Skip to content

phy: apple: atc: Serve USB2 for USB4/TBT on T8132 without CIO calibration - #1702

Open
techmore wants to merge 2 commits into
aurora-silicon:integration/aurorafrom
techmore:upstream/t8132-atcphy-usb2-fallback-20261111-v2
Open

techmore wants to merge 2 commits into
aurora-silicon:integration/aurorafrom
techmore:upstream/t8132-atcphy-usb2-fallback-20261111-v2

Conversation

@techmore

@techmore techmore commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

On a T8132 port without the CIO calibration, serve the partner's USB2 when the port controller enters USB4 or Thunderbolt 3, instead of refusing the mux state. The J773g (Mac mini M4) is such a machine: its device tree has no apple,tunable-cio-* properties.

What the hardware runs below show is that the mux update now succeeds. In USB4, the port's xHCI then starts. In Thunderbolt 3 it also needs the companion tipd patch. They do not show a USB2 device enumerating during a USB4 or Thunderbolt link: the one dock I have showed nothing on D+/D- while in USB4.

Why. The J773g's SN201202x port controllers enter USB4 by themselves with a USB4 dock or another Mac, and Thunderbolt 3 alt mode with a Thunderbolt 3 dock. atcphy_typec_target_mode() maps those states to the USB4 and TBT PHY modes, and atcphy_mux_check() refuses both with -ENODEV when the CIO calibration is missing. The failed mux update makes cd321x_update_work() retry before it sets the USB role, so the port ends with failed to configure Type-C connection: -19 and cable revalidation failed after 3 attempts, and its dwc3 never starts an xHCI. USB 2.0 on D+/D- does not depend on the USB4 or Thunderbolt link.

The change. In atcphy_typec_target_mode(), on T8132 without CIO calibration, map USB4 and TBT to USB2. This works the same way as the existing fallback for DisplayPort pin assignments that the PHY cannot carry. The new fallback has to run before the DisplayPort fallback picks its usb_mode. That fallback compares its result with target_mode, so if it started from USB4 it would put USB4 back. Ports with the CIO calibration (the J713/J715 device trees carry it) are unchanged. A KUnit case covers both cases.

Tested on one J773g, natively, with kernels based on #1192 plus local J773g test commits, not on integration/aurora. In those kernels, the atcphy_typec_target_mode() change is behaviour-identical to this patch. They also had a TEST-ONLY log line for each mux request, which is dropped here. The patch on this base has not been booted.

  • Run 53, without the change: a USB4 dock on Back Right and another Mac (USB4) on Back Left. On both ports the mux update failed with -19 three times, then cable revalidation failed after 3 attempts, and no rear xHCI came up.
  • Run 54: a first version ran the new fallback after usb_mode was taken. The DisplayPort fallback put USB4 back and the ports failed with -19 as before. That is why the fallback now runs first.
  • Run 55, this logic: the USB4 dock on Back Right got USB4/Thunderbolt is not available on this port: serving its USB2 only, the mux update returned 0 with the PHY in USB2, and the rear xHCI (41a280000.usb) registered. The dock enumerated nothing on that xHCI while it was in USB4. The Mac on Back Left (we are the UFP there, in USB4) took the same fallback and the mux update returned 0. A keyboard plugged into Back Right later enumerated on the rear xHCI.
  • Run 56, same PHY code: the dock entered Thunderbolt 3 alt mode (SVID 0x8087), and the PHY served USB2 (state mode 2 alt 0x8087 -> PHY mode 1: 0). No rear xHCI came up because the port controller driver sets no USB role in TBT3. That is fixed separately in "usb: typec: tipd: Set a USB role for TBT3 on T8132 ports without a router". The Mac on Back Left: USB4 to USB2 again, mux 0.
  • Run 57, same PHY code: the Mac on Back Left: USB4 to USB2, mux 0. The dock entered DP alt mode this time, so the fallback did not apply to it.
  • The front USB keyboard enumerated in every run.

Runs 56 and 57 also had two TEST-ONLY sn201202x module parameters on the command line, which are not part of this PR. They change which mode the port controller picks for the dock (no USB4 host capability in the DFP VDO, Thunderbolt mode disabled). They do not touch the PHY.

Checks on this branch: scripts/checkpatch.pl --strict reported only the missing Signed-off-by, which the commit now carries. arm64 W=1 build of drivers/phy/apple/ with GCC 15.2.0 / binutils 2.46, also with CONFIG_KUNIT=y and CONFIG_PHY_APPLE_ATC_KUNIT_TEST=y: no new warnings. The existing -Wparentheses in atc-j700-dp.h and the kernel-doc warnings on struct apple_atcphy are unrelated. I ran the phy-apple-atc-mux KUnit suite under UML (x86_64, CONFIG_COMPILE_TEST): the new atcphy_test_target_mode_t8132_cio and the other atcphy_test_target_mode* cases pass. Ten other cases fail the same way on the base, because UML's logic_iomem does not back their test registers. I did not run KUnit on arm64.

Not tested: this exact branch on hardware. Other T8132 boards. A TBT3 or USB4 dock that drives USB2 on D+/D- during the link (the one I have did not). J713/J715, where the path is unchanged.

No dependency: this applies to integration/aurora as is. The companion tipd PR, "usb: typec: tipd: Set a USB role for TBT3 on T8132 ports without a router", needs this one to have any effect on the J773g.

Written with help from an LLM; I reviewed it and ran the hardware tests above.

Hardware, 2026-11-19 (J773g, 27.2 stub; this PR's two commits on our display+Ethernet test tree, not this PR's own head; loader with the T8132 USB3 calibration copied, aurora-silicon/aura#113 at 53be8212; rear port, Thunderbolt 4 cable to a MacBook Pro; two boots, the second with one more dwc3-apple.c commit that did not take effect):

  • phy-apple-atc 40aa90000.phy: USB4/Thunderbolt is not available on this port: serving its USB2 only
  • the port takes the device role, 40a280000.usb mode=device, a gadget binds and stays configured at high speed, and the MacBook lists it at 480 Mb/s
  • no failed to configure Type-C connection, no cable revalidation failed

The same cable on the same tree without these two commits: sn201202x 1-0a: failed to configure Type-C connection: -19 three times, then cable revalidation failed after 3 attempts, and no USB device.

Not run: a Thunderbolt dock, host role against a USB4/TBT partner, replug.

@RyanTheTide

Copy link
Copy Markdown
Contributor

@techmore the change reads correctly and the KUnit case covers both paths. Two things before merge: the commit needs your Signed-off-by, and #1678 edits the line right after your hunk in atcphy_typec_target_mode(), so one of the two will need a rebase. Optional: the "serving its USB2 only" message prints on every USB4/TBT attach of such a port; dev_info would be enough.

@techmore
techmore force-pushed the upstream/t8132-atcphy-usb2-fallback-20261111-v2 branch from 4f9b5df to a4683bd Compare November 13, 2026 13:39
@techmore

Copy link
Copy Markdown
Contributor Author

Thanks, Ryan. I rebased this onto current integration/aurora after #1678 landed. There was one conflict, in atcphy_typec_target_mode(): the CIO fallback is kept first and #1678's DP condition remains directly below it. I also changed the "serving its USB2 only" message to dev_info as suggested, in the same commit.

Testing caveat: for the new head I have only compiled the changed file (atc.o). I have not completed a full J773g kernel build, and the rebased head has not been boot-tested on hardware.

…tion

The SN201202x port controllers of the Mac mini (J773g) enter USB4 by
themselves with a USB4 dock or another Mac, and Thunderbolt 3 alt mode
with a Thunderbolt 3 dock. atcphy_typec_target_mode() maps those states
to the USB4 and TBT PHY modes, which atcphy_mux_check() refuses with
-ENODEV on a T8132 PHY without the CIO calibration. The J773g device tree
has none. The failed mux update makes cd321x_update_work() retry before
it sets the USB role, so the port ends up with

  sn201202x 1-08: failed to configure Type-C connection: -19
  sn201202x 1-08: cable revalidation failed after 3 attempts

and its dwc3 controller never starts an xHCI.

USB 2.0 on D+/D- does not depend on the USB4 or Thunderbolt link. As is
already done for the DisplayPort pin assignments this PHY cannot carry,
serve the partner's USB2 instead of refusing the mode.

The CIO fallback has to come before the DisplayPort fallback picks its
USB mode: that one compares its result with target_mode, and started
from the USB4 mode it would put USB4 back.

Ports with the CIO calibration are unchanged. A KUnit case covers both.

Assisted-by: LLM checkpatch
Signed-off-by: Sean Dolbec <sdolbec1@gmail.com>
The ATC USB3 host test and the apple-dcp phy, pm, quiesce and dwc3
host tests compile functions taken from drivers/phy/apple/atc.c.
Define dev_info() as a no-op beside their existing logging shims so
the generated host C keeps compiling when one of those functions
prints an informational message.

Assisted-by: LLM
Signed-off-by: Sean Dolbec <sdolbec1@gmail.com>
@techmore

Copy link
Copy Markdown
Contributor Author

Hardware test, Mac mini M4 (J773g), 2026-11-14: PASS at this PR's head 0430675d5a04.

Kernel: aurora-silicon/linux integration/aurora 5d112b1d5493 plus the exact heads of #1192 36383187, #1718 5fbbff26, #1191 (first five commits) 3fa83613, #1865 9c71c29f, #1702 0430675d, #1703 9b0ca002 and #1872 2eb84498, merged without conflicts (release 7.3.0-prheads+, native arm64 build with GCC 15.2.0, 0 warnings). One boot on a Mac mini M4 (J773g, T8132) with a Fedora 44 userspace. The device tree came from the boot chain, not from this tree, so DT-only changes were not exercised.

This result comes from a second boot (2026-11-14 03:09 UTC) of that same kernel image (Image SHA-256 9f4d73f7de27).

Device: Apple Magic Keyboard with Touch ID and Numeric Keypad (05ac:029f, USB2 full speed) on a rear USB-C port, plugged in before boot (cold-plug).

Check Result Evidence
Rear-port device enumerates PASS usb 3-1: new full-speed USB device number 2 using xhci-hcd; usb 3-1: New USB device found, idVendor=05ac, idProduct=029f, bcdDevice= 4.20
On the rear controller and port PASS 3-1 05ac:029f speed=12 under 41a280000.usb/xhci-hcd.2.auto/usb3; typec port1: data_role: [host] device, partner present
Topology PASS Bus 003 root hub 480M → keyboard Driver=usbhid, 12M
Keyboard input PASS input: Apple Inc. Magic Keyboard with Touch ID and Numeric Keypad Keyboard as /devices/platform/soc/41a280000.usb/…/input1
USB2 path without calibration observed all three ATC PHYs (402a90000, 40aa90000, 41aa90000): T8132 calibration incomplete, USB2 only
USB/PHY errors PASS 0 error, fail or timeout lines from ATC, dwc3, xHCI, typec, tipd or USB

Not covered: hot-plug, USB 3 SuperSpeed and USB4/Thunderbolt docks on the rear ports.

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.

2 participants