For kernels that don't have commit ee71d16d22 ("s390/mm: make TASK_SIZE
independent from the number of page table levels") criu sets TASK_SIZE
to 4 TB on s390 (see compel_task_size()).

When dumping tasks >= 4 TB on such systems, we would silently loose
memory >= 4TB.

So add a check and refuse to dump the task in that case.

When restoring tasks >= 4 TB on such systems, the remap() call for
moving the vmas at the end of the restor process would fail
and a very unclear error message is printed.

So add a check and write a more speaking error message in that case.
Reviewed-by: Alice Frosi <alice@linux.vnet.ibm.com>
Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>