Booting Windows 11

[Heinrich Schuchardt]

On 8/16/23 09:41, Heinrich Schuchardt wrote:
> On 8/16/23 03:22, Simon Glass wrote:
>> Hi Heinrich,
>>
>> I thought I would try booting Windows on U-Boot but it does not get
>> very far. Have you tried that? I note a ew odd things below. Is there
>> a validation suite for it?
>>
>> https://github.com/sjg20/u-boot/tree/efi2
>
> qemu-x86_64_defconfig does not build with your tree (commit
> d4caae6bce9082d0a770047e8f3d3a68b7084740, sjg/efi2):
>
>    BINMAN  .binman_stamp
> Wrote map file './rom.map' to show errors
> binman: Node '/binman/rom/u-boot-spl-with-ucode-ptr': Offset 0xfffd8000
> (4294803456) overlaps with previous entry '/binman/rom/u-boot' ending at
> 0xfffd84e8 (4294804712)
> make: *** [Makefile:1115: .binman_stamp] Error 1
>
> ImagePos    Offset      Size  Name
> <none>    00000000  00100000  rom
> <none>     fff00000  000d84e8  u-boot
> <none>      00000000  000d79a0  u-boot-nodtb
> <none>      000d79a0  00000b48  u-boot-dtb
> <none>     fffd8000  00015f60  u-boot-spl-with-ucode-ptr
> <none>     fffd84e8  00000b48  u-boot-dtb
> <none>     fffd9030  000004c6  fdtmap
> <none>     fffd94f6  00000000  private-files
> <none>     fffedf60  00000b48  u-boot-dtb-with-ucode2
> <none>     fffff800  00000070  x86-start16-spl
> <none>     fffffff0  00000005  x86-reset16-spl
> <none>     fffffff8  00000008  image-header
>
>>
>> qemu-system-x86_64  -smp 4 -m 8G -serial mon:stdio -bios
>> /tmp/b/qemu-x86_64/u-boot.rom -drive
>> id=fisk,file=root.img,if=virtio,driver=raw -drive
>> id=disk,if=virtio,file=/vid/software/win/Win11_22H2_English_x64v2.iso
>
> This does not work with EDK II either.
>
> I was able to start the installer with:
>
> qemu-system-x86_64 \
> -M q35,smm=on,accel=kvm -smp 8 -m 4G -gdb tcp::1234 \
> -serial mon:stdio \
> -vga std \
> -device qemu-xhci \
> -device usb-kbd \
> -device usb-mouse \
> -global driver=cfi.pflash01,property=secure,value=on \
> -drive
> if=pflash,format=raw,unit=0,file=OVMF_CODE_4M.secboot.fd,readonly=on \
> -drive if=pflash,format=raw,unit=1,file=OVMF_VARS.ms.fd \
> -drive file=root.img,if=virtio,format=raw \
> -drive file=Win11_22H2_EnglishInternational_x64v2.iso,media=cdrom
>
> The Windows installer does not work without SMM but returns to the EFI
> shell. This might be what you have seen.
>
> U-Boot's qemu-x86_64_defconfig does not start with
> -M q35,smm=on,accel=kvm
>
> Instead of using KVM you could specify -cpu Skylake-Client-v4 which is
> supported by the Windows 11. But of course this is very slow.
>
> Please, retry with:
> qemu-system-x86_64 \
> -machine pc-i440fx-2.5,smm=on -cpu Skylake-Client-v4 \
> -vga std \
>
> Can we get support for -M q35 into U-Boot? Cf.
> https://wiki.qemu.org/Features/Q35. This would allow us to actually
> compare EDK II and U-Boot when running with the same settings.
>
> Why wouldn't U-Boot support kvm on an AMD Ryzen 5 2500U?

With KVM I either get:
"int_exception_handler() Exception 13 while executing option rom"
or the code hangs in cpu_call64.

Best regards

Heinrich

>
>>
>> Scanning bootdev 'virtio-blk#0.bootdev':
>> efi_install_multiple_protocol_interfaces_int() EFI: Call:
>> efi_locate_device_path(protocol, &dp, &old_handle)
>> efi_locate_device_path()   EFI: Entry efi_locate_device_path(Device
>> Path, 00000000becf6848, 00000000becf6840)
>> efi_locate_device_path()     EFI: Call:
>> efi_locate_handle_buffer(BY_PROTOCOL, protocol, NULL, &no_handles,
>> &handles)
>> efi_locate_handle_buffer()       EFI: Entry
>> efi_locate_handle_buffer(2, Device Path, 0000000000000000,
>> 00000000becf67b8, 00000000becf67b0)
>> efi_locate_handle_buffer()       EFI: Exit: efi_locate_handle_buffer: 0
>> efi_locate_device_path()     EFI: 0 returned by
>> efi_locate_handle_buffer(BY_PROTOCOL, protocol, NULL, &no_handles,
>> &handles)
>> efi_locate_device_path()   EFI: Exit: efi_locate_device_path: 14
>> efi_install_multiple_protocol_interfaces_int() EFI: 14 returned by
>>
>>
>> ^ This seems unhappy because it is a virtio device?
>
> Before a device path protocol is installed we check that the same device
> path has not been installed before (lib/efi_loader/efi_boottime.c:2685).
>
> EFI_NOT_FOUND signals that we are good to continue.
>
> Best regards
>
> Heinrich