swallace

I have a docker compose container configured with Compose Manger to run on a custom docker net with its own IP address, separate from that of my Unraid host. I have the web UI set to the correct IP/port etc, but it appears the Compose Manager prepends the Unraid Host IP to whatever is entered into this field. For example, here is my stack UI Label config: And here is the resulting URL I'm taken too when I click on the WebUI in the GUI: Is there any way to get around this? EDIT: Looks like I just need to add http:// before the IP. Duh

I'm having the same issue. Windows 11 VM, running WSL2 and Docker Desktop. CPU cores go to max, VM locks up. Have to force stop the VM to get it to work again.

I'm trying to utilize this container in Host Networking mode, such that I can assign it a to a tagged vlan interface (and have all my VPN clients end up on a particular subnet on my network). The container functions fine in Host mode, and clients can connect to the Wireguard server, but no traffic is able to be passed. Not quite sure why. Can anyone offer any advice? Thanks!

Thats what I thought. Yeah, Unturned is free, I just tried this in the course of troubleshooting. Thought it might be why the server would not start. Good to know! Appreciate your looking into it. Also glad I wasnt crazy when trying to setup. I'll update the container and post back

The unturned server does not seem to be working. I've set it up with a steam username and password on an account with the game purchased, and steam guard disabled. The server seems to loop itself though, never actually starting up. I've looked through the logs, but can't seem to identify anything specific that's going wrong. Maybe I'm missing a necessary game parameter? I kept everything other than the steam creds the default as they're provided by CA. Appreciate any input. Thanks!

Anyone know of a place where I can buy this? Been watching for months and it's out of stock everywhere, even before all the COVID19 stuff. Why would Rosewill just stop selling it so soon after release? Moreover, tons of their server cases are out of stock. Weird.

@meep Thanks for chiming in, I'm afraid you're right. I'll wait and see, but I'm afraid your're correct. Guess we'll see. I'll update when it comes in. Also, great blog post. I think I've just learned that I'm going to end up paying way too much money for a rack-mount PCI Expansion Chassis. All for some USB ports, lol.

@agarkauskas The Sonnet one you mentioned is what I purchased, the USB3-Pro-4P10-E. I am unable to get it functioning in Unraid. Also, even if I could, it only has 2 controllers onboard, whereas the old one had 4 (meaning you could pass through each individual port to a different VM, which, with Hubs, means you have lots of hot-pluggable USB devices to 4 VMs at once). The old one was pretty cool. It seems, however, that they may have just started manufacturing it again? I had ordered one from a thirdparty seller on Amazon for a bit over $100, logged on today to check the status of the order (shipping was going to take forever because of the pandemic) and, lo and behold, Sonnet Technologies on Amazon was selling them for $49.99. After a quick wtf on my part, I went to Sonnet's website, and, sure enough, they have it listed for sale. I swear, just 2 days ago I was looking for this part on their site and they had a listing saying it had been discontinued. I'm not really sure what is going on here, but it seems like they may have started reproducing this part. Maybe an updated version, I'm not sure, but far as I can tell the images look the same. They still have the purple PCB with 4 controller chips visible. I ordered 2 on Amazon, and I'm hoping for the best! I'll update in this thread when I find out if its for real or not.

So I just purchased the 3.1 card, the new one they still sell, thinking it was the card with 4 usb controllers onboard because I did not read this thread carefully enough. Bummer for me, only 2 controllers on the new card. I'm considering returning it and ordering the 4 controller version, found it for sale in a few places. Unfortunately, I'm not even able to get the new 3.1 show up as an "other PCI Device" for my VMs. It seems to be showing the Pericom PCI bridges in System Devices, but no USB controllers. I tried with the PCIe ACS Override setting enabled, but no dice. Anyone have any ideas about how to troubleshoot? Has anyone had any luck even using the new card? For anyone looking into this in the future, the card with 4 controllers has a purple PCB, and the model name is: USB3-PRO-4PM-E. Unfortunately, it is no longer in production. The new USB 3.1 card, with only 2 controllers, that I cant get working, has a black PCB and is model name: USB3-PRO-4P10-E.

More on my broken hardware acceleration saga: I tried to verify that hardware acceleration does work with this set of hardware if the VM is based in High Sierra, and not Mojave or Catalina. In short, I was not able to do so. I've described the behavior of mossman's system (from the passthrough post's discord) in earlier posts, but mine behaves differently in that as soon as GPU passthrough is configured for a High Sierra based VM, the OS will no longer load correctly. The GPU outputs a picture, clover loads, and boots into macOS, but the macOS bootup progress bar freezes at about 1/3 progress and the VM automatically pauses itself. It gets stuck here indefinitely, until I force stop the VM, and the behavior repeats if I start it up again. I'm confident it is not my configuration because my Catalina VM is still functioning fine and their configurations are effectively identical, just with separate config folders. With all that in mind, it seems like something about the way macOS High Sierra interacts with my particular GPU somehow halts the OS's bootup process. Bummer that mossman's experience was not repeatable for me. Below I've posted the errors I get in the VM log during the High Sierra OS loading process. 2020-03-09T03:39:04.813440Z qemu-system-x86_64: warning: host doesn't support requested feature: CPUID.01H:ECX.pcid [bit 17] 2020-03-09T03:39:04.814854Z qemu-system-x86_64: warning: host doesn't support requested feature: CPUID.01H:ECX.pcid [bit 17] 2020-03-09T03:39:04.815595Z qemu-system-x86_64: warning: host doesn't support requested feature: CPUID.01H:ECX.pcid [bit 17] 2020-03-09T03:39:04.816542Z qemu-system-x86_64: warning: host doesn't support requested feature: CPUID.01H:ECX.pcid [bit 17] 2020-03-09T03:39:04.817397Z qemu-system-x86_64: warning: host doesn't support requested feature: CPUID.01H:ECX.pcid [bit 17] 2020-03-09T03:39:04.818563Z qemu-system-x86_64: warning: host doesn't support requested feature: CPUID.01H:ECX.pcid [bit 17] 2020-03-09T03:39:06.945078Z qemu-system-x86_64: vfio_err_notifier_handler(0000:0e:00.1) Unrecoverable error detected. Please collect any data possible and then kill the guest 2020-03-09T03:39:06.945173Z qemu-system-x86_64: vfio_err_notifier_handler(0000:0e:00.0) Unrecoverable error detected. Please collect any data possible and then kill the guest 2020-03-09T03:39:33.088829Z qemu-system-x86_64: vfio_err_notifier_handler(0000:0e:00.0) Unrecoverable error detected. Please collect any data possible and then kill the guest 2020-03-09T03:39:33.095083Z qemu-system-x86_64: vfio_err_notifier_handler(0000:0e:00.1) Unrecoverable error detected. Please collect any data possible and then kill the guest I attempted to troubleshoot this process a bit, and stumbled upon some old posts from @Leoyzen regarding the second error. Specifically, this post: Unfortunately, I am not able able to install my RX580 into slot 1 (motherboard heat shield causes clearance issues with the card) and pci=noaer did not change anything for me. If anyone has any ideas about the root of the hardware acceleration issue based on everything up to this point, I'm all ears. Also would appreciate any input about getting High Sierra to load (possible I'm just being stupid and missing something) but I'm thinking the two issues may be related at this point as, again, the Catalina VM is configured similarly and the OS loads fine, just with no hardware acceleration.

@Leoyzen I tried every vbios listed as compatible for my card on techpowerup, and it nothing changed. Considering loading into Windows and extracting the vbios myself with GPUZ and see if that changes anything. Here's a list of all the vbios I used: https://www.techpowerup.com/vgabios/?did=1002-67DF-1DA2-E353-E7

@Leoyzen @david279 Just wanted to throw out there that I found someone else with this same issue on the Passthrough Post's discord channel. Mossman, whose messages I posted in screenshots eariler. He has experienced the same issue with the exact same GPU on a Ryzen 3700 Host, also with an ASRock motherboard. He also mentioned that he was able to successfully get pass though working in macOS High Sierra, but not in Mojave or Catalina. I'm going to setup a high sierra VM to verify that, and also still try setting up a VM from scratch using opencore, but the whole problem just seems to get weirder and weirder. He also reports that, if he starts up the High Sierra VM with functioning Hardware acceleration, then shuts it down, and starts up the Mojave or Catalina VM, that hardware acceleration will magically start working. But, if he stops that VM, and starts proceeds to start a Windows VM that utilizes all the same hardware for passthrough, it will crash the whole host. Weirder still. I dont even know where to begin debugging it anymore. If anyone has any input on how to pursue figuring this out, I'm all ears. Thanks! (Temporarily though, it seems like running a low resource high sierra VM, booting into that, shutting down, and then starting up the sierra VM could be a good middle ground solution. Smh)

@david279 Makes sense. I'm going to start one from scratch to try opencore with, but for testing in the meantime, it would be expected that if I just set the existing VM to use stock OVMF, it would stop working, right? If it was made with macinabox initially, then the VM needs that custom version? Or is it only by using apfs.efi and opencore that stock OVMF works, like Leoyzen said eariler?

@david279 so when you say setup the VM the regular way with standard OVMF, are you refering to this post from @Leoyzen? Or do you just mean setting up the VM manually, without using macinabox? Been considering doing that anyway so I can switch to opencore.

@david279 I have a quick question about setting up the VM the way you mentioned but OMG that switch. Totally forgot it existed. Any idea how to identify what it's currently set to? No indicator on the device itself and can't find anything in the manual. May just flip it and see if it works. Edit: Found it. My card is setup with the switch down, gaming mode, by default. Bummer (but also good, because if I had wasted this many hours just needing to flip a switch, I'd have been pretty pissed)

@david279 Unfortunately no dice. Same errors about AMD initialization even with those CPU flags you recommended in place. And The RX580 still shows without metal support. So weird.

@david279 Doing that now. Should the second line be indented as it is in the XML you posted? I assume so, but copy/paste does not capture that. Any idea if it should be tab, 2 spaces, or if it matters? I have whitespace nightmares from python and YAML. Edit: seems like not actually, if I try to save it with the the 2 spaces added, it undoes my indentation when I hit save. Trying now.

@david279 Was actually just reading through that as you replied, and looking into how to add the +fma flag. I'm trying what you advised now and I'll report back. Thanks!

@Leoyzen I think I've made some progress on this issue. I found someone on The Passthrough Post's discord channel who seems to have had a very similar issue with passing through their Pulse RX580 when using an AMD 3700 Host CPU (I'm using a 3950X). I've attached two images with his posts screenshotted for context, but the gist of it is he gets a very specific error about the Radeon driver in his system kernel log which I was also able to find in my install. He said he had some luck fixing it by adding the +fma Qemu cpu flag. Have yet to look into whether that could be helpful here (or try it for that matter) but thought it was worth passing the info along. Below are the error lines from my log that are the same as what he references. 2020-03-06 19:15:15.608472-0800 localhost kernel[0]: (AMDRadeonX4000HWLibs) <AMDRadeonX4000HWLibs`AmdCailServices::initializeAdapter()> AMD Error: 2020-03-06 19:15:15.608474-0800 localhost kernel[0]: (AMDRadeonX4000HWLibs) <AMDRadeonX4000HWLibs`AmdCailServices::initializeAdapter()> CAIL_ASICSetup() failed (CAILRESULT=1) 2020-03-06 19:15:15.608474-0800 localhost kernel[0]: (AMDRadeonX4000HWLibs) <AMDRadeonX4000HWLibs`AmdCailServices::initialize(_CailInitialize*)> AMD Error: 2020-03-06 19:15:15.608475-0800 localhost kernel[0]: (AMDRadeonX4000HWLibs) <AMDRadeonX4000HWLibs`AmdCailServices::initialize(_CailInitialize*)> initializeAdapter() failed!

@Leoyzen I've been thinking it was weird all along because the Pulse should work out of box. Glad you agree. EFI is attached. Thanks! EFI.zip

@Leoyzen I freshly downloaded Lilu 1.4.2 and WhateverGreen 1.3.7 from their releases pages on Github and installed to kext/others. This step on it's own did not fix anything, metal support is still not showing up at all. I updated the VM to use the specific vbios for my card. Specifically, the most recently compiled one listed here on techpowerup. This change also appears to have made no difference, metal support is still not showing up. Before completing the previous steps the Lilu Log in Hackintool was completely blank. I'm going to guess that means it was not loaded. After completing them, the Lilu Log still shows up blank. It would appear that it is not loading at all, or at least not functioning? Under System log, kernel process, and filtering for Lilu, I got the following line: 2020-03-05 09:53:41.25800-0800 localhost kernel[0]: Security policy loaded: Lilu Kernel Extention 1.4.2 (Lilu) So it does seem like it's at least loading at boot? But the Lilu log is still empty, so it seems like its not functioning. Unless I'm misunderstanding something, it seems very odd. Thanks for your help thus far- any more advice? Edit: Realized I was running an outdated Clover version. Updated from 5097 to 5104 and nothing changed. Probably not relevant, but just in case.

@testdasi Any ideas as to why 4.2 would not show up as an option for me? Did I miss an update somewhere? Edit: I definitely am. I'm on Unraid version 6.8.0 still, smh. Thanks!

I would also be very interested in this, if anyone has it working. I spent a good amount of time trying to get this working a bit ago and ran into similar issues.

@Sgtsmi1es Just wanted to also recommend the Q35 option. I had all kinds of issues getting rid of Error 43 before I switched it to Q35 4.1 (I believe the latest current Q35 version) and it worked immediately. Hopefully that fixes it for you!

What do you mean by restricted? Like Unraid will not even let you click on the options for more RAM in the first place, they're "grey-ed out?" Or do you mean a VM will not start after you configure it to use more than 1024 Mb and it gives you an execution error about memory allocation? If the latter, then you do not currently have enough free memory avalible on the system to provision to the VM in the first place. You may have other VM's or docker containers already using it that you'll need to shut down before this VM will start with the amount of memory you want for it. If it won't let you click on the other options in the first place, when you go to configure the VM, then I'm at a loss and hopefully someone else has some ideas.