Re: [RFC] RDMA/umem: pin_user_pages*() can temporarily fail due to migration glitches

[John Hubbard]

On 5/1/24 5:10 AM, Jason Gunthorpe wrote:
> On Tue, Apr 30, 2024 at 10:10:43PM -0700, Christoph Hellwig wrote:
..
> > This doesn't make sense.  IFF a blind retry is all that is needed it
> > should be done in the core functionality.  I fear it's not that easy,
> > though.

So do I. :)

> +1
>
> This migration retry weirdness is a GUP issue, it needs to be solved
> in the mm not exposed to every pin_user_pages caller.
>
> If it turns out ZONE_MOVEABLE pages can't actually be reliably moved
> then it is pretty broken..

OK, I'll work on finding out what is temporarily elevating the refcount
and preventing the migration. And see where that leads.


thanks,
--
John Hubbard
NVIDIA