Skip to content

Add Microsoft Surface Pro 11 board support - #14390

Merged
ryanrhughes merged 15 commits into
omacom:quattrofrom
turbineBMW:upstream/surface-pro-11
Oct 28, 2026
Merged

ryanrhughes merged 15 commits into
omacom:quattrofrom
turbineBMW:upstream/surface-pro-11

Conversation

@turbineBMW

@turbineBMW turbineBMW commented Oct 6, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Board setup for the Surface Pro 11 (Snapdragon X Elite, OLED, microsoft,denali-oled), in the pattern of yoga-slim7x.sh. The installer puts linux-omarchy on every aarch64 machine, and that kernel carries the Surface Pro 11 patches, so this script leaves the kernel alone.

  • Writes the early-boot drop-in: display, USB-C and keyboard modules plus GPU firmware before disk unlock, all optional (as in qualcomm/initramfs.sh) so any kernel's image builds.
  • Embeds the OLED model's device tree from linux-omarchy in place of dtb-uki.sh's list, which can't tell the OLED, LCD and X Plus apart (they share an SMBIOS identity). The tree is found under linux-omarchy* and the config is written only when it changes, flagging an image's rebuild, as dtb-uki.sh does. Without that tree the list stays and an earlier pin is removed, so the UKI still builds.
  • Adds surface-pro-11-support (Wi-Fi board data, audio, camera routing), iptsd (pen and touch) and surface-pro-11-sensors (ambient light), and enables the sensor service. All three are listed in omarchy-other.packages, so the offline ISO carries them.
  • Removes a surface_device_modules.conf left by installs from before Keep Intel Surface fixes off Snapdragon Surfaces #14389. It sorts after this drop-in and resets MODULES.
  • LCD and X Plus models are left alone until someone tests them.

Dependencies

All landed:

Test plan

  • test/shell.d/surface-pro-11-hardware-test.sh: rerun idempotence, the device-tree pin with and without linux-omarchy, the image rebuild flag, early firmware, offline package listing, other Denali variants untouched
  • surface, snapdragon, yoga-slim7x, thinkpad-t14s-camera, camera, mkinitcpio-hooks, platform and device-tree tests pass on current quattro
  • Fresh ISO install on a Surface Pro 11 (aarch64 ISO built from this branch, free-space install next to Windows): Flex Keyboard during install and at login, Wi-Fi, Bluetooth, speakers, microphone, both cameras, pen and touch, ambient light, charger state, lid suspend to deep sleep and wake
  • The final revision, run on that installed system, produces the same uki.conf and packages; the only change is the optional ? on the early modules, all present in linux-omarchy 7.2.8-23

@birkskyum
birkskyum changed the base branch from dragon to n1x October 9, 2026 22:28
@birkskyum
birkskyum force-pushed the upstream/surface-pro-11 branch from bf8493a to 5586e89 Compare October 9, 2026 22:31
@ryanrhughes
ryanrhughes force-pushed the n1x branch 2 times, most recently from dcdea0b to 2f83505 Compare October 10, 2026 04:01
@birkskyum
birkskyum force-pushed the upstream/surface-pro-11 branch from 5586e89 to 747e56f Compare October 10, 2026 16:10
@birkskyum

Copy link
Copy Markdown
Contributor

I rebased this onto today's n1x, since it was showing conflicts after the branch was rebuilt. Your two commits are unchanged, and the first one is the same commit as in #14389 again. One commit on top swaps omarchy-hw-qualcomm-soc for omarchy-hw-aarch64-qualcomm in two lines, because the old check is gone on n1x (#14834). The Surface Pro 11 test, the Surface test and the Snapdragon tests pass in an ARM container. Your branch was force-pushed, so reset your local copy before adding to it. The old head was 5586e89.

@birkskyum

Copy link
Copy Markdown
Contributor

Two notes for when the packages land, both about how n1x does things now.

The stock kernel is removed with pacman -Rdd. n1x.sh does the same swap with omarchy-pkg-drop linux-aarch64 linux-aarch64-headers, and AGENTS.md asks for that helper over raw pacman -R. If something here needs -Rdd, a comment with the reason would be enough.

The model check reads the device tree directly, where n1x sends feature code through a helper. The reason is sound, since the OLED, LCD and X Plus models share a DMI name. The Zenbook script in #13820 reads it the same way, so a small shared helper could serve both. I can add that separately.

Getting the Surface packages into the Snapdragon image is on the ISO side. We will pick that up in omarchy-iso once the packages are published.

@birkskyum
birkskyum force-pushed the upstream/surface-pro-11 branch from 747e56f to cde3105 Compare October 10, 2026 21:12
@birkskyum

Copy link
Copy Markdown
Contributor

I rebased this onto today's n1x now that #14389 is merged, since it was showing conflicts. The commit the two PRs shared is gone from here because it is in n1x. Your Surface Pro 11 commit is unchanged, and the one commit on top that swaps the helper is still there. The Surface Pro 11, Surface, Snapdragon and platform tests pass in an ARM container. Your branch was force-pushed, so reset your local copy before adding to it. The old head was 747e56f.

@turbineBMW

Copy link
Copy Markdown
Contributor Author

Thanks for the rebases, Birk. I reset my branch to cde3105 and added two commits on top.

@birkskyum

birkskyum commented Oct 11, 2026 •

Copy link
Copy Markdown
Contributor

The helper is up as #15064. With it the check becomes omarchy-hw-aarch64-qualcomm && omarchy-hw-match-device-tree microsoft,denali-oled, and the test can fake the device tree through OMARCHY_PROC_ROOT in place of the script's own path override.

One more thing since #14952 went in. The generic device tree list now takes Omarchy's kernel's own trees first, so the comment above the Denali block is out of date.

@turbineBMW

Copy link
Copy Markdown
Contributor Author

Thanks, Birk. The comment now says why the tree is still pinned: the list holds this kernel's trees, but systemd-stub picks among them by SMBIOS identity, which the OLED, LCD and X Plus models share (2273427). The switch to omarchy-hw-dt-match, with the test on OMARCHY_PROC_ROOT, is ready on top of #15064; I'll push it here once that's in.

@birkskyum

Copy link
Copy Markdown
Contributor

#15064 is merged, so omarchy-hw-match-device-tree is on n1x now. A rebase onto n1x picks it up.

@turbineBMW
turbineBMW force-pushed the upstream/surface-pro-11 branch from 2273427 to c54f583 Compare October 11, 2026 20:15
@turbineBMW

Copy link
Copy Markdown
Contributor Author

Rebased onto n1x, and the model check now uses omarchy-hw-match-device-tree microsoft,denali-oled, with the test faking the tree through OMARCHY_PROC_ROOT (c54f583).

@birkskyum

Copy link
Copy Markdown
Contributor

One line still has the old name, and that is my doing. I renamed the helper after my first note here, and the mail you got still said omarchy-hw-dt-match. Line 9 of surface-pro-11.sh calls that, and it is not on n1x, so the check is always false and the Surface test fails. With omarchy-hw-match-device-tree in that line the test passes here.

@birkskyum

Copy link
Copy Markdown
Contributor

I pushed the one-word fix on top as 4341a3f, since the mix-up came from my side. It is a normal push with nothing rewritten, so a git pull before your next commit is all it needs. The Surface Pro 11 test, the Snapdragon test and test/cli pass on it in an ARM container.

@turbineBMW

Copy link
Copy Markdown
Contributor Author

Thanks for pushing that, Birk. I've pulled it, so the branch is at 4341a3f here too.

I checked the detection on the Surface Pro 11 itself, booted on linux-omarchy-n1x 7.2.8-6. omarchy-hw-aarch64-qualcomm and omarchy-hw-match-device-tree microsoft,denali-oled both pass. The live compatible list is microsoft,denali-oled, microsoft,denali, qcom,x1e80100, the same one the test fakes, and the Surface test passes on the machine. The pinned device tree is at the path linux-omarchy-n1x installs it to.

I haven't run a full install yet. The script installs linux-omarchy-n1x and the Surface packages from omarchy-pkgs #825–#829, #972 and #975, and none of them are on edge aarch64 yet, so for now it would stop at the first omarchy-pkg-add. I'm marking this ready for review of the script. I'll run a full install on the Surface once those packages are published and report back here.

@turbineBMW
turbineBMW marked this pull request as ready for review October 11, 2026 22:23
@greptile-apps

greptile-apps Bot commented Oct 11, 2026 •

Copy link
Copy Markdown
Contributor

RetriggerView in GreptileConfidence Score: 5/5

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

Summary

Adds board-specific setup for the Surface Pro 11 OLED.

  • Surface Pro 11 OLED laptops get the right boot setup and hardware support.

The earlier failed-kernel-install concern no longer applies: this version does not replace the kernel and only pins a device-tree file that already exists.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Qualcomm hardware setup] --> B{Surface Pro 11 OLED?}
  B -->|No| C[Leave board unchanged]
  B -->|Yes| D[Remove stale Intel keyboard settings]
  D --> E[Pin installed OLED device tree]
  E --> F[Write early modules and firmware settings]
  F --> G[Install support packages and enable sensors]
  G --> H[Finish setup and rebuild boot image]
Loading

Reviews (9) · Last reviewed commit: "Keep the Surface Pro 11's boot setup saf..." · Reviewed by Greptile

@birkskyum

Copy link
Copy Markdown
Contributor

One small thing. The early-boot list names videocc_sm8550 without a ?, and the stock kernel does not have it (checked in linux-aarch64 7.2.10-1, where it is neither a module nor built in). With the list as written mkinitcpio 42.2 still writes the stock kernel's image but exits 1 with module not found. With videocc_sm8550? it exits 0.

Beyond that the script looks good to me. I'd merge it once the packages are published and your full install is in.

@turbineBMW

Copy link
Copy Markdown
Contributor Author

Thanks, fixed in 762a554: it's videocc_sm8550? now. linux-omarchy-n1x 7.2.8-6 has every module in the list, so nothing changes there. I'll post the full install here once the packages are published.

@greptile-apps

greptile-apps Bot commented Oct 13, 2026

Copy link
Copy Markdown
Contributor

Want your agent to iterate on Greptile's feedback? Try greploops.

@omarchybot

Copy link
Copy Markdown
Collaborator

Omabot looked at this pull request at 307f94b to classify it, and the classification is not settled yet, so it has no class label and nothing was run on a worker.

Two models classified it separately. Claude Opus 5.5 read it as an enhancement, because it adds setup for a board Omarchy does not support today: a new install script, a kernel swap and new packages. Codex Medium read it as compatibility, a hardware bug, because the work is making one specific machine run. The difference matters for what happens next. A hardware fix gets verified and brought towards ready, while a new feature waits until the maintainer asks for it.

Next it waits on the maintainer to say which it is. Nothing here is waiting on you, and the full install you are planning for once the omarchy-pkgs packages are published is still the test that counts.

@birkskyum
birkskyum changed the base branch from n1x to quattro October 17, 2026 03:20
@birkskyum
birkskyum force-pushed the upstream/surface-pro-11 branch from 307f94b to 17e7430 Compare October 17, 2026 03:25
@birkskyum

Copy link
Copy Markdown
Contributor

Rebased onto quattro, since the runtime n1x branch went into it with #14820. Your commits are unchanged. One commit on top installs the kernel by its name since the rename, linux-omarchy, including the device tree path, and the test passes with it.

This still waits on the Surface packages, omacom/omarchy-pkgs#825 (surface-pro-11-support), omacom/omarchy-pkgs#826 (iptsd) and omacom/omarchy-pkgs#828 (the light sensor), and on the kernel patches in omacom/omarchy-pkgs#975, which need to move to the renamed linux-omarchy-aarch64 folder. Until those are published, the setup stops on a Surface Pro 11 when it can't install them.

@birkskyum
birkskyum force-pushed the upstream/surface-pro-11 branch from 17e7430 to ab856dd Compare October 17, 2026 03:43
@omarchybot

Copy link
Copy Markdown
Collaborator

Omabot classified this again at ab856dd, after the rebase onto quattro and the switch to linux-omarchy, and the split is the same as at 307f94b: Claude Opus 5.5 reads it as an enhancement (setup for a board Omarchy does not support today), Codex Medium as compatibility (making one specific machine work). So it still has no class label, nothing was run on a worker, and it was not reviewed beyond classifying.

For the maintainer: #15031 takes a different route to the same machine, an aarch64-surface platform on Arch Linux Ports with the linux-sp11 kernel, where this one is a board script on linux-omarchy. The two are not duplicates, but they are unlikely to both land as written, so choosing between them is part of the same decision.

This waits on the maintainer to say which class it is. Nothing here is waiting on you beyond the full install you already planned once the omarchy-pkgs packages are published.

turbineBMW and others added 15 commits October 21, 2026 14:16
The Surface Pro 11 (Snapdragon X Elite, OLED, microsoft,denali-oled)
needs a kernel with its touch and pen digitizer, keyboard cover and
cameras, and its own device tree. The UKI's device tree setting applies
to every kernel, so linux-sp11 replaces linux-aarch64 rather than
sitting beside it.

Write the boot configuration first, so the Surface kernel's image is
built with it when the kernel installs: the Surface device tree in
uki.conf, early display, USB-C and keyboard modules plus firmware for
the unlock screen (Surface modules optional, so a stock kernel's image
still builds), and a boot order that prefers linux-sp11. Remove the
Intel Surface keyboard module file an earlier install may have written.
Then install linux-sp11 (and its headers if the stock headers were
present) with surface-pro-11-support, and only after that remove the
stock kernel, so a failed install leaves it in place. Install libcamera
with IMX681 support, replacing a stock libcamera if one is present,
then iptsd, the light sensor package and the power-profiles-daemon
build that reaches the Surface's platform profile, and enable the
sensor and CPU limit services. The LCD and X Plus models are left
alone until tested.

Validated with test/shell.d/surface-pro-11-hardware-test.sh and the
surface, yoga-slim7x and snapdragon hardware tests. An earlier revision
of this script set up the Surface Pro 11 in daily use from a dragon ISO
install; this revision reorders it and has only been validated by the
tests.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
n1x.sh drops linux-aarch64 the same way, and AGENTS.md asks for the helper
over raw pacman -R. -Rdd was not needed: linux-aarch64 depends only on
coreutils, kmod and mkinitcpio, which linux-sp11 needs too, so -Rns takes
nothing else with it.

Validated with test/shell.d/surface-pro-11-hardware-test.sh and the
surface, snapdragon and yoga-slim7x hardware tests.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
linux-omarchy-n1x now carries the Surface Pro 11 patches (omarchy-pkgs
omacom#975), so the board script installs it instead of linux-sp11, still in
place of the stock kernel, and points the UKI at that package's Denali
device tree. Omarchy's defaults already boot linux-omarchy-* first, so the
script no longer writes its own boot order and removes the linux-sp11 one
an earlier revision left.

Validated with test/shell.d/surface-pro-11-hardware-test.sh and the
surface and snapdragon hardware tests. A Surface Pro 11 booted
linux-omarchy-n1x with the Surface Pro 11 patches and this device tree in
its UKI.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ds the kernel's own

Since omacom#14952 the generic list takes Omarchy's kernel's trees first, so the
old reason, that its trees belonged to the stock kernel, no longer holds.
The tree is still pinned because systemd-stub chooses among the list by
SMBIOS identity, which the OLED, LCD and X Plus models share.

Validated with test/shell.d/surface-pro-11-hardware-test.sh.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The OLED, LCD and X Plus models share a DMI name, so the script read the
device tree itself. omarchy-hw-match-device-tree (omacom#15064) does that for any
such model, and the test fakes the tree through OMARCHY_PROC_ROOT like the
other platform tests, skipping as root where fixtures are ignored.

Validated with test/shell.d/surface-pro-11-hardware-test.sh,
hw-match-device-tree-test.sh, snapdragon-hardware-test.sh and
surface-hardware-test.sh.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
omarchy-hw-dt-match became omarchy-hw-match-device-tree before omacom#15064 went in, so the old name is not on n1x and the model check was always false.
… install

The Surface device tree is written before linux-omarchy-n1x installs, so the
kernel's image is built with it. If that install failed, the stock kernel was
kept but its next image rebuild pointed at a device tree that was never
installed. The previous configuration is now put back first.

Validated with the Surface Pro 11, Snapdragon and Surface tests on the
Surface Pro 11 (X1E OLED).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The stock kernel has no videocc_sm8550, so mkinitcpio exited 1 with
"module not found" when it built that kernel's image, while
linux-omarchy-n1x has it and still loads it early.

Validated with the Surface Pro 11 and Snapdragon tests on the Surface Pro 11
(X1E OLED), where linux-omarchy-n1x 7.2.8-6 has every module in the list.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
linux-omarchy-n1x's IMX681 driver now exposes a linear gain and
surface-pro-11-support ships the front camera's tuning, so Arch's
libcamera drives both cameras. Install pipewire-libcamera with
omarchy-pkg-add instead of replacing libcamera with
libcamera-surface-pro-11.

Validated on a Surface Pro 11 with Arch's libcamera 0.7.2-4: front and back
cameras in Snapshot, gain following the light, before and after suspend.
The Surface Pro 11, Surface, Snapdragon and Yoga hardware tests pass.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Surface-specific power-profiles-daemon build and the CPU cap service are
gone: omarchy-power-profiles-daemon caps the CPU clusters on every Snapdragon
X laptop (omacom/omarchy-pkgs#1029).

Validated with test/shell.d/surface-pro-11-hardware-test.sh.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The kernel has been linux-omarchy since 13 October, with its device trees
under /boot/dtbs/linux-omarchy, so the Surface Pro 11 setup installs and
pins it by that name.
Only its hardware setup installs surface-pro-11-support, iptsd and
surface-pro-11-sensors, so without a listing the ISO's offline mirror
lacks them and the install stops on "target not found".

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Leave the kernel to the installer, which puts linux-omarchy on every
aarch64 machine, and the camera backend to qualcomm/camera.sh. The OLED
model's device tree replaces the generic list only when linux-omarchy
provides it, and the change is written and flagged for an image's
rebuild the way dtb-uki.sh does. Cleanups for files that only earlier
test revisions of this script wrote go too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Without Omarchy's kernel the script now removes its own device-tree pin
instead of leaving one to a file that is gone, which would stop the UKI
from building. The tree is found under any linux-omarchy package
directory, as dtb-uki.sh finds it, and every early module is optional,
as in qualcomm/initramfs.sh, so a kernel without one still builds its
image. A keyboard module file from before the Intel Surface fixes were
limited to x86 is removed again: it sorts after this drop-in and resets
MODULES.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@turbineBMW
turbineBMW force-pushed the upstream/surface-pro-11 branch from ab856dd to feb215d Compare October 21, 2026 18:18
@birkskyum

Copy link
Copy Markdown
Contributor

@turbineBMW The script, its test and the module names on 7.3 check out here. Two things before this goes in. The three packages are aarch64-only, and omarchy-iso's x86_64 build only leaves out what's in aarch64_only in builder/build-iso.sh (Ryan's e6d817d, now just asus-proart-p14-speaker-firmware). With this merged, the next x86_64 ISO fails on target not found: iptsd. Add iptsd, surface-pro-11-sensors and surface-pro-11-support to that list in an omarchy-iso PR that lands first. Then post the full install run on the published packages you mentioned, and this is ready.

@turbineBMW

Copy link
Copy Markdown
Contributor Author

Thanks! The omarchy-iso change is up as omacom/omarchy-iso#308, so the x86_64 ISO leaves the three packages out.

The full install run is on the published packages: I built an aarch64 ISO from this branch against edge, which has linux-omarchy 7.2.8-23 and all four Surface packages, and did a fresh free-space install on my Surface Pro 11. Everything passed; the results are in the description's test plan.

@omarchybot

Copy link
Copy Markdown
Collaborator

Omabot classified this again at feb215d, and the two models split the same way as before: Claude Opus 5.5 reads it as an enhancement, setup for a board Omarchy does not support today, and Codex Medium reads it as compatibility, a fix for one specific machine. So it still has no class label, nothing was run on a worker, and it was not reviewed beyond classifying.

What has changed since the last note is on your side: the x86_64 ISO change is up as omacom/omarchy-iso#308 and the full install on the published packages is in the description. Nothing here is waiting on you now.

For the maintainer: whether this is a feature to take or a hardware fix to verify is the open question, and #15031, an aarch64-surface platform for the same machine, is still open beside it, so choosing between the two is part of the same decision.

@ryanrhughes
ryanrhughes merged commit 600ca88 into omacom:quattro Oct 28, 2026
2 checks passed
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.

4 participants