Skip to content

Lock screen retries fingerprint every 250 ms all night when no finger is enrolled #170

Closed

Description

@jkjoplin

On a Pixel 7 Pro with no finger enrolled (the fingerprint opt-in not done, omamobile-fp disabled), the lock screen starts a pam_fprintd attempt about 3.4 times a second for as long as the phone is locked. Every attempt fails at once, and the lock retries 250 ms later.

On this phone it ran all night. One boot, from 2026-10-28 01:05 to 12:05, logged 130,164 failed attempts: about 12,400 an hour from 02:00 to 10:00, all with the same error:

quickshell.service.pam.subprocess: Starting pam session for user "<user>" with config "omarchy-lock-fingerprint" in dir "/etc/pam.d"
quickshell.service.pam.subprocess: Error while authenticating: "Authentication service cannot retrieve authentication info" (code 9)

Each attempt also reaches fprintd and polkit. A 10 s D-Bus capture showed fprintd receiving GetDevices and ListEnrolledFingers 35 times, each followed by polkit's credential lookups. polkitd had used 2 minutes of CPU in 72 minutes of uptime.

Versions: omarchy 4.0.6-1, omamobile-fingerprint 0.1.8-1, omamobile-omarchy 0.1.12-1, omamobile-shell 0.2.32-1, fprintd 1.94.5-2, libfprint 1.94.100-1.6, kernel 7.3.0-rc6-omamobile-push1-g1cb4bd36f3be. omamobile-fingerprint 0.1.9 to 0.1.11 on main don't change the PAM file, the PKGBUILD or this part of install.

device/cheetah/NOTES.md (2026-10-17) records the same retry from another trigger. There fprintd refused the lock after a sleep, and the lock retried "about three times a second with PAM_AUTHINFO_UNAVAIL". libfprint 1.94.100-1.5 fixed that cause. This phone has 1.94.100-1.6, so this is a different trigger, through the same retry.

Why the lock thinks fingerprint is set up

Omarchy 4.0.6's lock decides fingerprintConfigured with this check (shell/plugins/lock/Service.qml:391 at v4.0.6):

if [[ -f /etc/pam.d/omarchy-lock-fingerprint ]] && command -v fprintd-list >/dev/null 2>&1 && fprintd-list "$USER" 2>/dev/null | grep -qi finger; then echo yes; else echo no; fi

With nothing enrolled, fprintd-list prints:

found 1 devices
Device at /net/reactivated/Fprint/Device/0
Using device /net/reactivated/Fprint/Device/0
User <user> has no fingers enrolled for Pixel 7 Pro under-display fingerprint sensor (Omarchy Mobile).

"has no fingers enrolled" matches grep -qi finger, so the check prints yes. Run as the user exactly as the lock runs it, it printed yes.

The retry: pam_authenticate's code 9 (PAM_AUTHINFO_UNAVAIL) reaches fingerprintPam.onError, which restarts fingerprintRetryTimer (lines 360-364). A failed match goes through handleFingerprintFinished instead (lines 230-238). Both restart the timer, which has interval: 250 (lines 368-369), so it retries for as long as the lock is up.

On a Pixel 7 Pro installed from the image, the first two parts of that check always hold:

  • omamobile-fingerprint installs /etc/pam.d/omarchy-lock-fingerprint unconditionally (install, line 24, still so at 0.1.11);
  • it depends on fprintd.

So only the grep decides. Upstream's 879d6583d (below) describes the desktop side of the same bug: "The lock screen and omarchy-apply-lock decided fingerprint was set up with fprintd-list | grep -qi finger, which also matches "has no fingers enrolled"…".

Your phone has a finger enrolled (FINGERPRINT-PLAN's status: "the lock screen unlocks with a finger"), so this false positive doesn't reach it.

What it costs

Measured on this phone with the cable out, the screen off and Wi-Fi connected. There were two 10-minute runs a few minutes apart, logging every 10 s. Between them:

  • the only configuration change was moving the PAM file aside;
  • the cable was briefly back in, and the screen woke once.

Both runs use the same rule: the first sample after the unplug is dropped and screen-on samples are left out (2 in the first run, 1 in the second). The charge-counter figure is a linear fit.

Loop running PAM file moved aside
Charge-counter slope 479 ± 2 mA 424 ± 2 mA
Mean current_now 480 mA 418 mA
CPU busy (/proc/stat) about 4-5 % under 1 %
Interrupts per second 2040 820
Fingerprint attempts per 10 s 34.4 0

That is about 55-62 mA, roughly 12 % of the phone's draw.

The phone doesn't sleep by itself yet, so the loop runs for as long as the lock is up. That night, upower's history went from 98 % at 01:07 to 0 % at 11:58 (10 h 51 min). That is roughly 445 mA on average, if its percentages are shares of the gauge's 4.9 Ah charge_full.

Upstream Omarchy has fixed it, in the 4.5 line only

Both commits are on quattro and in the v4.5.0-rc1 and -rc2 pre-releases. Neither is in a stable release: v4.0.6 is the latest, and GitHub's compare against v4.0.6 reads "diverged" for both.

  • omacom/omarchy@879d6583d (2026-10-04, "Fix fingerprint enrollment and lock-screen recovery", Fix fingerprint enrollment and lock-screen recovery omacom/omarchy#7158).
    • It adds FingerprintModel.js classifyProbe(): yes only for an enrolled-finger line (- #0: …), no for "has no fingers enrolled", unknown otherwise.
    • The probe becomes LC_ALL=C fprintd-list "$USER" 2>&1, read by applyFingerprintProbe().
    • It backs attempts off from 1 s doubling to 40 s when they never reach the reader or fail fast. Mismatches keep 250 ms.
    • It fixes omarchy-apply-lock's test the same way.
    • Upstream's tests feed the "has no fingers enrolled" text to both the classifier and the lock.
  • omacom/omarchy@765577331 (2026-10-14, "Leave the fingerprint reader alone while the lock has the displays off", Leave the fingerprint reader alone while the lock has the displays off omacom/omarchy#15294).
    • It starts no new attempt once the lock's own blank turns the displays off, and any wake resumes them.
    • It builds on 879d6583d's retry code, so it can't be applied to 4.0.6 alone.
    • I haven't checked whether the phone's screen-off goes through the lock's blank.

Workaround

Move /etc/pam.d/omarchy-lock-fingerprint out of /etc/pam.d, keeping a copy to put back when opting into fingerprint.

  • The lock's check then prints no, and beginLock() runs it at every lock.
  • Here the attempts stopped without a re-lock: 0 from the first sample after the move. Quickshell refuses to start a PAM context whose config file is missing, so start() returned false, and startFingerprint() schedules no retry when start() fails. The journal shows it once:
    ERROR: Cannot start PamContext(0x…) because specified config file "/etc/pam.d/omarchy-lock-fingerprint" is not a file.
    
  • An upgrade of omamobile-fingerprint will probably bring the file back.

A fix here

  1. A bridge patch in omamobile-omarchy, until the phone's Omarchy has 879d6583d. It would backport the probe classification: classifyProbe(), applyFingerprintProbe() and the LC_ALL=C fprintd-list … 2>&1 probe. Perhaps also the error backoff. That alone ends the loop on phones with nothing enrolled. The full 879d6583d is a 17-file change.
  2. Or move the PAM file out of omamobile-fingerprint and into the enrolment flow. For example, omamobile-fp-enroll would write it once ListEnrolledFingers returns a finger. This changes a documented decision: TOUCH-PLAN 0.1 has omamobile-fingerprint own the file, and omamobile-firstboot and lib/lock-pam keep omarchy-apply-lock away from it. So I'd do 1 first.

I can send a PR for 1, with a test that feeds the lock's probe the "has no fingers enrolled" text.

Activity

  1. derekross commented on Oct 30, 2026

    @derekross
    Owner

    Thank you, @jkjoplin. This is a superb report: the cause, the line numbers, the upstream fix and an A/B measurement. #171 backports the probe half of upstream's 879d6583d into our Omarchy patch series (omamobile-omarchy 0.1.15), so a phone with no finger enrolled no longer counts as set up and the lock stops retrying. If you can, it would help to try it on your phone with nothing enrolled before I merge.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions