Skip to content

device-panther 0.1.5: the screen sleeps with the display off, and the D3 plan - #281

Open
nesmeck wants to merge 3 commits into
derekross:mainfrom
nesmeck:panther-decon-pr
Open

nesmeck wants to merge 3 commits into
derekross:mainfrom
nesmeck:panther-decon-pr

Conversation

@nesmeck

@nesmeck nesmeck commented Nov 1, 2026

Copy link
Copy Markdown
Contributor

What and why

The Pixel 7's screen through gs-decon. The kernel side is derekross/linux-omarchy-mobile#181. This PR brings the rest:

  1. DISPLAY-PLAN (panther): D3, gs-decon drives the screen.
    • D3's changes and its staged runbook, every step with its Pass, Stop and Rollback: stage 0 (read-only), stage 1 (frame count), stage 2 (KMS), and the screen off with screen_off 0, 1 and 2.
    • The results on the phone, including the padded-row bug that stage 2 hit first and its fix.
    • The Pixel 7 row's display cell now says gs-decon and links the kernel PR.
  2. device-panther 0.1.5: the screen sleeps with the display off. omamobile-display.conf adds options gs_decon screen_off=2, as the Pixel 7 Pro's package does. With the screen off, DECON stops its frames and the panel goes to sleep; DPMS on wakes it through the panel's cold init. gs_decon needs no other option, since it goes straight to KMS (stage 2) by default.

Device and build

Checks (no phone needed)

  • tools/pr-check upstream/main passes
  • tests/all passes
  • tests/packages passes (the panther package's test/all)
  • Each changed package has its version bumped, and the new version is in the commit subject
  • Commits have area prefixes and bodies that say why; any AI help is a Co-Authored-By: trailer, never a Signed-off-by:

On the phone

  • I ran this on my own phone, booted into this build. 0.1.5 was built and installed on the phone; the database step is as in issue Contributor phones: pacman -U fails when [omamobile] is signed with a local key #168.
  • Runbook steps passed:
    • D3a, stage 0: gs-decon bound on the 1080x2400 set with ABL's state as expected.
    • D3b, stage 1: frames at 60.02 Hz, no storm.
    • D3c, stage 2: Hyprland at 1080x2400@60, smooth, with no flip timeouts.
    • D3d: screen_off 0, 1 and 2 all passed. With 2, the panel slept and woke through its cold init, also from the power key's lock and a tap.
    • 0.1.5 with no bring-up file: after a reboot, allow_tx read Y and stage and screen_off read 2 from the package alone, with everything above working.
  • What I did not test:
    • 90 Hz, which is the next step;
    • a suspend (s2idle) with gs-decon;
    • the Pixel 7 Pro, for the kernel's padded-row change.
  • Nothing in this change writes outside slot B and userdata, weakens the storage guard, or reads registers from userspace.
tools/phone-smoke report (that boot, 0.1.5)

phone-smoke: panther

Kernel 7.3.0-rc6-omamobile-push1-g53fab72d1489, tools at 05b472c, 2026-11-01T18:20Z.

Check Result Detail
boot (required) ok kernel 7.3.0-rc6-omamobile-push1-g53fab72d1489, slot b, up 2191s
storage guard (required) ok 55 partitions; only userdata and slot B's boot images writable
root (required) ok ext4 on userdata, 97G free
kernel health (required) ok no oops or BUG; 0 WARN, 9 error lines in dmesg
systemd units ok none failed
USB gadget (required) ok bound to 11210000.usb
display ok card2-DSI-1 connected, 1080x2400
GPU ok render node on panthor
session (required) ok Hyprland and the Omarchy shell running
lock ok PAM file present, the shell can lock (passwordPam true, locked false)
touch ok FocalTech FT3658
buttons ok gpio-keys
Wi-Fi skip no wireless interface
Bluetooth skip no controller
battery ok 91% Discharging, 31 °C
sound cards skip none
sensors skip no sensor proxy
cameras skip no libcamera tools
thermal ok hottest zone 45 °C

Packages: hyprland 0.56.2-5.1, libcamera 0.7.2-4.21, mesa 1:26.2.4-1.4, omacam 0.3.3-1, omamobile-apps 0.2.0-1, omamobile-device-panther 0.1.5-1, omamobile-firstboot 0.1.2-1, omamobile-keyring 20261026-1, omamobile-mirror 0.1.1-1, omamobile-omarchy 0.1.12-1, omamobile-owl-widgets 0.1.1-1, omamobile-shell 0.2.28-1, omarchy 4.0.6-1, quickshell 0.3.2.r7.g2b4501a-1

🤖 Generated with Claude Code

@derekross

Copy link
Copy Markdown
Owner

Thank you, @nesmeck. The D3 runbook is exactly what I'd hoped for: each stage has its Pass, Stop and Rollback, and the stage-0 bring-up file comes with the rule to delete it. The results read well too, including the failed first try of stage 2 and why it failed, and the screen_off table. The package change is one line with its test, the version is bumped, and the Pixel 7 row changes only its display cell. tools/pr-check is clean, CI is green, and the branch merges cleanly onto today's main.

The package is safe in either order with linux-omarchy-mobile#181, since screen_off already exists in the published gs_decon. I'll still merge #181 first, so that the plan and the matrix can say "merged" for the kernel when this lands. When I merge, I'll make two small edits, since you've allowed edits:

  1. "in review" becomes "merged" for omamobile-omarchy 0.1.17, omamobile-shell 0.2.41: the lock glance after hands-on testing #181, in the status line and the Pixel 7 row.
  2. The "Kernel D3" paragraph names only the first build (…-gdc3f0ce5e425). I'll add that D3c and D3d ran on …-g53fab72d1489, with the padded-row commit, as your results already say.

If you reorder the kernel series as I asked on #181 (the padded-row commit before the DECON0 enable), the numbered list under "The changes" should follow. Either push that here, or I'll do it with the edits above.

One question for the plan's next step, not for this PR. Suspend hasn't run on the Pixel 7 yet, and now gs-decon drives the screen at every boot. Before the first s2idle test, it would be good to add a D-step or a SLEEP gate that checks the screen comes back after a wake: DPMS on, a cold init if the panel was asleep, flips resuming. That way the first suspend doesn't also become the first test of gs-decon across a suspend.

Thanks again, this is a really nice step for the Pixel 7.

@nesmeck

nesmeck commented Nov 2, 2026

Copy link
Copy Markdown
Contributor Author

Thanks! I've pushed the reordered list of changes in D3: the padded-row commit is now number 4, with the reason it comes before the DECON0 enable, and the enable is number 5, matching linux-omarchy-mobile#181's new order. The "Kernel D3" paragraph and the "merged" edits are yours to make at merge, as you offered. Noted on the suspend gate: before the Pixel 7's first s2idle test I'll add a step that checks gs-decon across a wake (DPMS on, the panel's cold init if it slept, flips resuming).

🤖 Generated with Claude Code

@derekross

Copy link
Copy Markdown
Owner

Thank you, @nesmeck. The list of changes in D3 now matches linux-omarchy-mobile#181's new order, with the reason the padded-row commit comes first. The package commit is unchanged, and thank you for taking the suspend gate on board.

One thing moved under the PR since my last comment, not through anything you did. The release-channel change (#313) merged into main afterwards and bumped every device package, so main already has a device-panther 0.1.5 (the mirrorlist and the channel handling in the install script). Your branch still merges cleanly, and the package's test passes on the merged tree. But two different 0.1.5s would mean a phone with either one never sees the other, so this one needs to ship as 0.1.6.

Since you've allowed edits, I'll do that when I merge, with the edits from my last comment:

  1. The version becomes 0.1.6, as do the status line and the Pixel 7 row. The results paragraph keeps a note that the build you tested was numbered 0.1.5, so the record still says what ran.
  2. The "Kernel D3" paragraph names both kernels, worded for the new order: the first build had every commit but the padded-row one, and D3c with the fix and D3d ran on …-g53fab72d1489 with all five.
  3. Once linux-omarchy-mobile#181 has merged, "in review" becomes "merged" for it in the status line and the Pixel 7 row.

The squashed title will be "device-panther 0.1.6: the screen sleeps with the display off, and the D3 plan". I'll merge this after linux-omarchy-mobile#181, which waits for my Pixel 7 Pro run. Nothing more is needed from you here.

@nesmeck

nesmeck commented Nov 4, 2026

Copy link
Copy Markdown
Contributor Author

Updated, as asked on linux-omarchy-mobile#181. The new head is 2308680, rebased onto today's main:

  • angelrenefm's D3 run is in DISPLAY-PLAN as the second unit's results (2026-11-03), with his kernel release 7.3.0-rc6-omamobile-push1-g0a132eb92793, every step's result, and what he didn't check. The status line mentions it.
  • The two kernels: the D3a paragraph now says its kernel (…-gdc3f0ce5e425) lacked the padded-row commit, and D3c's (…-g53fab72d1489) that it had all five, for the series' new order.
  • 0.1.7: the version file, the status line and the Pixel 7 row. The results keep that the build tested was numbered 0.1.5. This assumes omarchy-mobile#396 (0.1.6) goes in first, as you planned.
  • The rebase: main had changed the rows around the Pixel 7's in docs/DEVICES.md, so the branch no longer merged cleanly. I took main's table with this branch's Pixel 7 row. The other two commits' content is unchanged.

tools/pr-check and tests/all pass, and it merges cleanly onto main. "in review" for #181 is still yours to flip at merge.

🤖 Generated with Claude Code

nesmeck and others added 3 commits November 4, 2026 05:46
Record D3's changes and runbook and its results on the phone: stage 0
and stage 1 (frame starts at 60.02 Hz), stage 2 first failing on
Hyprland's padded 1080-pixel rows and then passing with the fix
(1080x2400 at 60 Hz, smooth, no flip timeouts), and DPMS off/on with
screen_off 0, 1 and 2, the last putting the panel to sleep and waking it
through its cold init, also from the power key's lock.

The Pixel 7 row's display cell names gs-decon and links
linux-omarchy-mobile#181, which brings the kernel side.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Set gs_decon's screen_off to 2 in the display modprobe.d file, as the
Pixel 7 Pro's package does: with the screen off, DECON stops sending
frames and the panel goes to sleep, and DPMS on wakes it through the
panel driver's cold init. DISPLAY-PLAN gate D3d proved it on the Pixel 7,
from Hyprland's DPMS and from the power key's lock.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
angelrenefm's unit passed every D3 step on linux-omarchy-mobile#181's
head (issue derekross#342), as derekross asked on derekross#181: stage 0 found ABL's state
as on the first unit, 60 Hz with no storm, 871 flips with no timeouts,
and screen_off=2 six times out of six. The plan records it with his
kernel release.

The D3a paragraph now says its kernel lacked the padded-row commit, and
D3c's that it had all five, for the series' new order.

The package ships as 0.1.7: 0.1.6 is omarchy-mobile#396's AVB footers,
which merges first. The results keep that the build tested was numbered
0.1.5.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@nesmeck

nesmeck commented Nov 5, 2026

Copy link
Copy Markdown
Contributor Author

@derekross, the suspend gate you suggested has run: gs-decon across a wake on the Pixel 7, before any other suspend work.

The kernel. It isn't linux-omarchy-mobile#181 alone. It is my local daily stack: omamobile with the D3 branch (#181's commits), the 90 Hz branch (D4), the charger, Bluetooth (#196), Wi-Fi (#195) and the torch, with device-panther 0.1.7 plus Wi-Fi. gs-decon, gs-dsim-cmd and the panel driver are the same as in #181 and D4.

The gate (D5). Each sleep wrote /sys/power/state directly, through rtcwake -m freeze -s 30, so no hooks or sleep lock ran. Each ran detached and logged to UFS, with Wi-Fi unloaded first.

  • D5a, pm_test with the panel asleep: freezer, devices and platform each returned 0. The panel stayed asleep with the trigger masked. DPMS on then woke it through the cold init, at 90 Hz.
  • D5b, the first real s2idle, with the screen off: the RTC woke it after 30 s (through the PMIC's interrupt). DPMS on ran the cold init and brought the picture back at 90 Hz, with touch and the GPU working.
  • D5c, s2idle with the screen on: while asleep, the screen stayed lit on its last frame (DECON keeps scanning out). After the wake, frames and flips resumed with no DPMS step. A second run had a video playing at 30 flips a second. The flips in flight completed on time before the freeze, and playback went on after the wake.

All four sleeps had no flip timeouts or late flips, and no gs-decon, DSIM, panel or UFS errors. The one UFS line at each resume is the dme-set attr-id 0x4021 that the Pixel 9 Pro's clean resumes print too.

What it means for the suspend plan:

  • A lit sleep is safe for gs-decon. But it keeps the panel and DECON running, so the sleep path should still turn the screen off first: patch 0021 as on the Pixel 7 Pro, or gs-decon system-sleep callbacks later.
  • A flip cut by the sleep is hard to reach. Userspace freezes 20 to 30 ms before the device interrupts go off, and a pending flip completes within a frame. Those callbacks would close that case too.

The gate and its results are in panther's DISPLAY-PLAN as D5, on a local branch. I'll send them after this PR merges, rather than add them here.

🤖 Generated with Claude Code

derekross added a commit to angelrenefm/omarchy-mobile that referenced this pull request Nov 5, 2026
@derekross

Copy link
Copy Markdown
Owner

Thank you, @nesmeck, for the update, and thank you @angelrenefm for the D3 run on the second unit. The new table reads well: every step with its result and the kernel it ran on, and an honest "not checked" line. The rebase kept the Pixel 7 row as it was, the package change is the same one line with its test, and the version, status line and matrix now say 0.1.7, with the record still noting that the tested build was numbered 0.1.5. tools/pr-check passes, CI is green, and it merges cleanly onto main.

The plan stays as before: I'll merge linux-omarchy-mobile#181 once it has run on my Pixel 7 Pro, then #396 (0.1.6), then this as 0.1.7. At merge I'll flip "in review" to "merged" for #181 and write the squash title and body from the 0.1.7 version file, since the title here still says 0.1.5. One small thing I'll fix at the same time unless you get to it first: the "2026-11-03: D3 on a second Pixel 7" heading needs a blank line above it.

Thanks for keeping this current while the kernel side waits.

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