diff --git a/Documentation/locking/lockdep-design.rst b/Documentation/locking/lockdep-design.rst
index 82f36cab61bd..3bdc7a515df1 100644
--- a/Documentation/locking/lockdep-design.rst
+++ b/Documentation/locking/lockdep-design.rst
@@ -231,6 +231,31 @@ Note: When changing code to use the _nested() primitives, be careful and
 check really thoroughly that the hierarchy is correctly mapped; otherwise
 you can get false positives or false negatives.
 
+Pseudo locks
+------------
+
+Lockdep can detect some situations which lead to a deadlock by using
+a pseudo lock.  One such example is fs_reclaim.  In this situation,
+the deadlock is::
+
+  mutex_lock(A);
+  kmalloc(n, GFP_KERNEL);
+  (memory allocation enters filesystem reclaim)
+  mutex_lock(A);
+
+If every filesystem reclaim took mutex A, and every call to kmalloc()
+entered filesystem reclaim, lockdep would find this deadlock immediately
+and there would be no need for this special case.  However, most memory
+allocations will not enter reclaim at all, let alone filesystem reclaim,
+and not all filesystem reclaim paths take the same set of locks.  So we
+introduce a pseudo lock called fs_reclaim which is acquired for every
+memory allocation which might call fs reclaim.
+
+Previously, we used to treat fs reclaim as its own context like hardirq
+and softirq context, but it is more efficient for lockdep to treat it as
+a pseudo lock.  This approach can be used for other pseudo locks without
+needing to add more locking contexts.
+
 Annotations
 -----------