On Sat, 29 Oct 2011 19:56:43 -0700, Ben Widawsky wrote:
> Sigh. I started down that path, but it was becoming tedious with only
> one case where we actually want to not retire (I think), so I thought
> I'd see how this went down on the mailing list.
I don't even want to think about locking for t
On Sat, 29 Oct 2011 19:51:26 -0700
Keith Packard wrote:
> On Sat, 29 Oct 2011 19:07:23 -0700, Ben Widawsky wrote:
>
> > + /**
> > +* Flag if GTT ptes shouldn't be modified.
> > +*
> > +* This is set when graphics virtual address space
> > +
On Sat, 29 Oct 2011 19:07:23 -0700, Ben Widawsky wrote:
> + /**
> + * Flag if GTT ptes shouldn't be modified.
> + *
> + * This is set when graphics virtual address space
> + * should not be changed. It's currently only useful for
> +
After the ILK vt-d workaround patches it became clear that we had
introduced a bug. Chris tracked down the issue to recursive calls to
unmap. This happens because we try to optimize waiting on requests by
calling retire requests after the wait, which may drop the last
reference on an object and en
