libnvdimm/region: Enable MAP_SYNC for volatile regions

Some environments want to use a host tmpfs/ramdisk to back guest pmem.
While the data is not persisted relative to the host it *is* persisted
relative to guest crashes / reboots. The guest is free to use dax and
MAP_SYNC to keep filesystem metadata consistent with dax accesses
without requiring guest fsync(). The guest can also observe that the
region is volatile and skip cache flushing as global visibility is
enough to "persist" data relative to the host staying alive over guest
reset events.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Reviewed-by: Pankaj Gupta <pagupta@redhat.com>
Link: https://lore.kernel.org/r/20190924114327.14700-1-aneesh.kumar@linux.ibm.com
[djbw: reword the changelog]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>

diff --git a/drivers/nvdimm/region_devs.c b/drivers/nvdimm/region_devs.c
index ab91890f2486e01630b4ebca5801792a421f9676..ef423ba1a71163ee742dfca9795c4f758ab4cff9 100644 (file)
--- a/drivers/nvdimm/region_devs.c
+++ b/drivers/nvdimm/region_devs.c
@@ -1168,6 +1168,9 @@ EXPORT_SYMBOL_GPL(nvdimm_has_cache);
 
 bool is_nvdimm_sync(struct nd_region *nd_region)
 {
+       if (is_nd_volatile(&nd_region->dev))
+               return true;
+
        return is_nd_pmem(&nd_region->dev) &&
                !test_bit(ND_REGION_ASYNC, &nd_region->flags);
 }