https://lug.mtu.edu/w/index.php?action=history&feed=atom&title=Minutes_2026-10-01 2026-10-06T01:21:54Z Revision history for this page on the wiki MediaWiki 1.39.17 https://lug.mtu.edu/w/index.php?title=Minutes_2026-10-01&diff=8158&oldid=prev 2026-10-01T23:02:46Z

<p>Created page with &quot;=Virtual Function I/O= Presented by Simone ==Virtualization Primer== *Hypervisors let you run multiple OSes on one machine **KVM and QEMU on Linux **Xen, VMWare, HyperV on other platforms *Virtual Machines **Guest operating system **Almost native performance because CPU instructions are run directly **Other hardware is emulated *VMs don’t have direct hardware access, which is a problem **QEMU must emulate GPU, network, etc **Paravirtualization helps with some of these...&quot;</p> <p><b>New page</b></p><div>=Virtual Function I/O=<br /> Presented by Simone<br /> ==Virtualization Primer==<br /> *Hypervisors let you run multiple OSes on one machine<br /> **KVM and QEMU on Linux<br /> **Xen, VMWare, HyperV on other platforms<br /> *Virtual Machines<br /> **Guest operating system<br /> **Almost native performance because CPU instructions are run directly<br /> **Other hardware is emulated<br /> *VMs don’t have direct hardware access, which is a problem<br /> **QEMU must emulate GPU, network, etc<br /> **Paravirtualization helps with some of these<br /> ***Instead of emulating, a guest driver passes some things through<br /> ***Usually needs guest OS support<br /> ***Still plenty of limites<br /> **Dual booting is annoying<br /> ==What is VFIO?==<br /> *Gives a VM control over PCIE devices<br /> *Uses IOMMU (most devices in the past 10 years have this)<br /> *Remaps memory addresses and isolates devices<br /> **Prevents device as PCIe directly to guest<br /> **Low overhead<br /> **Little host interference ### Capabilities Generally allows faster access and removes overhead and quirks from the host OS and passthrough<br /> *GPU Access<br /> *Storage/NVMe<br /> *Network devices (good for high throughput, bypasses host network stack)<br /> *USB and audio<br /> ===Full Passthrough===<br /> *Assigns an entire PCIe device to a guest<br /> **Binds the device to vfio-pci on host side<br /> **Unbinds from normal host driver<br /> *Host can’t use device anymore<br /> *Works for almost every device<br /> ===SR-IOV===<br /> *Single-root I/O Virtualization<br /> *Shares PCIe deives between multiple guests<br /> *Host retains device access<br /> *More efficient sharing of resources<br /> *Used heavily in cloud computing<br /> *Virtual Function:<br /> **Single PCIe devices splits into multiple VFs<br /> **Can be assigned to guest<br /> *Only some hardware is capable<br /> **Mostly enterprise grade devices<br /> ==Demo==<br /> *Can write to a device file to add VFs<br /> *Adds more PCI devices in &lt;code&gt;lspci&lt;/code&gt; output<br /> *Intel GPU drivers are NOT READY for this, was not able to load a driver despite seeing the VF<br /> ==Use Cases==<br /> *Can run Windows games in a VM with good performance<br /> *Virtual workstations<br /> *Development/security<br /> **Driver testing<br /> **Malware analysis<br /> *Provisioning GPU resources to customer VMs on a server (cloud computing)</div>

Freya