blob: fcf41fc1f6f67e713eb71b6f196598d6afc14e8c [file] [log] [blame]
microvm Machine Type
``microvm`` is a machine type inspired by ``Firecracker`` and
constructed after its machine model.
It's a minimalist machine type without ``PCI`` nor ``ACPI`` support,
designed for short-lived guests. microvm also establishes a baseline
for benchmarking and optimizing both QEMU and guest operating systems,
since it is optimized for both boot time and footprint.
Supported devices
The microvm machine type supports the following devices:
- ISA bus
- i8259 PIC (optional)
- i8254 PIT (optional)
- MC146818 RTC (optional)
- One ISA serial port (optional)
- IOAPIC (with kernel-irqchip=split by default)
- kvmclock (if using KVM)
- fw_cfg
- Up to eight virtio-mmio devices (configured by the user)
Currently, microvm does *not* support the following features:
- PCI-only devices.
- Hotplug of any kind.
- Live migration across QEMU versions.
Using the microvm machine type
Machine-specific options
It supports the following machine-specific options:
- microvm.x-option-roms=bool (Set off to disable loading option ROMs)
- microvm.pit=OnOffAuto (Enable i8254 PIT)
- microvm.isa-serial=bool (Set off to disable the instantiation an ISA serial port)
- microvm.pic=OnOffAuto (Enable i8259 PIC)
- microvm.rtc=OnOffAuto (Enable MC146818 RTC)
- (Set off to disable adding virtio-mmio devices to the kernel cmdline)
Boot options
By default, microvm uses ``qboot`` as its BIOS, to obtain better boot
times, but it's also compatible with ``SeaBIOS``.
As no current FW is able to boot from a block device using
``virtio-mmio`` as its transport, a microvm-based VM needs to be run
using a host-side kernel and, optionally, an initrd image.
Running a microvm-based VM
By default, microvm aims for maximum compatibility, enabling both
legacy and non-legacy devices. In this example, a VM is created
without passing any additional machine-specific option, using the
legacy ``ISA serial`` device as console::
$ qemu-system-x86_64 -M microvm \
-enable-kvm -cpu host -m 512m -smp 2 \
-kernel vmlinux -append "earlyprintk=ttyS0 console=ttyS0 root=/dev/vda" \
-nodefaults -no-user-config -nographic \
-serial stdio \
-drive id=test,file=test.img,format=raw,if=none \
-device virtio-blk-device,drive=test \
-netdev tap,id=tap0,script=no,downscript=no \
-device virtio-net-device,netdev=tap0
While the example above works, you might be interested in reducing the
footprint further by disabling some legacy devices. If you're using
``KVM``, you can disable the ``RTC``, making the Guest rely on
``kvmclock`` exclusively. Additionally, if your host's CPUs have the
``TSC_DEADLINE`` feature, you can also disable both the i8259 PIC and
the i8254 PIT (make sure you're also emulating a CPU with such feature
in the guest).
This is an example of a VM with all optional legacy features
$ qemu-system-x86_64 \
-M microvm,x-option-roms=off,pit=off,pic=off,isa-serial=off,rtc=off \
-enable-kvm -cpu host -m 512m -smp 2 \
-kernel vmlinux -append "console=hvc0 root=/dev/vda" \
-nodefaults -no-user-config -nographic \
-chardev stdio,id=virtiocon0 \
-device virtio-serial-device \
-device virtconsole,chardev=virtiocon0 \
-drive id=test,file=test.img,format=raw,if=none \
-device virtio-blk-device,drive=test \
-netdev tap,id=tap0,script=no,downscript=no \
-device virtio-net-device,netdev=tap0
Triggering a guest-initiated shut down
As the microvm machine type includes just a small set of system
devices, some x86 mechanisms for rebooting or shutting down the
system, like sending a key sequence to the keyboard or writing to an
ACPI register, doesn't have any effect in the VM.
The recommended way to trigger a guest-initiated shut down is by
generating a ``triple-fault``, which will cause the VM to initiate a
reboot. Additionally, if the ``-no-reboot`` argument is present in the
command line, QEMU will detect this event and terminate its own
execution gracefully.
Linux does support this mechanism, but by default will only be used
after other options have been tried and failed, causing the reboot to
be delayed by a small number of seconds. It's possible to instruct it
to try the triple-fault mechanism first, by adding ``reboot=t`` to the
kernel's command line.