[PATCH v8 10/15] s390/mm: Cleanup pXXp_flush_lazy() routines

[Alexander Gordeev]

Avoid unnecessary pgtable entry read in ptep_flush_lazy() and
pmdp_flush_lazy() routines. Use the already obtained value of
a pgtable entry instead of dereferencing it again.

It might appear that the optimized out second dereferencing happens
under mm_context_t::flush_count lock and therefore a race condition
is possible. However, both routines write invalid entry bits to the
pgtable entries and therefore no reader is allowed to interpret the
rest of the value afterwards.

Signed-off-by: Alexander Gordeev <agordeev@xxxxxxxxxxxxx>
---
 arch/s390/mm/pgtable.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index 4acd8b140c4b..da076928b63b 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -107,7 +107,7 @@ static inline pte_t ptep_flush_lazy(struct mm_struct *mm,
 	atomic_inc(&mm->context.flush_count);
 	if (cpumask_equal(&mm->context.cpu_attach_mask,
 			  cpumask_of(smp_processor_id()))) {
-		set_pte(ptep, set_pte_bit(*ptep, __pgprot(_PAGE_INVALID)));
+		set_pte(ptep, set_pte_bit(old, __pgprot(_PAGE_INVALID)));
 		mm->context.flush_mm = 1;
 	} else
 		ptep_ipte_global(mm, addr, ptep, nodat);
@@ -227,7 +227,7 @@ static inline pmd_t pmdp_flush_lazy(struct mm_struct *mm,
 	atomic_inc(&mm->context.flush_count);
 	if (cpumask_equal(&mm->context.cpu_attach_mask,
 			  cpumask_of(smp_processor_id()))) {
-		set_pmd(pmdp, set_pmd_bit(*pmdp, __pgprot(_SEGMENT_ENTRY_INVALID)));
+		set_pmd(pmdp, set_pmd_bit(old, __pgprot(_SEGMENT_ENTRY_INVALID)));
 		mm->context.flush_mm = 1;
 	} else {
 		pmdp_idte_global(mm, addr, pmdp);
-- 
2.53.0