Skip to content

arm64: dts: apple: j716c: deep sleep - #2123

Merged
DjDeveloperr merged 14 commits into
integration/aurorafrom
j716c/dts-deep-sleep
Nov 17, 2026
Merged

DjDeveloperr merged 14 commits into
integration/aurorafrom
j716c/dts-deep-sleep

Conversation

@erikwb

@erikwb erikwb commented Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

What this does

The J716c (16" M5 Max) device tree for deep sleep (suspend-to-RAM):

  • the sleep resources, the PMU real-time clock and the SEP handoff;
  • t6050-j716c-sleep-devices.dtsi, the devices' settings for the sleep: the storage
    controller's restart spacing, the display coprocessors' firmware 27.0 sleep arguments, the port
    controllers' sleep power states (hpm0-2 and the MagSafe controller hpm5, now enabled), the
    right port's USB4 router sleeping in place (the clock-protection syscon, acio2's combo PHY and
    the clock-protection slices of its NHI and tunnelled-PCIe DARTs, 30 and 31 as on the J714s),
    the DARTs' GAPF slices and the internal panel's size;
  • t6050-j716c-sleep-full.dtsi, the full sleep description (the J714s's T6050 SoC data, which the
    J716c's ADT matches for pmgr, nub-spmi0 and its PMU, mcc, aic, wdt, smc, gpio0, apcie0 and the
    cpus), which includes the devices' settings and enables the sleep controller, with the Max's
    memory domains under automatic power, the masks of its two extra SPMI buses and the cores'
    power-gate enables (the boot loader leaves the boot core's clear, so without them the boot
    cluster never powers down and the sleep never completes).

Merge order: the commit that enables the full sleep ("enable the full deep sleep") turns the
J716c's deep sleep on, and needs the functional driver PRs first: #2119 (j716c/sleep-j716c-profile),
#2122 (j716c/sep-t6050-profile) and the driver PRs #2124 (j716c/drm-apple-sleep), #2125 (j716c/dart-sleep),
#2126 (j716c/thunderbolt-sleep), #2127 (j716c/usb-sleep), #2128 (j716c/phy-sleep), #2129 (j716c/pcie-sleep),
#2130 (j716c/aic-sleep), #2131 (j716c/asahi-sleep), #2132 (j716c/spmi-sleep), #2133 (j716c/macsmc-sleep) and
#2134 (j716c/nvme-sleep). This PR must merge after them.

Tested

  • arm64: dts: apple: j716c: deep sleep #2123 (j716c/dts-deep-sleep) builds on its own base (Image.gz, modules and dtbs, arm64, LLVM).
  • make dtbs builds t6050-j716c.dtb.
  • All 17 branches merged on integration/aurora were built as one kernel and run on a J716c (16" M5 Max): one suspend-to-RAM with a USB4 dock attached, about 30 s asleep, woken by the RTC alarm (wake source 76). After the wake all 18 CPUs were online, the dock's router came back in place with its Ethernet up, and the SEP answered RNG reads. The sleep's log showed "secondary clusters down first: cores 6+6, cluster 1 0x1a in 353 us, cluster 2 0x1c01a in 448 us". The PMP was not running in that build.

Dependencies

Branch #2123 (j716c/dts-deep-sleep). Based on integration/aurora. Its sleep controller compatible and SPMI bus count are documented by the dt-bindings change in #2119 (j716c/sleep-j716c-profile). Merge after the driver PRs listed above.

Part of the J716c deep sleep set: #2118 (j716c/spin-table-release-cpus), #2119 (j716c/sleep-j716c-profile), #2120 (j716c/sleep-wake-source-names), #2121 (j716c/sleep-sep-auto-wake), #2122 (j716c/sep-t6050-profile), #2123 (j716c/dts-deep-sleep), #2124 (j716c/drm-apple-sleep), #2125 (j716c/dart-sleep), #2126 (j716c/thunderbolt-sleep), #2127 (j716c/usb-sleep), #2128 (j716c/phy-sleep), #2129 (j716c/pcie-sleep), #2130 (j716c/aic-sleep), #2131 (j716c/asahi-sleep), #2132 (j716c/spmi-sleep), #2133 (j716c/macsmc-sleep), #2134 (j716c/nvme-sleep).

🤖 Generated with Claude Code

Describe the board resources for sleep bring-up while leaving the sleep
controller, alarm, and coprocessor nodes disabled. Do not inherit the
J714s transition sequences or enable suspend operations.

Developed with Claude Code.
Signed-off-by: Erik Bourget <erik@ebourget.net>
Enable the Abbey PMU counter and alarm on J716c while leaving the SoC
sleep controller and coprocessor sleep paths disabled. The generation 4
SPMI controller does not yet expose the PMU alarm interrupt to Linux.

Developed with Claude Code.
Signed-off-by: Erik Bourget <erik@ebourget.net>
Describe the SEP alias and enable its mailbox and power providers. Keep
the two fabric parents under firmware ownership during warm attachment.
The SEP and its DART remain disabled until the driver selects its policy;
the sensor and system sleep are not enabled by this change.

Developed with Claude Code.
Signed-off-by: Erik Bourget <erik@ebourget.net>
Add t6050-j716c-sleep-devices.dtsi for what the J716c's devices need
for a deep sleep besides the sleep controller, and include it: the
storage controller's restart spacing after a sleep and the display
coprocessors' firmware 27.0 sleep control arguments.

Developed with Claude Code.

Signed-off-by: Erik Bourget <erik@ebourget.net>
Give the USB-C port controllers hpm0-2 and the MagSafe controller hpm5
apple,sleep-ssps, so that each is switched to system power state S3
for a firmware suspend and back to S0 after it, and enable hpm5.

Developed with Claude Code.

Signed-off-by: Erik Bourget <erik@ebourget.net>
For the right port's USB4 router to sleep in place, as on the J714s,
describe the clock-protection syscon, give acio2 its combo PHY, and
give the router's NHI and tunnelled-PCIe DARTs their clock-protection
slices. The J716c's dart-acio2 and dart-apciec2 have slice indices 30
and 31, as on the J714s.

Developed with Claude Code.

Signed-off-by: Erik Bourget <erik@ebourget.net>
The J716c's ADT gives the SEP, PMP0, SIO, apcie0 and display DARTs the
same clock-protection slice indices as the J714s's, so their GAPF slices
are the same. The internal display coprocessor's DART claim refused the
sleep without them.

Developed with Claude Code.
Signed-off-by: Erik Bourget <erik@ebourget.net>
Without a size the display driver does not treat the internal display
as a panel: no backlight level is configured, and a panel the firmware
turned off for the sleep step is not lit again. Give the panel its
size: 3456 x 2234 pixels at a 0.1 mm pitch, 2160 rows below the camera
housing.

Developed with Claude Code.

Signed-off-by: Erik Bourget <erik@ebourget.net>
Add t6050-j716c-sleep-full.dtsi, the J716c's full deep sleep: it
includes t6050-j716c-sleep-devices.dtsi, with the devices' settings
for the sleep, and enables the sleep controller as the full one
(apple,t6050-j716c-sleep, apple,t6050-sleep) with its SoC data.

The controller takes the J714s's SoC data as it is. The J714s and J716c
ADTs of the same release have the same pmgr, nub-spmi0 and its PMU,
pmp0, mcc, aic, wdt, smc, gpio0, apcie0 and cpus nodes apart from
phandles; every power word the sequences and the wake order name is in
the J716c's own power inventory, and its own ADT gives the same AP wake
sources, the same codec reset pin (139), the same N1 warm reset
('bpli', 0x1000000) and the same SMC pcIO lines. The commit schedule,
the SMC's spacings and the clock-gating, CPM, PMC and PHY values are
the J714s's measurements, which no ADT holds.

As on the J714s, GFX and SPI_P stay on through the commit, the radio's
port gets its warm reset and the N1's SMC writes, the trackpad's
actuator its sleep report and the GPU its automatic power state after
the wake.

Developed with Claude Code.

Signed-off-by: Erik Bourget <erik@ebourget.net>
Include t6050-j716c-sleep-full.dtsi, which includes
t6050-j716c-sleep-devices.dtsi, so that the J716c's sleep controller
runs the full T6050 entry and wake.

Developed with Claude Code.

Signed-off-by: Erik Bourget <erik@ebourget.net>
…wer for the sleep

The J714s's boot sequence puts AMCC0-15 and their AFR/AFC fabric words
under automatic power: iBoot leaves them on with automatic power off, so
none can gate. The Max has AMCC16-31 and the fabric words in front of
them (FAB_AFR_AMCC_8-15, FAB_AFC_AMCC_4-7), and the full sleep left those
as iBoot had them: on, forced (AMCC16-31 read 0x0f0003ff on ewb154). A
real attempt never got through the sleep engine's entry. Put them under
automatic power too, group by group as the J714s does its own.

The second encoder needs nothing: its words read off.

FAB3_SOC, one of the ADT's no_ps PMGR devices, is enabled on this board
only as the left port's supply. genpd lowered it at the noirq suspend, and
it never reached 0 (requested 0xf0000f0, read 0x1f0003ff), leaving a
request off on a hardware-owned fabric word over the entry. Mark it
firmware-owned, as FAB0_PIOGW already is, so Linux no longer writes it.

Developed with Claude Code.
Signed-off-by: Erik Bourget <erik@ebourget.net>
…MI buses

The J716c's ADT has nub-spmi1 (a mosquito PMU) and nub-spmi4 (two jerez
PMUs) beyond the J714s's buses, and the sleep left their interrupt
enables as the running system had them. Give them the sleep's mask too,
after the J714s's five, in the same way.

Unverified: the J714s's mask (0x2000001) is assumed right for these PMUs,
and their order after the others is a choice, not a measurement.

Developed with Claude Code.
Signed-off-by: Erik Bourget <erik@ebourget.net>
… sleep

Describe the cores' power-gate enable words (/arm-io/pmgr reg[0] +
0x2c000, one word per core) for the sleep controller, the same node as
on the J714s. The boot loader leaves the boot core's power-gate enable
clear, so without the node the boot cluster never powers down: the
sleep never completes, and the engine's watchdog resets the machine
five seconds after the commit.

Developed with Claude Code.

Signed-off-by: Erik Bourget <erik@ebourget.net>
Take the merged T6050 bindings and driver prerequisites into the DT
validation branch. Keep the five J716c DTS blobs from the original
series byte-identical.

Signed-off-by: Diljit Singh <hey@djdev.me>
@DjDeveloperr
DjDeveloperr marked this pull request as ready for review November 17, 2026 19:44
@DjDeveloperr
DjDeveloperr merged commit cb57d4a into integration/aurora Nov 17, 2026
2 checks passed
@RyanTheTide
RyanTheTide deleted the j716c/dts-deep-sleep branch November 17, 2026 23:53
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