> From: "Theo de Raadt" <[email protected]>
> Date: Mon, 10 Nov 2025 06:51:52 -0700
> 
> Kirill A. Korinsky <[email protected]> wrote:
> 
> > On Sat, 01 Nov 2025 14:16:33 +0100,
> > Kirill A. Korinsky <[email protected]> wrote:
> > > 
> > > >Synopsis:        Remote machine blocked until KVM isn't connected on the 
> > > >first boot
> > > >Category:        install
> > > >Environment:
> > >   System      : OpenBSD 7.8
> > >   Details     : OpenBSD 7.8-current (GENERIC.MP) #80: Sat Nov  1 00:47:23 
> > > MDT 2025
> > >                    
> > > [email protected]:/usr/src/sys/arch/amd64/compile/GENERIC.MP
> > > 
> > >   Architecture: OpenBSD.amd64
> > >   Machine     : amd64
> > > >Description:
> > >   This is server from Hetzner AX102 which I use for a couple of
> > >   months. I've tried to replace the hardware to similar server, but it
> > >   doesn't fix a bug.
> > >   Almost each sysupgrade -s the machine is stuck if KVM isn't
> > >   connected. After first boot I can reboot it without any issue.
> > >   If I ask Hetzner to attach KVM on stucked machine... It boots.
> > >   If I have attached KVM, sysupgrade goes well. If no, machine goes to
> > >   it and don't response for ping, nothing. I had tried to wait about 1
> > >   hour before ask for KVM. When KVM is attached and I have access
> > >   details... in 20-30 seconds emails with update log came. Machine
> > >   fully booted and here no issue anymore.
> > > >How-To-Repeat:
> > >   Just run sysupgrade on this machine.
> > > >Fix:
> > >   Not idey
> > 
> > After some playing around. It smeels like BIOS bug, and without connected
> > screen gop->Mode->FrameBufferBase is 0.
> > 
> > If anyone encountered it, here a nice workaround: use com0 as serial.
> 
> I do not think this is a BIOS bug.
> 
> What it is, is a real-world configuration that causes us to not attach
> a console in some weird way, meaning we end up with no real console.
> later on we have some code which depends upon openbsd having a console,
> and when it doesn't find a console, it goes off the rails.

I suspect that init(8) tries to open /dev/console and fails.

> A workaround is not the way to handle this problem.

Do we need a "dummy" console device that we can fall back on as a last
resort?

Reply via email to