s390/mm: remove useless hugepage address alignment
authorGerald Schaefer <gerald.schaefer@linux.ibm.com>
Fri, 19 Aug 2022 17:40:11 +0000 (19:40 +0200)
committerVasily Gorbik <gor@linux.ibm.com>
Tue, 30 Aug 2022 19:57:07 +0000 (21:57 +0200)
commitff03b8846705e517f878585cf66c5d8fca1c38b2
tree65a96227a9421350aa62c152956f17fd3aae34ab
parentb90cb1053190353cc30f0fef0ef1f378ccc063c5
s390/mm: remove useless hugepage address alignment

The failing address alignment to HPAGE_MASK in do_exception(), for
hugetlb faults, was useless from the beginning. With 2 GB hugepage
support it became wrong, but w/o further negative impact. Now it
could have negative performance impact because it breaks the cacheline
optimization for process_huge_page().

Therefore, remove it.

Note that we still have failing address alignment by HW to PAGE_SIZE,
for all page faults, not just hugetlb faults. So this patch will not
fix UFFD_FEATURE_EXACT_ADDRESS for userfaultfd handling. It will just
move the failing address for hugetlb faults a bit closer to the real
address, at 4K page granularity, similar to normal page faults.

Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
arch/s390/mm/fault.c