Skip to content

omarchy-mac: wait for J514C native graphics before SDDM - #104

Merged
maralcbr merged 8 commits into
mainfrom
carry/j514c-sddm
Nov 7, 2026
Merged

maralcbr merged 8 commits into
mainfrom
carry/j514c-sddm

Conversation

@maralcbr

@maralcbr maralcbr commented Nov 7, 2026 •

Copy link
Copy Markdown
Member

Summary

Carries Chris Kearney's #92 (J514C waits for native graphics before SDDM), with my sign-off and one fix. Chris has left the project. Authorship is kept on his four commits. Replaces #92.

 wait-for-display
-  J514C + T6031                         -> wait <= 30 s for apple-drm card + asahi-g15 render node, settle <= 10 s
+  J514C + T6031 + handed over           -> wait <= 30 s for apple-drm card + asahi-g15 render node, settle <= 10 s
+    handed over = gpu alias node status "okay"
+                  && a dcp@* node carries apple,t6031-handoff
+                  && the last apple_t6031_display.enable is not false (kstrtobool: 0/n/f/d/off...)
+  J514C + T6031, not handed over        -> no wait (nothing will replace simpledrm)
   anything else                         -> existing <= 10 s display wait
  • Chris's change fixes a race on J514C's native path: SDDM opens the autologin VT before native DRM replaces simpledrm, so the first session gets SIGTERM.
  • The fix: on its own, omarchy-mac: wait for J514C graphics before SDDM #92 held every J514C greeter for 30 s whenever the boot loader had not handed graphics over. That covers an owner opt-out, a refused check, or an m1n1 from before the J514C defaults (kboot: default J514C native display and graphics m1n1#31). Now the long wait only runs when the device tree shows the handover, and a J514C without one exits at once instead of falling into the older 10 s wait.
  • The gate was checked on the lab J514C: aliases/gpu -> /soc/gpu@40a400000 with status okay, and apple,t6031-handoff on /soc/dcp@386c00000.

Version 0.2.5. Pairs with omacom/linux-aurora#363 (merged) and omacom/m1n1#31.

Evidence

  • Before (omarchy-mac: wait for J514C graphics before SDDM #92 alone): a J514C with the GPU node disabled, no handoff marker or apple_t6031_display.enable=0 enters the 30 s wait.
  • After (test/native-display-test.sh, Arch container):
    • As root, all five groups pass. As a user, the four that don't need root pass.
    • The new group fails without the gate (verified). It also covers the kernel spellings for off, the last value on the command line winning, and a present display-subsystem node.
  • test/display-test.sh still passes.

Merge Danger

Door: two-way

Blast Radius: J514C login

Other Macs keep the existing display wait. On a J514C the worst case is a 40 s greeter delay when a handover was made but native graphics never come up; login still proceeds.

chriskearney and others added 5 commits November 8, 2026 07:28
The live speaker-recovery fixture calls the staged system setup, which also checks optional Steam packages. Stub pacman in that fixture so an installed host Steam cannot invoke a real package transaction or make the test machine-dependent. This changes no runtime code.

Assisted-by: Codex:GPT-6
Signed-off-by: Marcelo Alcantara <maralc@gmail.com>
J514C native DRM can replace the boot framebuffer after SDDM has opened the login VT, terminating the first autologin session. The existing helper watches the older display-subsystem by-path name and misses this handoff.

For exact J514C/T6031 boots with the effective bring-up parameter equal to 1, wait at most 30 seconds for an apple-drm card plus an asahi-g15 render device, settle device events for at most 10 seconds and recheck. Other boots retain their existing display wait. Timeouts never fail the greeter, and no driver, owner setting or service is enabled or restarted.

A target-local prototype passed installed and ordinary automatic-login checks on the bring-up hardware. This packaged implementation has fixture evidence only; cold-boot package qualification remains required before release. Keep the experimental board outside general support.

Assisted-by: Codex:GPT-6
Signed-off-by: Marcelo Alcantara <maralc@gmail.com>
Normal J514C bring-up must start graphics and login without an enable parameter. Gate the readiness wait only on Apple Silicon and exact J514C/T6031 identity, matching the normal kernel and Stage2 defaults. Remove kernel-command-line parsing from the helper and document the automatic behavior.

The staged helper tests cover a normal empty command line, exact board/SoC exclusion, native DRM identities, late readiness, settle failure and a bounded missing-GPU timeout. The full package suite reaches the same inherited systemd261 Touch ID fixture failure reproduced on its unchanged base.

Assisted-by: Codex:GPT-6
Signed-off-by: Marcelo Alcantara <maralc@gmail.com>
The Touch ID unit-verification fixtures fail under systemd 261 because their isolated roots omit sysinit.target, which service default dependencies require. The same missing-target failure occurs on the untouched readiness base before the T6034 change.

Stage the minimal boot target through a shared fixture helper before the real systemd-analyze verification in both Touch ID suites. Keep production Touch ID behavior, unit files, verification commands and assertions unchanged.

The complete omarchy-mac package suite now passes on the controller, including native graphics readiness and both real staged-root Touch ID unit checks. No verification is skipped or ignored.

Assisted-by: Codex:GPT-6
Signed-off-by: Marcelo Alcantara <maralc@gmail.com>
The J514C wait held every greeter for 30 s whenever the boot loader had not
handed the GPU and display over: an owner opt-out, a refused check, or an m1n1
from before the J514C defaults. Wait only when the GPU node is enabled, the DCP
carries the T6031 handoff marker and apple_t6031_display.enable is not off;
otherwise keep the existing display wait.

Signed-off-by: Marcelo Alcantara <maralc@gmail.com>
@maralcbr
maralcbr requested a review from scottjones as a code owner November 7, 2026 21:31
@greptile-apps

greptile-apps Bot commented Nov 7, 2026 •

Copy link
Copy Markdown

RetriggerView in GreptileConfidence Score: 5/5

[Medium impact] The latest changes appear safe to merge.

Summary

Adds a bounded native-graphics wait before SDDM on J514C/T6031 when the boot loader handed graphics over. Other Macs keep the existing display wait.

  • J514C login waits for handed-over native graphics before SDDM starts.
  • Test fixtures provide their own systemd and package-command setup.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Before SDDM starts] --> B{J514C and T6031?}
  B -->|No| C[Existing display wait]
  B -->|Yes| D{Graphics handed over and display enabled?}
  D -->|No| E[Start SDDM]
  D -->|Yes| F[Wait for native display and GPU]
  F --> G[Bounded device-event settle and recheck]
  G --> E
  C --> E
Loading

Reviews (4) · Last reviewed commit: "omarchy-mac: tokenize the command line l..." · Reviewed by Greptile

Take the last apple_t6031_display.enable on the command line and read it as
kstrtobool does, so false, no, OFF and disable also skip the wait and a final
enable=1 still waits. A J514C without a handover now exits at once instead of
falling into the older 10 s display-subsystem wait.

Signed-off-by: Marcelo Alcantara <maralc@gmail.com>
Strip quotes, treat a bare apple_t6031_display.enable as on, and ignore a
value kstrtobool rejects, so the last valid setting decides as it does in the
kernel.

Signed-off-by: Marcelo Alcantara <maralc@gmail.com>
…switch

Group double-quoted text, stop at --, and match - and _ alike in the switch
name, so a switch quoted inside another parameter or passed to init is not
taken for the kernel's.

Signed-off-by: Marcelo Alcantara <maralc@gmail.com>

@scottjones scottjones left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approving. The new device-tree marker check hasn't had a packaged cold boot in our lab yet; I'd like to run that on the J514C before this reaches stable.

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.

3 participants