Skip to content

Pixel 8a (akita): enable UFS, as on the Pixel 8 Pro - #206

Merged
derekross merged 1 commit into
derekross:omamobilefrom
aninmucek:omamobile-zuma-akita-ufs
Nov 7, 2026
Merged

derekross merged 1 commit into
derekross:omamobilefrom
aninmucek:omamobile-zuma-akita-ufs

Conversation

@aninmucek

@aninmucek aninmucek commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

Part of derekross/omarchy-mobile#1. One commit, on today's omamobile (29a7a93):

  • arm64: dts: exynos: google: zuma-akita: Enable UFS adds the Pixel 8 Pro's three UFS nodes to akita's board file:
    • the UFS VCC regulator on gpp0-1, active high;
    • vcc-supply and status = "okay" on ufs_0;
    • ufs_0_phy switched on.

Why husky's nodes fit. akita's live device tree has every UFS value where husky's does:

  • the host at 0x13200000 (SPI 466) with the same six regions;
  • brd-for-cal 1;
  • HS-G4 on two lanes;
  • the reset and reference-clock pins, gph5-1 and gph5-0 (function 2);
  • the UFS sysreg at 0x13020000 (IOCC 0x710);
  • the PMU isolation at 0x3ec0;
  • VCC on a fixed regulator switched by gpp0-1, which Google's GPL zuma-ufs.dtsi describes for every zuma board.

The nodes are enabled in one commit under CONTRIBUTING §6.2's exception for a node proven on a same-SoC phone, now that it has passed on this one.

Tested on a Pixel 8a. The kernel was 7.3.0-rc6-omamobile-push1-g29a7a9326bf0-dirty: this commit's DTS on 29a7a93 before it was committed, byte-identical (blob bfa8b6f54e0e). It booted from slot B into the initramfs, with omamobile.ufs-probe omamobile.initramfs from a bring-up overlay, so nothing ran on userdata.

  • The link: Power mode changed to : FAST series_B G_4 L_2 0.24 s into boot, as on stock. The device is KLUDG4UHGC-B0E1, rev 3203.
  • The partitions: the four LUNs and all 62 partitions appear with stock's names on stock's device nodes.
  • The storage guard: it left exactly slot B's five boot images and userdata writable. sdb-sdd are read-only, and /dev/sda is removed.
  • Reads: a 64 MiB direct read of boot_b ran at 1.5 GB/s. About 2 GB of read-only backup reads followed (tools/nv-backup and a full partition set) with no UFS error, reset or I/O error.
  • The only other UFS lines: ufshcd_populate_vreg assumes vdd-hba, vccq and vccq2 are on, as on the Pixel 8 Pro.

The runbook and its results are in derekross/omarchy-mobile#456 (device/akita/BRINGUP-M2.md, M2.1-M2.4).

Checks: tools/kernel-pr-check --build OUT --merge origin/omamobile passes everything:

  • a clean merge onto 29a7a93;
  • the sign-off: 1 commit, signed off by its author, none by an AI;
  • checkpatch --strict with no errors and no warnings;
  • the 10 Google device trees with W=1 and dtbs_check, with no new warnings.

🤖 Generated with Claude Code

The Pixel 8a's UFS is wired as the Pixel 8 Pro's. Its live device tree
has the host at 0x13200000 (SPI 466) with the same six register
regions, the reset and reference-clock pins on gph5-1 and gph5-0
(function 2), the UFS sysreg at 0x13020000 (IOCC 0x710), the PMU
isolation at 0x3ec0, and VCC on a fixed regulator switched by gpp0-1,
active high, as Google's zuma-ufs.dtsi describes it for every zuma
board. ABL hands the device over with VCC off, so the driver has to
turn it on.

zuma.dtsi already describes the host, its PHY, the pins and the clock
stand-ins. Add the VCC regulator and enable the host and its PHY, the
same nodes zuma-husky.dts has.

Tested on a Pixel 8a, with the initramfs's storage guard: the link
comes up at HS-G4 on two lanes, rate series B, as on stock. The four
LUNs and all 62 partitions appear with stock's names on stock's device
nodes, a 64 MiB direct read runs at 1.5 GB/s, and the log has no UFS
error, reset or I/O error.

Link: https://android.googlesource.com/kernel/devices/google/zuma/+/refs/tags/android-16.0.0_r0.10/dts/zuma-ufs.dtsi
Assisted-by: Claude Opus 5.5
Signed-off-by: DarekM <149845949+aninmucek@users.noreply.github.com>
@derekross

Copy link
Copy Markdown
Owner

Thank you, @aninmucek. This one is easy to check: the three nodes are the Pixel 8 Pro's, your Link: shows the VCC regulator on gpp0-1 in Google's zuma-ufs.dtsi, and the DTS that booted in your M2.3 run (blob bfa8b6f54e0e) is exactly this commit's. Our gate passes on today's omamobile (clean merge, sign-off, checkpatch with no warnings, all ten Google device trees with no new warnings), and CI's dtbs_check is green. Enabling it in one commit is right under the same-SoC rule, since husky is zuma too and it passed on your phone first.

One small thing, if you're willing: the comment above &ufs_0 says the storage guard "runs before the disks appear", and your own M2.3 timing shows the opposite (disks at 0.26 s, /init at 0.28 s). The safety still holds, because nothing mounts or writes before the guard has flagged every partition. Something like "which sets every partition read-only before anything mounts them" would make the comment match your measurement. The sentence came from husky's file, and I'll fix that one myself.

I'll merge this first, then #456.

@derekross
derekross merged commit 75b9c0b into derekross:omamobile Nov 7, 2026
1 check passed
@aninmucek
aninmucek deleted the omamobile-zuma-akita-ufs branch November 7, 2026 13:57
@aninmucek

Copy link
Copy Markdown
Contributor Author

Thanks for merging! The comment fix is up as #214, with the wording you suggested. It is comment-only: on today's omamobile the DTB is byte-identical with and without it.

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