[dpdk-dev] VFIO no-iommu

[Jan Viktorin]

Hello,

I am not involved in the vfio very much, however, I was watching some
vfio-related code in last few weeks. It looks promising to me and
IMHO it seems to the best way to bring a support of integrated Ethernet
MACs into DPDK (related to many SoCs). Unfortunately, the ARMv7 SoCs (I
know) lacks of an IOMMU... The only protection there is the TrustZone
technology but I have no idea of its support in the kernel. It's also
far from being a replacement of an IOMMU. When using FPGAs, it is
possible to put an IOMMU engine there (I've got such a prototype
somewhere in my VHDL library) but nobody will probably do use because
of saving on-chip resources.

The X-Gene SoC (ARM 64) contains 2x 10 Gbps EMACs on the chip. I have no
idea about IOMMUs there. Thus, this platform can probably benefit of
such driver as well. The question is whether there is some interest to
have this kind of support in DPDK.

Thus, I'd like to have the vfio/no-iommu to support the ARMv7 (otherwise
it would be effectively dead in DPDK). Unfortunately, it's not my
primary job at the moment.

Regards
Jan

Note: as far as I know, it is discouraged to refer to lkml.org as
it is often very slow - my case today :).

On Fri, 11 Dec 2015 17:28:43 +0100
Thomas Monjalon <thomas.monjalon at 6wind.com> wrote:

> Recently there were some discussions to have an upstream replacement
> for our igb_uio module.
> Several solutions were discussed (new uio driver, uio_pci_generic, vfio):
> 	https://lkml.org/lkml/2015/10/16/700
> 
> Alex Williamson (maintainer of VFIO driver), submitted a solution
> and was waiting some feedback. Unfortunately, nobody caught it and
> he has reverted his work:
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=ae5515d
> 
> It is an important challenge to remove our out-of-tree modules and
> especially igb_uio. It is a long way to have a standard solution integrated
> in every distributions.
> The current cooking Linux kernel is 4.4 and will have a long term maintenance:
> 	https://kernel.org/releases.html
> So it is a pity to miss this opportunity.
> 
> Stephen has fixed a bug to use the IOMMU group zero:
> 	http://dpdk.org/browse/dpdk/commit/?id=22215f141b1
> 
> Is there someone interested to work on VFIO no-iommu and provide
> some feedbacks?
> We also need to prepare a documentation patch to explain its usage
> compared to the standard VFIO mode.
> 
> Thanks