Troubleshooting
Serial size is fixed at 80x24
In -nographic mode, QEMU serial uses vt220 and cannot automatically fit to the host’s terminal size. Thus commands such as vim and tmux may become inconvenient to use. We can manually set terminal size in the VM.
First get the host terminal size. The first value below is rows and the second is columns.
stty size
# 42 130
Then set the terminal size in VM. Put this in VM’s ~/.bash_login
if [ ! -f /tmp/ALREADYRESET ]; then
touch /tmp/ALREADYRESET
stty rows 42 cols 130
tput init
fi
Alternatively, setup ssh or telnet to connect to the VM.
Force-quit QEMU
Sometimes our modified kernel may panic, and we need to force halt the VM. We can use pkill in the host to kill QEMU. Alternatively, press Ctrl-a x in the QEMU serial console. This is useful in text mode (-nographic mode).
QEMU will terminate and print:
QEMU: Terminated
Side note: There are other functionalities in Ctrl-a. Press Ctrl-a h for help. For example, Ctrl-a c switches between console and QEMU Monitor. You may see something like below if you accidentally pressed Ctrl-a c. Press Ctrl-a c to switch back to VM.
QEMU 2.11.1 monitor - type 'help' for more information
(qemu)
To send a real Ctrl-a (go to start of line) to the VM, press Ctrl-a twice. To send Ctrl-a into screen in the VM, press Ctrl-a Ctrl-a a.
Fixing corrupted file system
Sometimes if you happen to run the VM and mount the disk image at the same time, the file system may be messed up.
First unmount the disk image and shutdown the VM. Then use fsck tools (e2fsck for ext4) to fix the file system.
e2fsck -f qemu-image.img
In most cases if the VM, only file system stats needs to be recalculated and no data will loss.
There may have some data corruption or data loss In some cases.
Extending disk size
First extend the disk image size with qemu-img, and then use file system utility (resize2fs for ext4) to resize the file system.
qemu-img resize qemu-image.img +1G
e2fsck -f qemu-image.img
resize2fs qemu-image.img
Caution for shrinking: We need to change the above command order for shrinking. File system resizing must be done before resizing disk image.
Reduce VM image size
The VM image size can grow into an explosive size after a while (e.g., 30+ GB), even though the total data you add to the VM is definitely not that big. This is because of the free space in the VM image is not reclaimed. You can use the virt-sparsify tool to reduce the image size, which is easy to use and safe. Please read the man page of virt-sparsify first: https://libguestfs.org/virt-sparsify.1.html
Shut down the VM, then run the command virt-sparsify indisk.img outdisk.img. You should see significantly reduced image size for the outdisk.img. Test the new image with QEMU and verify it does not loose data. After you double check it works, you can delete indisk.img and rename outdisk.img to indisk.img.
If the VM image format is qcow2, you can further add --compress option to compress the large image.
Loss of VM network after installing new kernel
When you install the custom-built kernel for the first time, you will likely encounter a loss of VM network after the VM reboots. The root cause is because the network interface changed with the custom kernel, so you need to update the network configuration to restore the network functionality.
Symptom
The VM’s network interface no longer has an IP address:
$ sudo ifconfig -a
enp0s2: flags=4098<BROADCAST,MULTICAST> mtu 1500
ether 52:54:00:f0:8b:6c txqueuelen 1000 (Ethernet)
RX packets 0 bytes 0 (0.0 B)
RX errors 0 dropped 0 overruns 0 frame 0
...
On the host, the DHCP lease is also gone:
$ virsh net-dhcp-leases default
Expiry Time MAC address Protocol IP address Hostname Client ID or DUID
-------------------------------------------------------------------------------------------------------------------
Diagnosis
Check the network interface configuration file /etc/network/interfaces inside the VM:
...
allow-hotplug ens2
iface ens2 inet dhcp
There is an ens2 in the configuration file, but there is no such interface in the output of ifconfig -a. In addition, the ifconfig -a shows an interface enp0s2, which does not appear in the configuration file. This discrepancy leads to our diagnosis hypothesis that the network interface ens2 from the VM creation time no longer exists and a new interface enp0s2 is added with the custom kernel. To validate this hypothesis, simply update the ens2 in the configuration:
- allow-hotplug ens2
- iface ens2 inet dhcp
+ allow-hotplug enp0s2
+ iface enp0s2 inet dhcp
Reboot the VM, then check the ifconfig output:
$ sudo ifconfig -a
enp0s2: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 192.168.123.2 netmask 255.255.255.0 broadcast 192.168.123.255
inet6 fe80::5054:ff:fef0:8b6c prefixlen 64 scopeid 0x20<link>
...
It gets assigned with the IP address now.