[Bug 207489] Kernel panic due to Lazy update IOAPIC EOI on an x86_64 *host*, when two (or more) PCI devices from different IOMMU groups are passed to Windows 10 guest, upon guest boot into Windows, with more than 4 VCPUs

[bugzilla-daemon@bugzilla.kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=207489

Yani Stoyanov (yaweb@mail.bg) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |yaweb@mail.bg

--- Comment #16 from Yani Stoyanov (yaweb@mail.bg) ---
Kernel 5.8+
on Fedora 32

I have the same issue with macos (sierra and catalina) virtual machines with
passed through gpus. 

What I do to identify the problem: 

Try adding different cards to the mac vm's no matter amd/nvidia the vm hangs
and if you try to restart/force restart it the whole libvirt process hangs. 

I get the same kernel overflow as mentioned in the bug. 

How I am sure its the same problem: 

I checked out the kernel source comment the problematic: 

if (edge && kvm_apicv_activated(ioapic->kvm))
                ioapic_lazy_update_eoi(ioapic, irq);

Rebuild the kernel and try with my version and it was working fine. I also try
kernel 5.5.19 where it was fine also since it didn't have the check. 

I assume macos have some special way of treating the interrupts but since I am
new to the linux kernel and linux as hole I can not identify what exactly case
the problem.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are watching the assignee of the bug.