mm/hmm: fault non-owner device private entries
authorRalph Campbell <rcampbell@nvidia.com>
Mon, 25 Jul 2022 18:36:14 +0000 (11:36 -0700)
committerAndrew Morton <akpm@linux-foundation.org>
Fri, 29 Jul 2022 18:33:37 +0000 (11:33 -0700)
commit8a295dbbaf7292c582a40ce469c326f472d51f66
tree07cb545d852fb7c56a2bb9796090b127b374126b
parent9282012fc0aa248b77a69f5eb802b67c5a16bb13
mm/hmm: fault non-owner device private entries

If hmm_range_fault() is called with the HMM_PFN_REQ_FAULT flag and a
device private PTE is found, the hmm_range::dev_private_owner page is used
to determine if the device private page should not be faulted in.
However, if the device private page is not owned by the caller,
hmm_range_fault() returns an error instead of calling migrate_to_ram() to
fault in the page.

For example, if a page is migrated to GPU private memory and a RDMA fault
capable NIC tries to read the migrated page, without this patch it will
get an error.  With this patch, the page will be migrated back to system
memory and the NIC will be able to read the data.

Link: https://lkml.kernel.org/r/20220727000837.4128709-2-rcampbell@nvidia.com
Link: https://lkml.kernel.org/r/20220725183615.4118795-2-rcampbell@nvidia.com
Fixes: 08ddddda667b ("mm/hmm: check the device private page owner in hmm_range_fault()")
Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
Reported-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Cc: Philip Yang <Philip.Yang@amd.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
mm/hmm.c