Skip to content

OE4T Meeting Notes 2024‐09‐12

Jump to bottom Edit New page
Dan Walkes edited this page Sep 14, 2024 · 1 revision

Video

https://youtu.be/NpBGdTHa1-s

Attendees

7

Topics

  • Latest status of oot kernel and linux-yocto kernel support.
    • See https://github.com/OE4T/meta-tegra/pull/1680
    • Issue with encryption module not playing nicely with kernel when wifi stack tries to use certain features. Wpa_supplicant crashing the kernel.
      • Removing crypto engine made it work again.
      • Suggestion to try disabling the problematic encryption algorithm as a workaround or lower priority of it.
      • Able to run docker containers.
      • Found issue with USB gadget drivers, initrd-flash method checks whether gadget is configured before moving to next step. May resolve issues with flash reprogramming which have been noticed by some users.
  • 35.6 update from NVIDIA and work in progress branch
    • Working branch at https://github.com/OE4T/meta-tegra/tree/wip-scarthgap-l4t-r35.6
    • Could use help testing or implementing
    • Updated kernel, new boot firmware.
    • Fixes in gstreamer plugins from patches on the dev forum.
    • No cuda or jetpack updates, fairly minor overall.
    • Hadn’t published UEFI sources yet, still pointing at 35.5 sources.
    • Plan is to put in scarthgap r35 branch, then backport to kirkstone.
  • Imminent styhead branch coming soon, non LTS branch
    • Will pull a styhead branch when meta-openembedded and others start a styhead branch.
  • LUKS disk encryption on Orin
    • Taking all the recipes from tegra-test-distro
    • Change to use the OP-TEE application from nvidia.
    • Either replicate the installer or use the initramfs.
    • Put the installer in the recovery partition and have the first boot be to the recovery partition. Don’t need to re-GPT the device.
    • Idea to add custom scripts to initrd flashing process. Could have the initrd flasher generate the passphrase and do the encryption, then instead of exporting the entire disk could export the LUKS opened dev mapper devices rather than raw underlying storage.
    • Cboot implementation was easier to understand than UEFI. Making sure you have your root slots matched up with boot partitions, not sure how this works in UEFI.
  • nvbootctrl generation, 37.0.1 is not aligned with jetpack.
  • Issue with retries, need to clean up UEFI variables after failure. Illies will create a PR for this.
    • If you have a successful boot it’s good to mark the other slot as unbootable, but after doing that you can’t update.
    • Need to check if UEFI variables are not correct and update them.
    • Probably will make sense to fix in l4tlauncher.
  • UEFI secureboot and signing
Add a custom footer
Add a custom sidebar
Clone this wiki locally