Rebase to HEAD.

Don't set interrupt description for all Mellanox drivers as part of request_irq() as per review request.

Fix as per code review comment, use name only if its not NULL.

We intentionally don't use the "name" passed to request_irq() because for the mlx5en driver it will overflow the very-short description field !

I think intr_event_describe_handler() handles the long name by not using it. I was bitten by it and then I had to shrink device name to get around it by using fewer characters.

Updating D21828: Linux KPI request_irq() doesn't set interrupt description.

Was submitted by r297806

Added CTASSERT check for new IVHD definitions.

Yes, I tested Win10 on my Ryzen box with 12 vcpus. With the patch I could install in < 5 mins.

Nice, could install Windows 10 with multiple vcpus.

Thanks a lot.

Change https://reviews.freebsd.org/D10049 include this and is in HEAD/12.x.

v2 extended feature is in "MMIO Offset 0030h IOMMU Extended Feature Register".

I think I have taken care of all comments in https://reviews.freebsd.org/D10049

Can we close this one since we have https://reviews.freebsd.org/D10049?

Thanks a lot for testing, review and code review. My new ssh key is in place, will get back to work.

Makefile is not part of review, please ignore.

Fix many of style(9) and removed multiple references to AMD docs.

Ignore the write to AMD microcode update MSR by guest.

Move a.anish entity after a.anholt