IB/mlx5: Add implicit MR support

Add implicit MR, covering entire user address space.
The MR is implemented as an indirect KSM MR consisting of
1GB direct MRs.
Pages and direct MRs are added/removed to MR by ODP.

Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>

drivers/infiniband/hw/mlx5/main.c

@@ -3583,6 +3583,8 @@ static int __init mlx5_ib_init(void)
 {
 	int err;
 
+	mlx5_ib_odp_init();
+
 	err = mlx5_register_interface(&mlx5_ib_interface);
 
 	return err;

drivers/infiniband/hw/mlx5/mlx5_ib.h

@@ -202,6 +202,7 @@ struct mlx5_ib_flow_db {
 #define MLX5_IB_UPD_XLT_ADDR	      BIT(3)
 #define MLX5_IB_UPD_XLT_PD	      BIT(4)
 #define MLX5_IB_UPD_XLT_ACCESS	      BIT(5)
+#define MLX5_IB_UPD_XLT_INDIRECT      BIT(6)
 
 /* Private QP creation flags to be passed in ib_qp_init_attr.create_flags.
  *
@@ -503,6 +504,10 @@ struct mlx5_ib_mr {
 	int			live;
 	void			*descs_alloc;
 	int			access_flags; /* Needed for rereg MR */
+
+	struct mlx5_ib_mr      *parent;
+	atomic_t		num_leaf_free;
+	wait_queue_head_t       q_leaf_free;
 };
 
 struct mlx5_ib_mw {
@@ -637,6 +642,7 @@ struct mlx5_ib_dev {
 	 * being used by a page fault handler.
 	 */
 	struct srcu_struct      mr_srcu;
+	u32			null_mkey;
 #endif
 	struct mlx5_ib_flow_db	flow_db;
 	/* protect resources needed as part of reset flow */
@@ -789,6 +795,9 @@ struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
 int mlx5_ib_dealloc_mw(struct ib_mw *mw);
 int mlx5_ib_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages,
 		       int page_shift, int flags);
+struct mlx5_ib_mr *mlx5_ib_alloc_implicit_mr(struct mlx5_ib_pd *pd,
+					     int access_flags);
+void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *mr);
 int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
 			  u64 length, u64 virt_addr, int access_flags,
 			  struct ib_pd *pd, struct ib_udata *udata);
@@ -868,16 +877,23 @@ int __init mlx5_ib_odp_init(void);
 void mlx5_ib_odp_cleanup(void);
 void mlx5_ib_invalidate_range(struct ib_umem *umem, unsigned long start,
 			      unsigned long end);
+void mlx5_odp_init_mr_cache_entry(struct mlx5_cache_ent *ent);
+void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset,
+			   size_t nentries, struct mlx5_ib_mr *mr, int flags);
 #else /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
 static inline void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev)
 {
 	return;
 }
 
 static inline int mlx5_ib_odp_init_one(struct mlx5_ib_dev *ibdev) { return 0; }
-static inline void mlx5_ib_odp_remove_one(struct mlx5_ib_dev *ibdev)	{}
+static inline void mlx5_ib_odp_remove_one(struct mlx5_ib_dev *ibdev)	    {}
 static inline int mlx5_ib_odp_init(void) { return 0; }
-static inline void mlx5_ib_odp_cleanup(void)				{}
+static inline void mlx5_ib_odp_cleanup(void)				    {}
+static inline void mlx5_odp_init_mr_cache_entry(struct mlx5_cache_ent *ent) {}
+static inline void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset,
+					 size_t nentries, struct mlx5_ib_mr *mr,
+					 int flags) {}
 
 #endif /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
 

drivers/infiniband/hw/mlx5/mr.c

@@ -469,7 +469,7 @@ struct mlx5_ib_mr *mlx5_mr_cache_alloc(struct mlx5_ib_dev *dev, int entry)
 			spin_unlock_irq(&ent->lock);
 
 			err = add_keys(dev, entry, 1);
-			if (err)
+			if (err && err != -EAGAIN)
 				return ERR_PTR(err);
 
 			wait_for_completion(&ent->compl);
@@ -669,8 +669,10 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
 		INIT_DELAYED_WORK(&ent->dwork, delayed_cache_work_func);
 		queue_work(cache->wq, &ent->work);
 
-		if (i > MAX_UMR_CACHE_ENTRY)
+		if (i > MAX_UMR_CACHE_ENTRY) {
+			mlx5_odp_init_mr_cache_entry(ent);
 			continue;
+		}
 
 		if (!use_umr(dev, ent->order))
 			continue;
@@ -935,6 +937,10 @@ static inline int populate_xlt(struct mlx5_ib_mr *mr, int idx, int npages,
 {
 	struct mlx5_ib_dev *dev = mr->dev;
 	struct ib_umem *umem = mr->umem;
+	if (flags & MLX5_IB_UPD_XLT_INDIRECT) {
+		mlx5_odp_populate_klm(xlt, idx, npages, mr, flags);
+		return npages;
+	}
 
 	npages = min_t(size_t, npages, ib_umem_num_pages(umem) - idx);
 
@@ -968,7 +974,9 @@ int mlx5_ib_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages,
 	struct mlx5_umr_wr wr;
 	struct ib_sge sg;
 	int err = 0;
-	int desc_size = sizeof(struct mlx5_mtt);
+	int desc_size = (flags & MLX5_IB_UPD_XLT_INDIRECT)
+			       ? sizeof(struct mlx5_klm)
+			       : sizeof(struct mlx5_mtt);
 	const int page_align = MLX5_UMR_MTT_ALIGNMENT / desc_size;
 	const int page_mask = page_align - 1;
 	size_t pages_mapped = 0;
@@ -1186,6 +1194,18 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 
 	mlx5_ib_dbg(dev, "start 0x%llx, virt_addr 0x%llx, length 0x%llx, access_flags 0x%x\n",
 		    start, virt_addr, length, access_flags);
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	if (!start && length == U64_MAX) {
+		if (!(access_flags & IB_ACCESS_ON_DEMAND) ||
+		    !(dev->odp_caps.general_caps & IB_ODP_SUPPORT_IMPLICIT))
+			return ERR_PTR(-EINVAL);
+
+		mr = mlx5_ib_alloc_implicit_mr(to_mpd(pd), access_flags);
+		return &mr->ibmr;
+	}
+#endif
+
 	err = mr_umem_get(pd, start, length, access_flags, &umem, &npages,
 			   &page_shift, &ncont, &order);
 
@@ -1471,8 +1491,11 @@ int mlx5_ib_dereg_mr(struct ib_mr *ibmr)
 		/* Wait for all running page-fault handlers to finish. */
 		synchronize_srcu(&dev->mr_srcu);
 		/* Destroy all page mappings */
-		mlx5_ib_invalidate_range(umem, ib_umem_start(umem),
-					 ib_umem_end(umem));
+		if (umem->odp_data->page_list)
+			mlx5_ib_invalidate_range(umem, ib_umem_start(umem),
+						 ib_umem_end(umem));
+		else
+			mlx5_ib_free_implicit_mr(mr);
 		/*
 		 * We kill the umem before the MR for ODP,
 		 * so that there will not be any invalidations in