diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index efe383c0ac86..eedba0d2ec4b 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -587,6 +587,7 @@ struct mlx5_ib_mr {
 	struct mlx5_ib_mr      *parent;
 	atomic_t		num_leaf_free;
 	wait_queue_head_t       q_leaf_free;
+	struct mlx5_async_work  cb_work;
 };
 
 struct mlx5_ib_mw {
@@ -944,6 +945,7 @@ struct mlx5_ib_dev {
 	struct mlx5_memic	memic;
 	u16			devx_whitelist_uid;
 	struct mlx5_srq_table   srq_table;
+	struct mlx5_async_ctx   async_ctx;
 };
 
 static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index fd6ea1f75085..bf2b6ea23851 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -123,9 +123,10 @@ static void update_odp_mr(struct mlx5_ib_mr *mr)
 }
 #endif
 
-static void reg_mr_callback(int status, void *context)
+static void reg_mr_callback(int status, struct mlx5_async_work *context)
 {
-	struct mlx5_ib_mr *mr = context;
+	struct mlx5_ib_mr *mr =
+		container_of(context, struct mlx5_ib_mr, cb_work);
 	struct mlx5_ib_dev *dev = mr->dev;
 	struct mlx5_mr_cache *cache = &dev->cache;
 	int c = order2idx(dev, mr->order);
@@ -216,9 +217,9 @@ static int add_keys(struct mlx5_ib_dev *dev, int c, int num)
 		ent->pending++;
 		spin_unlock_irq(&ent->lock);
 		err = mlx5_core_create_mkey_cb(dev->mdev, &mr->mmkey,
-					       in, inlen,
+					       &dev->async_ctx, in, inlen,
 					       mr->out, sizeof(mr->out),
-					       reg_mr_callback, mr);
+					       reg_mr_callback, &mr->cb_work);
 		if (err) {
 			spin_lock_irq(&ent->lock);
 			ent->pending--;
@@ -679,6 +680,7 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
 		return -ENOMEM;
 	}
 
+	mlx5_cmd_init_async_ctx(dev->mdev, &dev->async_ctx);
 	timer_setup(&dev->delay_timer, delay_time_func, 0);
 	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
 		ent = &cache->ent[i];
@@ -725,33 +727,6 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
 	return 0;
 }
 
-static void wait_for_async_commands(struct mlx5_ib_dev *dev)
-{
-	struct mlx5_mr_cache *cache = &dev->cache;
-	struct mlx5_cache_ent *ent;
-	int total = 0;
-	int i;
-	int j;
-
-	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
-		ent = &cache->ent[i];
-		for (j = 0 ; j < 1000; j++) {
-			if (!ent->pending)
-				break;
-			msleep(50);
-		}
-	}
-	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
-		ent = &cache->ent[i];
-		total += ent->pending;
-	}
-
-	if (total)
-		mlx5_ib_warn(dev, "aborted while there are %d pending mr requests\n", total);
-	else
-		mlx5_ib_warn(dev, "done with all pending requests\n");
-}
-
 int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev)
 {
 	int i;
@@ -763,12 +738,12 @@ int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev)
 	flush_workqueue(dev->cache.wq);
 
 	mlx5_mr_cache_debugfs_cleanup(dev);
+	mlx5_cmd_cleanup_async_ctx(&dev->async_ctx);
 
 	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++)
 		clean_keys(dev, i);
 
 	destroy_workqueue(dev->cache.wq);
-	wait_for_async_commands(dev);
 	del_timer_sync(&dev->delay_timer);
 
 	return 0;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
index 3e0fa8a8077b..a25a8c6f938e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
@@ -1711,12 +1711,57 @@ int mlx5_cmd_exec(struct mlx5_core_dev *dev, void *in, int in_size, void *out,
 }
 EXPORT_SYMBOL(mlx5_cmd_exec);
 
-int mlx5_cmd_exec_cb(struct mlx5_core_dev *dev, void *in, int in_size,
-		     void *out, int out_size, mlx5_cmd_cbk_t callback,
-		     void *context)
+void mlx5_cmd_init_async_ctx(struct mlx5_core_dev *dev,
+			     struct mlx5_async_ctx *ctx)
 {
-	return cmd_exec(dev, in, in_size, out, out_size, callback, context,
-			false);
+	ctx->dev = dev;
+	/* Starts at 1 to avoid doing wake_up if we are not cleaning up */
+	atomic_set(&ctx->num_inflight, 1);
+	init_waitqueue_head(&ctx->wait);
+}
+EXPORT_SYMBOL(mlx5_cmd_init_async_ctx);
+
+/**
+ * mlx5_cmd_cleanup_async_ctx - Clean up an async_ctx
+ * @ctx: The ctx to clean
+ *
+ * Upon return all callbacks given to mlx5_cmd_exec_cb() have been called. The
+ * caller must ensure that mlx5_cmd_exec_cb() is not called during or after
+ * the call mlx5_cleanup_async_ctx().
+ */
+void mlx5_cmd_cleanup_async_ctx(struct mlx5_async_ctx *ctx)
+{
+	atomic_dec(&ctx->num_inflight);
+	wait_event(ctx->wait, atomic_read(&ctx->num_inflight) == 0);
+}
+EXPORT_SYMBOL(mlx5_cmd_cleanup_async_ctx);
+
+static void mlx5_cmd_exec_cb_handler(int status, void *_work)
+{
+	struct mlx5_async_work *work = _work;
+	struct mlx5_async_ctx *ctx = work->ctx;
+
+	work->user_callback(status, work);
+	if (atomic_dec_and_test(&ctx->num_inflight))
+		wake_up(&ctx->wait);
+}
+
+int mlx5_cmd_exec_cb(struct mlx5_async_ctx *ctx, void *in, int in_size,
+		     void *out, int out_size, mlx5_async_cbk_t callback,
+		     struct mlx5_async_work *work)
+{
+	int ret;
+
+	work->ctx = ctx;
+	work->user_callback = callback;
+	if (WARN_ON(!atomic_inc_not_zero(&ctx->num_inflight)))
+		return -EIO;
+	ret = cmd_exec(ctx->dev, in, in_size, out, out_size,
+		       mlx5_cmd_exec_cb_handler, work, false);
+	if (ret && atomic_dec_and_test(&ctx->num_inflight))
+		wake_up(&ctx->wait);
+
+	return ret;
 }
 EXPORT_SYMBOL(mlx5_cmd_exec_cb);
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mr.c b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
index 0670165afd5f..ea744d8466ea 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mr.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
@@ -51,9 +51,10 @@ void mlx5_cleanup_mkey_table(struct mlx5_core_dev *dev)
 
 int mlx5_core_create_mkey_cb(struct mlx5_core_dev *dev,
 			     struct mlx5_core_mkey *mkey,
-			     u32 *in, int inlen,
-			     u32 *out, int outlen,
-			     mlx5_cmd_cbk_t callback, void *context)
+			     struct mlx5_async_ctx *async_ctx, u32 *in,
+			     int inlen, u32 *out, int outlen,
+			     mlx5_async_cbk_t callback,
+			     struct mlx5_async_work *context)
 {
 	struct mlx5_mkey_table *table = &dev->priv.mkey_table;
 	u32 lout[MLX5_ST_SZ_DW(create_mkey_out)] = {0};
@@ -71,7 +72,7 @@ int mlx5_core_create_mkey_cb(struct mlx5_core_dev *dev,
 	MLX5_SET(mkc, mkc, mkey_7_0, key);
 
 	if (callback)
-		return mlx5_cmd_exec_cb(dev, in, inlen, out, outlen,
+		return mlx5_cmd_exec_cb(async_ctx, in, inlen, out, outlen,
 					callback, context);
 
 	err = mlx5_cmd_exec(dev, in, inlen, lout, sizeof(lout));
@@ -105,7 +106,7 @@ int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
 			  struct mlx5_core_mkey *mkey,
 			  u32 *in, int inlen)
 {
-	return mlx5_core_create_mkey_cb(dev, mkey, in, inlen,
+	return mlx5_core_create_mkey_cb(dev, mkey, NULL, in, inlen,
 					NULL, 0, NULL, NULL);
 }
 EXPORT_SYMBOL(mlx5_core_create_mkey);
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 4e444863054a..039c9398614c 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -850,11 +850,30 @@ void mlx5_cmd_cleanup(struct mlx5_core_dev *dev);
 void mlx5_cmd_use_events(struct mlx5_core_dev *dev);
 void mlx5_cmd_use_polling(struct mlx5_core_dev *dev);
 
+struct mlx5_async_ctx {
+	struct mlx5_core_dev *dev;
+	atomic_t num_inflight;
+	struct wait_queue_head wait;
+};
+
+struct mlx5_async_work;
+
+typedef void (*mlx5_async_cbk_t)(int status, struct mlx5_async_work *context);
+
+struct mlx5_async_work {
+	struct mlx5_async_ctx *ctx;
+	mlx5_async_cbk_t user_callback;
+};
+
+void mlx5_cmd_init_async_ctx(struct mlx5_core_dev *dev,
+			     struct mlx5_async_ctx *ctx);
+void mlx5_cmd_cleanup_async_ctx(struct mlx5_async_ctx *ctx);
+int mlx5_cmd_exec_cb(struct mlx5_async_ctx *ctx, void *in, int in_size,
+		     void *out, int out_size, mlx5_async_cbk_t callback,
+		     struct mlx5_async_work *work);
+
 int mlx5_cmd_exec(struct mlx5_core_dev *dev, void *in, int in_size, void *out,
 		  int out_size);
-int mlx5_cmd_exec_cb(struct mlx5_core_dev *dev, void *in, int in_size,
-		     void *out, int out_size, mlx5_cmd_cbk_t callback,
-		     void *context);
 int mlx5_cmd_exec_polling(struct mlx5_core_dev *dev, void *in, int in_size,
 			  void *out, int out_size);
 void mlx5_cmd_mbox_status(void *out, u8 *status, u32 *syndrome);
@@ -885,9 +904,10 @@ void mlx5_init_mkey_table(struct mlx5_core_dev *dev);
 void mlx5_cleanup_mkey_table(struct mlx5_core_dev *dev);
 int mlx5_core_create_mkey_cb(struct mlx5_core_dev *dev,
 			     struct mlx5_core_mkey *mkey,
-			     u32 *in, int inlen,
-			     u32 *out, int outlen,
-			     mlx5_cmd_cbk_t callback, void *context);
+			     struct mlx5_async_ctx *async_ctx, u32 *in,
+			     int inlen, u32 *out, int outlen,
+			     mlx5_async_cbk_t callback,
+			     struct mlx5_async_work *context);
 int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
 			  struct mlx5_core_mkey *mkey,
 			  u32 *in, int inlen);