OE4T / meta-tegra

BSP layer for NVIDIA Jetson platforms, based on L4T
MIT License
385 stars 216 forks source link

tegra-kernel.cap not generated in "jetson-orin-nano-devkit-nvme" #1465

Closed haizzus closed 5 months ago

haizzus commented 5 months ago

Describe the bug Adding IMAGE_INSTALL:append = " tegra-uefi-capsules" in local.conf tegra-kernel.cap not generated in "jetson-orin-nano-devkit-nvme", only tegra-bl.cap; but they are generated in ""jetson-orin-nano-devkit" target. This applies to all the targets which are built for nvme I believe.

To Reproduce Steps to reproduce the behavior:

  1. Build meta-tegra branch 'kirkstone' with MACHINE based on 'jetson-orin-nano-devkit-nvme'
  2. Build with bitbake argument 'core-image-sato-dev'
  3. The capsules are missing in deploy/images folder.
madisongh commented 5 months ago

Yep, this would be the case for any machine config where the kernel resides on an external device, rather than internal storage. The BUP payload generator uses the flash layout file for the internal devices only, so no kernel BUP gets created in that case. Since the BUPs are the input for generating the capsules, no kernel capsule gets created, either.

TBH, I don't think using a UEFI capsule to update your kernel is a good idea. But I'll look into figuring out a way to solve this, if there are folks that really need this functionality.

dwalkes commented 5 months ago

TBH, I don't think using a UEFI capsule to update your kernel is a good idea.

See the strategy here which you can use as an alternative, and which I think matches how a few others are handling kernel updates without capsule update based on discussions in the device update channel on gitter/matrix

haizzus commented 5 months ago

TBH, I don't think using a UEFI capsule to update your kernel is a good idea.

See the strategy here which you can use as an alternative, and which I think matches how a few others are handling kernel updates without capsule update based on discussions in the device update channel on gitter/matrix

Thanks, this approach matches mine also.