On this page, we describe how to set up a [GUI domain](/news/2020/03/18/gui-domain/). In all the cases, the base underlying TemplateVM used is `Fedora` with `XFCE` flavor to match current desktop choice in `dom0`. That can be adapted very easily for other desktops and templates. By default, the configured GUI domain is a management qube with global admin permissions `rwx` but can be adjusted to `ro` (see [Introducing the Qubes Admin API](/news/2017/06/27/qubes-admin-api/)) in pillar data of the corresponding GUI domain to setup. For example, pillar data for `sys-gui` located at `/srv/pillar/base/qvm/sys-gui.sls`. Please note that each GUI domain has no `NetVM`.
> Note: The setup is done using `SaltStack` formulas with the `qubesctl` tool. When executing it, apply step can take time because it needs to download latest Fedora XFCE TemplateVM and install desktop dependencies.
Here, we describe how to setup `sys-gui` that we call *hybrid mode* or referenced as a *compromise solution* in [GUI domain](/news/2020/03/18/gui-domain/#the-compromise-solution).
At this point, you need to shutdown all your running qubes as the `default_guivm` qubes global property has been set to `sys-gui`. In order to use `sys-gui` as GUI domain, you need to logout and, in the top right corner, select `lightdm` session type to **GUI domain (sys-gui)**. Once logged, you are running `sys-gui` as fullscreen window and you can perform any operation as if you would be in `dom0` desktop.
Here, we describe how to setup `sys-gui-gpu` which is a GUI domain with *GPU passthrough* in [GUI domain](/news/2020/03/18/gui-domain/#gpu-passthrough-the-perfect-world-desktop-solution).
One more step is needed: attaching the actual GPU to `sys-gui-gpu`. This can be done either manually via `qvm-pci` (remember to enable permissive option), or via:
The latter option assumes Intel graphics card (it has hardcoded PCI address). If you don't have Intel graphics card, please use the former method with `qvm-pci` (see [How to use PCI devices](/doc/how-to-use-pci-devices/)).
> Note: Some platforms can have multiple GPU. For example on laptops, it is usual to have HDMI or DISPLAY port linked to the secondary GPU (generally called _discrete GPU_). In such case, you have to also attach the secondary GPU to `sys-gui-gpu` with permissive option.
> Note: For some platforms, it can be sufficient to shutdown all the running qubes and starting `sys-gui-gpu`. Unfortunately, it has been observed that detaching and attaching some GPU cards from `dom0` to `sys-gui-gpu` can freeze computer. We encourage reboot to prevent any data loss.
Once, `lightdm` is started, you can log as `user` where `user` refers to the first `dom0` user in `qubes` group and with corresponding `dom0` password. A better approach for handling password is currently discussed in [QubesOS/qubes-issues#6740](https://github.com/QubesOS/qubes-issues/issues/6740).
Here, we describe how to setup `sys-gui-vnc` that we call a *remote* GUI domain or referenced as *with a virtual server* in [GUI domain](/news/2020/03/18/gui-domain/#virtual-server-the-perfect-remote-solution).
At this point, you need to shutdown all your running qubes as the `default_guivm` qubes global property has been set to `sys-gui-vnc`. Then, you can start `sys-gui-vnc`:
```bash
qvm-start sys-gui-vnc
```
A VNC server session is running on `localhost:5900` in `sys-gui-vnc`. In order to reach the `VNC` server, we encourage to not connect `sys-gui-vnc` to a `NetVM` but rather to use another qube for remote access, say `sys-remote`. First, you need to bind port 5900 of `sys-gui-vnc` into a `sys-remote` local port (you may want to use another port than 5900 to reach `sys-remote` from the outside). For that, use `qubes.ConnectTCP` RPC service (see [Firewall](/doc/firewall). Then, you can use any `VNC` client to connect to you `sys-remote` on the chosen local port (5900 if you kept the default one). For the first connection, you will reach `lightdm` for which you can log as `user` where `user` refers to the first `dom0` user in `qubes` group and with corresponding `dom0` password.
> Note: `lightdm` session remains logged even if you disconnect your `VNC` client. Ensure to lock or log out before disconnecting your `VNC` client session.
> **WARNING**: This setup raises multiple security issues: 1) Anyone who can reach the `VNC` server, can take over the control of the Qubes OS machine, 2) A second client can connect even if a connection is already active and potentially get disconnected, 3) You can get disconnected by some unrelated network issues. Generally, if this `VNC` server is exposed to open network, it must be protected with some other (cryptographic) layer like `VPN`. The setup as is, is useful only for purely testing machine.
> Note: For the case of `sys-gui-gpu`, you need to prevent Qubes OS autostart of any qube to reach `dom0`. For that, you need to boot Qubes OS with `qubes.skip_autostart` GRUB parameter.
For any general GUI domain issues, please take a loot at existing issues `QubesOS/qubes-issues` under [C: gui-domain](https://github.com/QubesOS/qubes-issues/issues?q=is%3Aopen+is%3Aissue+label%3A%22C%3A+gui-domain%22) label.