Roger Pau Monné

The talk will explain the reasons of the ongoing work to add PCI emulation support inside of Xen. This purpose of the PCI emulation layer is to allow passthrough of PCI devices from Xen to guests without involving a third entity in the system (so far this was done using QEMU). This should allow transparent PCI passthrough to guests without requiring a QEMU device model.

Xen is a hypervisor using a microkernel design that allows running multiple concurrent operating systems on the same hardware. One of the key features of Xen is that it is OS agnostic, meaning that any OS (with proper support) can be used as a host. Xen has a long history going back to the 90s when it was designed and the early 2000s when it was released. As a consequence of this, many of the assumptions and virtualization techniques backed into it are now superseeded by new hardware features, that make virtualization more transparent from an OS point of view.

This talk provides an overview on the different kind of guests supported by Xen and how these new hardware features are used in order to improve and evolve them. It also describes the design and implementation of a new guest type, called PVHv2, and how it can be used as a control domain (Dom0).

It's been more than a year since FreeBSD gained support for running as a Xen host, but there's still a lot of ongoing work on both FreeBSD and Xen in order to improve the features and the performance of FreeBSD as a guest and a host.

This talk will cover the latest improvements in FreeBSD related to Xen support, and the new features being developed for Xen itself.

The goals are to provide information about the latest improvements in both FreeBSD and Xen:

• FreeBSD improvements:

  • Bug fixes.
  • Removal of the classic Xen PV port.
  • Indirect descriptors support for blkfront (Colin Percival).
  • Unmapped IO support for blkfront.
  • Code cleanup in preparation for ARM64 support (Julien Grall).
  • Rework of the netfront driver (Wei Liu).

• Xen improvements:

  • Rework of the PVH implementation inside of Xen.
  • Grant table performance improvements.
  • xSplice (hypervisor live patching).
  • Virtual Performance Monitoring Unit.

During the last year (2014) a lot of interesting development happened around the FreeBSD/Xen port. FreeBSD slowly moved from running as a PV with HVM drivers, to running as a PVHVM guests and finally as a PVH DomU. The last step into getting full Xen support into FreeBSD was to get it to run as a PVH Dom0, which is now finished and in a tech-preview status.

This talk will include a high-level description of the modifications needed in order to get FreeBSD running as a Dom0, and a demo of FreeBSD running as a PVH Xen Dom0.

The goal of this talk is to provide information about the Xen PVH and FreeBSD architecture, and to encourage other OS hackers to work on it.

• Basic Xen description and specific Xen concepts.
• How the Xen community works (compared to BSD communities).
• A look into new Xen features (PVH).
• Work being done in FreeBSD improving Xen support.
• Demo of a FreeBSD PVH Dom0.

Xen has recently gained a new guest type called PVH and it can run as both DomU and Dom0. This talk will focus on the architecture of PVH and the interface exposed to guest OSes in order to run under this mode.

Also, examples will be provided about how we ported FreeBSD to run under this new virtualization mode.

The goal of this talk is to provide information about the PVH architecture, and to encourage other OS hackers to port their OSes to PVH.

• Description of the PVH architecture:

  • Why was PVH introduced?
  • Differences between PV, PVHVM and PVH.
  • Description of the current PVH interface.

• Practical example: changes in FreeBSD in order to run as PVH:

  • How does FreeBSD make use of the PVH interfaces.
  • The road so far: from HVM to PVH, a progressive implementation.
  • General tips about best ways to add PVH support to an existing OS.

Xen paravirtualized guests have been around for a long time, but during that period only one BSD OS was able to fully support them, that is NetBSD. With the arrival of hardware extensions to help virtualization, the Xen community has started to develop a new kind of paravirtualized guest that make use of hardware extensions to provide better performance.

This new type of paravirtualized guest, called PVH, makes use of hardware-assisted pagetable virtualization to get rid of one of the most difficult parts to implement when porting a new Operating System to Xen, the paravirtualized MMU. This, apart from the obvious benefit of having a performance boost, also benefits Operating System that don't have a virtualization solution, with the aid of hardware extensions, the required code changes to port a new OS to Xen are reduced.

This talk will focus on the features of the new paravirtualization mode, and how it can simplify the port of new Operating Systems to Xen.