The Linux Thread - The Autist's OS of Choice

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.
View attachment 6525885
What do I have to do to get GPU acceleration working on this piece of shit? The Gentoo wiki says I have to install mesa-amber but that does nothing
Step 1: Stop using Gentoo
Did you really think that POS was going to be worthwhile

On the topic of GPU acceleration, has anyone checked out:

I haven't followed the guide yet. But not having to pass-trough your GPU, but still get acceleration on windows is pretty cool.

I refuse to run windows on bare-metal, and I don't have the hardware for pass-through.
My motherboard has a bug that prevents vfio-pci from exposing itself to guests correctly, so this is super helpful actually, thanks man.
 
for a vm, how much is a performance hit having windows installed into a .gcow2 image on a dedicated ext4 ssd mounted on the host compared to passing the entire drive directly to the vm? looks like passing the drive requires calling it by /dev/sde, which I'm squeamish about as that will change if drives are added or removed to the server so it won't survive a (admittedly very rare) reboot to autostart. or is there a more efficient third option? Cockpit doesn't give a very clear explanation of it's setup
 
  • Like
Reactions: The Anarki Main
for a vm, how much is a performance hit having windows installed into a .gcow2 image on a dedicated ext4 ssd mounted on the host compared to passing the entire drive directly to the vm? looks like passing the drive requires calling it by /dev/sde, which I'm squeamish about as that will change if drives are added or removed to the server so it won't survive a (admittedly very rare) reboot to autostart. or is there a more efficient third option? Cockpit doesn't give a very clear explanation of it's setup
Use GCOW2. It's more compact, easier to use, and generally works better. Also, snapshots.
 
  • Like
Reactions: The Anarki Main
for a vm, how much is a performance hit having windows installed into a .gcow2 image on a dedicated ext4 ssd mounted on the host compared to passing the entire drive directly to the vm? looks like passing the drive requires calling it by /dev/sde, which I'm squeamish about as that will change if drives are added or removed to the server so it won't survive a (admittedly very rare) reboot to autostart. or is there a more efficient third option? Cockpit doesn't give a very clear explanation of it's setup
Don't use GCOW2. It's really slow.

Rather than typing /dev/sde directly, try referring to it by one of the symlinks in /dev/disk, which should be persistent.
 
I ended up following this guide and it is much, much faster now
I run a Windows 11 virtual machine with good performance on my Chromebook with a 10th gen i5 and 16GB of RAM (similar specs as your laptop).

Looking at your libvirt XML, there are a few optimizations you can make:

Apply all available Hyper-V enlightenments - the <hyperv> section of your XML should look like this:

<hyperv>
<relaxed state='on'/>
<vapic state='on'/>
<spinlocks state='on' retries='8191'/>
<vpindex state='on'/>
<synic state='on'/>
<stimer state='on'>
<direct state='on'/>
</stimer>
<reset state='on'/>
<frequencies state='on'/>
<reenlightenment state='on'/>
<tlbflush state='on'/>
<ipi state='on'/>
</hyperv>
Disable all timers except for the hypervclock - the <clock> section of your XML should look like this:

<clock offset='localtime'>
<timer name='rtc' present='no' tickpolicy='catchup'/>
<timer name='pit' present='no' tickpolicy='delay'/>
<timer name='hpet' present='no'/>
<timer name='kvmclock' present='no'/>
<timer name='hypervclock' present='yes'/>
</clock>
Those two improvements alone should result in a massive speedup.

Further improvements can be made, though. I recommend using CPU pinning - this forces each virtual CPU to be pinned to a physical CPU core (or in this case, virtual Crostini core), reducing the performance overhead from the kernel constantly swapping the virtual CPUs to different threads. For example, I do the following (6 cores for the VM on an 8-core host):

<vcpu placement='static'>6</vcpu>
<iothreads>1</iothreads>
<cputune>
<vcpupin vcpu='0' cpuset='1'/>
<vcpupin vcpu='1' cpuset='5'/>
<vcpupin vcpu='2' cpuset='2'/>
<vcpupin vcpu='3' cpuset='6'/>
<vcpupin vcpu='4' cpuset='3'/>
<vcpupin vcpu='5' cpuset='7'/>
<emulatorpin cpuset='0,4'/>
<iothreadpin iothread='1' cpuset='0,4'/>
</cputune>
I highly recommend using virtio as your disk type, as this allows disk access to be paravirtualized, further reducing overhead. This requires driver support though on the Windows side - easiest way to enable this is to reinstall Windows, and when partitioning the disk, insert the virtio-win drivers ISO into your virtual machine so that the disk can be recognized during setup. I'm using this in my XML:

<disk type='file' device='disk'>
<driver name='qemu' type='qcow2' cache='none' io='threads' discard='unmap' iothread='1' queues='6'/>
<source file='/var/lib/libvirt/images/win11.qcow2'/>
<target dev='vda' bus='virtio'/>
<boot order='2'/>
<address type='pci' domain='0x0000' bus='0x05' slot='0x00' function='0x0'/>
</disk>
Finally, make sure you install the Spice guest tools to improve how the VM handles mouse input between the guest and host, and to automatically change the VM's resolution when the window resizes.
 
On the topic of GPU acceleration, has anyone checked out:

I haven't followed the guide yet. But not having to pass-trough your GPU, but still get acceleration on windows is pretty cool.

I refuse to run windows on bare-metal, and I don't have the hardware for pass-through.
Oh damn... that's very cool.

I am interested to see how well this would work with certain enterprise applications where the vendor is trying to push people onto their shitty web apps by hammering the GPU on their new desktop versions in a way that makes them impractical to run on terminal servers, which then means that if you want them to be hosted in a centralized way you need to pay out the ass for special NVIDIA drivers as well as the enterprise hardware and ever more expensive VMWare or other proprietary desktop virtualization software that works with the overpriced NVIDIA drivers. Which makes their absolutely criminal (and ever increasing) pricing look slightly more reasonable. Fuck Jack Dangermond and his fake charities I hope he gets raped to death by niggers.
 
But I was told that Linux is better for old computers and there are very few distros that still work on these old CPUs
You have a slower CPU, so you pick a distro where you are expected to frequently recompile packages... Sure.

Might I recommend either Puppy or Slackware?
 
  • Like
Reactions: Susanna
You have a slower CPU, so you pick a distro where you are expected to frequently recompile packages... Sure.
I don't mind waiting and Gentoo doesn't get in the way of my stupidity
Might I recommend either Puppy or Slackware?
using S15Pup32-22.12-241006.iso liveusb
1729062683245.png
1729062771330.png
this may actually be worse than Gentoo

I'll just try to find a good deal on an AGP HD 3000 series card and give up for now
 
There's a new hit piece titled: "The Stallman Report", written by people who wish to remain anonymous (of course, as they go after a man who is too busy fighting cancer).
The whole thing is just another attempt at removing Stallman from the FSF.

edit: Lunduke made a video about it
I was browsing some old stuff on Drew DeVault and realized he's reused the same claim that the GNU project is transphobic before: https://archive.ph/ZHoiN

Having a pretty good hunch who wrote the report now and why the author's name was redacted (because everyone knows DeVault and that he's a bumbling faggot), I dug some more and sure enough I was right. I'll let anon explain how he found out.
19338.jpg

I apologize in advance if this is old news, I haven't watched the Lunduke video yet.

EDIT: This is another one of DeVault's ramblings about Stallman, posted merely three weeks ago. Yet DeVault claimed not to write the report. Ok, sure bub.
Anyway this provides some additional insight into his views which are interesting to say the least. Without powerleveling too much, I'm autistic on the exact same level as RMS and in very similar ways so I can vouch for RMS being truthful whether he had official word from a medical professional or not. Of course, DeVault has to step in and claim that RMS is somehow being an ableist and poses an imminent cyberterrorist threat to all women or something. Because it's Drew, of course he fucking does.
 
Last edited:
I was browsing some old stuff on Drew DeVault and realized he's reused the same claim that the GNU project is transphobic before: https://archive.ph/ZHoiN
"using words whose meanings (as you understand them)", 'you' here referring to the speaker rather than the listener, was explicitly written to support Stallman's transphobic refusal[0] to use the singular "they"
This one is particularly stupid, as Stallman has long been an advocate of his own gender-neutral pronoun "per". I think it's even leaked out into the man pages at some point. Really the thing they hate him for the most is that he doesn't adhere to the current fashions.

But I was told that Linux is better for old computers and there are very few distros that still work on these old CPUs
If you want to run modern software on this box, Gentoo is the way to go. But you can't expect that drivers will have been maintained for all these years. At some point they will have changed one of the interfaces, and rather than spending the time fixing a driver for dead hardware they'll have removed it from the tree.

You might have better luck trying to find a SuSE or Debian ISO from 2003, and booting that.
 
The current kernel still has drivers for Amiga hardware, and they also still work. (You don't see them in your kernel configuration because you are configuring for x86_64. Try arch 68k)

That said, early ATI hardware was a nightmare in linux, a far cry from current AMD hardware. Needs more info, a full xorg log, useflags used around xorg and mesa, and kernel configuration in the graphics department. IIRC there was some change around how userspace was allowed to access memory, something like that, and it caused a lot of problems at least with the matrox drivers. If everything is correct, there's a non-zero chance you are stumbling across a bug nobody noticed because nobody is using this hardware. Gentoo is the way to go for weird and outdated hardware counter-intutively, because it lets you configure everything freely (and also patch, if necessary).
 
Don't use GCOW2. It's really slow.

Rather than typing /dev/sde directly, try referring to it by one of the symlinks in /dev/disk, which should be persistent.
it depends on the driver controller type. the sata controller is really slow, scsi is used more often. the virtuio controller is very fsst as it gives the VM more direct control over the storage drive - but it requires loading a driver during first install (or doing a workaround where you temporary add a virtuio drive, install the drivers, then switch the os drive over
 
View attachment 6525885
What do I have to do to get GPU acceleration working on this piece of shit? The Gentoo wiki says I have to install mesa-amber but that does nothing
Have you tried building your own kernel since you're now considered an "edge" usecase?
If everything is correct, there's a non-zero chance you are stumbling across a bug nobody noticed because nobody is using this hardware.
I seem to recall that mesa support on those ATI cards (A100-A300) was always a shitshow.

Also, he needs to build a custom kernel (not with genkernel either) to confirm that that dist-kernel isn't causing any issues. He might actually have to set "ATI Radeon display support"found in drivers/video/Kconfig (CONFIG_FB_RADEON) since that's not enabled by default and may be needed for the R200 set.
 
Last edited:
fuck I don't know how but Cockpit autobricks my network configuration, making the server inaccessible until i restart the network service, only for cockpit to immediately disable it again. I couldn't find a way to stop it from doing that so i had to uninstall cockpit.
 
  • Feels
Reactions: The Anarki Main
Can confirm my friend has tried multiple Linux distributions on his mid-2000s Radeon and GeForce computers and they've always lead to insane graphical bugs. Granted, some of them had hardware damage anyway, but these issues did not appear nearly as aggressively under Windows. Even after trying out different drivers and switching to lighter desktops, the problems didn't go away and it became clear it was from poor software support. I've had much better luck with my machines, fortunately, but I don't extensively use modern distributions on those older computers because I'm not going to unironically daily drive a Core 2 Duo in the current year. With how bloated the web has become it's not really pleasant. So most of my hours of Linux on those machines were older distros.
 
  • Like
Reactions: Chen Stirner
If I recall correctly, 3D acceleration on my old ATI Radeon 9550 was working best on Knoppix 5.0. On newer distros 3D performance was lower, so good luck.
 
  • Like
Reactions: The Anarki Main
But I was told that Linux is better for old computers and there are very few distros that still work on these old CPUs

Worth a shot for $5
Debian, at least for me, always seems to have the best support for weird hardware, usually even better than NetBSD. Mesa Amber is a shitshow, but Debian Bullseye is still in support and ships with Mesa 20 and the xorg driver for bullseye (https://manpages.debian.org/bullseye/xserver-xorg-video-radeon/radeon.4.en.html) looks like it may actually support your card. That's probably the direction I would go. I remember my Athlon XP 2400+ system, I really liked that thing, and it definitely pooped on the Pentium 4s that were available then.

A couple of years ago I managed to get a buster or bullseye system to come up with a truly cursed set of video cards attached to a system: two 3090s, an FX 5200 and whatever bullshit Matrox video card this supermicro board had on it, by blacklisting the nvidia and nouveau drivers appropriately so that both could be used where needed. If I can make insane hardware work, I'm sure you can too.
 
Back