Hello, all!
Finally I found what makes difference - regarding ATAPI timeouts (that
lead to kernel trap).
Both ACPI & HPET Timer must be enabled to avoid timeouts - so this
driver have to be used: https://man.openbsd.org/acpihpet.4
Under OpenBSD 7.3 Guest you have to verify that this message
appears in dmesg:
# dmesg | grep hpet
acpihpet0 at acpi0: 100000000 Hz
Once acpihpet0 is used by OpenBSD 7.3 guest the installation finish
complete sets installation from cd0 without single timeout and without trap.
How to verify that both ACPI & HPET is enabled in libvirt:
1. In <features> element there must exist <acpi/> sub-elemnt
(should be default):
<features>
<acpi/>
...
</features>
2. HPET must be enabled (disable by default!):
ensure that under <clock> element
is <timer name='hpet' present='yes'/>, for example:
<clock offset='utc'>
<timer name='rtc' tickpolicy='catchup'/>
<timer name='pit' tickpolicy='delay'/>
<timer name='hpet' present='yes'/>
</clock>
This was problem in my case - libvirt default is "<timer name='hpet'
present='no'/>" which hides HPET from OpenBSD guest causing this issue.
When running QEMU+KVM (you can verify command line from
/var/log/libvirt/qemu/VMNAME.log if using libvirt), verify that:
* "-no-acpi" is *absent*
* "-no-hpet" is *absent* - in my default configuration (that was also
published on this thread) it was present, causing these timeouts.
So now we know how to run OpenBSD 7.3 CD install under KVM without
timeouts and traps. Remaining mystery is:
* what causes memory corruption and trap on timeouts
* why older OpenBSD 7.2 CD works without problems even in such
configuration (as was reported by Winder Visen and confirmed on my
environment.)
So in short - ACPI enabled + HPET enabled is what make OpenBSD 7.3 work
with IDE CD under KVM.
Best regards
--Henryk Paluch
On 5/27/23 07:49, Mike Larkin wrote:
I don't know what's wrong with atapi CD emulation on wdc(4), my recommendation
would be to move the cd to a vioscsi device instead of wdc.
-ml
