Re: blocking vs. non-blocking mmu notifiers

[Jason Gunthorpe <jgg@xxxxxxxx>]

On Wed, Mar 23, 2022 at 05:49:43PM +0100, Michal Hocko wrote:
> > The bug here is that prior to commit a81461b0546c ("xen/gntdev: update
> > to new mmu_notifier semantic") wired the mn_invl_range_start() which
> > takes a mutex to invalidate_page, which is defined to run in an atomic
> > context.
> 
> Yeah, we have already identified that but quickly realized that the
> whole mmu notifier overhaul which this fix depends on would be no no for
> backporting to our older code base. So we are trying to find our way
> around that.

IMHO you don't need everything, just commit 369ea8242c0f ("mm/rmap:
update to new mmu_notifier semantic v2") which adds the missing
start/end outside the lock for the page callbacks.

Then you can take safely a8146 into gntdev.

Jason