Skip to content

omamobile-fingerprint 0.1.14, omamobile-shell 0.2.91: fingerprint on by default, guarded by the runner - #358

Merged
derekross merged 19 commits into
mainfrom
fingerprint-out-of-the-box
Nov 3, 2026
Merged

derekross merged 19 commits into
mainfrom
fingerprint-out-of-the-box

Conversation

@derekross

@derekross derekross commented Nov 2, 2026 •

Copy link
Copy Markdown
Owner

Fingerprint works out of the box. This follows the maintainer's 2026-11-02 decision: no consent step and no installer paragraph. The safety lives in the runner.

What changes

omamobile-fingerprint 0.1.14

The package now enables omamobile-fp itself, with a vendor wants link, so the only runner that can be enabled by default is this one, with its guards. The unit stays inert until tools/phone-provision has pushed Google's files, then starts at the reboot.

omamobile-fp-chroot preflight runs before storageproxyd:

  • Read-only copies: mfg_data is copied with dd, and persist:/ss/0 and /ss/nsp with debugfs, so persist is never mounted. Both partitions must read ro=1, and persist must not be mounted read-write. Each copy's sha256 goes into provenance. An existing mfg_data.img is kept only if it matches the partition; otherwise it is moved aside and copied again.
  • TDP refusal: the runner refuses to start when the TDP copy is missing, which it never recreates by itself. It also refuses when the copy or persist:/ss/0 differs from the recorded hash.
  • One storage proxy per boot: a /run marker blocks a second proxy. The marker holds a token, and the inner (namespace) start requires it, so OMAMOBILE_FP_IN_NS=1 alone can't skip preflight.
  • A start that never got running: if the last boot's start didn't reach running (for example, Trusty halted, taking the AoC with it), the next boot skips fingerprint once.
  • Test overrides: the overrides for the partition and marker paths only work with OMAMOBILE_FP_TEST=1.

Stopping and restarting:

  • The unit has no Restart= and uses KillMode=mixed.
  • Stop order: the bridge and HAL, then KeyMint and Gatekeeper, then storageproxyd. An EXIT trap keeps that order if the stack dies inside the namespace.
  • B1 fix: alive() now reads the namespace's procfs ($ROOT/proc). The test in test/all runs in a real PID namespace and fails on the old code.
  • post_upgrade sets the marker when an older runner's proxy is still running.

test/preflight covers each guard against a fake tree.

tools/install-phone, omamobile-firstboot: back to main's code: no new keys, no new paragraph, no version bump. A new installer still works with an older root image.

omamobile-shell 0.2.91: onboarding offers "Add a fingerprint" once, after the PIN step, when the sensor is ready. When it isn't, onboarding now tells three cases apart: the unit failed ("unavailable on this boot"), Google's files aren't there yet, or it's still starting.

tools/phone-provision: --fingerprint restores only the calibration and persist-ss/. --restore-td is opt-in, under the rule that no storageproxyd ran since the backup.

Docs:

  • FINGERPRINT-PLAN F19: the runbook and the 2026-11-02 results, plus status, the rules and the T1 checks. §4 now says persist is read with debugfs, and F4 carries a note that F19 supersedes it.
  • AGENTS.md: the fingerprint rule now matches what happens.
  • NOTES.md, INSTALL.md, DEVICES.md and the release-notes draft are updated, worded plainly.

Versions

omamobile-fingerprint 0.1.14 and omamobile-shell 0.2.91. Main is at 0.2.90 after #366. omamobile-firstboot is unchanged.

Phone checks before release (FINGERPRINT-PLAN F19)

  • P1, upgrade on the 7 Pro: install 0.1.14, then reboot. The journal should show "the existing TDP copy matches persist:/ss/0", provenance with persist-ss/0 at e0e1990d…, one storageproxyd, and "stage bridge running". A finger unlocks, and phone-smoke reports sound and sensors fine.
  • P2:
    • systemctl stop omamobile-fp should take a few seconds and follow the order above. systemctl start is then refused with "reboot", and the phone starts it again after a reboot.
    • Time a systemctl reboot: it should be no slower than on main.
  • T1, a fresh install: the unit is enabled and skipped until provisioning. Then: the copies and their hashes, Trusty's log showing tp/tdp/nsp loaded and td "clear requested", enrollment from onboarding with "write initial superblock", and unlocking. After a day of use, reboots and sleep cycles, TDP is unchanged and F19.0 reads SAME. With persist-ss/0 removed, the start is refused.
  • T2, back to stock, and T3, a reinstall after stock: as written in F19.

🤖 Generated with Claude Code

https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d

derekross and others added 8 commits November 2, 2026 17:50
…ne start per boot

Fingerprint is to work out of the box on a new install, so the runner, not
a person following a runbook, now carries the safety that F3/F4 gave:

- The storage stage refuses without the owner's install-time
  acknowledgement (/etc/omamobile/fingerprint.conf, CONSENT=yes; a missing
  file or key is no), unless this install's secure storage already exists
  (the bring-up phones, whose first run had the human's go).
- At the first start it makes the read-only copies itself: mfg_data with dd,
  persist:/ss/0 and /ss/nsp with debugfs (no mount, no journal replay), both
  partitions required ro=1 and persist not mounted read-write, and records
  their sha256 in /var/lib/omamobile-fp/provenance. Until now nothing on a
  new install made these copies, so a start served an empty TDP against
  stock's RPMB anchor.
- No storageproxyd without the TDP copy, and none once the copy or
  persist:/ss/0 differs from the recorded hash. This refuses rather than
  warns: once Linux (or a stock boot) has written TDP, every further start
  moves the RPMB anchor further from what stock would read, and only the
  maintainer can judge what is safe then. The design note left this open;
  refusing keeps the state as it was found.
- One storage proxy per boot: a marker in /run refuses a second start with
  "reboot". On 2026-11-02 a proxy stopped and restarted twice between TD's
  clear and its first commit asserted Trusty's storage app
  (block_get_cleared_super) and halted Trusty. The unit drops Restart= and
  the start limit, uses KillMode=mixed, and the runner stops the bridge and
  HAL first, KeyMint and Gatekeeper next, storageproxyd last.

test/preflight runs each case against a fake tree.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…ite, the codename acknowledges it

The maintainer decided on 2026-11-02 that fingerprint works out of the box,
with the install's one acknowledgement as the consent: no separate opt-in
later. For a device whose ROOTFS_PACKAGES has omamobile-fingerprint, the
warning before the codename gains one paragraph in plain words: the first
start writes Trusty's area of the RPMB, can't be undone, is the same kind of
write as an Android factory reset, touches neither persist, mfg_data, the
Titan M2 nor slot A, and going back to stock works as usual with fingers
enrolled again there.

The acknowledgement is recorded in firstboot.conf as FINGERPRINT=yes and
ACK_DATE (UTC). The file only reaches the phone after the codename is
typed. --no-fingerprint (developers) records FINGERPRINT=no; a device
without the package gets neither key. --help now prints every option.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…print's consent record

FINGERPRINT=yes (with ACK_DATE) in firstboot.conf is the owner's
acknowledgement of fingerprint's secure-storage write, typed in the
installer. Firstboot turns it into /etc/omamobile/fingerprint.conf
(CONSENT=yes, CONSENT_SOURCE=install-acknowledgement, CONSENT_DATE), the
record omamobile-fp-chroot requires before its first storage start, and
enables omamobile-fp.service. Nothing is started: the unit stays inert
until tools/phone-provision has pushed Google's files, and starts at the
reboot INSTALL.md already asks for.

A missing FINGERPRINT key (an older installer) is no, never consent;
FINGERPRINT=yes without a date, a date without yes, or any other value is
refused. The step is essential, so a failed write keeps firstboot.conf for
the next boot rather than losing the acknowledgement.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…store-td

--fingerprint restored the whole backup tree, Trusty's TD (data/vendor/ss)
and the bridge's Gatekeeper secret included, from a backup of any age. TD's
superblock lives in the RPMB, and every storageproxyd run moves it on, so a
TD older than the RPMB mounts but its files can't be read: the 10-13/10-14
backups were three weeks older than the RPMB at the 2026-11-02 reinstall,
and restoring them would have left a half-broken store.

By default --fingerprint now copies only the safe pieces, mfg_data.img and
the read-only persist:/ss copy (and the runner makes both itself on a new
install). --restore-td adds TD and the secret, states the rule that no
storageproxyd may have run between the backup and the restore, and refuses
to replace a TD already on the phone without --force.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…he PIN

With fingerprint on by default for new installs, setup should offer it.
After the PIN step, when the sensor is ready (fprintd lists the device
through omamobile-fp-enroll) and this user has no finger, onboarding asks
once: "Add a fingerprint" opens the fingerprint screen's enrollment (the
LHBM circle and polkit's PIN prompt), "Skip" says where to add one later
(Setup > Security > Fingerprint). On a new install the sensor is ready only
after tools/phone-provision and its restart, so setup then says it comes
later, and the first login after it is ready sends one notification
instead. Never before a PIN, never with a finger enrolled, never twice.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…ngerprint on by default

F19 records the 2026-11-02 reinstall and turns it into the runbook: why an
empty TD is safe (Trusty clears it, like a factory reset), why an old TD
isn't (the RPMB anchor moved on), and why TDP must be served from a fresh
read-only copy of persist:/ss/0. Corrected after the run: F19.4 to the
enrollment is one storageproxyd session (one start of the service, not the
stage-by-stage starts), a stop condition makes any proxy stop or restart in
the boot mean a reboot before enrolling, and the first enrollment should
show "write initial superblock".

Results on 7.3.0-rc6-omamobile-g2219cc6cda77: F19.0-F19.3 passed (TDP copy
e0e1990d, nsp 078a4f9c, mfg_data matching 10-13); F19.4 loaded tp, tdp and
nsp and cleared td; the first enrollment, after two stage restarts, asserted
Trusty's storage app and halted Trusty; after a reboot one continuous run
enrolled right-index-finger and matched; TDP unchanged, radio partitions
SAME. The rules section and status describe the default-on path and the
install acknowledgement as consent; the fresh-install checks T1-T3 are
listed as still to run. NOTES.md has the day's entry.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…rst storageproxyd run

The maintainer decided on 2026-11-02 that the owner's recorded install-time
acknowledgement (the installer's warning, the typed codename, firstboot's
/etc/omamobile/fingerprint.conf) counts as the human's go for that phone's
first storageproxyd run. Every other fingerprint rule stays: persist and
mfg_data are never written, the TDP copy stays read-only, and a phone
without the record still needs the human's go. Added: one storage proxy per
boot (reboot, never restart), never restore an old TD, and phone-provision's
new --fingerprint/--restore-td split.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…er provisioning

INSTALL.md replaces "not available on a new install yet" (#354) with what
happens now: after the provisioning restart the phone offers "Add a
fingerprint", with the install warning's risk text (the secure-storage
write, the same kind as a factory reset, nothing to persist, mfg_data, the
Titan M2 or slot A; fingerprint stays off if the read-only copies can't be
made or stop matching; restart the phone rather than the service). Going
back to Android says fingers are enrolled again on each side. DEVICES.md
and the release-notes draft follow.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
derekross and others added 5 commits November 2, 2026 18:30
# Conflicts:
#	packages/omamobile-shell/version
…a guards kept and hardened

The maintainer decided (2026-11-02) that fingerprint simply works out of the
box: no consent record. The runner's consent check (and its "TD already
exists" exception) goes, and the package enables omamobile-fp itself with a
vendor wants link, so only a runner with the guards is ever enabled by
default. The unit stays inert until tools/phone-provision has pushed
Google's files.

Kept: the read-only copies of mfg_data and persist:/ss/{0,nsp} (debugfs, no
mount, ro=1 checks), the recorded hashes, the refusal when the TDP copy is
missing or differs or persist:/ss/0 changed, one storage proxy per boot, no
Restart=, the stop order.

Fixed and hardened, from the round-1 review of #358:
- alive() read the host's /proc for namespace PIDs, so the ordered stop
  waited on unrelated host processes (or not at all). It now reads the
  namespace's procfs at $ROOT/proc; test/all checks it in a real PID
  namespace (and fails on the old code).
- The inner start refuses without the token this boot's preflight wrote
  into the /run marker, so OMAMOBILE_FP_IN_NS=1 can't skip the checks; the
  path overrides of the guard's inputs need OMAMOBILE_FP_TEST=1.
- An EXIT trap in the namespace keeps the stop order on a die.
- An existing mfg_data.img is adopted only if it matches the partition;
  otherwise it is moved aside and copied again.
- A start that never reached "running" (Trusty halted, which also takes
  the AoC's sound and sensors) leaves a marker in $STATE: the next boot
  skips fingerprint once, the one after tries again.
- post_upgrade sets the per-boot marker when an older runner's storage
  proxy is still running, so the upgrade boot can't start a second one.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…ter all

Fingerprint now works out of the box without a consent record (the
maintainer's decision, 2026-11-02): the installer's warning loses the
fingerprint paragraph and --no-fingerprint, firstboot.conf gains no
FINGERPRINT/ACK_DATE, and firstboot no longer writes fingerprint.conf or
enables the unit (omamobile-fingerprint enables it itself). Both are back to
main's, so a new installer and an older root image stay compatible, and the
firstboot bump to 0.1.4 is dropped.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…le, not "after the firmware"

When the sensor isn't ready, setup now tells the three cases apart: the unit
failed or was refused on this boot ("unavailable on this boot", with the
journal to read, and the PIN unlocks), Google's files aren't there yet
("ready once your phone's firmware is added"), or it is still starting
(offered at the next login). Before, a refused unit was blamed on missing
firmware. Renumbered from 0.2.85: main is at 0.2.86, #362 holds 0.2.88, and
0.2.87 is left for #357.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
… runner's guards keep stock's data safe

The consent step is gone (the maintainer's decision, 2026-11-02). AGENTS'
fingerprint rule now says what is true: omamobile-fp starts by default, its
storageproxyd writes Trusty's TD superblocks at every boot, and the runner's
guards (read-only copies, the TDP refusal, one proxy per boot) are what keep
stock's persistent data safe; never write persist or mfg_data, never restart
storageproxyd in a boot, never restore an old TD.

FINGERPRINT-PLAN: status, the rules, F19's default-on flow and the T1 checks
follow (the package enables the unit; a stop takes seconds in order; the
refused restart). §4 no longer says persist is read through a read-only
mount (it is debugfs), and F4 notes that the stage-by-stage starts are
superseded by one proxy per boot. INSTALL and the release notes say plainly
that fingerprint unlock works after provisioning, and that fingers are
enrolled again after a reinstall or on Android, without claims beyond what
has run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
@derekross derekross changed the title Fingerprint out of the box: install acknowledgement as consent, a safe runner, onboarding (fingerprint 0.1.14, firstboot 0.1.4, shell 0.2.85) omamobile-fingerprint 0.1.14, omamobile-shell 0.2.89: fingerprint on by default, guarded by the runner Nov 2, 2026
derekross and others added 5 commits November 2, 2026 18:59
…, data-check refusals named

From the round-2 review of #358:
- test/all removed $nsroot/proc with a bare rmdir after the SKIP branch, so
  where user namespaces are off (CI) set -e ended the script: rm -rf instead.
- The device and state overrides now need OMAMOBILE_FP_TEST=1 like the path
  overrides (the device comes from the board's compatible string; a forced
  "tegu" on a 7 Pro would skip the copies and hash checks). A read-only
  `paths` subcommand shows what is in use, and the test checks it instead of
  running preflight on the real paths.
- The once-per-boot marker is created with noclobber (O_EXCL), so two
  racing starts can't both pass.
- The inner start takes only a 32-hex token; post_upgrade's marker for an
  older runner's proxy carries none, so it only blocks.
- Data-check refusals (the TDP copy, persist:/ss/0, ro=1, the mounts) write
  /run/omamobile-fp-ui/off-reason, since they recur at every boot; onboarding
  says so instead of "unavailable on this boot".
- Tests now catch the three mutations that survived: the start-pending
  clear after "stage running", the namespace's EXIT trap, and post_upgrade's
  marker (run against a stub pgrep).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
A refusal of the runner's data checks (the TDP copy or persist:/ss/0 no
longer matching) recurs at every boot, so onboarding says fingerprint is off
until it is looked at, rather than "unavailable on this boot". It reads the
runner's /run/omamobile-fp-ui/off-reason.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
…rds per phone, plain wording

- The vendor wants link survives `systemctl disable`, so the rollbacks
  (cheetah FINGERPRINT-PLAN F10, FINGERPRINT-RESEARCH, tegu F10) and AGENTS
  say `systemctl mask omamobile-fp`, and INSTALL tells owners how to keep it
  off before the provisioning restart.
- AGENTS names the guards per phone (the 7 Pro's copies, hashes and
  persist:/ss/0 check; the 9a's tdp.img check; once per boot on both) and
  says storageproxyd "can write" TD superblocks at every start.
- INSTALL and the release notes say fingerprint unlock starts automatically
  after provisioning; on Android, its setup asks for fingerprints as usual.
  DEVICES marks the automatic path as being tested.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
# Conflicts:
#	docs/release-notes/2026-11-DRAFT.md
#	packages/omamobile-shell/version
derekross added a commit that referenced this pull request Nov 3, 2026
… the Open-Meteo copies

From the review of #366:

- Today showed visibility about 3.3 times too large in imperial mode:
  with precipitation_unit=inch Open-Meteo answers visibility in feet
  (hourly_units.visibility "ft"). Today now converts from the unit the
  answer reports; wind and rain come in the units asked for and pressure
  is always hPa. The test has a feet fixture and a metres one.
- "Town, Region" now picks the place as omamobile-weather's pick_place
  does (each part after a comma, US state codes read as states), since
  the lock screen reads Today's city too: "Springfield, IL" is Illinois
  in both. The test runs pick_place on the same answers.
- Not every copy asks for approval again: approvals compare command
  strings, and air and pollen and planes overhead kept theirs. The
  package README and the owl-review note now say which five ask.
- overrides/LICENSE carries the store's MIT licence, and each copy's
  first line names the store file and commit it was modified from.
- Omaprayer's "auto" asks for a city instead of showing Jakarta's
  prayer times.
- A comment says why the geocoding and first forecast requests always
  go out (the fixed host the approval sheet names).

omamobile-shell takes 0.2.90: main has 0.2.88 (#362) and #358 holds
0.2.89.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
# Conflicts:
#	packages/omamobile-shell/version
@derekross derekross changed the title omamobile-fingerprint 0.1.14, omamobile-shell 0.2.89: fingerprint on by default, guarded by the runner omamobile-fingerprint 0.1.14, omamobile-shell 0.2.91: fingerprint on by default, guarded by the runner Nov 3, 2026
@derekross
derekross merged commit cd293d6 into main Nov 3, 2026
2 checks passed
@derekross
derekross deleted the fingerprint-out-of-the-box branch November 3, 2026 00:54
derekross added a commit that referenced this pull request Nov 3, 2026
…0: review round 2, the modem-up guard bounded

The review of #398 asked for three changes and a few corrections.

The self-clearing modem-up guard was unbounded: a phone that hangs
because the modem is up went up, hang, refused, up, hang, refused, for
ever, with only a tile line on the boots in between. Now the refused
copy the script already keeps is the counter: one automatic retry, then
a hold. refuse_unclean keeps the marker and refuses "(held)" when a
modem-up.refused-* copy exists (no clean stop since the last refusal);
--stopped on this boot's marker removes the copies, so a clean stop
resets the count; --retry clears both markers and the copies. The test's
"four unclean boots in a row" block becomes "the second refusal holds, a
clean stop resets", and M4.6b gets drill B9b for it.

systemctl is-enabled reads "disabled" for a unit started by a vendor
wants link alone (it reports /etc and /run links only), so B6's pass and
stop lines expected the wrong answer and would have stopped a correct
install; B6 now expects "enabled" before and "disabled" after the
disable, and proves the link with ls -l of the /usr/lib link and
systemctl list-dependencies --plain multi-user.target. The README, the
target's header and the install message say the same, so nobody reads
"disabled" and skips the mask. The read-only NV checksum check (M3.5)
opens B6 as well as closing the runbook (AGENTS.md section 1). B8's
is-active expectation is "failed inactive inactive active" (mmd runs).

omamobile-shell takes 0.2.100: main is 0.2.98 after #397 and #373 is
taking 0.2.99. The DEVICES cheetah fingerprint cell reads "works": #358
made it on by default.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
derekross added a commit that referenced this pull request Nov 3, 2026
…0: the Pixel 7 Pro's modem stack on by default (#398)

* omamobile-modem 0.1.16, device-cheetah 0.1.43, omamobile-shell 0.2.98: the Pixel 7 Pro's modem stack on by default

The maintainer's decision of 2026-11-03: the next release ships the modem
working out of the box (calls, texts and data on the tested network, M4.6
through three reboots). The switch of M4.6, omamobile-modem-boot.target,
is now on by default on cheetah, and a boot that can't or shouldn't
start the modem costs as little as possible:

- omamobile-device-cheetah 0.1.43 enables the target with a vendor wants
  link under multi-user.target.wants, as it does for the AoC and Wi-Fi
  units; the four units the target wants get no link of their own, and
  omamobile-modem itself still enables nothing (its test holds it to
  that; tegu and raven are untouched). Off: systemctl mask the target.
  The stack's packages (omamobile-modem-sit, omamobile-mmd) join
  cheetah's ROOTFS_PACKAGES so the release image carries them.
- A phone that isn't provisioned costs nothing: the modem and inhibit
  units carry ConditionPathExists= for the chroot, the NV checksums and
  the board id (what tools/phone-provision puts there), so both are
  skipped, not failed, and the inhibitor isn't held. The inhibitor needs
  its own copy because the modem unit's Requires= pulls it in even when
  that unit's conditions skip it. sitd skips itself without
  /dev/umts_ipc0; sit-data now does too, through a device drop-in:
  without it a condition-skipped sitd counted as started for sit-data's
  Requires=, and the fallback would have failed after 60 s and restarted
  every 10 s all boot. The start script checks the same three paths
  first, for a start by hand.
- The modem-up guard clears itself after one refused boot: a stale
  marker still refuses that boot and logs where to look (journalctl -b
  -1), then moves to modem-up.refused-<time>-<pid> (the last three kept)
  so the next boot starts the modem again. A phone that hangs with the
  modem up every time loses the modem on every other boot at most, never
  for good. /etc/omamobile/modem-hold restores the old hard stop;
  --retry still clears both markers; the bring-up latch keeps its hard
  stop.
- Every refusal leaves one short line in /run/omamobile/modem-refused,
  and the shell's Mobile data tile shows it under its name instead of
  "No modem" (omamobile-shell 0.2.98, tile-state.sh): no new
  notification path, the journal has the rest.

Versions: #372 and #380 claim device-cheetah 0.1.42, #389 omamobile-modem
0.1.14 and #382 0.1.15, main has omamobile-shell 0.2.97 (#394); hence 0.1.43, 0.1.16
and 0.2.98. Docs: RELEASES (ROOTFS_PACKAGES, decision 12 amended,
checklist step 6), the release notes draft (the modem known issue becomes
what works and what still needs a hand), INSTALL, DEVICES, and MODEM-PLAN
M4.6b with the runbook: install, two reboots, a refused-boot drill by
planting a stale marker, the hold flag.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d

* omamobile-modem 0.1.16, device-cheetah 0.1.43, omamobile-shell 0.2.100: review round 2, the modem-up guard bounded

The review of #398 asked for three changes and a few corrections.

The self-clearing modem-up guard was unbounded: a phone that hangs
because the modem is up went up, hang, refused, up, hang, refused, for
ever, with only a tile line on the boots in between. Now the refused
copy the script already keeps is the counter: one automatic retry, then
a hold. refuse_unclean keeps the marker and refuses "(held)" when a
modem-up.refused-* copy exists (no clean stop since the last refusal);
--stopped on this boot's marker removes the copies, so a clean stop
resets the count; --retry clears both markers and the copies. The test's
"four unclean boots in a row" block becomes "the second refusal holds, a
clean stop resets", and M4.6b gets drill B9b for it.

systemctl is-enabled reads "disabled" for a unit started by a vendor
wants link alone (it reports /etc and /run links only), so B6's pass and
stop lines expected the wrong answer and would have stopped a correct
install; B6 now expects "enabled" before and "disabled" after the
disable, and proves the link with ls -l of the /usr/lib link and
systemctl list-dependencies --plain multi-user.target. The README, the
target's header and the install message say the same, so nobody reads
"disabled" and skips the mask. The read-only NV checksum check (M3.5)
opens B6 as well as closing the runbook (AGENTS.md section 1). B8's
is-active expectation is "failed inactive inactive active" (mmd runs).

omamobile-shell takes 0.2.100: main is 0.2.98 after #397 and #373 is
taking 0.2.99. The DEVICES cheetah fingerprint cell reads "works": #358
made it on by default.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d

* omamobile-modem 0.1.16: the start script's test stops after the marker decision

The case "the boot after a refused one starts the modem again" (and the
--retry case) ran lib/omamobile-modem-start past the marker refusals and
relied on the machine to stop it: on the laptop and in CI no sleep
inhibitor is held, so the script refused there before any driver. On the
phone, where tools/phone-install-package builds the package as an
unprivileged user, the inhibitor is held, so the test went on into the
real bring-up and failed on /sys/module/pcie_aspm/parameters/policy
(permission denied) and then on omamobile-vendor's "run as root".

The script takes a test-only OMAMOBILE_MODEM_TEST_STOP_AFTER_CHECKS=1
(documented in its header; no unit or install step sets it) and exits 0
right after the marker decision, before the link driver's sysfs read,
the inhibitor wait, the first modprobe and the profile. The two cases use
it and assert the decision itself: exit 0, the stale marker moved aside
with its refused copy kept, no latch written, the earlier reason left
for the tile (it is removed only once every refusal has passed, which
the stop point comes before). The test also checks that the stop point
sits between the marker decision and the sysfs read, the inhibitor and
the latch, and that no unit sets the variable.

Verified: packages/omamobile-modem/test/all as uid 1000, with and
without a systemd-inhibit stand-in that reports the inhibitor held (the
phone's case); tools/pr-check origin/main.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d

* release notes: the carrier APNs come from the SIM, not by hand

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…by default, guarded by the runner (derekross#358)

Fingerprint unlock is now on by default on the Pixel 7 Pro, with no consent step: omamobile-fingerprint enables omamobile-fp itself (a vendor wants link, so only this guarded runner is ever enabled), and it stays inert until phone-provision has pushed Google's fingerprint files, then starts at the reboot. Turning it off is `systemctl mask omamobile-fp`.

The runner's guards keep stock Android's data safe: at the first start it makes read-only copies of mfg_data (dd) and persist:/ss/0 and ss/nsp (debugfs, persist never mounted), records their hashes, and refuses to start storageproxyd if the TDP copy is missing or changed or persist:/ss/0 changed (it never writes persist or mfg_data). One storageproxyd per boot (a root-only /run token the inner start must present, created with O_EXCL): a proxy restart doesn't reset Trusty storage, and a restart between TD's clear and its first commit crashed Trusty's storage app on 2026-11-02, so a second start is refused until a reboot. No Restart=; the stop order is bridge and HAL, KeyMint and Gatekeeper, storageproxyd last, and alive() reads the namespace's procfs. A start that never reached "running" makes the next boot skip fingerprint once. Onboarding offers "Add a fingerprint" after the PIN step and says why when fingerprint is off. phone-provision --fingerprint restores only the safe pieces; --restore-td is opt-in. FINGERPRINT-PLAN F19 records the 2026-11-02 bring-up after the reinstall.

omamobile-fingerprint 0.1.14, omamobile-shell 0.2.91.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…0: the Pixel 7 Pro's modem stack on by default (derekross#398)

* omamobile-modem 0.1.16, device-cheetah 0.1.43, omamobile-shell 0.2.98: the Pixel 7 Pro's modem stack on by default

The maintainer's decision of 2026-11-03: the next release ships the modem
working out of the box (calls, texts and data on the tested network, M4.6
through three reboots). The switch of M4.6, omamobile-modem-boot.target,
is now on by default on cheetah, and a boot that can't or shouldn't
start the modem costs as little as possible:

- omamobile-device-cheetah 0.1.43 enables the target with a vendor wants
  link under multi-user.target.wants, as it does for the AoC and Wi-Fi
  units; the four units the target wants get no link of their own, and
  omamobile-modem itself still enables nothing (its test holds it to
  that; tegu and raven are untouched). Off: systemctl mask the target.
  The stack's packages (omamobile-modem-sit, omamobile-mmd) join
  cheetah's ROOTFS_PACKAGES so the release image carries them.
- A phone that isn't provisioned costs nothing: the modem and inhibit
  units carry ConditionPathExists= for the chroot, the NV checksums and
  the board id (what tools/phone-provision puts there), so both are
  skipped, not failed, and the inhibitor isn't held. The inhibitor needs
  its own copy because the modem unit's Requires= pulls it in even when
  that unit's conditions skip it. sitd skips itself without
  /dev/umts_ipc0; sit-data now does too, through a device drop-in:
  without it a condition-skipped sitd counted as started for sit-data's
  Requires=, and the fallback would have failed after 60 s and restarted
  every 10 s all boot. The start script checks the same three paths
  first, for a start by hand.
- The modem-up guard clears itself after one refused boot: a stale
  marker still refuses that boot and logs where to look (journalctl -b
  -1), then moves to modem-up.refused-<time>-<pid> (the last three kept)
  so the next boot starts the modem again. A phone that hangs with the
  modem up every time loses the modem on every other boot at most, never
  for good. /etc/omamobile/modem-hold restores the old hard stop;
  --retry still clears both markers; the bring-up latch keeps its hard
  stop.
- Every refusal leaves one short line in /run/omamobile/modem-refused,
  and the shell's Mobile data tile shows it under its name instead of
  "No modem" (omamobile-shell 0.2.98, tile-state.sh): no new
  notification path, the journal has the rest.

Versions: derekross#372 and derekross#380 claim device-cheetah 0.1.42, derekross#389 omamobile-modem
0.1.14 and derekross#382 0.1.15, main has omamobile-shell 0.2.97 (derekross#394); hence 0.1.43, 0.1.16
and 0.2.98. Docs: RELEASES (ROOTFS_PACKAGES, decision 12 amended,
checklist step 6), the release notes draft (the modem known issue becomes
what works and what still needs a hand), INSTALL, DEVICES, and MODEM-PLAN
M4.6b with the runbook: install, two reboots, a refused-boot drill by
planting a stale marker, the hold flag.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d

* omamobile-modem 0.1.16, device-cheetah 0.1.43, omamobile-shell 0.2.100: review round 2, the modem-up guard bounded

The review of derekross#398 asked for three changes and a few corrections.

The self-clearing modem-up guard was unbounded: a phone that hangs
because the modem is up went up, hang, refused, up, hang, refused, for
ever, with only a tile line on the boots in between. Now the refused
copy the script already keeps is the counter: one automatic retry, then
a hold. refuse_unclean keeps the marker and refuses "(held)" when a
modem-up.refused-* copy exists (no clean stop since the last refusal);
--stopped on this boot's marker removes the copies, so a clean stop
resets the count; --retry clears both markers and the copies. The test's
"four unclean boots in a row" block becomes "the second refusal holds, a
clean stop resets", and M4.6b gets drill B9b for it.

systemctl is-enabled reads "disabled" for a unit started by a vendor
wants link alone (it reports /etc and /run links only), so B6's pass and
stop lines expected the wrong answer and would have stopped a correct
install; B6 now expects "enabled" before and "disabled" after the
disable, and proves the link with ls -l of the /usr/lib link and
systemctl list-dependencies --plain multi-user.target. The README, the
target's header and the install message say the same, so nobody reads
"disabled" and skips the mask. The read-only NV checksum check (M3.5)
opens B6 as well as closing the runbook (AGENTS.md section 1). B8's
is-active expectation is "failed inactive inactive active" (mmd runs).

omamobile-shell takes 0.2.100: main is 0.2.98 after derekross#397 and derekross#373 is
taking 0.2.99. The DEVICES cheetah fingerprint cell reads "works": derekross#358
made it on by default.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d

* omamobile-modem 0.1.16: the start script's test stops after the marker decision

The case "the boot after a refused one starts the modem again" (and the
--retry case) ran lib/omamobile-modem-start past the marker refusals and
relied on the machine to stop it: on the laptop and in CI no sleep
inhibitor is held, so the script refused there before any driver. On the
phone, where tools/phone-install-package builds the package as an
unprivileged user, the inhibitor is held, so the test went on into the
real bring-up and failed on /sys/module/pcie_aspm/parameters/policy
(permission denied) and then on omamobile-vendor's "run as root".

The script takes a test-only OMAMOBILE_MODEM_TEST_STOP_AFTER_CHECKS=1
(documented in its header; no unit or install step sets it) and exits 0
right after the marker decision, before the link driver's sysfs read,
the inhibitor wait, the first modprobe and the profile. The two cases use
it and assert the decision itself: exit 0, the stale marker moved aside
with its refused copy kept, no latch written, the earlier reason left
for the tile (it is removed only once every refusal has passed, which
the stop point comes before). The test also checks that the stop point
sits between the marker decision and the sysfs read, the inhibitor and
the latch, and that no unit sets the variable.

Verified: packages/omamobile-modem/test/all as uid 1000, with and
without a systemd-inhibit stand-in that reports the inhibitor held (the
phone's case); tools/pr-check origin/main.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d

* release notes: the carrier APNs come from the SIM, not by hand

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d

---------

Co-authored-by: Claude Fable 5.1 <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.

1 participant