rx_item_t is ~800 bytes at RX_FRAME_MAX_LEN, and wifi_promisc_rx_cb declared one
as a local. That callback runs on the WiFi driver's own task, already several
frames deep in the driver's call chain, on a stack of roughly 3.5 KB
(CONFIG_ESP_WIFI_TASK_STACK_SIZE, left at its default). Putting a fifth of that
stack into a single local is a stack-overflow risk that only appears under real
traffic - in front of an RSU rather than on the bench - and would present as a
random panic rather than anything pointing at its cause.
Both instances are now static: one in the callback, one in rx_forward_task. Safe
because each is touched by exactly one task, so there is no re-entrancy to guard
against; the same reasoning serial_link.c already uses for its static send
buffers. xQueueSend copies the struct out before returning, so reusing the
callback's buffer on the next frame is fine.
Firmware-only, no protocol change, so it does not require a matching app install.
Re-verified against live traffic after flashing: 1094 frames over 125 s with zero
decode failures, USB errors, detaches, crashes or mutex timeouts. SPATEM capture
rose from 3.20/s to 3.98/s against a theoretical maximum of 4.00/s, which is the
direction relieving stack pressure would produce, though RF geometry moves
between runs and this is not proof.
Report updated with T9, the accepted 512-byte ceiling, and the decision to drop
Phase B: the intersection use case is CAM-driven and needs none of it.
Bench test on 2026-08-25 against live RSU and CiT One traffic found that 611 RSU
CAMs decoded correctly and none of them were ever displayed. Excluding RSUs from
UseCaseDetectionEngine - correct, since a permanently stationary station at a
fixed point trips the stopped-vehicle use case for as long as it is in range -
also removed them from the map and station list, because remotePositions is the
engine's own map.
RSUs are now tracked in a separate rsuStations flow and merged with the engine's
road users for display only. Expiry is clock-driven for the same reason as
hazards and signals: an RSU going out of range simply stops transmitting, and no
further emission would arrive to recompute the list. Cleared on link-down
alongside engine.reset(), so a stale RSU cannot outlive an unplug.
The kinematics line is suppressed for them. An RSU's CAM uses
rsuContainerHighFrequency, which carries no kinematics at all, so the zeroes in
the model are placeholders - printing "0.0 km/h - heading 0" would assert a
stationary vehicle pointing due north.
Also logs the ESP32's STATUS heartbeat counters whenever one changes. They
previously reached only the CAM Pinger card, so a bench run captured through
logcat had no record of whether the firmware dropped anything. Logged on change
rather than per beat: the interesting event is a drop appearing, and a
once-per-second line would bury it.
Test report in 05-obu-bench-test-2026-08-25.md.
Ashin Walpola