[PATCH RFC] mm: convert rcu_read_lock() to srcu_read_lock(), thus allowing to sleep in callbacks

[sagig@xxxxxxxxxxxx]

Callbacks: invalidate_page, invalidate_range_start/end, change_pte
Now that anon_vma lock and i_mmap_mutex are both sleepable mutex, it is possible to schedule inside invalidation cllabacks
This is essential for a scheduling HW sync in RDMA drivers which apply on demand paging methods

Signed-off-by: sagi grimberg <sagig@xxxxxxxxxxxxxx>
---
 mm/mmu_notifier.c |   63 ++++++++++++++++++++++++++++++++++++++++++++++------
 1 files changed, 55 insertions(+), 8 deletions(-)

diff --git a/mm/mmu_notifier.c b/mm/mmu_notifier.c
index 9a611d3..70dadd5 100644
--- a/mm/mmu_notifier.c
+++ b/mm/mmu_notifier.c
@@ -123,10 +123,16 @@ int __mmu_notifier_test_young(struct mm_struct *mm,
 void __mmu_notifier_change_pte(struct mm_struct *mm, unsigned long address,
 			       pte_t pte)
 {
+	int idx = -1;
+	struct srcu_struct srcu;
 	struct mmu_notifier *mn;
 	struct hlist_node *n;
 
-	rcu_read_lock();
+	if (init_srcu_struct(&srcu))
+		rcu_read_lock();
+	else
+		idx = srcu_read_lock(&srcu);
+
 	hlist_for_each_entry_rcu(mn, n, &mm->mmu_notifier_mm->list, hlist) {
 		if (mn->ops->change_pte)
 			mn->ops->change_pte(mn, mm, address, pte);
@@ -137,49 +143,90 @@ void __mmu_notifier_change_pte(struct mm_struct *mm, unsigned long address,
 		else if (mn->ops->invalidate_page)
 			mn->ops->invalidate_page(mn, mm, address);
 	}
-	rcu_read_unlock();
+
+	if (idx < 0)
+		rcu_read_unlock();
+	else
+		srcu_read_unlock(&srcu, idx);
+
+	cleanup_srcu_struct(&srcu);
 }
 
 void __mmu_notifier_invalidate_page(struct mm_struct *mm,
 					  unsigned long address)
 {
+	int idx = -1;
+	struct srcu_struct srcu;
 	struct mmu_notifier *mn;
 	struct hlist_node *n;
 
-	rcu_read_lock();
+	if (init_srcu_struct(&srcu))
+		rcu_read_lock();
+	else
+		idx = srcu_read_lock(&srcu);
+
 	hlist_for_each_entry_rcu(mn, n, &mm->mmu_notifier_mm->list, hlist) {
 		if (mn->ops->invalidate_page)
 			mn->ops->invalidate_page(mn, mm, address);
 	}
-	rcu_read_unlock();
+
+	if (idx < 0)
+		rcu_read_unlock();
+	else
+		srcu_read_unlock(&srcu, idx);
+
+	cleanup_srcu_struct(&srcu);
 }
 
 void __mmu_notifier_invalidate_range_start(struct mm_struct *mm,
 				  unsigned long start, unsigned long end)
 {
+	int idx = -1;
+	struct srcu_struct srcu;
 	struct mmu_notifier *mn;
 	struct hlist_node *n;
 
-	rcu_read_lock();
+	if (init_srcu_struct(&srcu))
+		rcu_read_lock();
+	else
+		idx = srcu_read_lock(&srcu);
+
 	hlist_for_each_entry_rcu(mn, n, &mm->mmu_notifier_mm->list, hlist) {
 		if (mn->ops->invalidate_range_start)
 			mn->ops->invalidate_range_start(mn, mm, start, end);
 	}
-	rcu_read_unlock();
+
+	if (idx < 0)
+		rcu_read_unlock();
+	else
+		srcu_read_unlock(&srcu, idx);
+
+	cleanup_srcu_struct(&srcu);
 }
 
 void __mmu_notifier_invalidate_range_end(struct mm_struct *mm,
 				  unsigned long start, unsigned long end)
 {
+	int idx = -1;
+	struct srcu_struct srcu;
 	struct mmu_notifier *mn;
 	struct hlist_node *n;
 
-	rcu_read_lock();
+	if (init_srcu_struct(&srcu))
+		rcu_read_lock();
+	else
+		idx = srcu_read_lock(&srcu);
 	hlist_for_each_entry_rcu(mn, n, &mm->mmu_notifier_mm->list, hlist) {
 		if (mn->ops->invalidate_range_end)
 			mn->ops->invalidate_range_end(mn, mm, start, end);
 	}
-	rcu_read_unlock();
+
+	if (idx < 0)
+		rcu_read_unlock();
+	else
+		srcu_read_unlock(&srcu, idx);
+
+	cleanup_srcu_struct(&srcu);
 }
 
 static int do_mmu_notifier_register(struct mmu_notifier *mn,
-- 
1.7.6.4

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>