[SLOF] [RFC PATCH slof 0/2] virtio: Enable IOMMU

Alexey Kardashevskiy aik at ozlabs.ru
Fri Nov 8 19:40:04 AEDT 2019


Virtio devices are PCI devices and in general should be using the same
DMA engine as any other PCI device; this helps with upcoming ultravisor
stuff as well. As part of the solution, we are going to enforce
iommu_platform on all virtio devices but once offered by QEMU,
this flag cannot be rejected and must be used by the driver.

This is an RFC for IOMMU support for virtio-net-pci for now.
1/2 is more or less mechanical; 2/2 adds the support.

Please comment. Thanks.



Alexey Kardashevskiy (2):
  virtio: Store queue descriptors in virtio_device
  RFC: virtio-net: Enable IOMMU

 lib/libvirtio/virtio-internal.h |  12 +-
 lib/libvirtio/virtio-net.h      |   2 -
 lib/libvirtio/virtio.h          |  29 ++---
 lib/libvirtio/virtio-9p.c       |  35 +++---
 lib/libvirtio/virtio-blk.c      |  52 ++++-----
 lib/libvirtio/virtio-net.c      |  57 +++++-----
 lib/libvirtio/virtio-scsi.c     |  67 ++++-------
 lib/libvirtio/virtio-serial.c   |  72 +++++-------
 lib/libvirtio/virtio.c          | 193 ++++++++++++++++++++------------
 board-qemu/slof/virtio-net.fs   |   1 +
 10 files changed, 263 insertions(+), 257 deletions(-)

-- 
2.17.1



More information about the SLOF mailing list