Michael Walle authored 1 year ago and Frieder Schrempf committed 1 year ago

This is supposed to fix a problem where you can end up with two
non-working images.

This can be triggered by the following actions:

  * Starting with rootfs A and fitImage_active
  * Doing an update via swupdate using a corrupted kernel image
  * fitImage_active gets renamed to fitImage_inactive.
  * The corrupted kernel now in fitImage_active fails to boot
  * A fallback to rootfs A occurs: rootfs A is started together
    with the working fitImage_inactive
  * On the next update via swupdate the corrupted fitImage_active is
    renamed to fitImage_inactive ending up with two corrupted kernel
    images
  * The now corrupted fitImage_inactive is copied to fitImage_active
    and the boot is bricked

To remedy this flaw, simply swap the fitImage in sync with the rootfs.
Instead of the active/inactive image we now use A/B for the kernel image
and the currently used image is selected by the sys_active file.

Adds console kernel command line settings for imx6 devices. The
dt chosen stdout-path is currently ignored for any reason

Use generic names for the boot scripts for booting from a single
partition and for doing A/B boots from two separate partitions.

Use the generic 'load' command instead of specific filesystem
commands.

Also sync the environment for i.MX6UL/ULL with the one for i.MX8MM.

In the future we will reference the BSP as meta-ked-bsp. Adjust names
of layers, files, etc. accordingly. Only some OE metadata like image,
distro or machine override names are keeping the 'ktn' for now.

KED stands for Kontron Electronics Deutschland GmbH where the BSP
is maintained. Nowadays it support only products of the KED branch
and products of the Kontron Europe GmbH or other branches are not
supported.