Mirrored_Repos/LookingGlass
1
0
Fork 0
mirror of https://github.com/gnif/LookingGlass.git synced 2024-09-19 08:51:33 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[doc] update and restructure installation documentation

Browse Source
This commit is contained in:
Geoffrey McRae 2024-03-08 18:22:07 +11:00
parent eb31815b46
commit 4a4f72ba38
9 changed files with 657 additions and 404 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
doc/build.rst
View File

@ -132,6 +132,15 @@ Fetching with APT
You can fetch these dependencies with the following command:
.. warning::
Do not just blindly install the list below, check if you are using PipeWire
or PulseAudio and adjust the list accordingly. Installing PipeWire libraries
on a PulseAudio system will result in a broken partial PipeWire install.
If you are not already using PipeWire we highly recommend you upgrade,
Looking Glass does not support audio input (microphone) with PulseAudio.
.. code:: bash
apt-get install binutils-dev cmake fonts-dejavu-core libfontconfig-dev \

2
doc/conf.py
View File

@ -22,7 +22,7 @@ from lgrelease import release
# -- Project information -----------------------------------------------------
project = 'Looking Glass'
copyright = '2022, Looking Glass team'
copyright = '2024, Looking Glass team'
author = 'Geoffrey McRae and the Looking Glass team'
rst_prolog = """

2
doc/index.rst
View File

@ -11,9 +11,9 @@ systems for legacy programs that require high-performance graphics.
:maxdepth: 2
requirements
build
install
usage
build
troubleshooting
obs
module

406
doc/install.rst
View File

@ -3,406 +3,8 @@
Installation
############
.. _libvirt:
libvirt/QEMU configuration
--------------------------
This article assumes you already have a fully functional libvirt domain with
PCI passthrough working.
If you use virt-manager, this guide also applies to you, since virt-manager uses
libvirt as its back-end.
.. _libvirt_ivshmem:
IVSHMEM
^^^^^^^
Configuration
~~~~~~~~~~~~~
.. note::
If your host GPU is either AMD or Intel it is better to set this up using the
KVMFR kernel module as this will allow you to make use of DMA transfers to
offload some of the memory transfers to the GPU.
See `VM->host` in :ref:`kernel_module`.
Add the following to your libvirt machine configuration inside the
'devices' section by running ``virsh edit <VM>`` where ``<VM>`` is the name of
your virtual machine.
.. code:: xml
<shmem name='looking-glass'>
<model type='ivshmem-plain'/>
<size unit='M'>32</size>
</shmem>
.. note::
If you are using QEMU directly without libvirt the following arguments are
required instead.
Add the following to the commands to your QEMU command line, adjusting
the ``bus`` parameter to suit your particular configuration:
.. code:: bash
-device ivshmem-plain,memdev=ivshmem,bus=pcie.0 \
-object memory-backend-file,id=ivshmem,share=on,mem-path=/dev/shm/looking-glass,size=32M
The memory size (show as 32 in the example above) may need to be
adjusted as per the :ref:`Determining memory <libvirt_determining_memory>`
section.
.. warning::
If you change the size of this after starting your virtual machine you may
need to remove the file `/dev/shm/looking-glass` to allow QEMU to re-create
it with the correct size. If you do this the permissions of the file may be
incorrect for your user to be able to access it and you will need to correct
this. See :ref:`libvirt_shmfile_permissions`
.. _libvirt_determining_memory:
Determining memory
~~~~~~~~~~~~~~~~~~
You will need to adjust the memory size to be suitable for your desired maximum
resolution, with the following formula:
.. code:: text
.. toctree::
width x height x pixel size x 2 = frame bytes
frame bytes / 1024 / 1024 = frame megabytes
frame megabytes + 10 MiB = total megabytes
Where `pixel size` is 4 for 32-bit RGB (SDR) or 8 for 64-bit
(HDR :ref:`* <libvirt_determining_memory_hdr>`).
Failure to do so will cause Looking Glass to truncate the bottom of the screen
and will trigger a message popup to inform you of the size you need to increase
the value to.
For example, for a resolution of 1920x1080 (1080p):
.. code:: text
1920 x 1080 x 4 x 2 = 16,588,800 bytes
16,588,800 / 1024 / 1024 = 15.82 MiB
15.82 MiB + 10 MiB = 25.82 MiB
You must round this value up to the nearest power of two, which for the
provided example is 32 MiB.
.. note::
Increasing this value beyond what you need does not yield any performance
improvements, it simply will block access to that RAM making it unusable by
your system.
.. list-table:: Common Values
:widths: 50 25 25
:header-rows: 1
* - Resolution
- Standard Dynamic Range
- High Dynamic Range (HDR) :ref:`* <libvirt_determining_memory_hdr>`
* - 1920x1080 (1080p)
- 32
- 64
* - 1920x1200 (1200p)
- 32
- 64
* - 2560x1440 (1440p)
- 64
- 128
* - 3840x2160 (2160p/4K)
- 128
- 256
.. _libvirt_determining_memory_hdr:
.. warning::
While Looking Glass can capture and display HDR, at the time of writing
neither Xorg or Wayland can make use of it and it will be converted by the
GPU drivers/hardware to SDR. Additionally using HDR doubles the amount of
memory, bandwidth, and CPU load and should generally not be used unless you
have a special reason to do so.
.. _libvirt_shmfile_permissions:
Permissions
~~~~~~~~~~~
The shared memory file used by IVSHMEM is found in ``/dev/shm/looking-glass``.
By default, it is owned by QEMU, and does not give read/write permissions to
your user, which are required for Looking Glass to run properly.
You can use ``systemd-tmpfiles`` to create the file before running your VM,
granting the necessary permissions which allow Looking Glass to use the file
properly.
Create a new file ``/etc/tmpfiles.d/10-looking-glass.conf``, and populate it
with the following::
# Type Path Mode UID GID Age Argument
f /dev/shm/looking-glass 0660 user kvm -
Change ``UID`` to the user name you will run Looking Glass with, usually your
own.
.. _libvirt_spice_server:
Keyboard/mouse/display/audio
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Looking Glass makes use of the SPICE protocol to provide keyboard and mouse
input, audio input and output, and display fallback.
.. note::
The default configuration that libvirt uses is not optimal and must be
adjusted. Failure to perform these changes will cause input issues along
with failure to support 5 button mice.
If you would like to use SPICE to give you keyboard and mouse input
along with clipboard sync support, make sure you have a
``<graphics type='spice'>`` device, then:
- Find your ``<video>`` device, and set ``<model type='vga'/>``
- If you can't find it, make sure you have a ``<graphics>``
device, save and edit again.
- Remove the ``<input type='tablet'/>`` device, if you have one.
- Create an ``<input type='mouse' bus='virtio'/>`` device, if you don't
already have one.
- Create an ``<input type='keyboard' bus='virtio'/>`` device to improve
keyboard usage.
.. note::
Be sure to install the the *vioinput* driver from
`virtio-win <https://fedorapeople.org/groups/virt/virtio-win/direct-downloads/stable-virtio/>`_
in the guest
To enable audio support add a standard Intel HDA audio device to your
configuration as per below:
.. code:: xml
<sound model='ich9'>
<audio id='1'/>
</sound>
<audio id='1' type='spice'/>
If you also want clipboard synchronization please see
:ref:`libvirt_clipboard_synchronization`
.. _libvirt_clipboard_synchronization:
Clipboard synchronization
^^^^^^^^^^^^^^^^^^^^^^^^^
Looking Glass can synchronize the clipboard between the host and guest using
the SPICE guest agent.
1. Install the SPICE guest tools from
https://www.spice-space.org/download.html#windows-binaries.
2. Configure your VM to enable the SPICE guest agent:
- QEMU
.. code:: bash
-device virtio-serial-pci \
-chardev spicevmc,id=vdagent,name=vdagent \
-device virtserialport,chardev=vdagent,name=com.redhat.spice.0
- libvirt
.. code:: xml
<channel type="spicevmc">
<target type="virtio" name="com.redhat.spice.0"/>
<address type="virtio-serial" controller="0" bus="0" port="1"/>
</channel>
<!-- No need to add a VirtIO Serial device, it will be added automatically -->
.. _libvirt_apparmor:
AppArmor
^^^^^^^^
For libvirt versions before **5.10.0**, if you are using AppArmor, you
need to add permissions for QEMU to access the shared memory file. This
can be done by adding the following to
``/etc/apparmor.d/local/abstractions/libvirt-qemu``::
/dev/shm/looking-glass rw,
then, restart AppArmor.
.. code:: bash
sudo systemctl restart apparmor
.. _libvirt_memballoon_tweak:
Memballoon
^^^^^^^^^^
The VirtIO memballoon device enables the host to dynamically reclaim memory
from your VM by growing the balloon inside the guest, reserving reclaimed
memory. Libvirt adds this device to guests by default.
However, this device causes major performance issues with VFIO passthrough
setups, and should be disabled.
Find the ``<memballoon>`` tag and set its type to ``none``:
.. code:: xml
<memballoon model="none"/>
.. _host_install:
Additional tuning
^^^^^^^^^^^^^^^^^
Looking Glass is latency sensitive and as such it may suffer microstutters if
you have not properly tuned your virtual machine. The physical display output
of your GPU will usually not show such issues due to the nature of the hardware
but be sure that if you are experiencing issues the following tuning is
required to obtain optimal performance.
1. Do not assign all your CPU cores to your guest VM, you must at minimum
reserve two CPU cores (4 threads) for your host system to use. For example,
if you have a 6 core CPU, only assign 4 cores (8 threads) to the guest.
2. Ensure you correctly pin your VMs vCPU threads to the correct cores for your
CPU architecture.
3. If you are on a NUMA architecture (dual CPU, or early Threadripper) be sure
that you pin the vCPU threads to the physical CPU/die attached to your GPU.
4. Just because your GPU is in a slot that is physically x16 in size, does not
mean your GPU is running at x16, this is dependent on how your motherboard
is physically wired and the physical slot may be limited to x4 or x8.
5. Be sure to set your CPU model type to `host-passthrough` so that your guest
operating system is aware of the acceleration features of your CPU and can
make full use of them.
6. AMD users be sure that you have the CPU feature flag `topoext` enabled or
your guest operating system will not be aware of which CPU cores are
hyper-thread pairs.
7. NVIDIA users may want to enable NvFBC as an alternative capture API in the
guest. Note that NvFBC is officially available on professional cards only
and methods to enable NvFBC on non-supported GPUs is against the NVIDIA
Capture API SDK License Agreement even though GeForce Experience and
Steam make use of it on any NVIDIA GPU.
How to perform these changes is left as an exercise to the reader.
Host application
----------------
The Looking Glass Host application captures frames from the guest OS using a
capture API, and sends them to the
:ref:`client <client_install>`—be it on the host OS (hypervisor) or another
Virtual Machine—through a low-latency transfer protocol over shared memory.
You can get the host program in two ways:
- Download a pre-built binary from https://looking-glass.io/downloads
(**recommended**)
- Download the source code as described in :ref:`building`, then
:ref:`build the host <host_building>`.
.. _host_install_linux:
For Linux
^^^^^^^^^
While the host application can be compiled and is somewhat functional for Linux
it is currently considered incomplete and not ready for usage. As such use at
your own risk and do not ask for support.
.. _host_install_osx:
For OSX
^^^^^^^
Currently there is no support or plans for support for OSX due to technical
limitations.
.. _host_install_windows:
For Windows
^^^^^^^^^^^
To begin, you must first run the Windows VM with the changes noted above in
either the :ref:`libvirt` section.
.. _installing_the_ivshmem_driver:
Installing the IVSHMEM driver
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Since B6 the host installer available on the official Looking Glass website
comes with the IVSHMEM driver and will install this for you. If you are running
an older version of Looking Glass please refer to the documentation for your
version.
.. _host_install_service:
Installing the Looking Glass service
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
After installing your IVSHMEM driver, we can now install the Looking Glass Host
onto our Windows Virtual Machine.
1. First, run ``looking-glass-host-setup.exe`` as an administrator
(:ref:`Why? <faq_host_admin_privs>`)
2. You will be greeted by an intro screen. Press ``Next`` to continue.
3. You are presented with the |license| license. Please read and agree to the
license by pressing ``Agree``.
4. You can change the install path if you wish, otherwise press ``Next`` to
continue.
5. You may enable or disable options on this screen to configure the
installation. The default values are recommended for most users.
Press ``Install`` to begin installation.
6. After a few moments, installation will complete, and you will have a
running instance of Looking Glass. If you experience failures, you can
see them in the install log appearing in the middle of the window.
7. Press ``Close`` to exit the installer.
Command line users can run ``looking-glass-host-setup.exe /S`` to execute a
silent install with default options selected. Further configuration from the
command line can be done with flags. You can list all available flags by
running ``looking-glass-host-setup.exe /?``.
.. _client_install:
Client application
------------------
The Looking Glass client receives frames from the :ref:`host <host_install>` to
display on your screen. It also handles input, and can optionally share the
system clipboard with your guest OS through SPICE.
First you must build the client from source, see :ref:`building`. Once you have
built the client, you can install it. Run the following as root::
make install
To install for the local user only, run::
cmake -DCMAKE_INSTALL_PREFIX=~/.local .. && make install
install_libvirt
install_host
install_client

22
doc/install_client.rst Normal file
View File

@ -0,0 +1,22 @@
.. _installing_client:
Client Application Installation
###############################
.. _client_install:
For Linux
---------
The Looking Glass client receives frames from the :ref:`host <host_install>` to
display on your screen. It also handles input, and can optionally share the
system clipboard with your guest OS through SPICE.
First you must build the client from source, see :ref:`building`. Once you have
built the client, you can install it. Run the following as root::
make install
To install for the local user only, run::
cmake -DCMAKE_INSTALL_PREFIX=~/.local .. && make install

81
doc/install_host.rst Normal file
View File

@ -0,0 +1,81 @@
.. _installing_host:
Host Application Installation
#############################
The Looking Glass Host application captures frames from the guest OS using a
capture API, and sends them to the
:ref:`client <client_install>`—be it on the host OS (hypervisor) or another
Virtual Machine—through a low-latency transfer protocol over shared memory.
You can get the host program in two ways:
- Download a pre-built binary from https://looking-glass.io/downloads
(**recommended**)
- Download the source code as described in :ref:`building`, then
:ref:`build the host <host_building>`.
.. _host_install_linux:
For Linux
^^^^^^^^^
While the host application can be compiled and is somewhat functional for Linux
it is currently considered incomplete and not ready for usage. As such use at
your own risk and do not ask for support.
.. _host_install_osx:
For OSX
^^^^^^^
Currently there is no support or plans for support for OSX due to technical
limitations.
.. _host_install_windows:
For Windows
^^^^^^^^^^^
To begin, you must first run the Windows VM with the changes noted above in
either the :ref:`libvirt` section.
.. _installing_the_ivshmem_driver:
Installing the IVSHMEM driver
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Since B6 the host installer available on the official Looking Glass website
comes with the IVSHMEM driver and will install this for you. If you are running
an older version of Looking Glass please refer to the documentation for your
version.
.. _host_install_service:
Installing the Looking Glass service
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
After installing your IVSHMEM driver, we can now install the Looking Glass Host
onto our Windows Virtual Machine.
1. First, run ``looking-glass-host-setup.exe`` as an administrator
(:ref:`Why? <faq_host_admin_privs>`)
2. You will be greeted by an intro screen. Press ``Next`` to continue.
3. You are presented with the |license| license. Please read and agree to the
license by pressing ``Agree``.
4. You can change the install path if you wish, otherwise press ``Next`` to
continue.
5. You may enable or disable options on this screen to configure the
installation. The default values are recommended for most users.
Press ``Install`` to begin installation.
6. After a few moments, installation will complete, and you will have a
running instance of Looking Glass. If you experience failures, you can
see them in the install log appearing in the middle of the window.
7. Press ``Close`` to exit the installer.
Command line users can run ``looking-glass-host-setup.exe /S`` to execute a
silent install with default options selected. Further configuration from the
command line can be done with flags. You can list all available flags by
running ``looking-glass-host-setup.exe /?``.

255
doc/install_libvirt.rst Normal file
View File

@ -0,0 +1,255 @@
.. _installing_libvirt:
libvirt/QEMU Installation
#########################
This article assumes you already have a fully functional `libvirt` domain with
PCI passthrough working. If you use `virt-manager`, this guide also applies to
you, since virt-manager uses `libvirt` as its back end.
.. _libvirt_determining_memory:
Determining memory
^^^^^^^^^^^^^^^^^^
You will first need to calculate the memory size to be suitable for your desired
maximum resolution using the following formula:
.. math::
\text{WIDTH} \times \text{HEIGHT} \times \text{BPP} \times 2 = \text{frame size in bytes}
\text{frame size in bytes} \div 1024 \div 1024 = \text{ frame size in MiB}
\text{frame size in MiB} + 10 = \text{ required size in MiB}
2^{\lceil \log_2(\text {required size in MiB}) \rceil} = \text{ total MiB}
Where `BPP` is 4 for 32-bit RGB (SDR) or 8 for 64-bit
(HDR :ref:`* <libvirt_determining_memory_hdr>`).
.. hint::
The final step in this calculation is simply rounding the value up to the
nearest power of two.
For example, for a resolution of 1920x1080 (1080p) SDR:
.. math::
1920 \times 1080 \times 4 \times 2 = 16,588,800 \text{ bytes}
16,588,800 \div 1024 \div 1024 = 15.82 \text{ MiB}
15.82 \text{ MiB} + 10 \text{ MiB} = 25.82 \text{ MiB}
2^{\lceil \log_2(25.82) \rceil} = 32 \text { MiB}
Failure to provide enough memory will cause Looking Glass to truncate the
bottom of the screen and will trigger a message popup to inform you of the size
you need to increase the value to.
.. note::
Increasing this value beyond what you need does not yield any performance
improvements, it simply will block access to that RAM making it unusable by
your system.
.. list-table:: Common Values
:widths: 50 25 25
:header-rows: 1
* - Resolution
- Standard Dynamic Range
- High Dynamic Range (HDR) :ref:`* <libvirt_determining_memory_hdr>`
* - 1920x1080 (1080p)
- 32
- 64
* - 1920x1200 (1200p)
- 32
- 64
* - 2560x1440 (1440p)
- 64
- 128
* - 3840x2160 (2160p/4K)
- 128
- 256
.. _libvirt_determining_memory_hdr:
.. warning::
While Looking Glass can capture and display HDR, at the time of writing
neither Xorg or Wayland can make use of it and it will be converted by the
GPU drivers/hardware to SDR. Additionally using HDR doubles the amount of
memory, bandwidth, and CPU load and as such should generally not be used
unless you have a special reason to do so.
.. _libvirt_ivshmem:
IVSHMEM
^^^^^^^
There are two methods of configuring IVSHMEM, using shared memory directly, or
using the KVMFR kernel module. While the KVMFR module is slightly more
complicated to configure, it substantially improves performace as it allows
Looking Glass to use your GPUs DMA engine to transfer the frame data.
.. toctree::
:maxdepth: 1
ivshmem_kvmfr
ivshmem_shm
.. _libvirt_spice_server:
Keyboard/mouse/display/audio
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Looking Glass makes use of the SPICE protocol to provide keyboard and mouse
input, audio input and output, and display fallback.
.. note::
The default configuration that libvirt uses is not optimal and must be
adjusted. Failure to perform these changes will cause input issues along
with failure to support 5 button mice.
If you would like to use SPICE to give you keyboard and mouse input
along with clipboard sync support, make sure you have a
``<graphics type='spice'>`` device, then:
- Find your ``<video>`` device, and set ``<model type='vga'/>``
- If you can't find it, make sure you have a ``<graphics>``
device, save and edit again.
- Remove the ``<input type='tablet'/>`` device, if you have one.
- Create an ``<input type='mouse' bus='virtio'/>`` device, if you don't
already have one.
- Create an ``<input type='keyboard' bus='virtio'/>`` device to improve
keyboard usage.
.. note::
Be sure to install the the *vioinput* driver from
`virtio-win <https://fedorapeople.org/groups/virt/virtio-win/direct-downloads/stable-virtio/>`_
in the guest
To enable audio support add a standard Intel HDA audio device to your
configuration as per below:
.. code:: xml
<sound model='ich9'>
<audio id='1'/>
</sound>
<audio id='1' type='spice'/>
If you also want clipboard synchronization please see
:ref:`libvirt_clipboard_synchronization`
.. _libvirt_clipboard_synchronization:
Clipboard synchronization
^^^^^^^^^^^^^^^^^^^^^^^^^
Looking Glass can synchronize the clipboard between the host and guest using
the SPICE guest agent.
1. Install the SPICE guest tools from
https://www.spice-space.org/download.html#windows-binaries.
2. Configure your VM to enable the SPICE guest agent:
- QEMU
.. code:: bash
-device virtio-serial-pci \
-chardev spicevmc,id=vdagent,name=vdagent \
-device virtserialport,chardev=vdagent,name=com.redhat.spice.0
- libvirt
.. code:: xml
<channel type="spicevmc">
<target type="virtio" name="com.redhat.spice.0"/>
<address type="virtio-serial" controller="0" bus="0" port="1"/>
</channel>
<!-- No need to add a VirtIO Serial device, it will be added automatically -->
.. _libvirt_apparmor:
AppArmor
^^^^^^^^
For libvirt versions before **5.10.0**, if you are using AppArmor, you
need to add permissions for QEMU to access the shared memory file. This
can be done by adding the following to
``/etc/apparmor.d/local/abstractions/libvirt-qemu``::
/dev/shm/looking-glass rw,
then, restart AppArmor.
.. code:: bash
sudo systemctl restart apparmor
.. _libvirt_memballoon_tweak:
Memballoon
^^^^^^^^^^
The VirtIO memballoon device enables the host to dynamically reclaim memory
from your VM by growing the balloon inside the guest, reserving reclaimed
memory. Libvirt adds this device to guests by default.
However, this device causes major performance issues with VFIO passthrough
setups, and should be disabled.
Find the ``<memballoon>`` tag and set its type to ``none``:
.. code:: xml
<memballoon model="none"/>
.. _host_install:
Additional tuning
^^^^^^^^^^^^^^^^^
Looking Glass is latency sensitive and as such it may suffer microstutters if
you have not properly tuned your virtual machine. The physical display output
of your GPU will usually not show such issues due to the nature of the hardware
but be sure that if you are experiencing issues the following tuning is
required to obtain optimal performance.
1. Do not assign all your CPU cores to your guest VM, you must at minimum
reserve two CPU cores (4 threads) for your host system to use. For example,
if you have a 6 core CPU, only assign 4 cores (8 threads) to the guest.
2. Ensure you correctly pin your VMs vCPU threads to the correct cores for your
CPU architecture.
3. If you are on a NUMA architecture (dual CPU, or early Threadripper) be sure
that you pin the vCPU threads to the physical CPU/die attached to your GPU.
4. Just because your GPU is in a slot that is physically x16 in size, does not
mean your GPU is running at x16, this is dependent on how your motherboard
is physically wired and the physical slot may be limited to x4 or x8.
5. Be sure to set your CPU model type to `host-passthrough` so that your guest
operating system is aware of the acceleration features of your CPU and can
make full use of them.
6. AMD users be sure that you have the CPU feature flag `topoext` enabled or
your guest operating system will not be aware of which CPU cores are
hyper-thread pairs.
7. NVIDIA users may want to enable NvFBC as an alternative capture API in the
guest. Note that NvFBC is officially available on professional cards only
and methods to enable NvFBC on non-supported GPUs is against the NVIDIA
Capture API SDK License Agreement even though GeForce Experience and
Steam make use of it on any NVIDIA GPU.
How to perform these changes is left as an exercise to the reader.

218
doc/ivshmem_kvmfr.rst Normal file
View File

@ -0,0 +1,218 @@
:orphan:
.. _ivshmem_kvmfr:
IVSHMEM with the KVMFR module (Recommended)
###########################################
The kernel module implements a basic interface to the IVSHMEM device
for Looking Glass allowing DMA GPU transfers.
.. _ivshmem_kvmfr_prereq:
Prerequisites
-------------
The Linux kernel headers for your kernel version are required for building.
Install them with ``apt-get``
.. code:: bash
apt-get install linux-headers-$(uname -r)
Then switch to the ``module/`` directory
.. code:: bash
cd module/
.. _ivshmem_kvmfr_dkms:
Using DKMS (recommended)
------------------------
You can use the kernel's DKMS feature to keep the module across upgrades.
``dkms`` must be installed.
.. code:: bash
apt-get install dkms
.. _ivshmem_kvmfr_installing:
Installing
~~~~~~~~~~
To install the module into DKMS, run
.. code:: bash
dkms install "."
.. _ivshmem_kvmfr_loading:
Loading
~~~~~~~
Using the value you should have already calculated as per
:ref:`Determining Memory <libvirt_determining_memory>`, simply use
``modprobe`` with the parameter ``static_size_mb``, for example:
.. code:: bash
modprobe kvmfr static_size_mb=32
Alternatively you can make this setting permanant by creating the file
``/etc/modprobe.d/kvmfr.conf`` with the following content.
.. code:: text
options kvmfr static_size_mb=32
After this has been done, simply running ``modprobe kvmfr`` is all that is
required.
.. note::
Don't forget to adjust ``static_size_mb`` to your needs.
.. _ivshmem_kvmfr_systemd:
systemd-modules-load
~~~~~~~~~~~~~~~~~~~~
For convenience, you may load the KVMFR module when starting your computer.
We can use the ``systemd-modules-load.service(8)`` service for this task.
Create the file ``/etc/modules-load.d/kvmfr.conf`` with the following
contents::
# KVMFR Looking Glass module
kvmfr
This will now run the next time you start your machine.
.. _ivshmem_kvmfr_verification:
Verification
~~~~~~~~~~~~
If everything has been done correctly you should see the following output in
dmesg:
.. code:: text
kvmfr: creating 1 static devices
You should now also have the character device ``/dev/kvmfr0``
.. warning::
If you start the VM prior to loading the module, QEMU will create the file
``/dev/kvmfr0`` as a regular file. You can confirm if this has happened by
running ``ls -l /dev/kvmfr0`` and checking if the filesize is greater then
zero, or the permissions do not start with ``c``. If this has occured, you
must delete the file and reload the module.
.. _ivhsmem_kvmfr_permissions:
Permissions
~~~~~~~~~~~
The module will create the ``/dev/kvmfr0`` node, which represents the KVMFR
interface. To use the interface, you need permission to access it by either
creating a udev rule to ensure your user can read and write to it, or simply
change its ownership manually, i.e.:
.. code:: bash
sudo chown user:user /dev/kvmfr0
As an example, you can create a new file in ``/etc/udev/rules.d/99-kvmfr.rules``
with the following contents::
SUBSYSTEM=="kvmfr", OWNER="user", GROUP="kvm", MODE="0660"
(replace ``user`` with your username)
.. _ivshmem_kvmfr_libvirt:
libvirt
^^^^^^^
Starting with QEMU 6.2 and libvirt 7.9, JSON style QEMU configuration is the
default syntax. Users running QEMU 6.2 or later **and** libvirt 7.9 or later,
should use this XML block to configure their VM for kvmfr:
.. code:: xml
<qemu:commandline>
<qemu:arg value='-device'/>
<qemu:arg value='{"driver":"ivshmem-plain","id":"shmem0","memdev":"looking-glass"}'/>
<qemu:arg value='-object'/>
<qemu:arg value='{"qom-type":"memory-backend-file","id":"looking-glass","mem-path":"/dev/kvmfr0","size":33554432,"share":true}'/>
</qemu:commandline>
.. note::
- The ``"size"`` tag represents the size of the shared memory device in
bytes. Once you determine the proper size of the device as per
:ref:`Determining Memory <libvirt_determining_memory>`, use the figure you
got to calculate the size in bytes:
``size_in_MB x 1024 x 1024 = size_in_bytes``
If you are running QEMU older than 6.2 or libvirt older than 7.9, please use
legacy syntax for IVSHMEM setup:
.. code:: xml
<qemu:commandline>
<qemu:arg value='-device'/>
<qemu:arg value='ivshmem-plain,id=shmem0,memdev=looking-glass'/>
<qemu:arg value='-object'/>
<qemu:arg value='memory-backend-file,id=looking-glass,mem-path=/dev/kvmfr0,size=32M,share=yes'/>
</qemu:commandline>
.. note::
- Using the legacy syntax on QEMU 6.2/libvirt 7.9 may cause QEMU to
abort with the following error message:
"``error: internal error: ... PCI: slot 1 function 0 not available for pcie-root-port, in use by ivshmem-plain``"
- Remember to add ``xmlns:qemu='http://libvirt.org/schemas/domain/qemu/1.0'``
to the ``<domain>`` tag.
Running libvirt this way violates AppArmor and cgroups policies, which will
block the VM from running. These policies must be amended to allow the VM
to start:
- For AppArmor, create ``/etc/apparmor.d/local/abstractions/libvirt-qemu`` if
it doesn't exist, and add the following::
# Looking Glass
/dev/kvmfr0 rw,
- For cgroups, edit ``/etc/libvirt/qemu.conf``, uncomment the
``cgroup_device_acl`` block, and add ``/dev/kvmfr0`` to the list.
Then restart ``libvirtd``:
.. code:: bash
sudo systemctl restart libvirtd.service
.. _ivshmem_kvmfr_qemu:
QEMU
^^^^
If you are using QEMU directly without libvirt, add the following arguments to your
``qemu`` command line::
-device ivshmem-plain,id=shmem0,memdev=looking-glass
-object memory-backend-file,id=looking-glass,mem-path=/dev/kvmfr0,size=32M,share=yes
.. note::
The ``size`` argument must be the same size you passed
to the ``static_size_mb`` argument when loading the kernel module.

66
doc/ivshmem_shm.rst Normal file
View File

@ -0,0 +1,66 @@
:orphan:
.. _ivshmem_shm:
IVSHMEM with standard shared memory
###################################
This method is here for those that can not use the KVMFR kernel module. Please
be aware that as a result you will not be able to take advantage of your GPUs
abillity to access memory via it's hardware DMA engine if you use this method.
Add the following to your libvirt machine configuration inside the
'devices' section by running ``virsh edit <VM>`` where ``<VM>`` is the name of
your virtual machine.
.. code:: xml
<shmem name='looking-glass'>
<model type='ivshmem-plain'/>
<size unit='M'>32</size>
</shmem>
.. note::
If you are using QEMU directly without libvirt the following arguments are
required instead.
Add the following to the commands to your QEMU command line, adjusting
the ``bus`` parameter to suit your particular configuration:
.. code:: bash
-device ivshmem-plain,memdev=ivshmem,bus=pcie.0 \
-object memory-backend-file,id=ivshmem,share=on,mem-path=/dev/shm/looking-glass,size=32M
The memory size (show as 32 in the example above) may need to be
adjusted as per the :ref:`Determining memory <libvirt_determining_memory>`
section.
.. warning::
If you change the size of this after starting your virtual machine you may
need to remove the file `/dev/shm/looking-glass` to allow QEMU to re-create
it with the correct size. If you do this the permissions of the file may be
incorrect for your user to be able to access it and you will need to correct
this. See :ref:`libvirt_shmfile_permissions`
.. _libvirt_shmfile_permissions:
Permissions
~~~~~~~~~~~
The shared memory file used by IVSHMEM is found in ``/dev/shm/looking-glass``.
By default, it is owned by QEMU, and does not give read/write permissions to
your user, which are required for Looking Glass to run properly.
You can use ``systemd-tmpfiles`` to create the file before running your VM,
granting the necessary permissions which allow Looking Glass to use the file
properly.
Create a new file ``/etc/tmpfiles.d/10-looking-glass.conf``, and populate it
with the following::
# Type Path Mode UID GID Age Argument
f /dev/shm/looking-glass 0660 user kvm -
Change ``UID`` to the user name you will run Looking Glass with, usually your
own.