Re: [OPAE] IOMMU in Xeon+FPGA and the perception of memory by OPAE apps

--===============5851918757245931283==
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

The summary of the detail from Enno's message is: fpgaGetIOAddress() will a=
lways return the address that an AFU should use to access the memory, wheth=
er or not an IOMMU is in the path. An IOMMU may introduce a second translat=
ion, but it will be invisible to the user-space software and to the AFU.

It would be ideal if, when an IOMMU is present, OPAE constructed an AFU gue=
st physical address space that matches the virtual addresses in the softwar=
e's virtual space. Unfortunately, this is impossible with current kernel in=
terfaces to the IOMMU.

-Michael



--===============5851918757245931283==--