nvidia – Vga pass-through fails with any kernel after 5.3.0-19 gtx 1660ti

I have gtx 1660ti.

17:00.0 VGA compatible controller: NVIDIA Corporation TU116 (GeForce GTX 1660 Ti) (rev a1)
17:00.1 Audio device: NVIDIA Corporation Device 1aeb (rev a1)
17:00.2 USB controller: NVIDIA Corporation Device 1aec (rev a1)
17:00.3 Serial bus controller (0c80): NVIDIA Corporation Device 1aed (rev a1)

When I try to disconnect the serial bus controller from the card

virsh nodedev-detach pci_0000_17_00_3

My system load is gradually increasing. The system freezes slowly. I have to force reset (turn off). Otherwise nothing helps.

Now it is kernel 5.6 and the problem is not solved. I'm looking for a workaround.

I am not using the 17: 00.3 bus in the VM. Because it's in the iommu group, I have to separate it. Is there a way for the kernel to ignore or disable the 17: 00.3 bus so the VM doesn't complain and force me to disconnect it?

Virtualization – Does VMware Workstation support PCI pass-through with VT-d technology?

Do it VMware Workstation Support for version 15.5.1 USB Passthrough on a host consisting of the Z68x chipset With Intel i7-3770 CPUwho has that VT-d Feature enabled (also known as: IOMMU virtualization)?

The USB controller is the one NEC/Renesas ┬ÁPD720202 Chip that is connected to the CPU via the PCI Express bus.

nvidia – Linux KVM guest with GPU pass-through does not work (black screen)

To outline my situation, I decided to run a gaming VM via KVM with Windows 10 to play the Master Chief collection. In the meantime, I realized that it will use its own monitors (I want to connect both my large monitors to my GPU) and input devices. In addition, my host operating system is significantly delayed when I run CPU-intensive applications such as a browser at the same time.

For this reason, I decided to use VMs that only use a small, dedicated host operating system (probably Debian with XFCE) instead, and to create an Ubuntu MATE VM next to the Windows VM, which as my "workstation" – PC acts. I was able to search and work on everything on my beautiful large screens and didn't have to worry about things like mirrors or tangled cables. I could also stop this VM to run the Windows VM and save resources, at least that's how I understand it. No powerful graphics card is required for the Linux VM since I will not be playing on it. I only use GPU pass-through to use my large monitors with ease. Please let me know if this is a bad idea or if there is a better way to do it.

I am assuming this works based on the assumption that it is okay to use GPU passthrough on multiple VMs and that it will work as long as the VMs are not running at the same time "). Please let me know whether this can cause problems.

I've worked Windows 10 pretty well (it stutters occasionally, overall performance is reasonable but could be faster), it shows up well on my large monitors, and I use my gaming mouse and keyboard quite well. It works well with GPU pass-through.

However, I have problems with my Workhorse VM.

I installed Ubuntu MATE on the guest and set up GPU pass-through in the Virtual Machine Manager. However, it never appears on my monitor. I can only view it from the graphical console in Virtual Machine Manager.

When I run screenfetch on the client, my GPU is reported as llvmpipe (LLVM 9.0, 256 bits)

lspci -k Indicates that my PCI-E graphics card is using it nvidia Kernel driver. I use nvidia proprietary driver metapackage Version 435 selected in "Additional Drivers". I have tried nouveau even without success.

My suspicion is that Ubuntu MATE is not using the correct display driver. In the display settings it only knows an "unknown" monitor with the name "Virtual 1".

When I run nvidia-settings I get the error ERROR: An internal driver error occurred The NVidia Control Panel window is completely blank and contains no controls. I originally suspected that NVIDIA intentionally refused to work on a VM, as is the case with Windows. So I added the standard section to the XML file to hide it from nvidia


      

Here is the XML file for my Ubuntu MATE guest computer:


  ubuntu-mate-workstation
  bd317e3a-61a0-468e-ac1b-aa260de83c67
  
    
      
    
  
  8192000
  8192000
  2
  
    hvm
    /usr/share/OVMF/OVMF_CODE.fd
    /var/lib/libvirt/qemu/nvram/ubuntu-mate-workstation_VARS.fd
    
  
  
    
    
    
      
    
    
  
  
    
  
  
    
    
    
  
  destroy
  restart
  destroy
  
    
    
  
  
    /usr/bin/qemu-system-x86_64
    
      
      
      
      
      

and here is the XML for my Windows machine


  win10
  dda109fe-ba6d-45b1-833f-4b479895262f
  
    
      
    
  
  12800000
  12800000
  4
  
    hvm
    /usr/share/OVMF/OVMF_CODE.fd
    /var/lib/libvirt/qemu/nvram/win10_VARS.fd
    
  
  
    
    
    
      
      
      
      
    
    
      
    
    
  
  
    
  
  
    
    
    
    
  
  destroy
  restart
  destroy
  
    
    
  
  
    /usr/bin/qemu-system-x86_64
    
      
      
      
      
      

ssl – Nginx selective TLS pass-through reverse proxy based on SNI

I have a system of IoT devices behind a NAT so that they are not accessible via the public internet (although this is desired).
To overcome this, I included them in a VPN, with one member exposed to the public internet to act as a gateway.
The VPN has set up an internal domain and each member of the network has a subdomain based on a unique ID (let's go to the MAC address) as follows: 12a4f81ead4e.vpn.example.com

I want to create a reverse proxy on that Gatway for proxy requests under nginx.

It is planned to create a DNS entry for the gateway. *.gateway.comand route traffic (Ahem, Proxy) from / to 12a4f81ead4e.gateway.com to 12a4f81ead4e.vpn.example.com, And so the end user would only have to type 12a4f81ead4e.gateway.com in your browser to access your device.
I would like to use nginx because the gateway is already running nginx for other purposes.

I expect HTTP requests to be simple and can be handled with a carefully crafted nginx proxy_pass Directive.

But what about HTTPS requests? As far as I know, the SNI-based TLS passthrough is now implemented by nginx, but all the examples I've seen so far create a static mapping for … mapping the incoming SNI to a destination in the upstream:

stream {
  map $ssl_preread_server_name $selected_upstream {
    example.org upstream_1;
    example.net upstream_2;
    example.com upstream_3;
    default upstream_4;
  }
  upstream upstream_1 { server 10.0.0.1:443; }
  upstream upstream_2 { server 10.0.0.2:443; }
  upstream upstream_3 { server 10.0.0.3:443; }
  upstream upstream_4 { server 10.0.0.4:443; }
  server {
    listen 10.0.0.5:443;
    proxy_pass $selected_upstream;
    ssl_preread on;
  }
}

The problem is that devices are dynamically added / removed from the VPN and I don't want to constantly rewrite the nginx configuration files. If reading the map from a file is possible, it is a step in the right direction, although I think that Nginx has to be reloaded every time something changes, which leads to problems with the permissions, which of course with the sudo rules could be bypassed, but not with the best solution.

I also only want to represent requests that are received *.gateway.comand usually send other https requests to the existing vhosts. If at all possible, I would like to avoid terminating the SSL connection. Not a really tough requirement, but would like to implement it if technically feasible. Only for the kicks.

Internally I can listen to an alternative port for the other vhosts. With HTTP I did a similar action when I wanted to set a "global" location and move all HTTP vhosts to port 81 and implement a catch all vhost for port 80 that served the "global" location and everything else Forwarded port 81. ­čÖé

So … what I would need something like this (obviously not working):

stream {
  map $ssl_preread_server_name $selected_upstream {
    (.*).gateway.com $1.vpn.example.com;
    default normal_serve;
  }

  upstream normal_serve { server 127.0.0.1:8443; }

  server {
    listen 0.0.0.0:443;
    proxy_pass $selected_upstream;
    ssl_preread on;
  }

  server {
    listen 127.0.0.1:8443;
    server_name other.website.com;

    (...)
  }
}

iptables – Transparent proxy with DHCP pass-through?

I am trying to replicate a behavior on a small Linux based router that I have observed elsewhere, e.g. client device and mostly just act as a bridge, except that the router has to use this DHCP address to go home for that Administration to make calls. In addition, the client device should be able to communicate with the router using a fixed IP address (e.g. 192.168.100.1) in order to correct local errors. Any external traffic that is not a response to a request from the router itself should be routed directly to the client device, except for a reserved management port.

The client device calls its own house for administration and specifies which IP address it has received so that it can be called back. otherwise I would just use NAT and be done.

I'm pretty sure I can do most of it with ebtables / tproxy, but I haven't figured out the right mantras yet. The WAN side has a single hardware-related reserved DHCP address that I cannot do anything about. My plan B is currently bridging (with MAC correction), but I would lose too much management functionality that way.

Can someone please give me a fully functional example that I can start with?

iphone – Set the default paper size for scanning with the pass-through camera

How can I set the default paper size when scanning with a through-camera with my iPhone?

I use the fantastic new Mojave feature to scan documents directly into the preview with my iPhone. My paper is a white A4 sheet. The iPhone automatically detects the edges against a darker background and I take a snapshot. However, when previewed, the image is slightly stretched and the dimensions (cmd + I) show that it has a letter size of 8.5 x 11.

How can I make sure that it is A4?

ms dos – Using QEMU to emulate DOS with passthrough SCSI access?

I would like to use QEMU to emulate an MS-DOS system running very specialized proprietary software required for communicating with a SCSI tape device.

I see that QEMU provides passthrough access to the host SCSI bus, and it looks like it could also emulate a SCSI adapter that the guest operating system can use. However, I have very little experience with setting up QEMU with SCSI, and the documentation is somewhat dense.

Here's my QEMU command line so far:

qemu-system-i386 -m 32 -drive-file = "dosdrive.img", format = raw 
-drive id = tape device, if = none, format = raw, readonly = off, file = / dev / sg2 
virtio-scsi-pci, id = scsi0 
-device scsi-generic, bus = scsi0.0, drive = tape device

… but my MS-DOS guest does not recognize the SCSI adapter. What am I doing wrong and is it possible at all?

[Quality made in Germany] KVM with CPU passthrough, all flash and lifetime discount

——————————-
Hi, we are the first root.

——————————-

All our nodes run on corporate hardware. These include Intel Data Center SSDs, hardware RAID controllers, dual Xeon processors, n + 1 network uplinks, and redundant power supplies!

NEW: Now you can privately network between all your VPS at no extra cost!

————————————————– —-
Boost your business: KVM Power Edition

————————————————– —-

Designed for all customers with high memory and CPU requirements. Now with 33% discount on the life.

KVM Power Edition Light – Configure now

CENTRAL PROCESSOR: 1 shared Intel Xeon Core
R.A.M: 1,024 MB ECC RAM
PLATE: 40 GB SSD
TRAFFIC: 5 TB at 1 Gbit / s (mirror)

PRICE: 3,00 EUR / month

Launch of the KVM Power Edition – Configure now

CENTRAL PROCESSOR: 2 shared Intel Xeon Core
R.A.M: 2.048 MB ECC RAM
PLATE: 80 GB SSD
TRAFFIC: 5 TB at 1 Gbit / s (mirror)

PRICE: 4,00 EUR / month

KVM Power Edition Pro – Configure now

CENTRAL PROCESSOR: 3 shared Intel Xeon Core
R.A.M: 4,096 MB ECC RAM
PLATE: 140 GB SSD
TRAFFIC: 5 TB at 1 Gbit / s (mirror)

PRICE: 7,00 EUR / month

KVM Power Edition Premium – Configure now

CENTRAL PROCESSOR: 4 shared Intel Xeon Core
R.A.M: 8.192 MB ECC RAM
PLATE: 300 GB SSD
TRAFFIC: 5 TB at 1 Gbit / s (mirror)

PRICE: 13.00 EUR / month

KVM Power Edition Ultra – Configure now

CENTRAL PROCESSOR: 6 shared Intel Xeon Core
R.A.M: 16,384 MB ECC RAM
PLATE: 600 GB SSD
TRAFFIC: 10 TB at 1 Gbit / s (mirror)

PRICE: 22.00 EUR / month

KVM Power Edition Ultimate – Configure now

CENTRAL PROCESSOR: 8 shared Intel Xeon Core
R.A.M: 32,768 MB ECC RAM
PLATE: 1000 GB SSD
TRAFFIC: 10 TB at 1 Gbit / s (mirror)

PRICE: 38.00 EUR / month


Prices shown exclude VAT. And for customers living outside the European Union. If you live in the European Union, your local VAT will be applied.

—————
F-Com Terminal

—————

Since the company was founded in 2011, we have been developing and using our own customer portal. Our vision is to create a place where you can handle all your needs conveniently and easily. We love the benefit of being able to develop and implement new features ourselves instead of relying on third parties.

LEARN MORE

—————
HARDWARE

—————

Nothing is more annoying than unreliable services or downtime. As one of our main goals is to provide first-class services and avoid downtime and problems at all costs, all our systems are based on carefully tested components with the highest level of reliability.

LEARN MORE

———————
NO RESELLER

———————

To provide the best services and availability, we own and operate all our server, switching and routing devices. In case of problems our office is located 5 minutes from our location, so that we can guarantee the shortest reaction times.

LEARN MORE

——————
AVAILABILITY

——————

Knowing that our customers rely on our services, we guarantee 99.9% availability throughout the year.

LEARN MORE

————————————————– ——————
TESTIMONIALS / FEEDBACK FROM OTHER CUSTOMERS

————————————————– ——————

Nobody is perfect, but we can try it. Read what our customers write about us.

SEE MORE

————————————————–
CONTRACT DETAILS. SIMPLE AND FAIR.

————————————————–

In the configuration process, choose from 4 contract models. These contract models are currently available:

1st contract model: 1 month – 0% discount on the life
2. Contract model: 3 months – 2% term rebate
3. Contract model: 6 months – 4% term rebate
4. Contract model: 12 months – 10% discount on the life

You can also deactivate the subscription so that the contract is automatically terminated after the expiration of the contract term.

—————
NETWORK

—————

All VPS have a dedicated IPv4 address and a dedicated / 64 IPv6 network.
We offer free BGP sessions (IPv4 and IPv6) for all KVM Power Edition VPS.

—————
INTERNAL NETWORK

—————

Get two or more VPS and create private connections between them for free.

—————
14 DAYS MONEY BACK GUARANTEE

—————

First Root offers a 14-day money-back guarantee for new customers.

————————–
PAYMENT METHODS

————————–

Our options for your payment.

– Bank transfer
– SEPA direct debit (for EU customers)
– PayPal
– Credit card (VISA, Mastercard, American Express)
– Bitcoin

All payment methods are free, we do not charge credit card fees!

———–
IMPRINT

———–

First root UG (limited liability)
M├╝nsterstr. 330
40470 Dusseldorf
Germany

E-Mail: info@first-root.com
Web: www.first-root.com
Phone: +49 211/41 74 12 23 (Mon – Fri 09:00 – 18:00 CE (D) T)

Registry Court: District Court D├╝sseldorf
Registration number: HRB 66599

Sales tax identification number: DE815299021
Managing Director: Michael Hoffrath

kvm virtualization – The KVM VGA passthrough guest has a mouse delay when using Youtube

I have a very specific and strange problem with my Windows 10 KVM guest. It uses VGA passthrough and can easily play AAA games.

However, if you watch a video on Youtube and jump to a random time, which caches the video, the mouse has a strange delay and jumps around, the CPU usage increases to about 40-50%, and then returns to normal. The video will be played again.

I tried updating all the virtio drivers in the Guest, updating libvirt, enabling MSI interrupts in the registry, and none of them could solve this problem. This is the only noticeable issue in the year I use this setup.

The host runs Gentoo, but is also reproducible on Fedora. It runs Kernel 4.14.83 and Libvirt 5.0.0.