The small linux problem thread

What distro?

CentOS7

The issue is, I'd like to have the VPN clients on the same subnet as the openvpn host and not in a subnet of the openvpn host. If that makes any sense. I've looked at the openvpn docs and am really struggling with it.

remember CentOS uses firewalld. (interface to iptables, but you still probably want to use this)

Might be a good idea to write down every connection, source destination, port. might help figure out what you need.

1 Like

I'm probably going to bite the bullet and just to get the VPN up NAT it, and do a deep dive later on how to do bridging... I think I may have successfully manually configured a KVM to bridge off the host. Now I just need to be able to apply that to connections that don't exist yet.

I'm looking for advice/best practices on this one here.

So I have a container running OpenVPN - the service cannot start. I feel like I've isolated the issue to the tun module not being loaded on the container. However, I can't load the module on the container, because I think the tun module isn't loaded on the host.

lsmod | grep tun 
    returns nothing on host

So, I fee like I'll need modprobe tun the host in order to get it to load in the kernel there and then I'll be able to continue forward with my project since the OpenVPN is container and shares the kernel with the host.

My question follows: is it best practice to load modules onto hosts for one off applications on container, like this situation - or is it better to create a complete new VM? Lastly, will I even be able to load it on the host machine? I ask because if I'm root on the container and tell it to load that module - shouldn't it? It is root, and does have access to the kernel (a shared one though) or is at a hypervisor level the kernel is controlled for containers?

Anyways - thoughts and ideas are appreciated - just looking before I leap.

So of the thing I've considered are - is loading a new module going to eat up more resources (than creating said VM)? Also, am I exposing my host to more security vulnerability by allowing tunnelling to place within the kernel?

Hi, I am new on this forum, I am Dutch (native, English is a foreign language to me). Level1Techs had this nice experiment with running Windows in a virtual machine for a Skylake CPU and a compatible motherboard.


I would like to set up this kind of system if it is possible with a Ryzen based system. I don't own the hardware yet, I intend to upgrade my i5-750 and a HD 7850 to a 1600(X)/1700 and some Vega card in the next few months. Given that I prefer a well tuned Linux system (by me) over a Windows system for all other tasks than Windows exclusive software and given that I am not too fond of some of Microsofts practices these days (like booting upt your computer without your permission to update and then not turn it off, or installing/uninstalling software without permission) and because I am a bit of a curious nerd I would like to try out to set up such a system myself.

My first question: is it possible at the moment to do the same for Ryzen? I hear Wendell talk a lot about IOMMU grouping problems. I don't know much about IOMMU and virtual machines but as far as I understand the IOMMU helps an OS running in a virtual machine to identify hardware and the PCIe-passthrough is necessary to allow the Virtual machine to use the graphics card almost as good as if the OS in the virtual machine would be the OS you are running at that moment. Does this IOMMU grouping problem (putting all that hardware in the same IOMMU group) make it impossible or more difficult to run Windows in a virtual machine under Linux with a Ryzen based system?
Second question. How does this PCIe passthrough work excactly? I am looking at this wiki: https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVMF
_"Provided you have a desktop computer with a spare GPU you can dedicate to the host (be it an integrated GPU or an old OEM card, the brands do not even need to match) and that your hardware supports it, it is possible to have a VM of any OS with its own dedicated GPU and near-native performance. _
Suppose that I would have a HD 7850 and a RX 580 and I would PCIe passthrough that RX 580 to the virtual machine, does that mean that I can't use the RX 580 outside the virtual machine while the virtual machine is running or would I even have to close some program or even change some setting, reboot the system and only then be able to use the RX 580 again outside the virtual machine?
Why do you have to PCIe passthrough a graphics card but can the virtual machine just access all the other hardware (CPU, RAM, SSD, keyboard...)?

I hope that the more experience guys don't mind to give a short explanation or reference even though I am sure that these are n00b-questions which have been asked before. I don't mind to dive into the 'literature' and look up all the details but at this moment I don't even know if it is possible to set up a Ryzen based system this way yet. I do have some experience with programming and working from a terminal but this whole Linux world is kind of new to me, I only have used Linux Mint a bit.

Hello, and welcome to the forums!

Bro, you have better English then me. And I'm American. :frowning:


As for your question, I don't know. I just wanted to say hi.

On Ryzen, there is no tangible advice anyone can give you how to make PCI passthrough work, everyone's pretty much still figuring shit out lol. It works on some systems, doesn't work on others, depending on UEFI updates and Windows problems. It'll probably still be a good couple of months before it would work.

Only advice I would give at this point for Ryzen is that, if you would buy the hardware before a pretty standardized solution is out there, you should not install Windows on the bare metal or use a harddisk with it that has/had Windows installed on it. But you can install linux on it and practice and try out kvm settings to see how far you get on your own based on different possible settings that turned out successful for others to do PCI passthrough. Just never ever install Windows on bare metal on new hardware ever again.

i'm on Ubuntu 16.10. is there any way to give a cli program [Ranger] it's own icon and button in the launcher so i can pin it and just run it from there instead of having to open a terminal first?

I believe that you can do this with a .desktop file

[Desktop Entry]
Version=1.0
Name=
Comment=
GenericName=
Exec=
Icon=
Type=Application
Categories=

for the command it will probably look something like this, I am writing for terminator, just change the arguments to what you need.

Exec=terminator -T 'Program Name' -e program

1 Like

cheers mate, will try.

Hey guys have a problem again. I managed to mount my NAS samba shares in Fedora 25 it work great... except steam. When i try to make a steam folder it returns an error Steam Library-Must be on a filesystem mounted with execute permissions.
The NAS runs ZFS on linux.

The file i made for steam has those permissions:
drwxrwxrwx. 2 alphazero1990 alphazero1990 0 May 6 19:48 'Steam Library-Linux'
What's wrong here?

well, as always. You find the solution 10 minutes after you make an embarrassing post somewhere, after you searched for 5 hours.

Solution, opensm might not listen to all the ports.
And it didn't run even though it was running. Don't know how that's possible but i digress.

Currently using this for wallpaper switcher occasionally extensions disappear until GDM reloads everything. When I was using LightDM it made all the extensions disappear.

GDM handles the switch(whenever switching wallpapers) better than LightDM, but the screen freezes and everything is redrawn(if that's the correct phrasing). Usually takes a second or two before everything is redrawn, but can get a little annoying.

On the page people suggested using [email protected], but apparently it is not as nice as random walls.

Any alternatives to this extension(they don't have to be limited to extensions only)?

Hi Guys, I'm in a bit of a jam. I need to create backups but the machine cannot touch the internet. There needs to be a one time backup of all data but not the full disk which is what dd would do from what I understand, after the full backup I need it to perform iterative back up every time it shuts down if possible. It is currently running Ubuntu 13.04 which cannot be updated or the machine will lose its certification status in the industry I am in. I initially was going to use Deja Dup but without being able to connect to the internet I am unsure how to install this application. I am not a heavy Linux user but I can fumble my way around and have a friend that is a little more versed in it. I can throw together a separate PC to download the application and transfer it via USB if that is an option. If there is already a built in function of Ubuntu that can handle this it's even better.

Thanks for any help you may have!

rsync might be installed by default? This will do what you want fairly well. If its not installed, its a small package so should be easy to get (i don't think there's many if any dependencies) and put it through any checks you need before putting it on your system.

you can set up a script to run at shutdown. I don't know what ubuntu 13.04 uses (newer ones use systemd for running things like that, you will need to check.

Thanks very much @Eden. I'll have a look today and see if I can get rsync to work.

By adding ,exec, to fstab and reboot, steam accepts the share. And now after all this effort and work Steam doesn't see shit 0MB Disk space available. Can't install cuz no disk space apparently... oh just a FU*k tone of 10.4 TB of spinning magnest...... i am so angry right now.
Anybody knows something that could help me? I do realise it's very specific even google doesn't know wtf to do when i search for this...

Question: If you have this SMB share mounted as a network drive, does it give you correct disk space metrics either with df or by viewing properties in your file manager? From what I understand, Steam's had problems with ZFS volumes in the past, and they may not be completely resolved as of yet.

Shot in the dark: If the folder you're attempting to get Steam to use is empty, try manually copying a file or two in there and see if Steam reports different disk space.

Edit: Your cifs-utils package is installed and up to date on your client machine, correct?

1 Like

i don't know if anyone uses Neofetch, but i can't seem to get Audacious to lend its info to my neofetch prints. i'm using the latest versions of both.