Reporting a working configuration for 2541:fa03, in case it helps narrow
down the regression.

Hardware: Realtek MoC sensor 2541:fa03, reported as "Realtek USB2.0
Finger Print Bridge (press)", built into a Minisforum AI X1 Pro.

System (UEFI/Secure Boot): Linux Mint 22.3 kernel:7.1.8-waydroid (Ubuntu
24.04 base), GCC 13.3.0.

Packages from the archive:
  fprintd 1.94.3-1
  libpam-fprintd 1.94.3-1
  libfprint-2-2 1:1.94.7+tod1-0ubuntu5~24.04.8

The ID 2541:fa03 was added upstream in December 2025 and first shipped
in libfprint 1.94.10, so the 1.94.7 package in noble does not detect the
device at all.

What I did: built the upstream tag v1.94.100 (released 2026-07-25) with
meson into --prefix=/usr/local and kept the archive fprintd. fprintd
picks up the self-built library rather than the archive one:

  $ ldd /usr/libexec/fprintd | grep fprint
  libfprint-2.so.2 => /usr/local/lib/libfprint-2.so.2

  $ sudo G_MESSAGES_DEBUG=all /usr/libexec/fprintd -t
  libfprint-context-DEBUG: Initializing FpContext (libfprint version 1.94.100)

Result: works. Enrollment through the desktop fingerprint dialog
completed for two fingers, fprintd-list shows the device and both
prints, verification matches. No segfault, no aborted enrollment.

  $ fprintd-list $USER
  found 1 devices
  Device at /net/reactivated/Fprint/Device/0
  Fingerprints for user <user> on Realtek USB2.0 Finger Print Bridge (press):
   - #0: left-index-finger
   - #1: right-thumb

The bundled test suite also passes for this driver. "meson test -C build
--suite drivers" gives 32/32 OK, including driver-realtek and driver-
realtek-5816. The realtek test exercises probe, open, clear-storage,
enroll, list, verify, identify, delete and close against the recorded
USB traces, all with error: none, and it runs under G_DEBUG=fatal-
warnings with MALLOC_PERTURB_ set, so a use-after-free or a GLib warning
in that path would have failed the test.

This does not prove the 1.95.1 crash is fixed, since I have not
reproduced it here to begin with. But it does suggest the enrollment
path in the realtek driver is sound as of 1.94.100, and that the problem
may sit in something 1.95.0 changed around it (SDCP support and the
reworked TOD driver API) rather than in the driver itself. The 1.94.100
release notes do list "synaptics, realtek, vfs101, vfs301, upekts, elan
and egismoc: Fix memory safety, bounds checking and error handling
issues", so a targeted fix in that series is also possible.

One caveat about my environment: GLib comes from a third-party PPA and
is slightly newer (glib-compile-resources --version 2.89.3) than what
24.04 ships, so this is not a stock system.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2151131

Title:
  Fingerprint enrollment segfaults on Realtek 2541:fa03 since libfprint
  1.95.1

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/libfprint/+bug/2151131/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to