On Tue, Nov 21, 2017 at 3:05 PM, Andrew Morton akpm@linux-foundation.org wrote:
On Tue, 14 Nov 2017 11:56:34 -0800 Dan Williams dan.j.williams@intel.com wrote:
Until there is a solution to the dma-to-dax vs truncate problem it is not safe to allow long standing memory registrations against filesytem-dax vmas. Device-dax vmas do not have this problem and are explicitly allowed.
This is temporary until a "memory registration with layout-lease" mechanism can be implemented for the affected sub-systems (RDMA and V4L2).
Sounds like that will be unpleasant. Do we really need it to be that complex? Can we get away with simply failing the get_user_pages() request? Or are there significant usecases for RDMA and V4L to play with DAX memory?
V4L plus DAX is indeed dubious, but RDMA to persistent memory is something the RDMA community is interested in supporting [1].
[1]: http://www.snia.org/sites/default/files/SDC15_presentations/persistant_mem/C...