1 |
On Sun, May 14, 2017 at 12:31 AM, Adam Carter <adamcarter3@×××××.com> wrote: |
2 |
> Tried kernels 4.10.13 and 4.11, with |
3 |
> CONFIG_GART_IOMMU=y |
4 |
> CONFIG_IOMMU_HELPER=y |
5 |
> CONFIG_IOMMU_API=y |
6 |
> CONFIG_IOMMU_SUPPORT=y |
7 |
> CONFIG_IOMMU_IOVA=y |
8 |
> CONFIG_AMD_IOMMU=y |
9 |
> CONFIG_AMD_IOMMU_V2=m |
10 |
> Chipset is 990FX, and AFAICT the V2 is for the APU (bdver3 and 4 vintage). |
11 |
> |
12 |
> When I enable the IOMMU in the BIOS I get stack traces. Is anyone using |
13 |
> AMD_IOMMU successfully? If so, any tips? |
14 |
|
15 |
I don't mean to suggest the problem is intractable, but on fairly new |
16 |
hardware (i7-4770K, MSI Z87-G45) there has been a regression in the |
17 |
kernel's nested KVM support that makes me view VT-x/AMD-V VT-d/AMD-Vi |
18 |
with a great deal of suspicion. In any case, devices are liable to |
19 |
just spontaneously not work, even if the hardware reports it has been |
20 |
properly initialized. This may require changes to the kernel and the |
21 |
information needed to make those changes may or may not be available. |