Move to zephyr v4.1

[petejohanson]

Opening this as a draft for now, as there is still work to be done here, along extensive testing, but at this point I want a place where progress, testing results, etc can be tracked.

This is considered beta level at this point, so basic support questions will not be answered on this PR/branch. I welcome testing by anyone willing to, given this level of readiness. TIA.

Prelim Blog Post

@nmunnich prepared an update post, which I am working to round out and finalize: https://deploy-preview-3060--zmk.netlify.app/blog/2025-xx-xx-zephyr-4-1

Testing steps

Local dev

1. If not building w/ VSCode docker integration, install https://github.com/zephyrproject-rtos/sdk-ng/releases/tag/v0.16.9
2. gh pr checkout 3060
3. west update
4. Use new build directory to build ZMK as usual.

GitHub Actions

Make two edits in the branch of your config repo:

1. Update .github/workflowr/build.yaml to contain:

   on: [push, pull_request, workflow_dispatch]
   
   jobs:
     build:
       uses: petejohanson/zmk/.github/workflows/build-user-config.yml@core/move-to-zephyr-4-1
   

2. Update config/west.yaml according to the https://zmk.dev/docs/features/beta-testing/ instructions to test this PR.

Outstanding Items

• Bring in blog post from docs(blog): Add Zephyr 4.1 upgrade post petejohanson/zmk#48
• Finish updating existing (nRF52 & RP2040) boards to support newer bootloader approach with retained mem + bootmode APIs from Zephyr
• Ensure blog post covers HWMv2 upgrades, bootloader changes, nRF52840 DCDC and NFC pin DTS changes, LVGL API changes, LED strip changes, input support for keyboard matrixes (testing only), and anything else discovered during testing
• Add note to blog post about RP2040 board upgrades: Need to enable CONFIG_CLOCK_CONTROL, adjust DTS include path, boot mode retention.
• Fix up display screens, including nice!view custom status screen + widgets
• Integrate split BLE API change code and test thoroughly.
• Test, test, more testing.
• Document CMake v3 requirements for some edge cases (nRF5340 where libmetal gets used)

PR check-list

• Branch has a clean commit history
• Additional tests are included, if changing behaviors/core code that is testable.
• Proper Copyright + License headers added to applicable files (Generally, we stick to "The ZMK Contributors" for copyrights to help avoid churn when files get edited)
• Pre-commit used to check formatting of files, commit messages, etc.
• Includes any necessary documentation changes.

[joelspadin]

Was deleting this intentional? I noticed this while editing all the boards using zmk,battery-voltage-divider for one of my two approaches for implementing fuel gauge API support.

[petejohanson]

Yes, because this board is moved over to using multiple versions, so the specific settings on that come from the version specific overlays. The prop is still set, it's just coming from somewhere else.

[galaxyeden]

Great to see this making progress, thanks for all your hard work on ZMK! I wanted to move to zephyr,input to get my own board ready for Zephyr 4.2 and I ran into a few issues and notes:

• There is a maximum of eight rows by default, users may need to select INPUT_KBD_MATRIX_16_BIT_ROW – I did
• The Zephyr documentation for gpio-kbd-matrix assumes the matrix is active low. In my previous testing, I found this provided a small but noticeable improvement over the active high ZMK-documented configuration, so my keyboard was already set up this way, but it may confuse some users looking at the Zephyr documentation and not understanding the difference, especially newbies
• gpio-kbd-matrix only supports scanning "columns" and listening on "rows" – this may also confuse some, especially if this doesn't match their physical layout
• Initially, I could only get five rows to work, @petejohanson suggested it was likely running out of hardware interrupts set on the falling (or rising, for active high) edge of the pin, as gpio-kbd-matrix sets. I'm on an nRF5340 with 8 interrupts available, which is comparable to most common ZMK users' hardware – so anyone with more than (hardware interrupts - defined buttons - other defined interrupt pins) is likely to encounter this. This can be fixed by emulating this behavior in software with port-based interrupts, more below.
• In physical_layouts.c there is a use of the ZMK_KSCAN_EVENT_QUEUE_SIZE that applies to the zephyr,input scan driver as well. I renamed this Kconfig symbol (for aesthetics) and removed the if ZMK_KSCAN conditional in my fork to resolve this, and it works as expected.

To emulate the falling edge interrupts in software, each GPIO port needs a sense-edge-mask defined. I set this to < 0xffffffff > for both ports and that resolved the issue and my nRF5340-based board is working as expected with zephyr,input.

On the very good news side, I've seen dramatically better power consumption – about 25% lower at idle than before total (≈240uA to ≈180uA), about 40% lower with keys held forcing scanning (at the same 1ms scan interval, my initial, much better results were due to the 5ms default scan interval), and similar – hard to tell but slightly higher if anything – during my "average" typing – with the amount of time spent idle for most users, I expect this to be a significant, though variable, overall power savings if it holds true across multiple SoCs and matrix configurations.

[wlo2]

Can confirm &bootloader is now correctly puts dongle into the bootloader if called from macro (Central locality) for Seeed XIAO nRF52840 with a Adafruit bootloader, solid work!