create_disk: Use target EFI binaries #901
Conversation
Which in turn is part of the dependency chain for coreos/mantle#1060 |
|
(I have some bigger questions around this...I notice we've been omitting e.g. |
cgwalters
force-pushed
the
uefi-target
branch
from
0050133 to
91cd59d
Compare
I think we talked about this yesterday but didn't root cause it? Clearly unrelated though, someone mind taking a look at this PR? |
Yeah, I'll take this. |
|
OK, that bug took me on a long journey 🚌. |
This is analogous to coreos#652 We were hardcoding `fedora` here, pulling the EFI binaries from the coreos-assembler container. Change to installing the *target* EFI binaries instead. First, I verified this produces the same output and `cosa run --uefi-secure` still works with FCOS. But, this fixes `cosa run --uefi-secure` with RHCOS, and ensures we're installing the RHEL8 EFI binaries. Conceptually now, the only "traces" or "influence" that the coreos-assembler content has on the target system will (should) be in the created filesystem structure (i.e. what `mkfs.xfs` does on disk), but not the actual *contents* of any files. Closes: openshift/os#383
cgwalters
force-pushed
the
uefi-target
branch
from
91cd59d to
75f0a5f
Compare
|
Rebased 🏄♂️ |
This is analogous to
#652
We were hardcoding
fedorahere, pulling the EFI binariesfrom the coreos-assembler container. Change to installing
the target EFI binaries instead.
First, I verified this produces the same output
and
cosa run --uefi-securestill works with FCOS.But, this fixes
cosa run --uefi-securewith RHCOS, and ensureswe're installing the RHEL8 EFI binaries.
Conceptually now, the only "traces" or "influence" that the
coreos-assembler content has on the target system will (should) be
in the created filesystem structure (i.e. what
mkfs.xfsdoes ondisk), but not the actual contents of any files.