power-up and power-down get the new field.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
A rear USB-C port's ATC PHY is also its USB4 PHY, run by the ATC PHY
driver, which powers the PHY. Give each usb-c port node the port's PHY
in DisplayPort use (PHY_TYPE_DP): a DCP powers it while it drives the
PHY, from before a cold start until the PHY is powered down, and the
PHY driver keeps the PHY powered and out of USB4 meanwhile.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The T8152 DCPs powered a rear USB-C port's ATC PHY themselves: before
a cold start they set the port's USB_AON, DP_0 and DP_1 PMGR power
states (PS +0x100, +0, +0x20) to 0xf and its common PSD service to 0,
and after a power-down they set back those that had been off. The same
PHY now runs CIO mode for USB4, through the ATC PHY driver, which
powers it through the PMGR and PSD power domain drivers. With two
writers of the same states, one would switch off what the other uses,
or leave the power domain drivers believing a domain on that is off.

Take the PHY through the port's DP PHY instead (the usb-c node's phys):
power it on before the cold start's first PHY access and off once the
PHY is powered down. The PHY driver then holds the PHY's power domains
and keeps it out of USB4; while the PHY runs USB4 the power-on fails
with -EBUSY and the cold start with it. A display the bootloader left
running is kept powered by the PHY driver from its probe; the DCP that
adopts it powers the DP PHY on at once and so takes that over. A port
whose common block runs but whose DP PHY cannot be powered is in USB4,
not a bootloader display: the power and reset block (CFG +0x60000)
reads 0x1b in both. The PMGR and PSD state registers are still read to
tell whether the PHY can be looked at.

The DisplayPort programming itself is unchanged.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
The three rear USB-C ports' ATC PHYs (ADT atc-phy0, 1 and 3) were only
known to the display controllers, as register windows they program and
power themselves. Describe them as T8152 Type-C PHYs, which the ATC PHY
driver runs in CIO mode for USB4:

- their core windows (0x403000000, 0x40b000000, 0x413000000, 0x84000
  each, also mapped by the display controllers);
- the orientation and mode switches, linked to the SuperSpeed port
  (port@1) of the port's usb-c-connector;
- their power domains, in power-up order: the ATC's USB_AON, its common
  PSD service (55, 56, 58) and its COMMON_DP_0/1 power states, new
  nodes next to ATC2's. The USB_AON states are kept on, as ATC2's: they
  hold the rear USB controllers' state the loader set up;
- the CIO tunables of the ADT's atc-phyN groups, at the PHY blocks they
  apply to: the common defaults before the lanes are powered up, the
  fabric, PLL and remaining common groups after them, then each lane's
  groups. They are the lists the earlier port monitor applied, with
  which a rear port trained USB4.

The display controllers' usb-c port nodes take the PHYs through their
DP PHY, which powers them while a controller drives the port.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
A reboot or power-off shuts the PCIe-C host and the router's and the
tunnel's DARTs down before the router, while a started router still
runs over them. With the J873g's rear ports, the earlier port monitor
stopped the routers from a reboot notifier for this. Without a port
monitor, do that in the router driver: for routers whose match data
asks for it (T8152), stop a started router from a reboot notifier, as
the port driver does for an unplugged cable, and refuse a new start
from then on.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
Describe the three rear USB4 routers (ADT acio0, acio1, acio3) with
their mailboxes, NHI, ACIO DARTs, PCIe-C tunnels and their DARTs, the
CIO reset block, the PSD services they need and the LSTX pins.
Generated from the J873g ADT, including the router, NHI, DART and
PCIe-C tunables and the router DROMs, whose UID field is the port
index.

Each router coprocessor's mailbox has apple,t8152-smc-mailbox as its
fallback and the two interrupts of the ADT's acio-cpuN node, which list
send-empty first, as the SMC mailbox's do.

Each router's Thunderbolt switch (its port@1) is linked to the USB4
port (port@4) of its port's usb-c-connector, as on the J813: the PD
controller driver sets the port's ATC PHY to CIO mode and then starts
the router with the cable's EUDO and orientation, and stops the router
before it takes the PHY out of CIO mode.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
Every enabled DART must name the power state of the engine it serves
(scripts/dtc/apple-dart-pd-check.py). The three rear routers' NHI DARTs
named none.

A DART sits inside its router, so it takes the router's first PSD
service (59, 60 and 62: ADT acioN power-gates), the one the router
powers up first and down last. The router populates the DART only once
it is powered and its firmware is ready, and removes it before it
powers down, so the DART's vote never holds the router on after a stop.

Signed-off-by: Thomas Glanzmann <thomas@glanzmann.de>
@sithglan

Copy link
Copy Markdown
Contributor Author

Right: the back-right router (cio_psd62) hasn't been exercised at this head. I'll run one enclosure replug on that port as soon as the Ugreen enclosure is back on the M6. Right now it holds a backup and isn't attached, and a monitor is on the back-right port. Restacked to 5f89281 for #1797's usb -> usb_aon parents, with no other change.

Assisted-by: LLM (Claude:Opus 5.5)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants