maschine = q35 and hostpci passthrough pcie

[two56bytes]

when creating a new qemu vm the "machine" parameter is not noticed which breakes hostpci passthrough for pcie devices:

machine = "q35"
bios = "ovmf"

hostpci {
device = "hostpci0"
id = "0000:03:00"
xvga = true
pcie = true
}

produces:

proxmox_virtual_environment_vm.gaming: Creating...
╷
│ Error: error waiting for VM start: task "UPID:server:0000550D:000780EC:67C30E2D:qmstart:106:root@pam:" failed to complete with exit code: q35 machine model is not enabled at /usr/share/perl5/PVE/QemuServer/PCI.pm line 491.
│
│ with proxmox_virtual_environment_vm.gaming,
│ on qemu-gaming.tf line 1, in resource "proxmox_virtual_environment_vm" "gaming":
│ 1: resource "proxmox_virtual_environment_vm" "gaming" {
│

[bpg]

Hey @two56bytes! 👋🏼

Are you creating a new VM from scratch, or cloning from a template?

Could you fill out the bug template to provide more details? That would help with the investigation.

[two56bytes]

i am cloning from an ubuntu-24-lts cloud image template:

clone {
node_name = "server"
datastore_id = "vg0"
vm_id = 9501
}

[two56bytes]

And here the full bug report:

when creating a new qemu vm the "machine" via cloning a cloud-intit (ubuntu24-lts-amd64) template the machine parameter is not noticed which breakes hostpci passthrough for pcie devices.

changing the machine type from i440 to q35 in the template i used for cloning seems to solve the problem:

resource "proxmox_virtual_environment_vm" "gaming" {
  name      = "gaming.homelab.internal"
  node_name = "server"
  tags      = ["terraform", "gaming", "ubuntu-24"]
  on_boot   = false

  machine = "q35"
  bios    = "ovmf"

  initialization {
    datastore_id = "vg0"

    ip_config {
      ipv4 {
        address = "192.168.1.20/24"
        gateway = "192.168.1.1"
      }
    }

    dns {
      servers = [
        "192.168.1.16",
      ]
    }

    user_account {
      keys = [
        XXXXXXXXX-MYKEYS-XXXXXXXXX
      ]
      username = "user"
    }
  }

  agent {
    enabled = true
  }

  operating_system {
    type = "l26"
  }

  cpu {
    cores = 4
    type  = "host"
  }

  memory {
    dedicated = 16384
  }

  vga {
    memory = 128
    type   = "virtio"
  }

  clone {
    node_name    = "server"
    datastore_id = "vg0"
    vm_id        = 9501
  }

  hostpci {
    device = "hostpci0"
    id     = "0000:03:00"
    xvga   = true
    pcie   = true
    rombar = true
  }

  efi_disk {
    file_format  = "raw"
    datastore_id = "vg0"
    type         = "4m"
  }

  scsi_hardware  = "virtio-scsi-pci"
  disk {
    file_format  = "raw"
    ssd          = true
    discard      = "on"
    cache        = "writeback"
    datastore_id = "vg0"
    interface    = "scsi0"
    size         = 100
  }

  network_device {
    bridge = "vmbr0"
    model  = "virtio"
  }

}

Expected behavior:

even when cloning a template it should set the machine type accordingly (like the rest of the hardware can be changed when cloning from a template)

Details:

Single or clustered Proxmox: single
Proxmox version:  8.3.4
Provider version (ideally it should be the latest version): 0.73
Terraform/OpenTofu version: 1.9.0 / gitlab component 0.55.0
OS (where you run Terraform/OpenTofu from): debian bookworm