[PATCH v8 01/15] x86/boot: Place kernel_info at a fixed offset

[H. Peter Anvin]

On March 21, 2024 6:45:48 AM PDT, "Daniel P. Smith" <dpsmith at apertussolutions.com> wrote:
>Hi Ard!
>
>On 2/15/24 02:56, Ard Biesheuvel wrote:
>> On Wed, 14 Feb 2024 at 23:31, Ross Philipson <ross.philipson at oracle.com> wrote:
>>> 
>>> From: Arvind Sankar <nivedita at alum.mit.edu>
>>> 
>>> There are use cases for storing the offset of a symbol in kernel_info.
>>> For example, the trenchboot series [0] needs to store the offset of the
>>> Measured Launch Environment header in kernel_info.
>>> 
>> 
>> Why? Is this information consumed by the bootloader?
>
>Yes, the bootloader needs a standardized means to find the offset of the MLE header, which communicates a set of meta-data needed by the DCE in order to set up for and start the loaded kernel. Arm will also need to provide a similar metadata structure and alternative entry point (or a complete rewrite of the existing entry point), as the current Arm entry point is in direct conflict with Arm DRTM specification.
>
>> I'd like to get away from x86 specific hacks for boot code and boot
>> images, so I would like to explore if we can avoid kernel_info, or at
>> least expose it in a generic way. We might just add a 32-bit offset
>> somewhere in the first 64 bytes of the bootable image: this could
>> co-exist with EFI bootable images, and can be implemented on arm64,
>> RISC-V and LoongArch as well.
>
>With all due respect, I would not refer to boot params and the kern_info extension designed by the x86 maintainers as a hack. It is the well-defined boot protocol for x86, just as Arm has its own boot protocol around Device Tree.
>
>We would gladly adopt a cross arch/cross image type, zImage and bzImage, means to embedded meta-data about the kernel that can be discovered by a bootloader. Otherwise, we are relegated to doing a per arch/per image type discovery mechanism. If you have any suggestions that are cross arch/cross image type that we could explore, we would be grateful and willing to investigate how to adopt such a method.
>
>V/r,
>Daniel

To be fair, the way things are going UEFI, i.e. PE/COFF, is becoming the new standard format. Yes, ELF would have been better, but...