Update build script to facilitate iso builds

[rahulbabu95]

Description

Linuxkit also offers an option to build kernel as an iso file. Update the build script to be able Hook as an ISO.

Why is this needed

Apart from the default build, this change provides the users to be able to mount Hook as an ISO and boot using the virtual media option.

Fixes: #

How Has This Been Tested?

Set LINUXKIT_ISO to true and built hook as an ISO. Mounted the ISO on a Dell R340 as a virtual media device and booted using the ISO. As this flow kicks in only when the above env var is set, it should affect any existing user workflow.

Checklist:

If this gets approved, I will update the Readme appropriately to document the option to build as an ISO.

I have:

• updated the documentation and/or roadmap (if required)
• added unit or e2e tests
• provided instructions on how to upgrade

[rahulbabu95]

/hold

[rpardini]

Hey @rahulbabu95, interesting; my main question is that by booting the .ISO directly (and thus directly into HookOS), there's no chance for iPXE to add the kernel command-line arguments that are required (eg, the tink address, worker image, etc) -- how would one define those in the .ISO scenario?

[rahulbabu95]

@rpardini That's a great question. So end goal of getting an ISO build is to be able to have Tinkerbell stack running on L3. I don't have a concrete design yet but will trying out some POCs with an Hook ISO. But if the worker image is embedded then the only thing to be dynamically arrived at is the tink address. Main ideas I have been thinking of is to either provide the address during the build time or dynamically feed in the address at run time to Hook.

[rahulbabu95]

@rpardini just following up on the earlier discussion, I used tink-worker image with embedded images and tried up two scenarios. 1. Provide syslog_host and grpc_authority host as kernel commandLine args at buildtime.
2. Set them dynamically by patching the Hook ISO at run time.
I was able to boot a DELL R340 for my POC using Hook ISO.

Could you take another look on this when you get a chance? Let me know if you have more questions/ideas.!

[jacobweinstock]

@rpardini , @rahulbabu95 . I like where this is going as it leads us to the possibility for layer 3 provisioning. There is still lots to do to enable layer 3 provisioning and @rahulbabu95 we'll need a proper design doc for that so we can all discuss but I don't see this impacting any exist code. I'm going to merge this. If there are concerns, please bring them up and we can iterate on resolving them. Thanks All!