GNU Linux/Workarounds

A lot of distributions come fairly close to being perfect, but sometimes default configurations or behaviors need to be adjusted to work for certain scenarios. Please consult the help material that accompanies the tools referenced below, if you need a better understanding of their functionalities or behaviors, and whether or not these fixes are appropriate for your use cases.

Networking

WiFi Hotspot Issues

Clients Unable to Get Internet Access

I recommend just having iptables as a firewall alone, and not in combination with, other firewalls. After writing this, I found that no rules in iptables chains were necessary for my client to connect to my WiFi hotspot. Add the rules below, as a last resort.

In my case, after installing Docker on endeavourOS, I found that my hotspot could be connected to, but, the client was unable to obtain an IP address. I was using GNOME's settings applet, which is a wrapper for NetworkManager.

Luckily, iptables, the firewall built into Linux, can be used. Based on hints provided by (https://wiki.archlinux.org/title/Internet_sharing#With_iptables)[1], I could add a link level packet filter with NAT, using something similar to the these commands (with superuser privileges):

iptables -t nat -A POSTROUTING -o <ethernet interface name> -j MASQUERADE
iptables -A FORWARD -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT
iptables -A FORWARD -i <wifi adapter name> -o <ethernet interface name> -j ACCEPT

Note: Network interfaces can be listed with the "ifconfig" program.

If clients are still unable to gain internet access after connecting to the hotspot, try disabling firewalld (this is because, as a RedHat blog article points out, it can conflict with iptables: https://www.redhat.com/en/blog/iptables).

GPU's

Wayland User Experience

Lag with NVIDIA GPU's

If you have an AMD graphics card, either integrated or dedicated, you will likely be accustomed to having a better user experience than those with NVIDIA graphics cards.

Wayland works pretty well, but memory and other clock settings stay at too low of a frequency, causing lag. This is noticeable with the GNOME desktop shell. I use this code below to increase the minimum clock speeds on my dedicated GPU.

sudo nvidia-smi -pm 1
sudo nvidia-smi -lmc 1000,6000
sudo nvidia-smi -lgc 1000,2000

The first command enables the persistence mode daemon, which keeps the kernel mode driver running (even when no other application is actively using it), the second locks the memory clock between 1000 and 6000 MHz, and the last, locks the graphics clock to be between 1000 and 2000 MHz. The command line arguments cannot be combined into one invocation, as nvidia-smi will not allow it.

I've found it useful to have a systemd service to do this automatically, when the system starts. I have written a file located here: /etc/systemd/system/nvidia-fix-clocks.service:

[Unit]
Description=Boost NVIDIA GPU memory and graphics clock speeds

[Service]
Type=oneshot
ExecStart=/opt/ScriptFixes/nv_fix_clocks.sh

[Install]
WantedBy=multi-user.target

Don't forget to reload systemd's daemon with systemctl reload-daemon, and to enable the service with systemctl enable nvidia-fix-clocks.service. The changes will apply on the next reboot.

OpenGL/X11 Application Not Running With a Dedicated NVIDIA GPU

I encountered this issue when running Freecad, which, in my case, could not use my dedicated GPU due to initialization issues with the OpenGL Extension for X11 (GLX). I use this code below, within a script, to ensure Freecad's rendering path goes a different way:

Xwayland -decorate -geometry 1920x1080 :2 &
QT_QPA_PLATFORM=xcb prime-run freecad -display :2