Skip to content

device-akita 0.1.0: the Pixel 8a's package, and its storage plan (M2) - #456

Merged
derekross merged 2 commits into
derekross:mainfrom
aninmucek:device-akita-m2
Nov 8, 2026
Merged

derekross merged 2 commits into
derekross:mainfrom
aninmucek:device-akita-m2

Conversation

@aninmucek

Copy link
Copy Markdown
Contributor

Follows #414. The Pixel 8a's storage now comes up behind the guard, and this adds what the Arch root needs next, in one PR as the Pixel 8 Pro's #42 did. The kernel side is derekross/linux-omarchy-mobile#206.

  • device/akita/BRINGUP-M2.md (new): stage 2's runbook on husky's UFS-PLAN path (M2.1-M2.7), with the results so far:

    • M2.1, the port: husky's three DTS nodes, which akita's live tree matches value for value.
    • M2.3, the first UFS boot, read-only (gate c): HS-G4 on two lanes; 62 partitions on stock's nodes; only UFS_RW writable.
    • M2.4, a fresh read-only backup taken from our own kernel. Stock ran for about 43 h after stage 0's backup, and Magisk's root went with slot B at M1.3. The backup has 58 items. 11 changed since stage 0, all of them partitions stock or the bootloader writes.
  • device/akita/overlays/ufs-probe.dtso (new): the bring-up overlay for M2.3's boot, never default. It sets the board's bootargs plus omamobile.ufs-probe omamobile.initramfs.

  • device/akita/device.conf: NV_PARTS and NV_BACKUP_EXTRA, with cheetah's values, so that tools/nv-backup -d akita works. All seven partitions exist on akita, and none is on UFS_RW. NOTES.md's provenance table gets the row.

  • packages/omamobile-device-akita 0.1.0 (new): the generic set of the Pixel 8 Pro's first package (device-husky 0.1.0: the Pixel 8 Pro's package, and its UFS plan #42):

    • the storage rule, the USB gadget network and NetworkManager's drop-in;
    • the power key;
    • the mirror list and its pacman hook;
    • reboot-to;
    • the descriptor.

    The shared files are byte-identical to cheetah's and husky's, and its test checks that. install stages nine files, none of which marks a slot or starts the radio, the charger or a camera.

  • device/akita/build-u3-root (new): husky's test-root builder with akita's names, for M2.6.

M2.7, Omarchy's image, needs omamobile-device-akita in the signed [omamobile] repository, which only you can publish.

Two observations:

  • The guard's timing. With UFS built in, the disks attach before PID 1: at 0.26 s, with /init starting at 0.28 s. The guard flags every partition by 0.61 s, and nothing is mounted or written in between. husky's UFS-PLAN (Design 6) says the guard runs before the disks appear; it's probably this way on husky too.
  • The radio backup over USB hit one SSH connection timeout and four chunk re-reads, which nv-backup's retries absorbed. A plain stream afterwards ran at 36 MB/s.

Checks:

  • tests/all: 424 ok, 0 not ok.
  • tests/packages: every package passes, the new one included (15 ok). Under the load of a kernel gate running alongside, timing tests in five other packages failed: omamobile-mirror, -mmd, -shell, -audio (its 30-minute timeout) and -sensors. All of them pass with the CPU free, and mmd and sensors also when run on their own.
  • tools/pr-check: passes.
  • tools/nv-backup -d akita --list, the loader check (DEVICE=akita bin/omamobile initramfs-config) and tests/device-overlays pass.
  • tools/phone-smoke: not run, since it needs the Arch root and a session, which come in M2.6 and M2.7.

🤖 Generated with Claude Code

The Pixel 8a's UFS comes up behind the storage guard (gate c), and a
fresh read-only backup is taken from our own kernel.

- device/akita/BRINGUP-M2.md: stage 2's runbook (M2.1-M2.7) and the
  results of M2.1-M2.4.
- device/akita/overlays/ufs-probe.dtso: the bring-up overlay for the
  first UFS boot, never default.
- device.conf: NV_PARTS and NV_BACKUP_EXTRA, for tools/nv-backup, with
  their row in NOTES.md's provenance table.
- packages/omamobile-device-akita 0.1.0: the generic set of the Pixel
  8 Pro's first package.
- device/akita/build-u3-root: husky's test-root builder, for M2.6.

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

Copy link
Copy Markdown
Owner

Thank you, @aninmucek, this is another clean stage. The read-only first UFS boot, the guard's listing checked partition by partition against stock, and a fresh backup taken from your own kernel with both ends hashed are exactly the order I'd want. Explaining the eleven changed items by what stock and the bootloader write is what makes the new baseline trustworthy. The package is tidy too: every shared file is compared with its source, the test guards those comparisons so the package still builds on its own, and nothing in it touches a slot or the radio. CI is green.

Two small fixes in BRINGUP-M2.md before I merge:

  1. The status text predates the results. Line 4 still says "nothing is built or flashed yet", and the M2.1 result says the kernel change is "still uncommitted". Both are now linux-omarchy-mobile#206.
  2. 2d26e9ece doesn't exist anywhere I can see. The M2.4 results say the change was committed as 2d26e9ece, while the status line and the gate paragraph say 31d8b229d (extract-vendor-firmware, device-tegu README: the Wi-Fi check names the Wi-Fi chip, the README the suffixed files #206's head). Since kernel PRs land by rebase, extract-vendor-firmware, device-tegu README: the Wi-Fi check names the Wi-Fi chip, the README the suffixed files #206 will get a new id when it merges anyway. Could you cite extract-vendor-firmware, device-tegu README: the Wi-Fi check names the Wi-Fi chip, the README the suffixed files #206 and the DTS blob bfa8b6f54e0e instead? I checked that the blob is extract-vendor-firmware, device-tegu README: the Wi-Fi check names the Wi-Fi chip, the README the suffixed files #206's zuma-akita.dts, so that ties the M2.3 boot to what lands.

Suggestions, none blocking:

  • In M2.6, please read /sys/power/mem_sleep. Neither this package nor the Pixel 8 Pro's has the suspend-to-idle drop-in. If the 8a offers deep, Omarchy's Suspend would use it, and we'll want the drop-in before M2.7's session.
  • NV_PARTS takes the 7 Pro's list. If akita has another identity or calibration partition that list doesn't name, a line in NOTES would say so (your full backup already has it).
  • ufs-probe.dtso carries a copy of the board's bootargs, so a comment saying which kernel it was copied from would stop a later reuse from silently reverting them.

I'll merge #206 first, then this one. Thanks also for the note on the guard's timing. You're right, and it means husky's plan has the same wrong sentence. I'll fix that one.

@ShoreBits

Copy link
Copy Markdown

Tested on another retail Pixel 8a: M2.3, M2.4 and M2.6 pass on device-akita-m2 (93de3d9), kernel 7.3.0-rc6-omamobile-push1-gfe12fd24f7f8 (omamobile at fe12fd24f, UFS from #206 included, no ufs-probe overlay).

Starting point: stock CP3A.261005.005, bootloader akita-17.0-15681527, locked, running from slot B with slot A slot-unbootable:yes. Unlocked, then sideloaded the full OTA for the same build once, which installed to slot A and booted it (slot-successful:a yes). Then bin/omamobile flash of the M1 images to slot B.

M2.3 (guard), checked from the M1 initramfs over SSH:

  • Power mode changed to : FAST series_B G_4 L_2; SAMSUNG KLUDG4UHGC-B0E1 rev 3203, as on the first unit.
  • 62 partitions, every name on the same node as stock's /dev/block/by-name listing (diff empty).
  • Writable: exactly UFS_RW (sda23-sda27, sda34); sdb-sdd ro; no /dev/sda.

M2.4 (backups), with the guard on and nothing mounted:

  • tools/nv-backup -d akita: 7 partitions, each verified. The same stalls as test 10 (Connection timed out, 2-3 chunks re-read per partition), absorbed by the retries.
  • The full set (every partition except userdata and super, plus the sdb-sdd LUNs): 63 items, 1.6 GB, each streamed over one connection and checked against the phone's hash; the phone's hashes were the same before and after. The seven NV images are byte-identical to nv-backup's copies. mfg_data needed 5 tries because of connection timeouts.

M2.6 (test root on userdata):

  • bin/omamobile modules, device/akita/build-u3-root, then tools/install-phone (dry run first). Five _b images and userdata, then --set-active=b.
  • Login prompt on the panel; SSH as root with the installed key; / is sda34 grown to 108 GB; systemctl is-system-running → running; systemctl --failed empty; the guard's flags unchanged under Arch.
  • reboot-to bootloader reached fastboot in 16 s. After --set-active=b (retries back to 3) and fastboot reboot, the second boot came up the same: network after 24 s, running, no failed units, guard unchanged.

Findings:

  1. The test root has no depmod indexes. The build container has no kmod, so modules_install warns requires depmod and the tarball has no modules.dep; build-u3-root doesn't run depmod and the test root runs no pacman, so kmod's hook never makes them either. Arch boots and M2.6 passes (what it needs is built in), but modprobe finds nothing until depmod -a $(uname -r) is run on the phone; after that, modprobe -n -v aes-neon-blk resolves and the indexes survive a reboot. Running depmod in build-u3-root (or adding kmod to the build image) would close this. husky's builder is the same, so it likely has it too.
  2. The M2.6 runbook command needs --reinstall when the phone runs from slot B, as it does after M1/M2.3 (install-phone refuses otherwise). The same for M2.7.
  3. Host side: INSTALL.md §5's profile uses ipv4.method shared, which needs dnsmasq; on a laptop without it NetworkManager fails activation in a loop and the address flaps. ipv4.method manual with 172.16.42.2/24, as CONTRIBUTING describes, works. Maybe worth a line in INSTALL.md.
  4. The clock starts in July 2026 at every boot (no RTC driver, no network), as the M2.7 watch-for says.

Nothing outside UFS_RW was written. Logs and backups are private and kept off-repo.

…ew notes

- packages/omamobile-device-akita: the suspend-to-idle drop-in. The 8a's
  /sys/power/mem_sleep reads "s2idle [deep]", and Linux has no zuma SoC
  sleep sequence to wake from deep. It uses cheetah's settings, with a
  comment that names zuma; the test compares the settings.
- device/akita/build-u3-root: install kmod and run depmod for the slot B
  kernel's release, so modprobe finds modules in the test root (found by
  a second tester on derekross#456).
- BRINGUP-M2.md: the status and the kernel citations now name
  linux-omarchy-mobile#206 and its DTS blob; --reinstall in the M2.6 and
  M2.7 commands; M2.6's results (gate d).
- NOTES.md: the storage findings, and the per-phone partitions that
  NV_PARTS doesn't name.
- ufs-probe.dtso: where its copied bootargs come from.

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

Copy link
Copy Markdown
Contributor Author

Thanks for the review! Both fixes are in the new commit, along with M2.6's results:

  1. The status text now says M2.1-M2.6 are done, and M2.1's result names linux-omarchy-mobile#206.
  2. 2d26e9ece was the commit before my sign-off rebase, so it never existed outside this laptop. Wherever the runbook named a kernel commit it now cites extract-vendor-firmware, device-tegu README: the Wi-Fi check names the Wi-Fi chip, the README the suffixed files #206, merged as 75b9c0b73, and the DTS blob bfa8b6f54e0e.

The suggestions:

M2.6 (gate d) passed on this phone too:

  • Arch runs from userdata, and the guard's flags held under Arch.
  • reboot-to bootloader reached fastboot in 25 s.
  • A second boot came up the same, with SSH in 28 s (4.1 s by systemd-analyze).

@ShoreBits, thank you for testing on a second 8a! Your findings are in:

  • depmod: confirmed here too (no modules.dep). build-u3-root now installs kmod, runs depmod for the slot B kernel's release, and fails the build if no index results. The rebuilt image has the indexes. husky's builder has the same gap.
  • --reinstall: the M2.6 and M2.7 commands now carry it, with the reason.
  • dnsmasq: we hit the same on this laptop, and installed it. I've left INSTALL.md to the maintainer, since it isn't akita's.
  • The clock: noted in M2.6, for M2.7's keyring step.

Your results are cited in BRINGUP-M2 and NOTES as the second unit.

@derekross

Copy link
Copy Markdown
Owner

Thank you, @aninmucek. Both fixes are in, and I checked each one. The status line now matches the results. The plan cites linux-omarchy-mobile#206: 75b9c0b73 is on omamobile with zuma-akita.dts blob bfa8b6f54e0e, and the kernel you ran M2.6 on (31d8b229d, #206's head) has exactly the merged tree, so gate (d) ran what landed. All three suggestions are in too. The suspend-to-idle drop-in matters more than it looked: the board's bootargs don't set mem_sleep_default, so this file is what keeps Suspend off deep. The M2.6 record (the dry run, the root grown, the guard's flags under Arch, a second boot) is exactly the evidence I'd want for gate (d).

@ShoreBits, thank you for running M2.3, M2.4 and M2.6 on a second 8a. Your phone was on the same build and bootloader, which also confirms the descriptor's AVB values for a second owner. The depmod and --reinstall findings were real, and both are fixed here. I'll fix the same depmod gap in the Pixel 8 Pro's builder, and add a note to INSTALL.md about dnsmasq for the shared connection.

Checks: tools/pr-check passes. CI is green on this head. The branch merges cleanly onto today's main, and on that merge the package test (16 ok, also from a copy of the package alone), the storage-guard and nv-backup tests, nv-backup -d akita --list and the initramfs config all pass, with UFS_RW unchanged.

I'm happy to merge this as it is. A few small things, none blocking, for this PR or the next one:

  • NV_BACKUP_EXTRA: the Pixel 8 Pro's package (omamobile-device-husky 0.1.8: the radio partitions for tools/nv-backup #449, merged this afternoon) now adds trusty_persist and trusty_userdata next to devinfo, so nv-backup --check shows when Android has written Trusty's storage. The 8a has both, and trusty_persist already changed while stock ran, so I'd use the same list here. NOTES' new list could name trusty_userdata as well.
  • ufs-probe.dtso: "If the board file's change" should be "If the board file's bootargs change".
  • tools/phone-smoke: it is read-only, and with the M2.6 root it can already report the slot, the guard, the root and USB. The session rows will fail for now, which is expected. A report from that root, or at M2.7 as the plan says, would be good to have.
  • Suspend in M2.7: with the drop-in, Suspend now means s2idle, which hasn't run on Tensor G3 yet. A line in M2.7's "Watch for" saying not to use Suspend until the Pixel 8 Pro's U6 would keep the session test from turning into a suspend test.
  • BRINGUP-M2.md line 275 says "extract-vendor-firmware, device-tegu README: the Wi-Fi check names the Wi-Fi chip, the README the suffixed files #206's head". Writing linux-omarchy-mobile#206, as in the rest of the file, avoids reading it as a PR here.

I'll update the 8a's row in DEVICES.md for the Arch root after this merges, unless you'd rather add it here. Thanks again for another careful stage.

@derekross
derekross merged commit 6565c7d into derekross:main Nov 8, 2026
2 checks passed
@aninmucek
aninmucek deleted the device-akita-m2 branch November 8, 2026 04:04
derekross pushed a commit that referenced this pull request Nov 8, 2026
…up, and #456's notes (#509)

From the review of #456:

- device.conf (and the package's copy): NV_BACKUP_EXTRA adds
  trusty_persist and trusty_userdata next to devinfo, as husky's
  package does since #449, so tools/nv-backup --check shows when
  Android has written Trusty's storage.
- NOTES.md: trusty_userdata in the list of per-phone partitions, and
  the provenance row for the new value.
- BRINGUP-M2.md: don't use Suspend in M2.7 until husky's U6;
  linux-omarchy-mobile#206 spelled out; tools/phone-smoke's report from
  the M2.6 root (test 18).
- ufs-probe.dtso: "If the board file's bootargs change".

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
derekross added a commit that referenced this pull request Nov 12, 2026
…i 8 lines #505 left (#680)

The Pixel 8a's row still described first light. Since then Omarchy's
image runs from userdata (BRINGUP-M2 gate e, #456 and kernel #206
merged), and the buttons and the touchscreen passed gates f and g
(kernel #238 and #241 merged): Boot, Storage, Touch and Buttons say so.

For the Mi 8, three lines promised on #505: BRINGUP-M1's [NOTES.md] link
had no definition, the board DTS (kernel #220) is merged, not in review,
and the status line names the 2026-11-11 boot NOTES records.

The POCO F3 row said #25 was in review; it was closed as paused on
2026-11-11.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…derekross#456)

The Pixel 8a's UFS comes up behind the storage guard (gate c), and a
fresh read-only backup is taken from our own kernel.

- device/akita/BRINGUP-M2.md: stage 2's runbook (M2.1-M2.7) and the
  results of M2.1-M2.6 (gate d ran on linux-omarchy-mobile#206).
- device/akita/overlays/ufs-probe.dtso: the bring-up overlay for the
  first UFS boot, never default.
- device.conf: NV_PARTS and NV_BACKUP_EXTRA, for tools/nv-backup, with
  their row in NOTES.md's provenance table, and the per-phone partitions
  NV_PARTS doesn't name.
- packages/omamobile-device-akita 0.1.0: the generic set of the Pixel
  8 Pro's first package, plus the suspend-to-idle drop-in (the 8a offers
  "deep", and Linux has no zuma SoC sleep sequence to wake from it).
- device/akita/build-u3-root: husky's test-root builder for M2.6, with
  kmod and depmod for the slot B kernel's release (found by a second
  tester).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…up, and derekross#456's notes (derekross#509)

From the review of derekross#456:

- device.conf (and the package's copy): NV_BACKUP_EXTRA adds
  trusty_persist and trusty_userdata next to devinfo, as husky's
  package does since derekross#449, so tools/nv-backup --check shows when
  Android has written Trusty's storage.
- NOTES.md: trusty_userdata in the list of per-phone partitions, and
  the provenance row for the new value.
- BRINGUP-M2.md: don't use Suspend in M2.7 until husky's U6;
  linux-omarchy-mobile#206 spelled out; tools/phone-smoke's report from
  the M2.6 root (test 18).
- ufs-probe.dtso: "If the board file's bootargs change".

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…i 8 lines derekross#505 left (derekross#680)

The Pixel 8a's row still described first light. Since then Omarchy's
image runs from userdata (BRINGUP-M2 gate e, derekross#456 and kernel derekross#206
merged), and the buttons and the touchscreen passed gates f and g
(kernel derekross#238 and derekross#241 merged): Boot, Storage, Touch and Buttons say so.

For the Mi 8, three lines promised on derekross#505: BRINGUP-M1's [NOTES.md] link
had no definition, the board DTS (kernel derekross#220) is merged, not in review,
and the status line names the 2026-11-11 boot NOTES records.

The POCO F3 row said derekross#25 was in review; it was closed as paused on
2026-11-11.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
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