ghost82

Are you sure the names are correct? From what I found they should be: kvm-amd.avic= kvm-amd.npt= kvm-amd.nested= Check your syslog when booting with your arguments and if they are wrong you should see some logged messages about invalid parameters or something similar; if parameters are wrong they are simply ignored by the kernel.

The change is not persistent because unraid loads in ram, all the system files are in the compressed bzroot file and extracted into ram at each boot. Moreover, as far as I know it is not possible to load a custom rom (romfile) in libvirt for the <video></video> block. The only solution I see is to automate the copy/replace of that file when the array starts with a user script.

Most probably related to corrupted usb flash drive. Forced shutdown should be the only option; you should not damage any hd. If you can get access to the usb in another os copy the config folder; redo/change the usb flash drive, make it bootable and copy the config folder on it.

It should be possible if you add an emulated gpu as primary in the xml. vmvga should be preferred as model for compatibility (it's like 'vmware compatible'). This will result in something like this: <graphics type='vnc' port='-1' autoport='yes' websocket='-1' listen='0.0.0.0' keymap='en-us'> <listen type='address' address='0.0.0.0'/> </graphics> <video> <model type='vmvga' vram='9216' heads='1' primary='yes'/> <address type='pci' domain='0x0000' bus='0x00' slot='0x02' function='0x0'/> </video> vram can be tweaked if necessary. address 00:02.0 should be used for the emulated primary gpu. Note that this will set the primary gpu to the emulated one, so the os will use it by default, defeating the purpose of the discrete gpu passthrough..but this depends on the user case. I would try to avoid this and solve the issues with a vnc server/teamviewer or whatever you want installed inside the vm.

Yes and no the "issue" here is that the primary gpu is flagged as boot vga and when this happens the bios makes a copy of the gpu rom (however this happens or doesn't happen depending on the bios too!), and this copy can be freely modified in the memory; however if this happens, when it comes to vfio, you could have invalid pci rom errors because you are reading the shadow copy of the rom image that is no more a 1:1 copy. That's why passing a copy of the raw vbios fixes the issue. I would say it's not because vfio happens too late but because of the native way it works for the primary gpu (at the bios level).

Hi, I don't understand what you are trying to achieve..Ok to save time because you wont install games twice, but how are you thinking to save space if you are going to clone the disk?And why having 2 identical vms running? I can think about 2 solutions: one (preferred in my opinion) is to delete/copy the vdisk from the master vm, 2 is to sync the files. Note that if you have 2 vdisks for each vm, let's say one for the os and one for installed games (like a secondary vdisk), the game could install things on the os drive too, so you should clone both. Usually data files are saved in the users folder; if no data is present the game will start "clean". For steam, game data should be saved into the Program Files (x86)\Steam\userdata directory. However each game can save data in different locations, so you should need to check the saved location for each particular case.

Otherwise, bind the gpu to vfio at boot and the nvidia plugin should ignore that gpu (useful is you have other nvidia gpus that require the plugin).

You can't do it from the os itself if the disk is the only one (or even if you can do it, it could be very risky). You can boot a gparted live iso and mount the vdisk in there and you will be able to extend the partition. You can make a new vm booting from the gparted iso with the additional vdisk attached.

Hi, 6600 xt is supported from mac os monterey 12.1, when apple introduced the drivers. To make it work just passthrough the gpu (audio and video), edit the xml to set them as a multifunction device. Then, you need to add: agdpmod=pikera to your boot-args in opencore config.plist, otherwise black screen. As far as passing through the ssd you have several option, and general settings applies also to the mac os vm; if you do custom modifications other than the ones macinabox know, don't use the scripts otherwise they will mess the xml, you need to manual edit it. You can passthrough the nvme controller, or the sata controller to which the ssd is attached, or use a type=block disk device to pass the id of the disk (by-id) emulating a virtual controller.

I had a similar issue in my mac os vm in the past, with a 2 screens setup, the vm froze at boot if the second monitor was connected at boot (if I remember well...anyway I remember that the issue was a vm freeze with a monitor connected at boot (hdmi)). However I was using a custom compiled ovmf bios and I was able to identify the issue to the toolchain I was using to compile ovmf. To troubleshoot, you may try to replace the ovmf files you are using with latest ones, attached. With the vm stopped, rename the files to that you are using and replace them, or change in the xml the paths that point to ovmf code and vars. Use the tpm version if you are running windows 11. ovmf-tpm-stable-202205.zip ovmf-stable-202205.zip

as far as I know yes, better convert to raw image using 'qemu-img convert' command, then resize with 'qemu-img resize', then use gparted or the os itself to expand the partition.

Dumping the vbios for integrated gpu can be very tricky or not possible at all in some cases. In this case, it is suggested to download your motherboard bios from the manufacturer, or dump the mb bios (same version, same revision) and then extract the vbios from it. You could follow these steps: https://forums.unraid.net/topic/112649-amd-apu-ryzen-5700g-igpu-passthrough-on-692/?do=findComment&comment=1134762

This is ok, you will get some video output on boot, because vfio driver attaches after some other things. You should have last lines on the monitor outputting something like: vfio-pci 0000:xx:xx.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=io+mem:owns=io+mem where vfio takes control over the gpu. This depends on something else.

Did you format the disk to apfs before attempting at the installation (disk utility in recovery before install mac os)?

It is suggested to edit it manually in the xml view, so to be able to set the gpu correctly as multifunction. As far as the code, take as an example these blocks, referring to my 6900 xt: <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x06' slot='0x00' function='0x0'/> </source> <address type='pci' domain='0x0000' bus='0x03' slot='0x00' function='0x0' multifunction='on'/> </hostdev> <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x06' slot='0x00' function='0x1'/> </source> <address type='pci' domain='0x0000' bus='0x03' slot='0x00' function='0x1'/> </hostdev> This includes 2 hostdev blocks, one for the video and one for the digital audio. As far as buses/slots/functions: each hostdev block has 2 lines with "address"; the address line inside <source></source> is the address of the gpu in the host, the address unraid sees, in this case video is at source address 06:00.0 (video), audio is at 06:00.1 (audio). This address depends only the physical slot (it's the output of the lspci command). The address line outside <source></source> is the address of the gpu you set in the vm: so you can change this as you want; in this case I set it to be at address 03:00.0 (video) and 03:00.1 (audio). In the vm, for a q35 machine type you attach devices usually to pcie-root-port(s), this means that in this example to attach the gpu in the vm at address 03:00.0/1 you need a pcie-root-port defined with index 3 (same number as the target bus number): <controller type='pci' index='3' model='pcie-root-port'> <model name='pcie-root-port'/> <target chassis='3' port='0xc' hotplug='off'/> <address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x2'/> </controller>

It seems it's an issue with virtio nic drivers and localized windows isos, found a couple threads about german and hungarian windows. You could: 1. change machine version to 5.1 in the xml, or 2. install the english iso, the virtio driver should work, then install the language pack and set all the other localization settings

You need disk type='block' (not file) for the passed through disk (by-id).

As far as I know unraid doesn't include the qemu-system-ppc binary, so it shouldn't be possible. However, a workaround to emulate powerpc could be make a mac os monterey vm (with host-passthrough cpu) and install in it utm to run a powerpc based os; nested virtualization must be enabled and this should work with intel cpu in the host.

syslinux config, add to the append line: video=efifb:off

AMD cards are good, but unfortunately they suffer the so called amd gpu reset bug unless you have one of the latest 6000 series. There are workarounds (somewhat fixes in the kernel) that may work or not, or work partially.

Sounds like corrupted iso.

Hi, how many physical nics do you have in the system? Easiest and fastest way, if you have 2 nics (at least). Configure both nics in unraid for bridge (br0 and br1). Let's say you have eth0 and eth1: eth0 bridged to br0, eth1, bridged to br1. eth0 having internet access, br0 will have internet access too, so use br0 in the vm; configure eth0/br0 (eth0 in the host, br0 in the vm) with dhcp from router, or assign manually ips in the network 192.168.172.0/24. eth1 without internet access (no cable plugged in the adapter), br1 will not have internet access, use additional br1 in the vm; configure eth1/br1 (eth1 in the host, br1 in the vm) manually to have ips in the network 10.1.1.0/24. If you have only one nic (eth0): eth0 having internet access, br0 will have internet access too, so use br0 in the vm; configure eth0/br0 (eth0 in the host, br0 in the vm) with dhcp from router, or assign manually ips in the network 192.168.172.0/24. For the second nic I think you can create a virtual network (vnet)?you could use also virbr0 which has ips 192.168.122.0/24; for custom ip addresses you need to define the new network in a new xml and enable it. Or For the second local network (10.1.1.0/24) you may create a dummy nic in the host (dummy1) and bridge it (br1), and assign manually the ips: I never tried in unraid (I don't know if unraid has included the dummy kernel module), but in other generic linux oses it's feasible. Depending on your case I can try to see if it works in unraid too. For this second case, in a generic linux host, it works like this with systemd-networkd: in /etc/systemd/network/ file bridge1.netdev: [NetDev] Name=br1 Kind=bridge file bridge1.network: [Match] Name=br1 [Link] MACAddress=4e:c0:b1:12:13:a2 [Network] Address=10.1.1.1/24 [Route] Gateway=10.1.1.1 Metric=2048 file dummy1.netdev: [NetDev] Name=dummy1 Kind=dummy file dummy1.network: [Match] Name=dummy1 [Network] Bridge=br1 DHCP=No

It should be the second one, efifb is not in use if the os is booted in legacy mode.

Hi, you need to: 1. setup the gpu as multifunction in the vm <-- to be done 2. it should be isolated (bound to vfio) <-- done 3. allow unsafe interrupts may be required in unraid <-- done 4. newest drivers should be installed <-- cannot say anything on this 5. modification to syslinux config may be required (es: video=efifb:off) <-- (to be done) 6. q35+ovmf should be preferred <-- give it a try 7. video rom should be dumped from your gpu and not downloaded somewhere <-- cannot say anything on this Setup a q35+ovmf virtual machine with vnc, with all the advices above, enable remote desktop inside the vm, shutdown the vm, enable gpu passthrough, boot and connect directly to the vm with remote desktop with a second external device to install the drivers; look at the system devices for errors if it doesn't work.

It depends on the source where one downloads the virtio iso. Fedora source, which is that integrated into unraid, provides signed drivers, other sources may not have the drivers signed.