fix(provision): pin camera userspace stack to 36.4.4#108
Merged
Conversation
The 36.4.7/36.5.0 Argus userspace regressed: every nvarguscamerasrc teardown errors, and ~1 in 60-120 relaunches drops all frames and segfaults nvargus-daemon. Bench-validated on PAB_V3/Orin NX that the 36.4.4 quartet on the unchanged 36.5.0 kernel/RCE firmware is clean (250/250 launches), so provision repacks and holds those debs. Full evidence and repro harness in docs/argus_relaunch_regression.md.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Pin the camera userspace stack (
nvidia-l4t-gstreamer/-camera/-multimedia/-multimedia-utils) to 36.4.4 at provision time, sidestepping the JetPack 6.2.2 Argus regression. Bench-reproduced, root-caused, and validated on PAB_V3 / Orin NX 16GB with dual IMX219 — full data and repro harness indocs/argus_relaunch_regression.md.Problem
The 36.4.7/36.5.0 NVIDIA camera userspace regressed (#107): every
nvarguscamerasrcteardown errors (CANCELLED / Argus Correctable Error Status), and roughly 1 in 60–120 camera relaunches delivers no frames and segfaults nvargus-daemon (Camera HwEvents waittimeout → EGLStreamProducerMutex not initialized→ SIGSEGV), a ~40 s video outage per event and a permanent wedge on setups where the daemon doesn't restart. It reproduces with stock sensors, is independent of the encoder and of rtcpu runtime PM (NVIDIA's forum workaround does not help on Orin NX), and lives entirely in the closed userspace: the same relaunch cycling is 250/250 clean with 36.4.x userspace on the unchanged 36.5.0 kernel, OOT modules, and RCE firmware.Solution
--provisionnow installs the four camera-stack debs atNV_CAMERA_STACK_VERSION(36.4.4 — the newest bench-clean release) instead of the BSP stamp. The 36.4.x debs are repacked (relax_l4t_deps) to lift theirnvidia-l4t-core (<< 36.5-0)cap and out-of-set exact-version deps, installed as one plainapt-get install --allow-downgradestransaction (no--force-depends, dpkg/apt state stays consistent), andapt-mark holdso an on-device upgrade can't restore the regressed stamp. Validated end-to-end on the bench in exactly this repacked form:apt-get checkclean, 20/20 clean launches (plus 100/100 on the raw 36.4.4 quartet). Docs carry the bench data and the retest-and-drop-the-pin guidance for future BSP bumps.