Guests using the qxl-vga are freezing

[Leonardo Müller]

I have noticed that guests using qxl-vga are freezing. They may freeze after a few minutes or after many hours. The freeze consists of the entire system hanging, except the cursor, but the cursor animation stops too. Changing to tty is not possible after this. There are three things noticed in common on the guests when they freeze:

-The guest is using the QXL VGA (freezes weren't observed with other VGAs);
-A new workload is starting;
-The mouse cursor is the animated as the one of loading. For example, https://i.imgur.com/raQFteG.png

The host is Xubuntu 18.04 amd64, QEMU version is 3.0.0-dirty. The guests tested were:

-openSUSE Tumbleweed;
-openSUSE Leap 15;
-Xubuntu 18.04 Bionic Beaver;
-CentOS 7.

With openSUSE guests, the install process couldn't even be finished, as the installer would freeze. There were 2 GB of available memory (checked in a tty before the freeze) and netconsole was enabled. Unfortunately, it was impossible to obtain any information from them. This is an image of one openSUSE guest frozen: https://i.imgur.com/ZP0eQKq.png

The command line used was:

qemu-system-x86_64 -nodefaults -m 3072 -M pc,usb=true -accel kvm -cpu host -smp cores=2,threads=2 -device qemu-xhci -drive id=centusb,if=none,file=leap.qcow2 -device usb-storage,id=centusb,drive=centusb -netdev user,id=n0 -device usb-tablet,id=usbtablet -device e1000,netdev=n0 -device usb-audio,id=usbaudio -device qxl-vga,xres=1366,yres=768 -display gtk -monitor vc -serial vc -cdrom "openSUSE-Leap-15.0-DVD-x86_64.iso" -boot d

With CentOS guests, the install process fail sometimes, but sometimes it's able to install. However, on the yum update, it would freeze too. In one instance it froze while updating glibc, which made the guest unbootable. https://i.imgur.com/B3WhSDX.png

The command line used was:

qemu-system-x86_64 -nodefaults -m 2048 -M pc,usb=true -accel kvm -cpu host -smp cores=2,threads=2 -device qemu-xhci -drive id=centusb,if=none,file=centos.qcow2 -device usb-storage,id=centusb,drive=centusb -netdev user,id=n0 -device usb-tablet,id=usbtablet -device e1000,netdev=n0 -device usb-audio,id=usbaudio -device virtio-vga,virgl=true -display gtk -monitor vc -serial vc -cdrom "CentOS-7-livecd-GNOME-x86_64.iso" -bios /usr/share/ovmf/OVMF.fd

With Xubuntu 18.04 guests, the system worked for many hours until the freeze happened. On this case it happened when opening Audacious. Fortunately, the logging services worked for some time, which allowed me to get a relevant message which can be seen at http://termbin.com/nuof . It repeated a few times, but then the logging stopped. https://i.imgur.com/2zckqj5.png shows the guest screen in the moment it froze.

The command line used was:

qemu-system-x86_64 -nodefaults -m 1024 -M pc,usb=true -accel kvm -cpu host -smp cores=2,threads=2 -device qemu-xhci -drive id=centusb,if=none,file=xubmini -device usb-storage,id=centusb,drive=centusb -netdev user,id=n0 -device usb-tablet,id=usbtablet -device e1000,netdev=n0 -device usb-audio,id=usbaudio -device qxl-vga,xres=1366,yres=768 -display gtk -monitor vc -serial vc

I'm sorry for not having more detailed information but, even setting netconsole, openSUSE and CentOS guests were unable to print any information.

[Gerd Hoffmann (kraxel-redhat)]

Can you attach gdb to qemu when the guest hangs, then run "thread apply all bt"?

[Leonardo Müller (leozinho29-eu)]

I have run the openSUSE guest again, the output of the gdb command can be seen at https://paste.ubuntu.com/p/mCRMvkNWYq/

[Gerd Hoffmann (kraxel-redhat)]

Hmm, nothing obvious in the stack trace.
Also doesn't reproduce here.
What qemu version is this?
If older than 3.0: Does it still happen for you with qemu 3.0?

[Leonardo Müller (leozinho29-eu)]

QEMU version I'm using now is v3.0.0-65-g1d746ee95d-dirty and has the problem, the first version I observed the problem was 2.11 from Ubuntu's repositories.

The fact you are unable to reproduce the bug is unsurprising to me: this is not the first time I have a problem with QXL VGA and others are unable to reproduce. I don't understand why QXL fails so consistently here.

Another QXL bug I reported:

https://bugs.launchpad.net/ubuntu/bionic/+source/qemu/+bug/1755912

[Thomas Huth (th-huth)]

The QEMU project is currently considering to move its bug tracking to another system. For this we need to know which bugs are still valid and which could be closed already. Thus we are setting older bugs to "Incomplete" now.
If you still think this bug report here is valid, then please switch the state back to "New" within the next 60 days, otherwise this report will be marked as "Expired". Or mark it as "Fix Released" if the problem has been solved with a newer version of QEMU already. Thank you and sorry for the inconvenience.

[Launchpad Janitor (janitor)]

[Expired for QEMU because there has been no activity for 60 days.]