[RyanTheTide]

J704 and J815 lack board trees, and J813 lacks descriptions for many SoC peripherals. Share the common T8142 boot resources, add conservative J704/J815 trees and record disabled peripheral resources with the shared IP bindings. Known I2C controllers and audio codec identities use their existing bindings.

This is a resource-description draft. Parent power domains, mixed IRQ routing, firmware-service children and per-board handoff remain external audit work. Newly added codecs and unqualified power control remain disabled. Audit ledgers are maintained outside the kernel repository.

Depends on the shared binding foundation and its existing T6050/PROLED prerequisites. Validation with DT2's all-Apple differential W=1/schema checker is pending buildbox capacity; no successful validation or hardware boot is claimed yet.

[dmitzsaz]

Note

🤖 Claude Opus 5.5 responding on behalf of @dmitzsaz

I have a J704 tree that boots on hardware, and some of it can be enabled rather than left disabled. Should I send the J704 enablement on top of this PR once it lands, or would you rather take it another way?

Tested on the J704 (macOS 27.0 stub) on 2026-11-07, with aurora-wip 57222e0 plus #1060, #1122, #1123 and #1124:

• cpufreq and PMC work. All 10 CPUs come up: 4 P at up to 4608 MHz, 6 E at up to 3048 MHz, schedutil. They're disabled here.
• NVMe. The ANS firmware on this stub ends the nvme block at 0x10000 (ADT ans reg[9]), with the linear SQ doorbells in the NVMMU block. My J704 DT describes it that way, and nvme: apple: find the linear SQ doorbells from the nvme window #1060 makes the driver follow the window size.
• PCIe. It is described the way J813 describes its N1: the loader trains the links and Linux uses a generic ECAM host at 0x1cb0000000.
  • Port 0: BCM4388 Wi-Fi and Bluetooth, DART 0x488000000 (IRQ 1463).
  • Port 2: GL9755 SD reader, DART 0x48a000000 (IRQ 1481).
  • Both use stream 0. A single apple,t8142-pcie-msi lists both ports' MSI maps (PCI: apple-msi: program the MSI map of every listed port #1124), because they share one AIC range.
  • Wi-Fi, BT and SD (SDR104, hot-plug) all work, including after s2idle.
• MTP. My U-Boot hands the MTP over live (mtp-live-v1) for its keyboard. The kernel side is iommu: apple-dart: keep a T8142 DART that a live coprocessor uses #1122 (DART) and soc: apple: rtkit-helper: preserve an adopted T8142 MTP session #1123 (rtkit-helper). Without it, the keyboard is dead in Linux.
• Also working: keyboard and trackpad, USB on the three Type-C ports, SMC (battery, AC, RTC, fan, temperatures), and s2idle.
• Not working. The internal DCP stays on the boot framebuffer on this firmware, and its error path then hits a devm_clk_put WARN in dcp_comp_unbind.