[Qemu-devel] [RFC 0/9] Virtio-mmio refactoring.

[Evgeny Voevodin]

This patchset is derived from patchset provided by Peter Maydell 
<address@hidden>:
http://lists.gnu.org/archive/html/qemu-devel/2011-11/msg01870.html
Also this patchset includes some fixes for bugs spotted by Ying-Shiuan Pan 
<address@hidden>.
Still isue pointed by Peter presents here I think:
 * extra vring alignment field not saved/restored (because virtio
   layer isn't using VMState and doesn't allow the virtio base layer
   to specify a version for its data so back-compat would be tricky)
I've tested blk and net parts. Work good to me.

In this patchset refactoring of virtio-mmio layer is made.
Instead of creating virtio-blk-mmio, virtio-net-mmio, etc on the system bus
we create virtio-blk, virtio-net, etc devices on the virtio-transport bus.
To create virtio-transport bus virtio-mmio-transport device provided. 
Transport device plugs into virtio-mmio bus.
To create virtio-mmio bus virtio-mmio-bridge device provided. 
Bridge device is borrowed from s390 platform and plugs into system bus.
So, we have:

system bus----> |MMIO Bridge|----virtio-mmio bus---->----> |MMIO 
Transport|----virtio-transport-bus(name:virtio-mmio.0)----> |Virtio Backend 
(blk, net, etc)|
                                                     |
                                                     |---> |MMIO 
Transport|----virtio-transport-bus(name:virtio-mmio.1)----> |Virtio Backend 
(blk, net, etc)|

The interface to platforms consists of:
 - create a virtio bus (at this time virtio-mmio only)
 - create a transport device (at this time virtio-mmio only)

Then from command line user can plug back-end (virtio-blk, net, etc) into 
desired bus (at this time virtio-mmio.x only).

The next step of refactoring will touch virtio-pci layer.
There PCI Transport device will be created which provides the same transport 
bus as MMIO transport device.
I guess that it will look like:

pci bus----> |PCI Bridge|----virtio-pci bus---->----> |PCI 
Transport|----virtio-transport-bus(name:virtio-pci.0)----> |Virtio Backend 
(blk, net, etc)|
                                                |
                                                |---> |PCI 
Transport|----virtio-transport-bus(name:virtio-pci.1)----> |Virtio Backend 
(blk, net, etc)|

Also, there should be some back compatibility to keep pci backends able to be 
plugged into "virtio" bus as it done now 
(if I'm not wrong) and to not break present command lines.

Evgeny Voevodin (6):
  Virtio: Add transport bindings.
  hw/virtio-serial-bus.c: Add virtio-serial device.
  hw/virtio-balloon.c: Add virtio-balloon device.
  hw/virtio-net.c: Add virtio-net device.
  hw/virtio-blk.c: Add virtio-blk device.
  hw/exynos4210.c: Create two virtio-mmio transport instances.

Peter Maydell (3):
  virtio: Add support for guest setting of queue size
  virtio: Support transports which can specify the vring alignment
  Add MMIO based virtio transport

 Makefile.objs          |    2 +
 hw/exynos4210.c        |   12 ++
 hw/virtio-balloon.c    |   34 ++++
 hw/virtio-balloon.h    |    7 +
 hw/virtio-blk.c        |   42 +++++
 hw/virtio-blk.h        |   11 +
 hw/virtio-mmio.c       |  480 ++++++++++++++++++++++++++++++++++++++++++++++++
 hw/virtio-mmio.h       |   31 +++
 hw/virtio-net.c        |   46 +++++
 hw/virtio-net.h        |   11 +
 hw/virtio-serial-bus.c |   42 +++++
 hw/virtio-serial.h     |    9 +
 hw/virtio-transport.c  |   46 +++++
 hw/virtio-transport.h  |   39 ++++
 hw/virtio.c            |   20 ++-
 hw/virtio.h            |    2 +
 16 files changed, 832 insertions(+), 2 deletions(-)
 create mode 100644 hw/virtio-mmio.c
 create mode 100644 hw/virtio-mmio.h
 create mode 100644 hw/virtio-transport.c
 create mode 100644 hw/virtio-transport.h

-- 
1.7.5.4