b17693453e
At least for now cpu-topology is implemented only for KVM. We already say this, but this tries to be more explicit, and also show it in the examples. This adds a new reference in the introduction that we can point to, whenever we need to reference accelerators and how to select them. Signed-off-by: Claudio Fontana <cfontana@suse.de> Message-ID: <20240314172218.16478-1-cfontana@suse.de> Reviewed-by: Nina Schoetterl-Glausch <nsg@linux.ibm.com> Tested-by: Nina Schoetterl-Glausch <nsg@linux.ibm.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Thomas Huth <thuth@redhat.com>
220 lines
7.5 KiB
ReStructuredText
220 lines
7.5 KiB
ReStructuredText
Introduction
|
|
============
|
|
|
|
.. _Accelerators:
|
|
|
|
Virtualisation Accelerators
|
|
---------------------------
|
|
|
|
QEMU's system emulation provides a virtual model of a machine (CPU,
|
|
memory and emulated devices) to run a guest OS. It supports a number
|
|
of hypervisors (known as accelerators) as well as a JIT known as the
|
|
Tiny Code Generator (TCG) capable of emulating many CPUs.
|
|
|
|
.. list-table:: Supported Accelerators
|
|
:header-rows: 1
|
|
|
|
* - Accelerator
|
|
- Host OS
|
|
- Host Architectures
|
|
* - KVM
|
|
- Linux
|
|
- Arm (64 bit only), MIPS, PPC, RISC-V, s390x, x86
|
|
* - Xen
|
|
- Linux (as dom0)
|
|
- Arm, x86
|
|
* - Hypervisor Framework (hvf)
|
|
- MacOS
|
|
- x86 (64 bit only), Arm (64 bit only)
|
|
* - Windows Hypervisor Platform (whpx)
|
|
- Windows
|
|
- x86
|
|
* - NetBSD Virtual Machine Monitor (nvmm)
|
|
- NetBSD
|
|
- x86
|
|
* - Tiny Code Generator (tcg)
|
|
- Linux, other POSIX, Windows, MacOS
|
|
- Arm, x86, Loongarch64, MIPS, PPC, s390x, Sparc64
|
|
|
|
Feature Overview
|
|
----------------
|
|
|
|
System emulation provides a wide range of device models to emulate
|
|
various hardware components you may want to add to your machine. This
|
|
includes a wide number of VirtIO devices which are specifically tuned
|
|
for efficient operation under virtualisation. Some of the device
|
|
emulation can be offloaded from the main QEMU process using either
|
|
vhost-user (for VirtIO) or :ref:`Multi-process QEMU`. If the platform
|
|
supports it QEMU also supports directly passing devices through to
|
|
guest VMs to eliminate the device emulation overhead. See
|
|
:ref:`device-emulation` for more details.
|
|
|
|
There is a full :ref:`featured block layer<Live Block Operations>`
|
|
which allows for construction of complex storage topology which can be
|
|
stacked across multiple layers supporting redirection, networking,
|
|
snapshots and migration support.
|
|
|
|
The flexible ``chardev`` system allows for handling IO from character
|
|
like devices using stdio, files, unix sockets and TCP networking.
|
|
|
|
QEMU provides a number of management interfaces including a line based
|
|
:ref:`Human Monitor Protocol (HMP)<QEMU monitor>` that allows you to
|
|
dynamically add and remove devices as well as introspect the system
|
|
state. The :ref:`QEMU Monitor Protocol<QMP Ref>` (QMP) is a well
|
|
defined, versioned, machine usable API that presents a rich interface
|
|
to other tools to create, control and manage Virtual Machines. This is
|
|
the interface used by higher level tools interfaces such as `Virt
|
|
Manager <https://virt-manager.org/>`_ using the `libvirt framework
|
|
<https://libvirt.org>`_.
|
|
|
|
For the common accelerators QEMU, supported debugging with its
|
|
:ref:`gdbstub<GDB usage>` which allows users to connect GDB and debug
|
|
system software images.
|
|
|
|
Running
|
|
-------
|
|
|
|
QEMU provides a rich and complex API which can be overwhelming to
|
|
understand. While some architectures can boot something with just a
|
|
disk image, those examples elide a lot of details with defaults that
|
|
may not be optimal for modern systems.
|
|
|
|
For a non-x86 system where we emulate a broad range of machine types,
|
|
the command lines are generally more explicit in defining the machine
|
|
and boot behaviour. You will find often find example command lines in
|
|
the :ref:`system-targets-ref` section of the manual.
|
|
|
|
While the project doesn't want to discourage users from using the
|
|
command line to launch VMs, we do want to highlight that there are a
|
|
number of projects dedicated to providing a more user friendly
|
|
experience. Those built around the ``libvirt`` framework can make use
|
|
of feature probing to build modern VM images tailored to run on the
|
|
hardware you have.
|
|
|
|
That said, the general form of a QEMU command line can be expressed
|
|
as:
|
|
|
|
.. parsed-literal::
|
|
|
|
$ |qemu_system| [machine opts] \\
|
|
[cpu opts] \\
|
|
[accelerator opts] \\
|
|
[device opts] \\
|
|
[backend opts] \\
|
|
[interface opts] \\
|
|
[boot opts]
|
|
|
|
Most options will generate some help information. So for example:
|
|
|
|
.. parsed-literal::
|
|
|
|
$ |qemu_system| -M help
|
|
|
|
will list the machine types supported by that QEMU binary. ``help``
|
|
can also be passed as an argument to another option. For example:
|
|
|
|
.. parsed-literal::
|
|
|
|
$ |qemu_system| -device scsi-hd,help
|
|
|
|
will list the arguments and their default values of additional options
|
|
that can control the behaviour of the ``scsi-hd`` device.
|
|
|
|
.. list-table:: Options Overview
|
|
:header-rows: 1
|
|
:widths: 10, 90
|
|
|
|
* - Options
|
|
-
|
|
* - Machine
|
|
- Define the machine type, amount of memory etc
|
|
* - CPU
|
|
- Type and number/topology of vCPUs. Most accelerators offer
|
|
a ``host`` cpu option which simply passes through your host CPU
|
|
configuration without filtering out any features.
|
|
* - Accelerator
|
|
- This will depend on the hypervisor you run. Note that the
|
|
default is TCG, which is purely emulated, so you must specify an
|
|
accelerator type to take advantage of hardware virtualization.
|
|
* - Devices
|
|
- Additional devices that are not defined by default with the
|
|
machine type.
|
|
* - Backends
|
|
- Backends are how QEMU deals with the guest's data, for example
|
|
how a block device is stored, how network devices see the
|
|
network or how a serial device is directed to the outside world.
|
|
* - Interfaces
|
|
- How the system is displayed, how it is managed and controlled or
|
|
debugged.
|
|
* - Boot
|
|
- How the system boots, via firmware or direct kernel boot.
|
|
|
|
In the following example we first define a ``virt`` machine which is a
|
|
general purpose platform for running Aarch64 guests. We enable
|
|
virtualisation so we can use KVM inside the emulated guest. As the
|
|
``virt`` machine comes with some built in pflash devices we give them
|
|
names so we can override the defaults later.
|
|
|
|
.. code::
|
|
|
|
$ qemu-system-aarch64 \
|
|
-machine type=virt,virtualization=on,pflash0=rom,pflash1=efivars \
|
|
-m 4096 \
|
|
|
|
We then define the 4 vCPUs using the ``max`` option which gives us all
|
|
the Arm features QEMU is capable of emulating. We enable a more
|
|
emulation friendly implementation of Arm's pointer authentication
|
|
algorithm. We explicitly specify TCG acceleration even though QEMU
|
|
would default to it anyway.
|
|
|
|
.. code::
|
|
|
|
-cpu max,pauth-impdef=on \
|
|
-smp 4 \
|
|
-accel tcg \
|
|
|
|
As the ``virt`` platform doesn't have any default network or storage
|
|
devices we need to define them. We give them ids so we can link them
|
|
with the backend later on.
|
|
|
|
.. code::
|
|
|
|
-device virtio-net-pci,netdev=unet \
|
|
-device virtio-scsi-pci \
|
|
-device scsi-hd,drive=hd \
|
|
|
|
We connect the user-mode networking to our network device. As
|
|
user-mode networking isn't directly accessible from the outside world
|
|
we forward localhost port 2222 to the ssh port on the guest.
|
|
|
|
.. code::
|
|
|
|
-netdev user,id=unet,hostfwd=tcp::2222-:22 \
|
|
|
|
We connect the guest visible block device to an LVM partition we have
|
|
set aside for our guest.
|
|
|
|
.. code::
|
|
|
|
-blockdev driver=raw,node-name=hd,file.driver=host_device,file.filename=/dev/lvm-disk/debian-bullseye-arm64 \
|
|
|
|
We then tell QEMU to multiplex the :ref:`QEMU monitor` with the serial
|
|
port output (we can switch between the two using :ref:`keys in the
|
|
character backend multiplexer`). As there is no default graphical
|
|
device we disable the display as we can work entirely in the terminal.
|
|
|
|
.. code::
|
|
|
|
-serial mon:stdio \
|
|
-display none \
|
|
|
|
Finally we override the default firmware to ensure we have some
|
|
storage for EFI to persist its configuration. That firmware is
|
|
responsible for finding the disk, booting grub and eventually running
|
|
our system.
|
|
|
|
.. code::
|
|
|
|
-blockdev node-name=rom,driver=file,filename=(pwd)/pc-bios/edk2-aarch64-code.fd,read-only=true \
|
|
-blockdev node-name=efivars,driver=file,filename=$HOME/images/qemu-arm64-efivars
|