Skip to content

omamobile-gnss 0.2.11: a Qualcomm modem's GNSS over QMI LOC (opt-in) - #933

Open
johannSo wants to merge 1 commit into
derekross:mainfrom
johannSo:hotdog-gnss
Open

johannSo wants to merge 1 commit into
derekross:mainfrom
johannSo:hotdog-gnss

Conversation

@johannSo

Copy link
Copy Markdown

What and why

omamobile-gnss 0.2.11: a Qualcomm modem's GNSS over QMI LOC, for the OnePlus 7T Pro (#931, device/hotdog/GNSS-PLAN.md).

The modem's engine answers QMI LOC over QRTR, and its own NMEA is only the proprietary $PQ sentences. The new omamobile-gnss-qmi-loc does three things:

  • runs a LOC session;
  • writes GGA, RMC, GSA and GSV into /run/omamobile-gnss/nmea_out;
  • leaves the existing bridge to serve that FIFO to GeoClue, as it serves the Pixels'.

Inert on every other phone. It runs only where the device package ships /usr/share/omamobile/gnss/qmi-loc.conf. Without that file:

  • every path is the Pixels' as before;
  • the bridge's FIFO list and its start wait only gain the new FIFO at the end;
  • omamobile-gnss.service only gains the file as a third alternative condition.

ENGINE_UNLOCK=1 clears an engine lock that stock left in the modem's EFS (QMI error 46). It is a LOC request, so the modem writes its own EFS. On hotdog, rmtfs runs read-only (-r) and keeps those writes in RAM. The daemon's doc says a device without that must leave it off.

Device and build

  • Phone: OnePlus 7T Pro (hotdog)
  • Kernel release (uname -r): 7.3.0-rc6-omamobile-push1-g7539ca3e06f6
  • Packages changed, with versions: omamobile-gnss 0.2.11

Checks (no phone needed)

  • tools/pr-check origin/main passes
  • tests/packages: the touched package's test/all passes (omamobile-gnss, with new tests for the QMI LOC path)
  • Each changed package has its version bumped, and the new version is in the commit subject
  • Commits have area prefixes and bodies that say why; AI help is a Co-Authored-By: trailer, never a Signed-off-by:

On the phone

  • I ran this on my own phone (it runs omamobile-gnss 0.2.11 from these files)
  • Runbook steps passed: GNSS-PLAN G0–G5 (2026-11-15). Outdoors, the first fix came 52 s after the request (10 satellites used, 36 tracked), and GeoClue's where-am-i got it.
  • What I did not test: XTRA injection and a reference position (not implemented); any Pixel (path unchanged)
  • Nothing in this change writes outside slot B and userdata, weakens the storage guard, or reads registers from userspace

The phone-smoke report from the OnePlus 7T Pro, running the port's branch, is in #931.

🤖 Generated with Claude Code

On a Qualcomm phone the GNSS engine runs in the modem and answers QMI LOC
over QRTR, and its own NMEA is only the proprietary $PQ sentences, which
ModemManager's gps-nmea and GeoClue's modem source can't use. The new
omamobile-gnss-qmi-loc opens a LOC client, runs a periodic session and
writes GGA, RMC, GSA and GSV into /run/omamobile-gnss/nmea_out, which the
existing bridge serves to GeoClue as it does the Pixels' FIFOs.

It runs only on a phone whose device package ships
/usr/share/omamobile/gnss/qmi-loc.conf: omamobile-gnss-start execs it
then, and omamobile-gnss.service gains that file as a third alternative
condition. Without the file every path is the Pixels' as before; the
bridge's FIFO list and its start wait only gain the new FIFO at the end.

ENGINE_UNLOCK=1 clears an engine lock the stock system left in the
modem's EFS (QMI error 46 on every start). It is a LOC request, so the
modem writes its EFS; the OnePlus 7T Pro's rmtfs keeps those writes in
RAM (rmtfs -r), and a device without that must leave it off.

On the OnePlus 7T Pro (device/hotdog/GNSS-PLAN.md G0-G5, 2026-11-15,
7.3.0-rc6-omamobile-push1 builds of its port): outdoors, the first fix
52 s after the request (10 satellites used, 36 tracked), and GeoClue's
where-am-i got it. test/all
passes on the laptop, with new tests for the QMI LOC path.

Co-Authored-By: Claude Opus 5.5 <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