Skip to content

T6050 sleep: select drm/apple v27 transport and asahi g17 provider by hardware, not board - #2442

Draft
RyanTheTide wants to merge 3 commits into
integration/aurorafrom
spec-m5neo/t6050-sleep-gates-drm
Draft

RyanTheTide wants to merge 3 commits into
integration/aurorafrom
spec-m5neo/t6050-sleep-gates-drm

Conversation

@RyanTheTide

@RyanTheTide RyanTheTide commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

Problem: the drm/apple firmware 27 sleep transport for the J716c and the drm/asahi G17 provider translation are switched on by board name. That is where they were run, not a hardware property.

Change: both now use apple_t6050_sleep_described(), which is true when the DT has an available apple,t6050-sleep node. Only the J714s and J716c trees have one, so on every board in the tree today nothing changes. Builds on the helper commit at the base of #2440.

Testing: W=1 compile of drm/apple dcp.o and drm/asahi g17/sleep.o (release config, clang) at base and with this branch: both build, no code warnings before or after. The only drm/asahi file touched is C; the asahi Rust objects are unchanged and were not rebuilt.

Still to run: @DjDeveloperr @erikwb, could you give a J714s and J716c row (deep sleep with the panel on and with a USB-C display, then resume)? Draft until then.

Related: #2384 also edits dcp.c (the J813 system-sleep stop, around :9405, and the s2r_untested text). The hunks do not overlap, and git merge-tree with its head 8407d40 merges cleanly.

Details
File:line (0abc2f3) Old New Before After
drm/apple dcp.c:5791 j716c and t6050 (after the J714s 27.0 path) apple,t6050-sleep J716c (plus J714s through its own path) same
drm/asahi g17/sleep.c:54 t6050 and (j714s or j716c) and agx-g17s apple,t6050-sleep and agx-g17s J714s, J716c J714s, J716c

dcp.c:5791 on the J714s: with firmware 27.0, v27_j714s returns first, as before. With any other firmware, the generic branch's own fw_compat == 27.0 test refuses it, as the board check did. So the result is the same on every board.

Kept (J714s only): dcp_j714s_ext() (dcp.c:473), v27_j714s (dcp.c:7003), the fabric's sticky pipeline (dcp-fabric.c:3363) and the peer limit (dcp-peer-v27.inc:72). The J716c runs a different IOMFB method table and panel profile and has not run these. Widening them would change the J716c, so they need a J716c row first.

Hardware rows: J714s/J716c, echo mem > /sys/power/state with deep sleep. The panel comes back, and the GPU commit-stop lines in dmesg match the current tip.

Original work: DjDeveloperr (T6050 display/GPU sleep), Erik Bourget (J716c).

Assisted-by: LLM (Claude:Opus 5.5)

Several drivers select their T6050 (M5 Pro and Max) deep-sleep handling
with a list of boards, the ones the sleep has run on (J714s, J716c). What
those boards have in common is in the device tree: an available
sleep-controller node, compatible "apple,t6050-sleep". No other board in
the tree has one; a board gains it with its sleep description.

Add apple_sleep_described(compatible), true when an available node with
that compatible exists, and apple_t6050_sleep_described() for
"apple,t6050-sleep". Both are static inline and independent of
CONFIG_APPLE_SOC_SLEEP, as the board checks they replace are.

Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
Besides the J714s's own firmware 27.0 path, the generic 27.0 sleep
transport was taken on a machine that is "apple,j716c" and
"apple,t6050". It belongs to the T6050 sleep, so take it where the
device tree describes that sleep (apple,t6050-sleep).

Nothing changes in the tree today: the J716c is as before; the J714s
takes its own path first when the firmware is 27.0, and with any other
firmware the generic transport's own 27.0 check refuses it, as the
board check did; no other board describes the sleep.

The J714s-only rules (dcp_j714s_ext(), v27_j714s, the fabric's sticky
pipeline and the peer limit) are left as they are: the J716c runs a
different method table and panel profile and has not run them.

Fixes: fd5cc59 ("drm/apple: v27: the T6050 sleep transport on the J716c")
Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
…iption

The wrapped provider translation for the GPU's commit stop was used on
a machine that is "apple,t6050" and "apple,j714s" or "apple,j716c". It
belongs to the T6050 deep sleep, so use it for an "apple,agx-g17s"
where the device tree describes that sleep (apple,t6050-sleep): the
same two boards in the tree today, so nothing changes.

Fixes: 541d9be ("drm/asahi: g17: the M5 Pro's commit stop on the J716c too")
Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
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.

1 participant