On Wed, 3 Oct 2018 10:21:33 -0700 Kash Pande <k...@tripleback.net> wrote:
> On 2018-10-03 9:42 a.m., Alex Williamson wrote: > > In general, AMD folks tend to deny that there > > are reset issues, so except for the Bonaire/Hawaii quirks above where > > they admit there's a silicon issue, we haven't gotten much support for > > additional quirks. > > > Does the PCIe reset bug on Threadripper apply here? Because I know the > symptoms of that problem are different, for different users/systems. My > TR4 wouldn't start a VFIO VM without the PCIe bus reset patch being > applied to the kernel. Some people can't reboot a VM, but it'll start > just fine. > > > Also, no idea if that patch has made it upstream or not, I haven't dealt > with it in a few months. AFAIK, the Threadripper bus reset issue is fixed in AGESA firmware. There is no upstream software fix for this, nor does it seem that one is necessary given the firmware fix. That issued affected any device requiring a bus reset (ie. FLR or PM reset is not available on the endpoint), not just GPUs, though GPUs are a predominant class of devices falling into this category. Thanks, Alex _______________________________________________ vfio-users mailing list vfio-users@redhat.com https://www.redhat.com/mailman/listinfo/vfio-users
