platforms: [qemu] Also use hvc0 for x86_64/aarch64 #2785
Conversation
Pairs with coreos/coreos-assembler#3689 Basically hvc0 is faster and more modern, but requires us to opt-in specifically. We already had a few usages of hvc0 from cloud providers that default to this, no reason for us not to do it for our own qemu provider. (One thing TODO here: I believe s390x qemu could do this too; as of right now we don't have any references to s390x here)
|
Draft since this needs some more testing... |
|
Ahh yes, Fedora (and c9s) currently uses In contrast...a quick check of opensuse Leap ( Though Debian ( I'm pretty strongly of the opinion though that we should change this to be built-in; we already have |
|
I agree it should be built-in. My old friend |
|
so we just need to submit a PR to the kernel configs? or is it more complicated than that? |
| kernel_arguments: | ||
| - console=hvc0 | ||
| - console=ttyAMA0,115200 |
There was a problem hiding this comment.
I'm not sure if this will change the behavior on aarch64 qemu today. Right now it autodetects. So will this change it so that VGA console isn't the primary console?
There was a problem hiding this comment.
I believe so, yes. In our use cases we basically never use a graphical console (right?)
That said, honestly since podman is targeting applehv now as of recently it matters much less what we do with qemu. So I'm just going to close this for now.
It probably does make sense for us to switch to hvc for qemu, but doesn't seem worth the churn for now.
Motivated by making it easier to debug problems like the applehv issues we're seeing today. This is also easy to do with qemu. However note that in practice this really wants coreos/fedora-coreos-config#2785 so that we can get the full console from the first boot on applehv. For now, this is just a debugging aid for us. Signed-off-by: Colin Walters <[email protected]>
Pairs with coreos/coreos-assembler#3689
Basically hvc0 is faster and more modern, but requires us to opt-in specifically. We already had a few usages of hvc0 from cloud providers that default to this, no reason for us not to do it for our own qemu provider.
(One thing TODO here: I believe s390x qemu could do this too;
as of right now we don't have any references to s390x here)