Skip to content

Detect Broadcom fingerprint readers by vendor ID - #8486

Open
gbillium143 wants to merge 1 commit into
omacom:quattrofrom
gbillium143:fix-broadcom-fingerprint-detection
Open

gbillium143 wants to merge 1 commit into
omacom:quattrofrom
gbillium143:fix-broadcom-fingerprint-detection

Conversation

@gbillium143

Copy link
Copy Markdown

Summary

  • omarchy-hw-fingerprint's vendor allowlist doesn't include Broadcom (0a5c), so Broadcom fingerprint readers are never detected — the wizard bails with "No fingerprint sensor detected" even though fprintd/libfprint-tod fully support the device once installed.
  • Broadcom readers report a bare model number as their product string (e.g. "58200"), so they can't fall back to the product-name match either — they only match on vendor ID.
  • Adds 0a5c to fingerprint_vendors and a regression test mirroring the existing vendor-match coverage.

Fixes #8483

Test plan

  • ./test/shell.d/hw-fingerprint-test.sh — all cases pass, including the new Broadcom case
  • ./test/all — same 5 pre-existing failures as on a clean quattro checkout (environment-dependent: missing omarchy-pkgs checkout, sandbox permission bits, snapper/pacman DB access), unrelated to this change
  • Verified on hardware: a Broadcom sensor (idVendor=0a5c, idProduct=5843) at /sys/bus/usb/devices/3-8, previously rejected by the detector, now passes; ran the full omarchy setup security fingerprint flow end to end (enroll, verify, PAM for sudo/polkit/lock screen) successfully with this fix applied.

Broadcom (0a5c) sensors report a bare model number as their product
string (e.g. "58200"), so they never match the product-name branch
and the detector reports no reader present. Add 0a5c to the vendor
allowlist so these devices are recognized the same way as the other
vendor-only matches.

Fixes omacom#8483
@gbillium143
gbillium143 marked this pull request as ready for review August 27, 2026 00:10
Copilot AI balanced review requested due to automatic review settings August 27, 2026 00:10

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Adds Broadcom vendor detection so supported fingerprint readers with numeric product strings are recognized.

Changes:

  • Adds Broadcom vendor ID 0a5c to the fingerprint allowlist.
  • Adds regression coverage for Broadcom reader detection.

Tip

If you aren't ready for review, convert to a draft PR.
Click "Convert to draft" or run gh pr ready --undo.
Click "Ready for review" or run gh pr ready to reengage.

Reviewed changes

Copilot reviewed 1 out of 2 changed files in this pull request and generated no comments.

File Description
bin/omarchy-hw-fingerprint Recognizes Broadcom fingerprint readers by vendor ID.
test/shell.d/hw-fingerprint-test.sh Tests detection with a numeric product string.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

@omarchybot

Copy link
Copy Markdown
Collaborator

Reviewed on a disposable VM: test/shell.d/hw-fingerprint-test.sh (12 passed) and test/shell.d/fingerprint-invitation-test.sh (1 passed), plus a purpose-built probe that builds fake sysfs trees for real Broadcom hardware and runs the detector against them. Second opinion from Codex at xhigh reasoning; where it agreed with a conclusion already reached its independence is not currently guaranteed, so I have marked the one thing it contributed on its own.

The obvious worry about this change turns out not to fire, and that is worth saying plainly. 0a5c is overwhelmingly a Bluetooth vendor — 51 of the 81 0a5c product entries in usb.ids are Bluetooth controllers, against 5 that mention a fingerprint sensor — so the fear is that every Dell, HP and Lenovo laptop with a BCM20702 starts getting nagged to set up a reader it does not have. It does not happen. btusb binds at least one interface on all of them, so has_kernel_driver rejects them. Verified against fake sysfs trees for 0a5c:21e8, 0a5c:216f (with two unbound vendor-specific interfaces alongside the bound ones, to check that one bound interface is enough), 0a5c:21e1, and the 0a5c:4500 hub: all rejected. The guard added in 6ebdfa3 for Synaptics does its job here too.

Where it does misfire is Dell ControlVault configurations that have no fingerprint sensor. 0a5c:5800 (BCM5880 Secure Applications Processor), 0a5c:5803 (with secure keyboard) and 0a5c:5832 (Smartcard reader) are all detected as fingerprint readers. Linux has no in-kernel CCID driver — pcsc-lite drives smartcard readers from userspace over libusb — so nothing binds at idle. Codex added the half I had missed: when pcscd is holding the reader, the interface shows the synthetic usbfs driver, which 970ec26 deliberately makes not count, so these match whether the smartcard is in use or not. Verified for all three ids. On those machines the result is a menu entry, a first-run notification, and a wizard that cannot succeed.

The bigger issue is what happens after detection succeeds. Stock libfprint has no Broadcom driver at all. I installed libfprint 1.94.100 on a clean worker and enumerated its driver table: aes*, egis0570, elan, elanmoc, elanspi, etes603, focaltech_moc, goodixmoc, nb1010, realtek, synaptics, upek*, uru4000, vcom5s, vfs*. No Broadcom, no ControlVault. 0a5c:5843 works only through the proprietary TOD blob (libfprint-2-tod1-broadcom, which depends on libfprint-tod). Since omarchy-setup-security-fingerprint installs libfprint fprintd usbutils, that gives two outcomes and neither is a working reader:

So the detection change is right as far as it goes, but on its own it converts "No fingerprint sensor detected" into "Enrollment failed" on a clean machine, or into a pacman conflict on a machine set up to use the reader, while also claiming some Dell machines that have no reader at all. #7993 is building vid:pid-based TOD routing that this hardware would need, and its design note argues for exact vid:pid over vendor matching for precisely this reason.

Worth knowing before this lands: #7726 detects the same readers by matching fingerprint/biometric in the interface descriptor instead of widening the vendor list. On this hardware that interface is named "Broadcom ControlVault 3+ w/FingerPrint" (see #7424), so it would find the reader without also claiming the 76 0a5c products that are not fingerprint sensors. The two approaches overlap and both could land, but which one to take is the maintainer's call rather than mine.

On the new test: the case at test/shell.d/hw-fingerprint-test.sh:77 passes without modelling any USB interface, so has_kernel_driver returns on an unmatched glob before it checks anything. It is not vacuous — dropping 0a5c fails it — but it does not exercise the real 5843 topology, and there is no case pinning the property the whole vendor list rests on, that a Broadcom Bluetooth adapter is still rejected.

On the issue links: #8483 and #8562 are the same device, 0a5c:5843 with product string 58200, and so is #7662, filed five days before #8483 with the same proposed fix. #7424 is the same reader family at 0a5c:5865. The body here only says "Fixes #8483".

Nothing pushed to your branch. Narrowing to vid:pid pairs, or holding this until TOD routing exists, is a judgement about which hardware Omarchy claims to support, and that belongs to the maintainer. Waiting on them.

@omarchybot

Copy link
Copy Markdown
Collaborator

Looked at this again because two things have changed since the last review. The head is still 71f9fdb and nothing was run this time. This pass had only one reviewer, Opus 5.5 in Claude Code: the second opinion, Codex at medium, was not available because the bot's daily review budget is spent.

A narrower competing fix now exists. #9524 fixes the same 0a5c:5843 reader, but it matches the exact vendor and product ID pair instead of the whole Broadcom vendor. It keeps the has_kernel_driver guard, and its tests cover an unrelated Broadcom device being rejected and a ControlVault bound to a kernel driver being rejected. The last review here found that matching all of 0a5c also claims the ControlVault configurations that have no fingerprint sensor (0a5c:5800, 5803, 5832). Nothing binds those at idle, so the kernel-driver guard does not stop them. #9524 does not have that problem. #7726 takes a third approach: it matches a USB interface named fingerprint/biometric. That covers the ControlVault 3+ at 0a5c:5865 from #7424, which neither ID-based change catches. #9524 and #7726 do not overlap, and between them they cover both reported IDs without claiming the rest of Broadcom's catalog. I have not confirmed on hardware whether the 5843 also names its fingerprint interface.

The linked issue has moved. #8483 was closed as a duplicate of #7424, so this pull request's Fixes #8483 no longer points at an open issue.

Which detection approach to take is the maintainer's decision, so nothing was pushed here. My read is that #9524, possibly together with #7726, is the better-scoped fix. Waiting on the maintainer.

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

Labels

bug Something isn't working compatibility Make Omarchy work better on everything!

Projects

None yet

Development

Successfully merging this pull request may close these issues.

omarchy-hw-fingerprint doesn't detect Broadcom (0a5c) fingerprint sensors

3 participants