I've tried to follow the trail of commits and comments you provided but
find myself even more confused now.
While iPXE code may technically be running, from my end-user perspective
I don't see any interaction with iPXE's prompts or protocols, so it may
as well not be.
I haven't been able to directly network boot a UEFI VM for years, and
with these changes in the UX it doesn't look like I will be able to any
time soon. From [2]: "The default UEFI user experience for network boot
is somewhat excremental" is an understatement, but the choice to use it
appears to have already been made, so I'll continue doing what I've been
doing.
Let's go ahead and close this report out.
On 1/30/25 14:44, Sven Geuer wrote:
On Thu, 2025-01-30 at 11:18 -0500, Kevin Otte wrote:
I'm confused. If the VM won't launch iPXE from ROM, doesn't that make
the EFI builds in this package moot?
Yes, if I boot an iPXE upstream build from USB it works just fine.
Again, that kind of defeats the purpose of having it in the ROM in
the first place.
In fact iPXE _is_ launched from ROM but hands over control to EFI BIOS
during its course of execution. The comment in the upstream sources [1]
says that's done to support the PXE base code protocol, iPXE supports
it natively only in Legacy BIOS.
Please find more details in upstream's commit message [2] and in bug
1789319 [3] + [4] at Ubuntu which ended up in activating this behaviour
in the ROMs.
I opted to mirror what the Ubuntu guys did, and apparently also the
people at RHEL/Fedora [4].
[1] https://github.com/ipxe/ipxe/blob/master/src/config/qemu/general.h
[2] https://github.com/ipxe/ipxe/commit/a15c0d7e868a
[3] https://bugs.launchpad.net/maas/+bug/1789319
[4] https://bugs.launchpad.net/maas/+bug/1789319/comments/16
