ghost82

Members
  • Content Count

    997
  • Joined

  • Last visited

  • Days Won

    3

ghost82 last won the day on July 15

ghost82 had the most liked content!

Community Reputation

183 Very Good

6 Followers

About ghost82

  • Rank
    Member

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

  1. All is written in the message you copied You are trying to passthrough audio at 00:1f.3, which is not the only device in your iommu group (10). When you passthrough a device, in general you must passthrough all the iommu group, sometimes it's possible, sometimes not. As I wrote:
  2. Audio is part of a multifunction device 00:1f.x; if you already used the acs override (downstream or downstream,multifunction ('both')) and unsafe_interrupts=1 arguments and it didn't split, it's not possible.
  3. Well, in theory nothing changes..I'm using the emulated ethernet to connect to my host with smb, so the transfer speed is limited by the writing performance of my sata ssd. However iperf3 tests are much better with virtio. Here were my test results: This is without any doubt a valid solution, together with device properties inside config.plist, but I prefer also to have the most correct topology in the xml; again this is not a must, it's more my "obsession" for keeping things in their more correct places. Then it seems it's my issue, it's very strange that i
  4. Hi, after reading the readme.md here are some suggestions: 1. Not sure I understood well, it's true that 3.0 qemu XHCI is not supported by mac os, but ich9-ehci and ich9-uhci emulated usb 2/1 controllers are perfectly fine. 2. this is very minor, but virtio instead of virtio-net, though it's older, performs better, at least in iperf3 tests. 3. Referring to the above image, I think there's something wrong: multifunction='on' should not be inside <source></source> but inside <address> Also, function 0x01 should be i
  5. There's the vendor-reset patch: https://github.com/gnif/vendor-reset/ It's a patch for the kernel, there's also a kernel helper/builder for unraid to build a kernel with the included patch: However note that this may not work for mac os guests, depending on the gpu.
  6. It's your choice. I have ShowPicker=false too, if you need the picker to show you can just press (let it pressed) esc or 0 "zero" when the vm boots and the picker will show (0 (zero) was added later because if you press esc too early it will show the ovmf bios settings, so better to press 0 (zero)). Otherwise you can let ShowPicker=true and opencanopy will always show, and the default boot entry will automatically boot after xx seconds, defined in Timeout. In both ways, if something goes wrong, you will always able to show the picker.
  7. happy you solved it. To hide opencanopy (the bootloader gui) and automatically boot mount the efi in mac os, go to EFI/OC open config.plist with a text editor, search for ShowPicker and set it to false.
  8. Apply my pull request as described here: https://github.com/SpaceinvaderOne/Macinabox/pull/49#issuecomment-873745079 You may need also this:
  9. It doesn't seem you are binding to vfio the gpu, in your vfio-pci.cfg there's only: BIND=0000:03:00.0|1b21:2142 0000:04:00.0|1106:3483 which is something related to usb controllers. In unraid you need to bind to vfio the gpu and the audio part, from your diagnostic the iommu group should include also a pci bridge which should not be an issue to passthrough with the gpu (video+audio). If the bridge is an issue, use pcie acs override downstream,multifunction ("both") to split more your iommu groups. You can do it from the unraid gui. Once you bind the gpu this will not be a
  10. I don't know...I'm sorry I can't be of any help..If I were you I would try to change the pcie slot as a last chance..
  11. Maybe...the unrecoverable error doesn't sound good. The hdmi sound must not have multifunction, multifunction is applied to video and function is changed to audio as it is in the snipped I attached. Do you have only one video card? Does unraid uses this gpu and then that gpu tries to switch to the vm? Can you attach a diagnostic unraid file after you start the vm, so that it includes in the log that unrecoverable error? Does the vm work with vnc (no gpu passthrough)? Can you also try with no rom file at all?
  12. All I can see from the data you attached is wrong topology: You have video on bus 3 and audio on bus 4. Replace with this and make sure to dump your own rom and maybe the unrecoverable errors will go away (or may be related to amd reset bug, but starting from a fresh unraid boot you should see at least some video output on first vm startup) : <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x0e' slot='0x00' function='0x0'/> </source> <rom
  13. Share you diagnostic file: Tools tab, click on the Diagnostics icon, then click on the Download button, save the file and attach it here; so we can see full xml of your vm, your iommu groups and hardware. You can start by dumping your own rom (gpu-z in windows if you have it installed, or unraid), again, do not download it from internet.