1 files changed, 22 insertions, 16 deletions

diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.h b/drivers/infiniband/hw/hfi1/mmu_rb.h
index 7417be2b9dc8..3fa50dd64db6 100644
--- a/drivers/infiniband/hw/hfi1/mmu_rb.h
+++ b/drivers/infiniband/hw/hfi1/mmu_rb.h
@@ -1,4 +1,4 @@
-/* SPDX-License-Identifier: GPL-2.0 or BSD-3-Clause */
+/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */
 /*
  * Copyright(c) 2020 Cornelis Networks, Inc.
  * Copyright(c) 2016 Intel Corporation.
@@ -16,46 +16,52 @@ struct mmu_rb_node {
 	struct rb_node node;
 	struct mmu_rb_handler *handler;
 	struct list_head list;
+	struct kref refcount;
 };
 
-/*
- * NOTE: filter, insert, invalidate, and evict must not sleep.  Only remove is
- * allowed to sleep.
- */
+/* filter and evict must not sleep. Only remove is allowed to sleep. */
 struct mmu_rb_ops {
 	bool (*filter)(struct mmu_rb_node *node, unsigned long addr,
 		       unsigned long len);
-	int (*insert)(void *ops_arg, struct mmu_rb_node *mnode);
 	void (*remove)(void *ops_arg, struct mmu_rb_node *mnode);
-	int (*invalidate)(void *ops_arg, struct mmu_rb_node *node);
 	int (*evict)(void *ops_arg, struct mmu_rb_node *mnode,
 		     void *evict_arg, bool *stop);
 };
 
 struct mmu_rb_handler {
+	/*
+	 * struct mmu_notifier is 56 bytes, and spinlock_t is 4 bytes, so
+	 * they fit together in one cache line.  mn is relatively rarely
+	 * accessed, so co-locating the spinlock with it achieves much of
+	 * the cacheline contention reduction of giving the spinlock its own
+	 * cacheline without the overhead of doing so.
+	 */
 	struct mmu_notifier mn;
-	struct rb_root_cached root;
-	void *ops_arg;
 	spinlock_t lock;        /* protect the RB tree */
-	struct mmu_rb_ops *ops;
+
+	/* Begin on a new cachline boundary here */
+	struct rb_root_cached root ____cacheline_aligned_in_smp;
+	void *ops_arg;
+	const struct mmu_rb_ops *ops;
 	struct list_head lru_list;
 	struct work_struct del_work;
 	struct list_head del_list;
 	struct workqueue_struct *wq;
+	void *free_ptr;
 };
 
 int hfi1_mmu_rb_register(void *ops_arg,
-			 struct mmu_rb_ops *ops,
+			 const struct mmu_rb_ops *ops,
 			 struct workqueue_struct *wq,
 			 struct mmu_rb_handler **handler);
 void hfi1_mmu_rb_unregister(struct mmu_rb_handler *handler);
 int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler,
 		       struct mmu_rb_node *mnode);
+void hfi1_mmu_rb_release(struct kref *refcount);
+
 void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg);
-void hfi1_mmu_rb_remove(struct mmu_rb_handler *handler,
-			struct mmu_rb_node *mnode);
-bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler,
-				     unsigned long addr, unsigned long len,
-				     struct mmu_rb_node **rb_node);
+struct mmu_rb_node *hfi1_mmu_rb_get_first(struct mmu_rb_handler *handler,
+					  unsigned long addr,
+					  unsigned long len);
 
 #endif /* _HFI1_MMU_RB_H */