Hi dann, >OK, then I think there may be multiple conflated issues here. Let's >focus on the original use case you described - a VM created with >virt-manager using a SCSI controller doesn't work. You tried
Nonono, not quite. The original use case: an *EFI* VM with a SCSI controller does not work. This is what I reported. >I just tried this on latest sid, and was able to reproduce. "lsilogic" >does indeed not work. Oh, interesting. That’s got to be a new/different bug. We were on bullseye, not sid, in case that helps. >I also didn't find a way to tell virt-manager to use anything other >that "lsilogic". But, when I edited the XML and changed "lsilogic" to >"virtio-scsi" (and ran virsh define), the system booted fine. Yeah well virt-manager… it does have an XML editor built in, but sometimes… I did just test a BIOS case: I took an existing VM I had but was not using (an OpenBSD VM), dropped the IDE disc, added an lsilogic SCSI controller, added a SCSI disc with the same backing LV that the IDE disc had, booted it, and it works. Also on bullseye. >Thorsten - do you have reason that you prefer lsilogic to virtio-scsi? Yes: I mostly run operating systems with no or insufficient support for virtio over the hypervisor interface. (There’s also virtio over PCI, but my inquiries to the qemu developers how to even access this led to them eventually agreeing it probably isn’t even implemented fully yet.) In the specific case, it was a VM “appliance” imported from some other virtualisation tools that had a preconfigured Windows, and the other VM hosts all use lsilogic for that. >Now, Vincent me too'd this bug to say that virtio-scsi wasn't working >for them, but the version in bullseye did work. Thorsten reported >using the version of ovmf that *is* in bullseye and wasn't using >virtio-scsi. So whatever Vincent is/was seeing seems like a separate >issue. If you are still having a problem Vincent, please report a >separate issue. I’m not too sure about this either. I also had a grml-efi VM lying around, which incidentally already had a virtio-scsi configured, so I did the same thing: drop the SATA CD, re-add it as an SCSI HDD, change the boot order, start. It switches from “the guest has not initialised the display yet” to “viewer was disconnected” very quickly. (I also did a test with the NIC in the boot order enabled, and it does netboot, so the problem is with, again, SCSI.) So I can state, with reasonable confidence, that EFI booting in bullseye works with neither lsilogic nor virtio-scsi. This makes it mostly unsuitable for running most Windows VMs. bye, //mirabilos -- 21:12⎜<Vutral> sogar bei opensolaris haben die von der community so ziemlich jeden mist eingebaut │ man sollte unices nich so machen das desktopuser zuviel intresse kriegen │ das macht die code base kaputt 21:13⎜<Vutral:#MirBSD> linux war früher auch mal besser :D