Category Archives: Hardware

Threadripper / Epyc processor core optimization

I had a pet project (folding@home) where I wanted to maximize computing power. I became frustrated with default CPU scheduling of my folding@home threads. Ideal performance would keep similar threads on the same CPU, but the threads were jumping all over the place, which was impacting performance.

Step one was to figure out which threads belonged to which physical cores. I found on this site that you can use cat to find out what your “sibling threads” are:

cat /sys/devices/system/cpu/cpu{0..15}/topology/thread_siblings_list

The above command is for my Threadripper & Epyc systems, which each have 16 cores hyperthreaded to 32 cores. Adjust the {0..15} number to match your number of cores (core 0 being the fist core.) This was my output:

cat /sys/devices/system/cpu/cpu{0..15}/topology/thread_siblings_list

0,16
1,17
2,18
3,19
4,20
5,21
6,22
7,23
8,24
9,25
10,26
11,27
12,28
13,29
14,30
15,31

Now that I know the sibling threads are offset by 16, I can use this information to optimize my folding@home VMs. I modified my CPU pinning script to take this into consideration. The script ensures that each VM is pinned to only use sibling threads (ensuring they all stay on the same physical CPU.)

This script should be used with caution. It pins processes to specific CPUs, which limits the kernel scheduler’s ability to move things around if needed. If configured badly this can cause the machine to lock up or VMs to be terminated.

I saw some impressive results spinning up four separate 8 core VMs and pinning them to sibling cores using this script. It almost doubled the rate at which I completed folding@home work units.

And now, the script:

#!/bin/bash
#Properly assign CPU cores to their respective die for EPYC/Threadripper systems
#Based on how hyperthreads are done in these systems
#cat /sys/devices/system/cpu/cpu{0..15}/topology/thread_siblings_list

#The script takes two arguments - the ID of the Proxmox VM to modify, and the core to begin the VM on
#If running this against multiple VMs, make sure to increment this second number by half of the cores of the previous VM
#For example, if I have one 8 core VM and I run this script specifying 0 for the offset, if I spin up a second VM, the second argument would be 4
#this would ensure the second VM starts on core 4 (the 5th core) and assigns sibling cores to match

set -eo pipefail

#take First argument as which VMID to pin CPU cores to, the second argument is which core to start pinning to
VMID=$1
OFFSET=$2

#Determine offset for sibling threads
SIBLING_THREAD_OFFSET=$(cat /sys/devices/system/cpu/cpu0/topology/thread_siblings_list| sed 's/,/ /g' | awk '{print $2}')

#Function to determine number of CPU cores a VM has
cpu_tasks() {
	expect <<EOF | sed -n 's/^.* CPU .*thread_id=\(.*\)$/\1/p' | tr -d '\r' || true
spawn qm monitor $VMID
expect ">"
send "info cpus\r"
expect ">"
EOF
}

#Only act if VMID & OFFSET are set
if [[ -z $VMID  || -z $OFFSET ]]
then
	echo "Usage: cpupin.sh <VMID> <OFFSET>"
	exit 1
else
	#Get PIDs of each CPU core for VM, count number of VM cores, and get even/odd PIDs for assignment
	VCPUS=($(cpu_tasks))
	VCPU_COUNT="${#VCPUS[@]}"
	VCPU_EVEN_THREADS=($(for EVEN_THREAD in "${VCPUS[@]}"; do echo $EVEN_THREAD; done | awk '!(NR%2)'))
	VCPU_ODD_THREADS=($(for ODD_THREAD in "${VCPUS[@]}"; do echo $ODD_THREAD; done | awk '(NR%2)'))

	if [[ $VCPU_COUNT -eq 0 ]]; then
		echo "* No VCPUS for VM$VMID"
		exit 1
	fi

	echo "* Detected ${#VCPUS[@]} assigned to VM$VMID..."
	echo "* Resetting cpu shield..."

	#Start at offset CPU number, assign odd numbered PIDs to their own CPU thread, then increment CPU core number
	#0-3 if offset is 0, 4-7 if offset is 4, etc
	ODD_CPU_INDEX=$OFFSET
	for PID in "${VCPU_ODD_THREADS[@]}"
	do
		echo "* Assigning ODD thread $ODD_CPU_INDEX to $PID..."
		taskset -pc "$ODD_CPU_INDEX" "$PID"
		((ODD_CPU_INDEX+=1))
	done

	#Start at offset + CPU count, assign even number PIDs to their own CPU thread, then increment CPU core number
	#16-19 if offset is 0,	20-23 if offset is 4, etc
	EVEN_CPU_INDEX=$(($OFFSET + $SIBLING_THREAD_OFFSET))
	for PID in "${VCPU_EVEN_THREADS[@]}"
	do
		echo "* Assigning EVEN thread $EVEN_CPU_INDEX to $PID..."
		taskset -pc "$EVEN_CPU_INDEX" "$PID"
		((EVEN_CPU_INDEX+=1))
	done
fi

Folding@home opencl error fix

I decided to contribute my GPU on my Ubuntu-based system to the Folding@Home effort for COVID-19. I kept getting this error message for my NVIDIA GeForce GTX 1050 TI when I tried:

ERROR:WU00:FS00:Failed to start core: OpenCL device matching slot 0 not found, make sure the OpenCL driver is installed or try setting 'opencl-index' manually

I had the nvidia opencl packages installed but apparently missed something. I finally found on the folding at home forum what I was missing – ocl-icd-opencl-dev

sudo apt install ocl-icd-opencl-dev

After running the above command and restarting the FAHClient service, the GPU started folding. For science!


EDIT 5/6/2020: After a re-install I had the issue where the GPU wouldn’t show up at all. It addition to ocl-icd-opencl-dev, it looks like you also need nvidia-cuda-dev.

sudo apt install ocl-icd-opencl-dev nvidia-cuda-dev

WD*EZRZ NAS array spindown fix

I recently acquired some 5TB Western Digital Blue drives (WD50EZRZ.) These particular drives were shucked from external USB enclosures. When I tried to add them into my ZFS raid array, though, I ran into constant problems. I would continually get errors like this from the kernel:

[155069.298001] sd 0:0:10:0: attempting task abort! scmd(ffff8f0678887100)
[155069.298005] sd 0:0:10:0: [sdk] tag#5 CDB: Write(16) 8a 00 00 00 00 01 a8 1e 77 10 00 00 00 58 00 00
[155069.298008] scsi target0:0:10: handle(0x0014), sas_address(0x5001438023a93296), phy(22)
[155069.298010] scsi target0:0:10: enclosure logical id(0x5001438023a932a5), slot(53) 
[155069.298012] sd 0:0:10:0: task abort: SUCCESS scmd(ffff8f0678887100)
[155069.298016] sd 0:0:10:0: [sdk] tag#5 FAILED Result: hostbyte=DID_TIME_OUT driverbyte=DRIVER_OK
[155069.298018] sd 0:0:10:0: [sdk] tag#5 CDB: Write(16) 8a 00 00 00 00 01 a8 1e 77 10 00 00 00 58 00 00
[155069.298020] blk_update_request: I/O error, dev sdk, sector 7115536144
[155069.298023] zio pool=storage vdev=/dev/disk/by-id/ata-WDC_WD50EZRZ-32RWYB1_WD-WX31XXXXXVA-part1 error=5 type=2 offset=3643153457152 size=45056 flags=180880

After a couple of said errors, the drive would be marked as bad and taken out of the array. A battery of tests on a different system revealed the drives to be fine. It did not matter where I inserted these drives on my NAS, they did the same thing, even on ports I knew had working drives. It wasn’t a cabling or other hardware issue.

The drives would resilver back into the array just fine, and then pop out again at random intervals – sometimes minutes later, other times hours later. After a lot of research I came across this post that got me thinking – this sounds like a drive spindown issue! The random nature of it could simply be the drives not being used and then powering themselves down.

I tried using hdparm to set the spindown timer but was greeted with this lovely error:

sudo hdparm -B /dev/sdk
/dev/sdk:
 APM_level	= not supported

I eventually found this post complaining about their Western Digital drives spinning down aggressively.

idle3 to the rescue

The above post mentions apmtimer which did not help me, however more searches reveled this godsend: idle3-tools

idle3-tools is an open source utility to handle spindown on Western Digital drives themselves (not the OS level.)

Download & compile idle3:

wget https://sourceforge.net/projects/idle3-tools/files/latest/download
cd idle3-tools-0.9.1/
make
sudo make install

Use idle3 to query current spindown status (update drive letters to suit your needs)

for drive in {a..p}; do echo /dev/sd$drive; sudo idle3ctl -g /dev/sd$drive; done

For anything that doesn’t say Idle3 timer is disabled run the following:

sudo idle3ctl -s 0 /dev/sd(DRIVE_LETTER)

No more drive spindown!

proxmox bond not present fix

I really banged my head on the wall on this one. I recently decided to re-architect my networking setup in proxmox to utilize bonded network configuration. I followed this writeup exactly. The problem is it didn’t work.

I would copy the example exactly, only changing the interface name, and yet every time I tried to start the networking service I would get this lovely error:

rawdevice bond0 not present

I finally found on the Debian Wiki one critical line :

First install the ifenslave package, necessary to enable bonding

For some reason the ProxMox howtos don’t speak of this – I guess because it comes installed by default. I discovered, however, that if you install ifupdown2 it removes ifenslave. I had installed ifupdown2 in the past to reload network configuration without rebooting. Aha!

I re-installed ifenslave (which removed ifupdown2 and re-installed ifupdown) and suddenly, the bond worked!

Bond not falling back to primary intrerface

I had configured my bond in active – backup mode. I wanted it to prefer the faster link, but if there was a failure in that link it wouldn’t switch back automatically (thanks to this site for showing me the command to check:

cat /proc/net/bonding/bond0

I read again in Debian bonding wiki that I needed to add this directive to the bond:

        bond-primary enp2s0

Here is my complete working active-backup configuration, assigning vlan 2 to the host, and making enp2s0 (the 10gig nic) the primary, with a 1gig backup (eno1)

auto bond0
iface bond0 inet manual
        slaves enp2s0 eno1
        bond-primary enp2s0
        bond_miimon 100
        bond_mode active-backup

iface bond0.2 inet manual

auto vmbr0v2
iface vmbr0v2 inet static
        address 192.168.2.2
        netmask 255.255.255.0
        gateway 192.168.2.1
        bridge_ports bond0.2
        bridge_stp off
        bridge_fd 0

auto vmbr0
iface vmbr0 inet manual
        bridge_ports bond0
        brideg_stp off
        bridge_fd 0

migrate between amd & intel in proxmox

I recently acquired an Intel based server and plugged it into my AMD-based Proxmox cluster. I ran into an issue transferring from AMD to Intel boxes (the other direction worked fine.) After a few moments, every VM that moved from AMD to Intel would kernel panic.

Fortunately I found here that the fix is to add a few custom CPU flags to your VMs. Once I did this they could move back and forth freely (assuming they had the kvm64 CPU assigned to them – host obviously won’t work.)

qm set *VMID* --args "-cpu 'kvm64,+ssse3,+sse4.1,+sse4.2,+x2apic'"

Fix proxmox iommu operation not permitted

In trying to passthrough some LSI SAS cards to a VM I kept receiving this error:

kvm: -device vfio-pci,host=0000:03:00.0,id=hostpci0,bus=ich9-pcie-port-1,addr=0x0,rombar=0: vfio 0000:03:00.0: failed to setup container for group 7: Failed to set iommu for container: Operation not permitted

I found on this post that the fix is to add a line to /etc/modprobe.d/vfio.conf with the following:

options vfio_iommu_type1 allow_unsafe_interrupts=1

then reboot the host. It worked in my case!

Dell LSI SAS2008 2TB drive fix

I just recently got a $40 external SAS adapter for my new storage server. The plan was to create a DAS device from my old NAS chassis and have it be driven by my new storage server (new to me anyway – a Dell PowerEdge R610.) I ordered what was listed simply as “Dell SAS External Dual Ports PCI-E 6GB/S Host Bus Server Adapter 12DNW 342-0910 Consumer Electronics” from Amazon for $40 to accomplish this goal.

When I plugged everything in, to my dismay none of my disks with greater than 2TB capacity showed up. Well, they sort of showed up – they all reported capacities of exactly 2TB. I was clearly running into some sort of firmware issue.

lspci revealed this card uses the LSI SAS2008 chipset, which from what I’ve read is capable of drives greater than 2TB in size. I later found the model number of my card – Dell PERC H200E – which proved to be quite vital information. After hours of digging around in unholy corners of the internet I finally arrived on this Dell Support page. It had exactly what I was hoping for:

ENHANCEMENTS:
– Added support for SAS HDDs larger than 2TB

To flash this I chose to create a bootable dos ISO as per the instructions here. First, download the Windows installer, open with your archive program of choice and extract to the folder you’re going to build your ISO from. Then follow the instructions linked to above of downloading a freeDOS ISO, extracting it to the same folder you extracted the firmware to, then running the command to build your ISO (adjust as needed)

mkisofs -o <ISO_OUTPUT_LOCATION -q -l -N -boot-info-table -iso-level 4 -no-emul-boot -b isolinux/isolinux.bin -publisher "FreeDOS - www.freedos.org" -A "FreeDOS beta9 Distribution" -V FDOS_BETA9 -v .

I got so far and yet tripped at the finish line. If you simply run flash.bat you’ll be greeted with a message saying no compatible adapters were found. Fortunately that’s a LIE. My savior was this writeup on how to flash certain versions of these cards to IT mode. I didn’t care about IT mode (my card is not a RAID card) but it had the information I needed. Here are the magic commands!

sas2flsh -listall

#Use the number in the first column to get the SAS Address for the card.
sas2flsh -c 0 -list
#Write down the SAS Address and continue to the next steps.
sas2flsh -o -f 6GBPSAS.FW
sas2flsh -o -sasadd 5xxxxxxxxxxxxxxx (replace this address with the one you wrote down in the first steps).

Reboot, and finally, after hours of banging my head on the wall… success!!!

These 4 drives were only being reported as 2TB before

I didn’t end up using it but in my internet travels I came across this. Broadcom offers a neat utility called the LSI pre-boot USB tool that I didn’t end up using: https://www.broadcom.com/support/knowledgebase/1211161499804/lsi-pre-boot-usb-tool-download

Update 3/7/2020

I had some issues with my 4tb+ drives dropping out of my zpools. I found better firmware to flash in order to fix it. It was very frustrating to flash, however. I tried following the instructions as laid out here but I was met with this lovely message:

"Cannot Flash IT Firmware over IR Firmware"..

I found this guide on how to use the megarec utility to wipe the firmware in order to flash over properly. I was able to find the megarec utility here.

I very frustratingly found I couldn’t use the megarec utility on my Dell server; megarec would simply hang

I ended up taking the card out and putting it into my desktop to run megarec commands. Comically, my desktop had a chipset that caused sas2flash not to work!! It would fail with the message

Failed to initialize PAL

Instructions per this page were to boot to EFI and run the flash utilities there, but that desktop didn’t have an EFI shell and I couldn’t get it to boot one from USB.

My final resort: an even older desktop (my Dad’s old PC, circa 2008.) It did the deed!

FINALLY

With both utilities working I was still having trouble with sas2flash erroring out on me. I finally found the wise words from fourlynx on this homelab reddit discussion on the final song and dance I had to perform to get my Dell H200 card to work with the LSI firmware I wanted

  1. Flash to Dell 6GBPSAS.FW
    1. I used megarec to wipe the card first before it would let me install that firmware
  2. Erase the card
    1. sas2flsh -o -e 7 -c 0
  3. Flash to 6GBPSAS.FW again
    1. sas2flsh -o -f 6GBPSAS.FW
    2. If asked me to state a firmware, I entered 6GBPSAS.FW, waited for it to finish, then ran the sas2flsh command (flashed a total of 3 times the same firmware.)
  4. Reboot
  5. Finally flash LSI firmware
    1. sas2flsh -o -f 2118it.bin

No need to flash BIOS (-b flag) if not going to boot from that controller. Also no need to set SAS address if it’s the only card in the server.

Words of wisdom from fourlynx:

For what concerns your case, I’d try to flash it to the Dell firmware first (any of your choice, for H200I, H200A or with the 6GBPSAS.fw). From there, clear it completely sas2flsh -o -e 7 -c 0 and flash the 6GBPSAS.fw before rebooting. You should now have better luck in crossflashing that to the LSI firmware. Note that you’ll need to use the v5 or v7 version of the flasher to do this step as newer versions will refuse to crossflash. You can then flash the bootloader for EFI (x64sas2.rom) or for BIOS (mptsas2.rom) at your leisure according to what you’re going to use, or flash both, or none if you’re not going to boot from those drives at all but instead use an USB key.

megarec -cleanflash 0 is equivalent to sas2flsh -o -e 7, btw, and the megarec -writesbr sbrempty.bin command that is often found in guides is only relevant when coming from a M1015 afaik, so not being able to use megarec is not a show stopper.

I feel I should add that, contrary to what seems the popular opinion in the various guides, these cards aren’t really easy to brick and I haven’t managed to achieve that despite all the experiments I’ve subjected them to 🙂

Update 3/8/2020

I still had issues with a drive popping out of the array so I found this page with an even better firmware for my card:

https://www.ixsystems.com/community/threads/crossflash-dell-h200e-to-lsi-9200-8e.41307/

Things seem more stable now!

Java IDRAC 6 Ubuntu 18.04 setup

I recently acquired a Dell PowerEdge R610 and had a hard time getting its iDRAC to work properly on my ElementaryOS setup (Ubuntu 18.04 derivative.) I had two problems: Connection failed error and keyboard not working.

Connection Failed

After much searching I finally found this post:

The post explains the problem is with the security settings of Java 8+ preventing the connection. I didn’t know where my security file was so I first ran a quick find command to find it:

sudo find / -name java.security

In my case it was located in /etc/java-11-openjdk/security/java.security

The last step was to remove RC4 from the list of blacklisted ciphers, as this is the cause of the problem.

sudo vim /etc/java-11-openjdk/security/java.security

#change jdk.tls.disabledAlgorithms=SSLv3, RC4, DES, MD5withRSA, DH keySize < 1024, \ to be:
jdk.tls.disabledAlgorithms=SSLv3, DES, MD5withRSA, DH keySize < 1024, \

Save and exit, and iDRAC will now load!

Except now…

Keyboard doesn’t work

My system was defaulting to using JRE 11, which apparently causes the keyboard to not function at all. I found on this reddit post that you really need an older version of Java. To do so on Ubuntu 18.04 you need to install it along with the icedtea plugin and run update-alternatives

sudo apt install openjdk-8-jre icedtea-8-plugin

Edit /etc/java-8-openjdk/security/java.security and remove the restriction on the RC4 algorhythm. Then configure the system to run java 8:

sudo update-alternatives --config java
#select java 8

Lastly, configure the icedtea plugin to run Java 8 instead of 11, because for some reason this plugin ignores the system java settings. Launch the IcedTea Web Control panel (find it in your system menu) and then Navigate to JVM settings. Enter /usr/ in the section “Set JVM for IcedTea-Web – working best with OpenJDK” section. Then hit Apply / OK

Phew. FINALLY you should be able to use iDRAC 6 on your modern Ubuntu system.

proxmox suspend & resume scripts

Update 12/17/2019: Added logic to wait for VM to be suspended before suspending the shypervisor

Update 12/8/2019: After switching VMs I needed to tweak the pair of scripts. I modified it to make all the magic happen on the hypervisor; the VM simply needs to SSH into the hypervisor and call the script. The hypervisor now also needs access to SSH via public key to the VM to tell it to suspend.

#!/bin/sh
#ProxMox suspend script part 1 of 2
#To be run on the VM 
#All this does is call the suspend script on the hypervisor
#This could also just be a bash alias

####### Variables #########
HYPERVISOR=        #Name / IP of the hypervisor
SSH_USER=          #User to SSH into hypervisor as
HYPERVISOR_SCRIPT= #Path to part 2 of the script on the hypervisor

####### End Variables ######

#Execute server suspend script
ssh $SSH_USER@$HYPERVISOR "$HYPERVISOR_SCRIPT" &
#!/bin/bash
#ProxMox suspend script part 2 of 2
#Script to run on the hypervisor, it waits for VM to suspend and then suspends itself
#It relies on passwordless sudo configured on the VM as well as SSH keys to allow passwordless SSH access to the VM from the hypervisor
#It resumes the VM after it resumes itself
#Called from the VM

########### Variables ###############

VM=             #Name/IP of VM to SSH into
VM_SSH_USER=    #User to ssh into the vm with
VMID=           #VMID of VM you wish to suspend

########### End Variables############

#Tell guest VM to suspend
ssh $VM_SSH_USER@$VM "sudo systemctl suspend"

#Wait until guest VM is suspended, wait 5 seconds between attempts
while [ "$(qm status $VMID)" != "status: suspended" ]
do 
    echo "Waiting for VM to suspend"
    sleep 5 
done

#Suspend hypervisor
systemctl suspend

#Resume after shutdown
qm resume $VMID

I have a desktop running ProxMox. My GUI is handled via a virtual machine with physical hardware passed through it. The challenge with this setup is getting suspend & resume to work properly. I got it to work by suspending the VM first, then the host; on resume, I power up the host first, then resume the VM. Doing anything else would cause hardware passthrough problems that would force me to reboot the VM.

I automated the suspend process by using two scripts: one for the VM, and one for the hypervisor. The first script is run on the VM. It makes an SSH command to the hypervisor (thanks to this post) to instruct it to run the second half of the script; then initiates a suspend of the VM.

The second half of the script waits a few seconds to allow the VM to suspend itself, then instructs the hypervisor to also go into suspend. I had to split these into two scripts because once the VM is suspended, it can’t issue any more commands. Suspending the hypervisor must happen after the VM itself is suspended.

Here is script #1 (to be run on the VM) It assumes you have already set up a private/public key pair to allow for passwordless login into the hypervisor from the VM.

#!/bin/sh
#ProxMox suspend script part 1 of 2
#Tto be run on the VM so it suspends before the hypervisor does

####### Variables #########
HYPERVISOR=HYPERVISOR_NAME_OR_IP
SSH_USER=SSH_USER_ON_HYPERVISOR
HYPERVISOR_SCRIPT_LOCATION=NAME_AND_LOCATION_OF_PART2_OF_SCRIPT

####### End Variables ######

#Execute server suspend script, then suspend VM
ssh $SSH_USER@$HYPERVISOR  $HYPERVISOR_SCRIPT_LOCATION &

#Suspend
systemctl suspend

Here is script #2 (which script #1 calls), to be run on the hypervisor

#!/bin/bash
#ProxMox suspend script part 2 of 2
#Script to run on the hypervisor, it waits for VM to suspend and then suspends itself
#It resumes the VM after it resumes itself

########### Variables ###############

#Specify VMid you wish to suspend
VMID=VMID_OF_VM_YOU_WANT_TO_SUSPEND

########### End Variables############

#Wait 5 seconds before doing anything to allow for VM to suspend
sleep 5

#Suspend hypervisor
systemctl suspend

#Resume after shutdown
qm resume $VMID

It works on my machine 🙂

Primary VGA passthrough in ProxMox

I recently decided to amplify my VFIO experience by experimenting with passing my primary display adapter to a VM in proxmox. Previously I had just run tasksel on the proxmox host itself to install a GUI. I wanted better separation from the server side of proxmox and the client side. I also wanted to be able to distro-hop while maintaining the proxmox backend.

Initially I tried following my guide for passing through a secondary graphics card but ran into a snag. It did not work with my primary card and kept outputting these errors:

device vfio-pci,host=09:00.0,id=hostdev0,bus=pci.4,addr=0x0: Failed to mmap 0000:09:00.0 BAR 1. Performance may be slow

After much digging I finally found this post which explained I needed to unbind a few things for it to work properly:

echo 0 > /sys/class/vtconsole/vtcon0/bind
echo 0 > /sys/class/vtconsole/vtcon1/bind
echo efi-framebuffer.0 > /sys/bus/platform/drivers/efi-framebuffer/unbind

After more searching I found this post on reddit which had a nifty script for automating this when VM startup is desired. I tweaked it a bit to suit my needs.

Find your IDs for GPU by doing lspci and looking for your adapter. Find the IDs by running lspci -n -s <GPU location discovered with lspci>. Lastly VMID is the promxox ID for the VM you wish to start.

#!/bin/sh
#Script to launch Linux desktop
#Adapted from from https://www.reddit.com/r/VFIO/comments/abfjs8/cant_seem_to_get_vfio_working_with_qemu/?utm_medium=android_app&utm_source=share

GPU=09:00
GPU_ID="10de 1c82"
GPU_AUDIO="10de 0fb9"
VMID=116

# Remove the framebuffer and console
echo 0 > /sys/class/vtconsole/vtcon0/bind
echo 0 > /sys/class/vtconsole/vtcon1/bind
echo efi-framebuffer.0 > /sys/bus/platform/drivers/efi-framebuffer/unbind

# Unload the Kernel Modules that use the GPU
modprobe -r nvidia_drm
modprobe -r nvidia_modeset
modprobe -r nvidia
modprobe -r snd_hda_intel

# Load the vfio kernel module
modprobe vfio
modprobe vfio_iommu_type1
modprobe vfio-pci

#Assign card to vfio-pci
echo -n "${GPU_ID}" > /sys/bus/pci/drivers/vfio-pci/new_id
echo -n "${GPU_AUDIO}" > /sys/bus/pci/drivers/vfio-pci/new_id

#Start desktop
sudo qm start $VMID

#Wait here until the VM is turned off
while [ "$(qm status $VMID)" != "status: stopped" ] 
do
 sleep 5
done

#Reassign primary graphics card back to host
echo -n "0000:${GPU}.0" > /sys/bus/pci/drivers/vfio-pci/unbind
echo -n "0000:${GPU}.1" > /sys/bus/pci/drivers/vfio-pci/unbind
echo -n "${GPU_ID}" > /sys/bus/pci/drivers/vfio-pci/remove_id
echo -n "${GPU_AUDIO}" > /sys/bus/pci/drivers/vfio-pci/remove_id
rmmod vfio-pci
modprobe nvidia
modprobe nvidia_drm
modprobe nvidia_modeset
modprobe snd_hda_intel
sleep 1
echo -n "0000:${GPU}.0" > /sys/bus/pci/drivers/nvidia/bind
echo -n "0000:${GPU}.1" > /sys/bus/pci/drivers/snd_hda_intel/bind
sleep 1
echo efi-framebuffer.0 > /sys/bus/platform/drivers/efi-framebuffer/bind
echo 1 > /sys/class/vtconsole/vtcon0/bind
echo 1 > /sys/class/vtconsole/vtcon1/bind

With my primary adapter passed through I realized I also want other things passed through, mainly USB. I tried Proxmox’s USB device passthrough options but it doesn’t work well with USB audio (stutters and choppy.) I wanted to pass through my whole USB controller to the VM.

This didn’t work as well as I had planned due to IOMMU groups. A great explanation of IOMMU groups can be found here. I had to figure out which of my USB controllers were in which IOMMU group to see if I could pass the whole thing through or not (some of them were in the same IOMMU group as SATA & network controllers, which I did not want to pass through to the VM.)

Fortunately I was able to discover which USB controllers I could safely pass through first by running lspci to see the device ID, then running find to see which IOMMU group it was in, then checking against lspci to see what other devices were in that group. The whole group comes over together when you pass through to a VM.

First determine the IDs of your USB controllers

lspci | grep USB

01:00.0 USB controller: Advanced Micro Devices, Inc. [AMD] Device 43ba (rev 02)
08:00.0 USB controller: Renesas Technology Corp. uPD720201 USB 3.0 Host Controller (rev 03)
0a:00.3 USB controller: Advanced Micro Devices, Inc. [AMD] Device 145c
43:00.3 USB controller: Advanced Micro Devices, Inc. [AMD] Device 145c

Next get which IOMMU group these devices belong to

find /sys/kernel/iommu_groups/ -type l|sort -h|grep '01:00.0\|08:00.0\|0a:00.3\|43:00.3'

/sys/kernel/iommu_groups/14/devices/0000:01:00.0
/sys/kernel/iommu_groups/15/devices/0000:08:00.0
/sys/kernel/iommu_groups/19/devices/0000:0a:00.3
/sys/kernel/iommu_groups/37/devices/0000:43:00.3

Then see what other devices use the same IOMMU group (the group is the number after /sys/kernel/iommu_groups/)

find /sys/kernel/iommu_groups/ -type l|sort -h | grep '/14\|/15\|/19\|/37'

/sys/kernel/iommu_groups/14/devices/0000:01:00.0
/sys/kernel/iommu_groups/14/devices/0000:01:00.1
/sys/kernel/iommu_groups/14/devices/0000:01:00.2
/sys/kernel/iommu_groups/14/devices/0000:02:00.0
/sys/kernel/iommu_groups/14/devices/0000:02:04.0
/sys/kernel/iommu_groups/14/devices/0000:02:05.0
/sys/kernel/iommu_groups/14/devices/0000:02:06.0
/sys/kernel/iommu_groups/14/devices/0000:02:07.0
/sys/kernel/iommu_groups/14/devices/0000:04:00.0
/sys/kernel/iommu_groups/14/devices/0000:05:00.0
/sys/kernel/iommu_groups/14/devices/0000:06:00.0
/sys/kernel/iommu_groups/15/devices/0000:08:00.0
/sys/kernel/iommu_groups/19/devices/0000:0a:00.3
/sys/kernel/iommu_groups/37/devices/0000:43:00.3

As you can see one of my USB controllers (01:00.0) has a whole bunch of stuff in its IOMMU group, so I don’t want to use it lest I bring all those other things into the VM with it. The other three, though, are isolated in their groups and thus are perfect for passthrough.

In my case I passed through 0a:00.3 & 43:00.3 as 08:00.0 is a PCI card I want passed through to my Windows VM. This passed through about 2/3 of the USB ports on my system to my guest VM.