Net-booting Raspberrypi4-64

After booting to U-boot:

U-boot>setenv serverip 192.168.1.60
U-boot>tftp ${kernel_addr_r} Image
U-boot>tftp ${fdt_addr_r} bcm2711-rpi-4-b.dtb
U-boot>booti ${kernel_addr_r} - ${fdt_addr_r}

Using the eSDK container

Create a volume with the name of the SDK (I keep separate volumes for each SDK)

podman volume create cortexta72-poky-linux

Install the eSDK to the volume:

[edtwardy@Hackintosh ~]$ podman run -it --rm \
       -v cortexa72-poky-linux:/home/yocto/poky_sdk \
       -v $PWD:/home/yocto/installer \
       localhost/esdk-container
yocto@ea5526d4ee6e:~$ ./installer/poky-glibc-*.sh

Finally, add the SSTATE mirror, by manually editing ~/poky_sdk/conf/local.conf to include the following:

SSTATE_MIRRORS = "\
...
file://(.*) http://edtwardy-yocto.local/sstate-cache/\1 \
"

Also, make sure that there's an entry in /etc/hosts for edtwardy-yocto.local. Or, the URL https://edtwardy.hopto.org/sstate-cache can be used to allow remote access.

After this, the container is configured to source the environment setup script automatically, so all that's necessary is to spin up a container and develop!

podman run -it --rm -v cortexa72-poky-linux:/home/yocto/poky_sdk \
    localhost/esdk-container

One could even maintain different SDK's in different volumes and select the desired SDK at container creation time.

In the past, I've had problems with UID ownership of files/directories while using the container. To resolve this, I have to chown the necessary filesystem items on the host to the UID/GID (which are the same) of the yocto user in the container. To discover this user's UID/GID, try taking a look at the SDK volume using podman inspect cortexa72-poky-linux and then running ls -l on the directory where the volume resides on the host.

I find it's also useful to mount my .gitconfig inside the container. So, the entire invocation looks like:

podman run -it --rm \
    -v cortexa72-poky-linux:/home/yocto/poky_sdk \
    -v $HOME/.gitconfig:/home/yocto/.gitconfig \
    -v $PWD:/home/yocto/app \
    localhost/esdk-container

Building and Booting a QEMU x86_64 Image

The qemu-debug-kirkstone.yaml kas file is provided to build a rootfs that can be used to boot a custom kernel under QEMU. The kas file will not attempt to populate a bootloader or kernel, so those must both be provided by the user.

Both a .wic image and a .cpio.gz image are built, so that the user can choose to boot using a disk image, or an initrd.