Greetings Daniel,
Sent: Tuesday, August 03, 2021 at 4:12 PM From: "Daniel P. Berrange" <dan@berrange.com> To: "daggs" <daggs@gmx.com> Cc: "Martin Kletzander" <mkletzan@redhat.com>, libvirt-users@redhat.com Subject: Re: issues with vm after upgrade
On Tue, Aug 03, 2021 at 02:54:39PM +0200, daggs wrote:
Greetings Martin,
Sent: Tuesday, August 03, 2021 at 11:37 AM From: "Martin Kletzander" <mkletzan@redhat.com> To: "daggs" <daggs@gmx.com> Cc: libvirt-users@redhat.com Subject: Re: issues with vm after upgrade
On Mon, Aug 02, 2021 at 04:33:20PM +0200, daggs wrote:
Greetings Martin,
Sent: Monday, August 02, 2021 at 12:17 PM From: "Martin Kletzander" <mkletzan@redhat.com> To: "daggs" <daggs@gmx.com> Cc: libvirt-users@redhat.com Subject: Re: issues with vm after upgrade
On Sat, Jul 31, 2021 at 12:19:59PM +0200, daggs wrote:
Greetings,
a few weeks ago I've upgraded my system, this resulted with qemu and libvirt being upgraded to 6.0.0 and ~7.5.0 respectfully. I have two vms running on my system, router and streamer. the router vm works great, the streamer vm doesn't. after the streamer vm start, the monitor screen gets black and thats it. no relevant error are found in the log, see: http://dpaste.com/ERWDEJQPC the xml ca be found at https://dpaste.com/FQDN6NTN2 in contrast the following oneliner works: qemu-system-x86_64 \ -machine pc-q35-5.0,accel=kvm,usb=off,smm=on,dump-guest-core=off \ -cpu host,migratable=on \ -m 15360 \ -smp 4,sockets=1,dies=1,cores=2,threads=2 \ -drive file=/home/streamer/streamer.img.qcow2.new,if=virtio,format=qcow2 \ -device vfio-pci,host=0000:00:02.0,romfile=/home/streamer/gpu-8086:5912-uefi.rom,multifunction=on \ -device vfio-pci,host=0000:00:1f.3,multifunction=on \ -usb \ -device usb-host,vendorid=0x046d,productid=0xc52e \ -device usb-host,vendorid=0x2548,productid=0x1002 \ -display none \ -netdev tap,id=hostnet0,ifname=virtsw-streamer,script=no,downscript=no \ -device e1000e,netdev=hostnet0,id=net0,mac=52:54:00:5a:4c:8c \ -blockdev '{"driver":"file","filename":"/usr/share/edk2-ovmf/OVMF_CODE.secboot.fd","node-name":"libvirt-pflash0-storage","auto-read-only":true,"discard":"unmap"}' \ -blockdev '{"driver":"file","filename":"/var/lib/libvirt/qemu/nvram/streamer-vm-q35_VARS.fd","node-name":"libvirt-pflash1-storage","auto-read-only":true,"discard":"unmap"}'
another issue encountered caused by this param: "-audiodev id=audio1,driver=none" which is auto added to the qemu line. if I add the line to the above oneliner, the guest doesn't detects the pt sound card. without it it does and it works.
The links have expired, so I cannot look at the XML. However do you have any <audio/> in the XML at all? I believe that without it we are disabling any audio backends since no audio HW was requested.
here are the links: 1. xml: https://dpaste.com/D6JANX3Z3 2. log: https://dpaste.com/FMUDZY9PD
as for the audio entry, " <audio id='1' type='none'/>" is in the xml I didn't added it
So let get this straight. You want the VM to use the HW GPU that you assigned to it and also use the audio on that GPU (e.g. HDMI or DP output). When you add the "-audiodev driver=none" the VM cannot see the audio device on that GPU card? Or is that unrelated?
There were some strict definitions added for audio devices lately, I presume to make sure that no data gets leaked in/out the VM and in order to get audio anywhere else than from/to VNC/spice you'd need to configure that. It should not affect device assignment as far as I understand it, although there might be some misunderstanding.
I'll try to explain, I'm passing the computer's O/B gpu and O/B sound card to the vm so I can use them inside the vm which serves as streamer. all was working well until I've upgraded both qemu and libvirt (target versions are started in the original mail) since then, the vm doesn't boot at all. I was able to bisect the qemu cmd libvirt generates partially and go the vm up. this points to a config issue. as part of the testing I've found out that when the vm boots, it has no sound card. following your suggestion, I've found that the xml has this: <audio id='1' type='none'/> I have never added it to the xml so I assume that it was added as part of the upgrade. based on this I assume that there is another config which prevents my vm from booting. unfortunately, I don't have the xml prior to the upgrade to diff them.
The <audio> element just refers to the *host* backend used for audio playback. It would not affect guest hardware. Further, this has always existed - it just wasn't exposed in the XML previously.
the upgrade changed something, here is the qemu cmd before the upgrade: https://dpaste.com/F2N5T8CT8 here is after https://dpaste.com/F2N5T8CT8 notice diff in audio. btw, can the efi fw file change prevent the vm from booting? I din't change that too Dagg