The Linux Questions Thread: a bunch of pitfalls, but technically it's possible

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Linux Questions Thread: a bunch of pitfalls, but technically it's possible

«‹›1033 »

Storm One: Jan 12, 2011

some kinda jackal posted:

I'm trying to think of how to explain what I'm trying to do conceptually and there's no great analogy, but imagine that you've got a CD iso that you want to present to a VM as a read-write hard disk image, to which the VM can make all the subsequent modifications it wants and they persist in the form of a different backing file.

Unless I'm missing something, that's what XFS and Btrfs do by default with cp --reflink=always (or =auto, which I also think is the default behaviour of cp now).

IOW, take the original iso and make how many copies of it you need and no extra space will be used. Mark the original read-only, it has its own inode so it won't affect the copies like hardlinks would, but the extents are still 100% shared. As each of the copies is written to, only the modified extents of each get CoW'd back to disk.

# ? Nov 23, 2023 13:33

Adbot: ADBOT LOVES YOU

# ? May 24, 2024 16:44

Computer viking: May 30, 2011; Now with less breakage.

Yeah, that's the "block cloning" I mentioned in practice. Useful stuff.

# ? Nov 23, 2023 17:13

Computer viking: May 30, 2011; Now with less breakage.

I don't want help with this, I just want to complain:

We're trying to install ... I guess you could call it an edge solution; it's a tool to pull down and run containers on your own locally stored sensitive data, built for hospital use by a large German company. We're a research group, but we're in a European consortium that has gotten a free license so we can run each other's experimental AI models on our own data. As a bonus, it makes it easy to move it to the clinical side if any of the models prove useful, since they already use this.

Of course, the deployment is un-fun. Our budget is "I guess we can find a workstation", and their first instructions started with "in one of the VMWare VMs, install this windows tool we've made - btw it requires Windows Datacenter; we assume you have some licenses laying around".

We somehow talked them down from that, but now they want two Linux VMs, at least one of them with [nvidia] GPU acceleration. Without much hope, I forwarded a 3090 into one of them, and it actually fires up with noveau and seems to work*. Somewhat more hopeful, I thought I'd see what the nvidia drivers make of the situation - and was immediately reminded that they gave us an RHEL 8 ISO that we definitely don't have a support contract for, while the relevant instructions start with "If not done already, register the system using subscription-manager". I don't even think I can go the manual installation route; I think the tools I need to build a DKMS module are in a subscription-only repo.

I ... will see what the Germans say about this tomorrow. Worst case I can hack something together with third party repos, but I'd kind of like the configuration to be vanilla enough that it's not my fault when it breaks.

* I would not recommend a dual monitor setup split between a monitor connected to a GPU and a VMWare console on a separate laptop; it's somehow the worst of every component involved. Unexpected and fun the first time, though.

Computer viking fucked around with this message at 17:34 on Nov 23, 2023

# ? Nov 23, 2023 17:31

Wibla: Feb 16, 2011

Here's some help anyway :haw:

# ? Nov 23, 2023 19:13

ExcessBLarg!: Sep 1, 2001

some kinda jackal posted:

Is anyone aware of any userland app or daemon that will take a read-only file as an argumet and interpose itself as that file (through a socket file or something, I would guess) to a consuming application and write any changes to a different delta file?

You can use qemu-img to create a qcow2 of your "CD iso", then create more qcow2s referencing the original as a backing file. The qcow2s are directly usable in QEMU, but if you want to access them as block devices in the "host" OS you can use qemu-nbd to mount them as /dev/nbd* devices. This does require that the nbd module is loaded in the kernel, and you may have to manually modprobe it if it isn't.

There's other ways too, but the qemu tools are entirely userspace and don't have fancy underlying FS requirements.

ExcessBLarg! fucked around with this message at 21:49 on Nov 23, 2023

# ? Nov 23, 2023 21:45

Computer viking: May 30, 2011; Now with less breakage.

Wibla posted:

Here's some help anyway

Yeah that looks about right.

Also, Islay gin? Color me intrigued.

# ? Nov 23, 2023 21:53

AlexDeGruven: Jun 29, 2007; Watch me pull my dongle out of this tiny box

The Botanist is one of my absolute favorites.

It's not peated, though, if that's what is intriguing you. It's just lovely and floral.

# ? Nov 23, 2023 21:59

Computer viking: May 30, 2011; Now with less breakage.

AlexDeGruven posted:

The Botanist is one of my absolute favorites.

It's not peated, though, if that's what is intriguing you. It's just lovely and floral.

Yeah I guessed - it was more about the branching out.

Though I wouldn't mind a Caol Ila.

# ? Nov 23, 2023 22:02

Wibla: Feb 16, 2011

It's lovely, floral and kicks like a mule.

I need to buy another bottle, and of course polet here doesn't have it in stock :smith:

# ? Nov 23, 2023 22:09

West SAAB Story: Mar 13, 2014; by Athanatos
(and can't post for 222 days!)

code:

# dd if=/dev/vda | ssh root@x.x.x.x dd of=/dev/vda

https://www.youtube.com/watch?v=180YLV4b5Nw

# ? Nov 23, 2023 23:33

ziasquinn: Jan 1, 2006; Fallen Rib

West SAAB Story posted:

code:
# dd if=/dev/vda | ssh root@x.x.x.x dd of=/dev/vda
https://www.youtube.com/watch?v=180YLV4b5Nw

https://www.youtube.com/watch?v=xuUbubFKT0I

# ? Nov 23, 2023 23:48

ExcessBLarg!: Sep 1, 2001

West SAAB Story posted:

code:

# dd if=/dev/vda | ssh root@x.x.x.x dd of=/dev/vda

The trick is to preceed that with a

code:

echo u > /proc/sysrq-trigger

so you can reimage the active root volume when you don't have an A/B setup.

# ? Nov 24, 2023 00:01

West SAAB Story: Mar 13, 2014; by Athanatos
(and can't post for 222 days!)

ExcessBLarg! posted:

The trick is to preceed that with a
code:
echo u > /proc/sysrq-trigger
so you can reimage the active root volume when you don't have an A/B setup.

There is no need. netboot iPXE.

# ? Nov 24, 2023 00:22

Computer viking: May 30, 2011; Now with less breakage.

Wibla posted:

It's lovely, floral and kicks like a mule.

I need to buy another bottle, and of course polet here doesn't have it in stock

You could always order it for pickup, but of course the delay is unfortunate.

I haven't really had a lot of gin since a series of incidents as a student, but I think it's about time to give it another more measured try.

# ? Nov 24, 2023 01:17

Tesseraction: Apr 5, 2009

Computer viking posted:

I haven't really had a lot of gin since a series of incidents as a student, but I think it's about time to give it another more measured try.

A bunch of pitfalls, but biologically it's possible

# ? Nov 24, 2023 01:21

Subjunctive: Sep 12, 2006; ✨sparkle and shine✨

cheers, goons

# ? Nov 24, 2023 01:32

West SAAB Story: Mar 13, 2014; by Athanatos
(and can't post for 222 days!)

code:

251658240+0 records in
251658240+0 records out
128849018880 bytes (129 GB, 120 GiB) copied, 11033.8 s, 11.7 MB/s

# ? Nov 24, 2023 03:08

keep punching joe: Jan 22, 2006; Die Satan!

I've encountered a quite annoying issue on Arch that I am at a loss on how to troubleshoot/fix.

For some reason in the last 24 hours my system has stopped recognising usb devices (external drives, controllers etc.) I plug them in and nothing registers.

However, if they are plugged in before I boot the system all work and as long as one is plugged in I can disconnect/reconnect others and they are recognised. Then if I disconnect all USB devices/unmount drives the issue arises again, and nothing is recognised until I reboot with them plugged in.

The only change that has happened in my system over the past 24 hours was doing a system update through pacman (which I assume is where I broke it by not reading/understanding the console output).

All my system files are on a seperate /root partition so I am tempted just to nuke the lot to a fresh install, but it seems like such a ballache when I assume it's basically just a config file that I need to fix or restart a daemon.

keep punching joe fucked around with this message at 12:30 on Nov 24, 2023

# ? Nov 24, 2023 12:26

Tesseraction: Apr 5, 2009

Try running lsmod when it's working and when it's not and see if you need to modprobe a USB driver.

# ? Nov 24, 2023 12:36

VictualSquid: Feb 29, 2012; Gently enveloping the target with indiscriminate love.

keep punching joe posted:

I've encountered a quite annoying issue on Arch that I am at a loss on how to troubleshoot/fix.

For some reason in the last 24 hours my system has stopped recognising usb devices (external drives, controllers etc.) I plug them in and nothing registers.

However, if they are plugged in before I boot the system all work and as long as one is plugged in I can disconnect/reconnect others and they are recognised. Then if I disconnect all USB devices/unmount drives the issue arises again, and nothing is recognised until I reboot with them plugged in.

The only change that has happened in my system over the past 24 hours was doing a system update through pacman (which I assume is where I broke it by not reading/understanding the console output).

All my system files are on a seperate /root partition so I am tempted just to nuke the lot to a fresh install, but it seems like such a ballache when I assume it's basically just a config file that I need to fix or restart a daemon.

Are you using an usb-keyboard? I assume it also only works if plugged in at boot.
Is there some syslog output when you unplug the last device on a hub? And does "lsusb -tv" show anything interesting about your hubs?
The most likely thing is that one of your internal hubs gets deactivated for some reason if it is empty, no idea why it would do that though.

# ? Nov 24, 2023 14:03

Klyith: Aug 3, 2007; GBS Pledge Week

keep punching joe posted:

The only change that has happened in my system over the past 24 hours was doing a system update through pacman (which I assume is where I broke it by not reading/understanding the console output).

/var/log/pacman.log is a copy of the console log, so you can go back and review what it did and if there were any errors that you can dig into

# ? Nov 24, 2023 14:08

keep punching joe: Jan 22, 2006; Die Satan!

It actually seems to be fixed now but I have no idea how. Installed tlp and goofed around with some power settings, switched back to the lts kernel and done some reboots. :iiam:

# ? Nov 24, 2023 14:09

Watermelon Daiquiri: Jul 10, 2010; I TRIED TO BAIT THE TXPOL THREAD WITH THE WORLD'S WORST POSSIBLE TAKE AND ALL I GOT WAS THIS STUPID AVATAR.

Ok, I'm trying to get a windows virtual machine working in kvm/libvirt/qemu/et al with full gpu passthrough of the dedicated quadro on my thinkpad p53. I've got everything needed in initramfs and grub.cfg to pass the iommu group the card is in, but is there a way to easily check that vfio is working properly?

One other thing is that since the card is, of course, a mobile version the nvidia drivers apparently look for the battery. Someone made a solution by passing this SSDT acpi table to the vm, but I'm having problems getting it to work. I keep getting a permission denied error no matter if i run virt-manager as root or give the file itself full permissions/owned by my user name:kvm, libvirtd-qemu:kvm, root:root.

Someone who had the problem too got it to work by adding an exception to a apparmor config file, but it did nothing for me. How the heck can i get this to work?

Also, i can usually only run the vm once, sometimes twice per boot before the vm just fails to start and eventually fully freeze my laptop. Is this related to some vfio thing i missed?

# ? Nov 24, 2023 19:18

Computer viking: May 30, 2011; Now with less breakage.

Watermelon Daiquiri posted:

Ok, I'm trying to get a windows virtual machine working in kvm/libvirt/qemu/et al with full gpu passthrough of the dedicated quadro on my thinkpad p53. I've got everything needed in initramfs and grub.cfg to pass the iommu group the card is in, but is there a way to easily check that vfio is working properly?

Wibla posted:

Here's some help anyway

# ? Nov 24, 2023 19:28

ziasquinn: Jan 1, 2006; Fallen Rib

Watermelon Daiquiri posted:

Ok, I'm trying to get a windows virtual machine working in kvm/libvirt/qemu/et al with full gpu passthrough of the dedicated quadro on my thinkpad p53. I've got everything needed in initramfs and grub.cfg to pass the iommu group the card is in, but is there a way to easily check that vfio is working properly?

One other thing is that since the card is, of course, a mobile version the nvidia drivers apparently look for the battery. Someone made a solution by passing this SSDT acpi table to the vm, but I'm having problems getting it to work. I keep getting a permission denied error no matter if i run virt-manager as root or give the file itself full permissions/owned by my user name:kvm, libvirtd-qemu:kvm, root:root.

Someone who had the problem too got it to work by adding an exception to a apparmor config file, but it did nothing for me. How the heck can i get this to work?

Also, i can usually only run the vm once, sometimes twice per boot before the vm just fails to start and eventually fully freeze my laptop. Is this related to some vfio thing i missed?

did you start the systemd service? https://wiki.archlinux.org/title/Virt-manager
I never saw that section and kept going through all the extra configuration to allow permissions and poo poo when it wasn't even necessary if you just start the service...

As for checking, you can grep lspci to see if the main system is loaded with the igpu or the dgpu?

# ? Nov 24, 2023 19:54

Subjunctive: Sep 12, 2006; ✨sparkle and shine✨

The virtualization megathread may also have people who know about this: https://forums.somethingawful.com/showthread.php?threadid=3467608

# ? Nov 24, 2023 20:46

Watermelon Daiquiri: Jul 10, 2010; I TRIED TO BAIT THE TXPOL THREAD WITH THE WORLD'S WORST POSSIBLE TAKE AND ALL I GOT WAS THIS STUPID AVATAR.

Yeah, i started the socket/service. I'll ask that thread, though it's not that active...

# ? Nov 25, 2023 00:54

Watermelon Daiquiri: Jul 10, 2010; I TRIED TO BAIT THE TXPOL THREAD WITH THE WORLD'S WORST POSSIBLE TAKE AND ALL I GOT WAS THIS STUPID AVATAR.

Ok, it seems to work when starting it from virsh start vm... now to see if the battery things works

# ? Nov 25, 2023 01:51

Watermelon Daiquiri: Jul 10, 2010; I TRIED TO BAIT THE TXPOL THREAD WITH THE WORLD'S WORST POSSIBLE TAKE AND ALL I GOT WAS THIS STUPID AVATAR.

I found this guide too late, but it has everything one needs to accomplish in a relatively up to date manner: https://lantian.pub/en/article/modify-computer/laptop-muxed-nvidia-passthrough.lantian/ there are a couple things that need changing for ubuntu, but the basics are there

# ? Nov 25, 2023 18:44

Pablo Bluth: Sep 7, 2007; I've made a huge mistake.

I put a 2.5GBE pcie card in to my Fedora machine. It shows up in lspci, it shows up in dmesg, there are no obvious errors. But there's no interface shown anywere (eg ip addr, networkmanager, etc). Any suggestions for logs to check? From dmesg, it seems to be given the name enp3s0; there's a Link is Down message but no Link Is Up message as there is for enp5s0 (the motherboard interface). Using ip link to try and set enp3s0 to up returns an unknown device error.

edit: I upgraded to the latest Fedora. Now in dmesg I see

code:

[    2.098505] r8169 0000:03:00.0 enp3s0: renamed from eth0
[   10.073999] r8169 0000:03:00.0 enp3s0: Link is Down
[   12.502681] r8169 0000:03:00.0 enp3s0: Link is Up - 2.5Gbps/Full - flow control off
[   13.680175] r8169 0000:03:00.0 enp3s0: Link is Down

So for some reason it's briefly appearing then doing down and disappearing.

Pablo Bluth fucked around with this message at 00:56 on Nov 26, 2023

# ? Nov 25, 2023 23:59

Pablo Bluth: Sep 7, 2007; I've made a huge mistake.

So i figured it out. Installing the pcie card caused a renumbering of all the pcie devices, which broke two passthrough to VMs; I'd noticed straight away the first one which was the sata controller being passed to my storage VM (it wouldn't boot), but I hadn't noticed the VM that was meant to be given an old TV capture card was instead receiving the new Ethernet card. It took me a while to notice!

I can now backup from my windows machine to the ZFS pool at 280MB/s. As the windows pc motherboard doesn't have the pcie lanes for some 10G fibre, I'll consider a x2.25 speedup a good win for the price.

# ? Nov 26, 2023 19:17

pseudorandom name: May 6, 2007

I was wondering why your ethernet cards had such suspiciously low PCI bus numbers...

Weird that your motherboard doesn't have stable numbering, there's literally no reason to waste any effort on making that dense.

# ? Nov 26, 2023 19:56

Tacos Al Pastor: Jun 20, 2003

Had something weird randomly popup in my terminal. Wasnt running anything at the time, this just randomly popped up. Dont see anything online about this.

code:


tacos@mx:~
$ 
::1              fe00::0          ff00::0          ff02::1          ff02::2          ip6-allnodes     ip6-allrouters   ip6-localhost    ip6-localnet     ip6-loopback     ip6-mcastprefix  localhost        mx
tacos@mx:~

I'm running:

Linux mx 6.1.0-10-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.38-2 (2023-07-27) x86_64 GNU/Linux

Anyone know what might have happened? Im confused :iiam:

# ? Nov 26, 2023 22:49

spiritual bypass: Feb 19, 2008; Grimey Drawer

Are those ip6 loopback addrs?

e: yes, after scrolling sideways :pwn:

# ? Nov 26, 2023 22:54

Tacos Al Pastor: Jun 20, 2003

spiritual bypass posted:

Are those ip6 loopback addrs?

e: yes, after scrolling sideways

Yeah, I'm just not sure how that stuff just randomly pops up in the terminal.

edit: going to look through some logs...

# ? Nov 26, 2023 23:10

pseudorandom name: May 6, 2007

Programs like e.g. wall can write messages to all login terminals and depending how your xterm/whatever is configured it can count as a login terminal. Why anything would do this in 2023 is an open question.

# ? Nov 26, 2023 23:34

cruft: Oct 25, 2007

Tacos Al Pastor posted:

Had something weird randomly popup in my terminal. Wasnt running anything at the time, this just randomly popped up. Dont see anything online about this.
code:
tacos@mx:~
$ 
::1              fe00::0          ff00::0          ff02::1          ff02::2          ip6-allnodes     ip6-allrouters   ip6-localhost    ip6-localnet     ip6-loopback     ip6-mcastprefix  localhost        mx
tacos@mx:~
I'm running:

Linux mx 6.1.0-10-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.38-2 (2023-07-27) x86_64 GNU/Linux

Anyone know what might have happened? Im confused

Bash can complete hostnames. Do these match up with what's in /etc/hosts perchance?

If this were something writing to your tty, you wouldn't expect to see the prompt again.

# ? Nov 27, 2023 04:12

Combat Pretzel: Jun 23, 2004; No, seriously... what kurds?!

Hmm, I'm just toying with the Large Text accessibility option in Gnome in a VM. Seems like this might make most of the UI look similar to 150% HiDPI on Windows, i.e. the larger UI surfaces in Gnome approach same dimensions, while browsers and Electron type apps get 125% on top of that, where applicable, to scale appropriately, and I might be set. I guess my weekends planned out already. I hope NVidia downclocking works better nowadays.

--edit:
Fractional scaling now in Gnome 45 and GTK4, at least for Wayland, or not? I find conflicting information about this. Last I looked into it, Gnome devs were their usual and dismissing it and targeting GTK5. Now it seems like there was development done to make it work in GTK4 already?

Combat Pretzel fucked around with this message at 21:14 on Nov 28, 2023

# ? Nov 28, 2023 20:46

Combat Pretzel: Jun 23, 2004; No, seriously... what kurds?!

Took a bit of mucking about, because neither installing the NVidia driver went smoothly, nor installing Chrome. Year of desktop Linux.

That said, my power usage grievances with NVidia are gone. The card is running cool and downclocked even with two 4K displays at 120hz. And just browsing the web, the thing uses less power on the socket than with Windows, lol. And the experimental fractional scaling thing seems to work OK so far.

# ? Nov 28, 2023 22:14

Adbot: ADBOT LOVES YOU

# ? May 24, 2024 16:44

Subjunctive: Sep 12, 2006; ✨sparkle and shine✨

Tacos Al Pastor posted:

Had something weird randomly popup in my terminal. Wasnt running anything at the time, this just randomly popped up. Dont see anything online about this.
code:
tacos@mx:~
$ 
::1              fe00::0          ff00::0          ff02::1          ff02::2          ip6-allnodes     ip6-allrouters   ip6-localhost    ip6-localnet     ip6-loopback     ip6-mcastprefix  localhost        mx
tacos@mx:~
I'm running:

Linux mx 6.1.0-10-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.38-2 (2023-07-27) x86_64 GNU/Linux

Anyone know what might have happened? Im confused

it looks like you put something into the background via prompt$ whatever & and then it emitted that. if you run jobs does it show anything? otherwise run tty to find out what terminal device you�re on, and then look through the output of ps aux | less to find ones with the same terminal listed

# ? Nov 28, 2023 22:19

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Linux Questions Thread: a bunch of pitfalls, but technically it's possible

«‹›1033 »