Re: [Xen-devel] [PATCH RFC 6/7] xen/x86/efi: Verify dom0 kernel with SHIM_LOCK protocol in efi_multiboot2()

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 14.05.18 at 18:56, <daniel.kiper@xxxxxxxxxx> wrote:
> On Mon, May 14, 2018 at 04:43:13AM -0600, Jan Beulich wrote:
>> >>> On 08.05.18 at 15:09, <daniel.kiper@xxxxxxxxxx> wrote:
>> > On Fri, May 04, 2018 at 09:46:33AM -0600, Jan Beulich wrote:
>> >> >>> On 08.07.17 at 23:53, <daniel.kiper@xxxxxxxxxx> wrote:
>> >> > @@ -484,9 +497,12 @@ __efi64_mb2_start:
>> >> >          /* Keep the stack aligned. Do not pop a single item off it. */
>> >> >          mov     (%rsp),%rdi
>> >> >
>> >> > +        mov     %r14d,%edx
>> >> > +
>> >> >          /*
>> >> >           * efi_multiboot2() is called according to System V AMD64 ABI:
>> >> > -         *   - IN:  %rdi - EFI ImageHandle, %rsi - EFI SystemTable.
>> >> > +         *   - IN: %rdi - EFI ImageHandle, %rsi - EFI SystemTable,
>> >> > +         *         %rdx - dom0 kernel module struct address.
>> >>
>> >> How come everything further up treats this as a 32-bit quantity only?
>> >
>> > According to the Multiboot2 spec the bootloader is not allowed to
>> > put the kernel (xen.gz) and the modules above 4 GiB boundary.
>>
>> Interesting - how would they load a 1Gb initrd on a system with just 1Gb
>> RAM below 4Gb? Not to speak of a 4Gb initrd ...
> 
> That is not possible right now. This requires changes in the boot protocol.
> Anyway, have you seen such setups in the wild today?

Years ago we've already had to make our XenoLinux forward port cope with
512Mb+ initrd-s - see the commit introducing XEN_ELFNOTE_MOD_START_PFN,
which tells you that Xen itself needed to be changed for this as well. Those
folks wanted to be able to boot a full fledged distro without loading anything
from disk (or network) post-boot.

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel