[ofa-general] Re: mapping kernel memory to userspace in <= 2.6.14

[Michael S. Tsirkin]

> Quoting Steve Wise <swise at opengridcomputing.com>:
> I figured out why I was hitting the BUG_ON() in page_remove_rmap().  Its
> because my library was destroying the QP or CQ object _before_ unmaping
> the objects.  I changed it to unmap first, and things work as expected.

I guess kernel side needs to be fixed then to prevent a buggy
userspace from crashing the kernel.

-- 
MST