Attendees

5

Video

https://youtu.be/dFJgFucy23g

Discussion

• r36.4.3 PR status at https://github.com/OE4T/meta-tegra/pull/1837
  • See test sheet.
  • Some issues noted with cuda on orin-nano-nvme, could use help with this if anyone is available to help.
• mender dunfell TX2 PR
  • See https://github.com/mendersoftware/meta-mender-community/pull/438 - not sure if it will be ultimately merged.
  • Important for anyone on mender + u-boot and dunfell/kirkstone branches.
• swupdate install-if-different feature at https://github.com/OE4T/tegra-demo-distro/pull/320
  • Concern that some of the devicetree may get bundled in with mb2 or with UEFI.
  • Likely are situations where it won’t be safe to rely on L4T Version to skip capsule updates, especially if devicetree changes.
  • Useful as a development tool if you know you shouldn’t need the capsule update and all changes are in the rootfs.
• meta-swupdate hash mismatch issues on master
  • See branch at https://github.com/Trellis-Logic/meta-swupdate/tree/fix-hash-issues
  • See thread at https://groups.google.com/g/swupdate/c/ZrDn9hVCet8/m/4lZhxoFfAgAJ. Could use some additional comments on this thread if anyone else validates the problem/fix.
• scarthgap migration and capsule update issues on signed/encrypted builds
  • On any scarthgap capsule updates, including scarthgap->scarthgap, the swupdate + capsule update completes, capsules apply
  • However at the earliest level see MB1 continuing to boot from slot 0. Don’t see any indication as to why that is.
  • Added UEFI debugging to confirm capsule update process completed successfully
  • Built tegra-demo-distro as a reference on devkit and was unsigned. Doesn’t reproduce the problem.
  • Turned on UEFI debugging. Doesn’t have signing or encryption, however.
  • Boot ROM is loading up a few things… did one experiment to see if it was a problem with the BCT partition storing the boot chain sequence.
  • Showed status in progress. Marked slot B as unbootable after several boots.
  • Manually moved slot to slot B, applying capsule from there also falls back to slot B.
  • MB1 has a different reset source in the failure case rst_source: 0x17, rst_level: 0x1
  • Discussed updating the L4T Launcher, looking at this for the source of the problem since this is where the switching rootfs should happen.
  • Also can’t force switch to a different slot, marks the slot as unbootable.
  • Discussed forcing signing or encryption errors along with slot switches to see if the MB1 boot response is the same to identify which is the problem.
• ESP partition updates for L4T Launcher
  • Still missing in the reference implementation for swupdate.
  • Would be good to understand redundancy aspects/limitations and optionally include in the swupdate demo, potentially using install-if-different logic.
• Orin NX 8GB and 16GB combined MACHINE
  • Briefly discussed element thread at https://matrix.to/#/!mKaltPMAHUgQwPUKcH:gitter.im/$YNFxAcQpdRFIVorWvrFwSWzQIALgOLkKxYmeKtL3C3g?via=gitter.im&via=matrix.org&via=3dvisionlabs.com regarding combining Orin Nano 8GB and 16GB machines in capsule and flash programming steps.
  • Several people are interested but nobody has spent time on this yet.