4 files changed, 132 insertions, 39 deletions

diff --git a/drivers/vdpa/mlx5/core/mlx5_vdpa.h b/drivers/vdpa/mlx5/core/mlx5_vdpa.h
index b6cc53ba980c..0002b2136b48 100644
--- a/drivers/vdpa/mlx5/core/mlx5_vdpa.h
+++ b/drivers/vdpa/mlx5/core/mlx5_vdpa.h
@@ -35,12 +35,14 @@ struct mlx5_vdpa_mr {
 
 	/* serialize mkey creation and destruction */
 	struct mutex mkey_mtx;
+	bool user_mr;
 };
 
 struct mlx5_vdpa_resources {
 	u32 pdn;
 	struct mlx5_uars_page *uar;
 	void __iomem *kick_addr;
+	u64 phys_kick_addr;
 	u16 uid;
 	u32 null_mkey;
 	bool valid;
diff --git a/drivers/vdpa/mlx5/core/mr.c b/drivers/vdpa/mlx5/core/mr.c
index 800cfd1967ad..dcee6039e966 100644
--- a/drivers/vdpa/mlx5/core/mr.c
+++ b/drivers/vdpa/mlx5/core/mr.c
@@ -219,11 +219,6 @@ static void destroy_indirect_key(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_m
 	mlx5_vdpa_destroy_mkey(mvdev, &mkey->mkey);
 }
 
-static struct device *get_dma_device(struct mlx5_vdpa_dev *mvdev)
-{
-	return &mvdev->mdev->pdev->dev;
-}
-
 static int map_direct_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_direct_mr *mr,
 			 struct vhost_iotlb *iotlb)
 {
@@ -239,7 +234,7 @@ static int map_direct_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_direct_mr
 	u64 pa;
 	u64 paend;
 	struct scatterlist *sg;
-	struct device *dma = get_dma_device(mvdev);
+	struct device *dma = mvdev->vdev.dma_dev;
 
 	for (map = vhost_iotlb_itree_first(iotlb, mr->start, mr->end - 1);
 	     map; map = vhost_iotlb_itree_next(map, start, mr->end - 1)) {
@@ -298,7 +293,7 @@ err_map:
 
 static void unmap_direct_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_direct_mr *mr)
 {
-	struct device *dma = get_dma_device(mvdev);
+	struct device *dma = mvdev->vdev.dma_dev;
 
 	destroy_direct_mr(mvdev, mr);
 	dma_unmap_sg_attrs(dma, mr->sg_head.sgl, mr->nsg, DMA_BIDIRECTIONAL, 0);
@@ -360,7 +355,7 @@ err_alloc:
  * indirect memory key that provides access to the enitre address space given
  * by iotlb.
  */
-static int _mlx5_vdpa_create_mr(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *iotlb)
+static int create_user_mr(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *iotlb)
 {
 	struct mlx5_vdpa_mr *mr = &mvdev->mr;
 	struct mlx5_vdpa_direct_mr *dmr;
@@ -374,9 +369,6 @@ static int _mlx5_vdpa_create_mr(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb
 	int err = 0;
 	int nnuls;
 
-	if (mr->initialized)
-		return 0;
-
 	INIT_LIST_HEAD(&mr->head);
 	for (map = vhost_iotlb_itree_first(iotlb, start, last); map;
 	     map = vhost_iotlb_itree_next(map, start, last)) {
@@ -414,7 +406,7 @@ static int _mlx5_vdpa_create_mr(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb
 	if (err)
 		goto err_chain;
 
-	mr->initialized = true;
+	mr->user_mr = true;
 	return 0;
 
 err_chain:
@@ -426,33 +418,94 @@ err_chain:
 	return err;
 }
 
-int mlx5_vdpa_create_mr(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *iotlb)
+static int create_dma_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_mr *mr)
+{
+	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+	void *mkc;
+	u32 *in;
+	int err;
+
+	in = kzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_PA);
+	MLX5_SET(mkc, mkc, length64, 1);
+	MLX5_SET(mkc, mkc, lw, 1);
+	MLX5_SET(mkc, mkc, lr, 1);
+	MLX5_SET(mkc, mkc, pd, mvdev->res.pdn);
+	MLX5_SET(mkc, mkc, qpn, 0xffffff);
+
+	err = mlx5_vdpa_create_mkey(mvdev, &mr->mkey, in, inlen);
+	if (!err)
+		mr->user_mr = false;
+
+	kfree(in);
+	return err;
+}
+
+static void destroy_dma_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_mr *mr)
+{
+	mlx5_vdpa_destroy_mkey(mvdev, &mr->mkey);
+}
+
+static int _mlx5_vdpa_create_mr(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *iotlb)
 {
 	struct mlx5_vdpa_mr *mr = &mvdev->mr;
 	int err;
 
-	mutex_lock(&mr->mkey_mtx);
+	if (mr->initialized)
+		return 0;
+
+	if (iotlb)
+		err = create_user_mr(mvdev, iotlb);
+	else
+		err = create_dma_mr(mvdev, mr);
+
+	if (!err)
+		mr->initialized = true;
+
+	return err;
+}
+
+int mlx5_vdpa_create_mr(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *iotlb)
+{
+	int err;
+
+	mutex_lock(&mvdev->mr.mkey_mtx);
 	err = _mlx5_vdpa_create_mr(mvdev, iotlb);
-	mutex_unlock(&mr->mkey_mtx);
+	mutex_unlock(&mvdev->mr.mkey_mtx);
 	return err;
 }
 
-void mlx5_vdpa_destroy_mr(struct mlx5_vdpa_dev *mvdev)
+static void destroy_user_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_mr *mr)
 {
-	struct mlx5_vdpa_mr *mr = &mvdev->mr;
 	struct mlx5_vdpa_direct_mr *dmr;
 	struct mlx5_vdpa_direct_mr *n;
 
-	mutex_lock(&mr->mkey_mtx);
-	if (!mr->initialized)
-		goto out;
-
 	destroy_indirect_key(mvdev, mr);
 	list_for_each_entry_safe_reverse(dmr, n, &mr->head, list) {
 		list_del_init(&dmr->list);
 		unmap_direct_mr(mvdev, dmr);
 		kfree(dmr);
 	}
+}
+
+void mlx5_vdpa_destroy_mr(struct mlx5_vdpa_dev *mvdev)
+{
+	struct mlx5_vdpa_mr *mr = &mvdev->mr;
+
+	mutex_lock(&mr->mkey_mtx);
+	if (!mr->initialized)
+		goto out;
+
+	if (mr->user_mr)
+		destroy_user_mr(mvdev, mr);
+	else
+		destroy_dma_mr(mvdev, mr);
+
 	memset(mr, 0, sizeof(*mr));
 	mr->initialized = false;
 out:
diff --git a/drivers/vdpa/mlx5/core/resources.c b/drivers/vdpa/mlx5/core/resources.c
index 6521cbd0f5c2..d4606213f88a 100644
--- a/drivers/vdpa/mlx5/core/resources.c
+++ b/drivers/vdpa/mlx5/core/resources.c
@@ -54,6 +54,9 @@ static int create_uctx(struct mlx5_vdpa_dev *mvdev, u16 *uid)
 	void *in;
 	int err;
 
+	if (MLX5_CAP_GEN(mvdev->mdev, umem_uid_0))
+		return 0;
+
 	/* 0 means not supported */
 	if (!MLX5_CAP_GEN(mvdev->mdev, log_max_uctx))
 		return -EOPNOTSUPP;
@@ -79,6 +82,9 @@ static void destroy_uctx(struct mlx5_vdpa_dev *mvdev, u32 uid)
 	u32 out[MLX5_ST_SZ_DW(destroy_uctx_out)] = {};
 	u32 in[MLX5_ST_SZ_DW(destroy_uctx_in)] = {};
 
+	if (!uid)
+		return;
+
 	MLX5_SET(destroy_uctx_in, in, opcode, MLX5_CMD_OP_DESTROY_UCTX);
 	MLX5_SET(destroy_uctx_in, in, uid, uid);
 
@@ -247,6 +253,7 @@ int mlx5_vdpa_alloc_resources(struct mlx5_vdpa_dev *mvdev)
 		goto err_key;
 
 	kick_addr = mdev->bar_addr + offset;
+	res->phys_kick_addr = kick_addr;
 
 	res->kick_addr = ioremap(kick_addr, PAGE_SIZE);
 	if (!res->kick_addr) {
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
index dda5dc6f7737..2a31467f7ac5 100644
--- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
+++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
@@ -611,8 +611,8 @@ static void cq_destroy(struct mlx5_vdpa_net *ndev, u16 idx)
 	mlx5_db_free(ndev->mvdev.mdev, &vcq->db);
 }
 
-static int umem_size(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtqueue *mvq, int num,
-		     struct mlx5_vdpa_umem **umemp)
+static void set_umem_size(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtqueue *mvq, int num,
+			  struct mlx5_vdpa_umem **umemp)
 {
 	struct mlx5_core_dev *mdev = ndev->mvdev.mdev;
 	int p_a;
@@ -635,7 +635,7 @@ static int umem_size(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtqueue *mvq
 		*umemp = &mvq->umem3;
 		break;
 	}
-	return p_a * mvq->num_ent + p_b;
+	(*umemp)->size = p_a * mvq->num_ent + p_b;
 }
 
 static void umem_frag_buf_free(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_umem *umem)
@@ -651,15 +651,10 @@ static int create_umem(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtqueue *m
 	void *in;
 	int err;
 	__be64 *pas;
-	int size;
 	struct mlx5_vdpa_umem *umem;
 
-	size = umem_size(ndev, mvq, num, &umem);
-	if (size < 0)
-		return size;
-
-	umem->size = size;
-	err = umem_frag_buf_alloc(ndev, umem, size);
+	set_umem_size(ndev, mvq, num, &umem);
+	err = umem_frag_buf_alloc(ndev, umem, umem->size);
 	if (err)
 		return err;
 
@@ -829,9 +824,9 @@ static int create_virtqueue(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtque
 	MLX5_SET(virtio_q, vq_ctx, umem_1_id, mvq->umem1.id);
 	MLX5_SET(virtio_q, vq_ctx, umem_1_size, mvq->umem1.size);
 	MLX5_SET(virtio_q, vq_ctx, umem_2_id, mvq->umem2.id);
-	MLX5_SET(virtio_q, vq_ctx, umem_2_size, mvq->umem1.size);
+	MLX5_SET(virtio_q, vq_ctx, umem_2_size, mvq->umem2.size);
 	MLX5_SET(virtio_q, vq_ctx, umem_3_id, mvq->umem3.id);
-	MLX5_SET(virtio_q, vq_ctx, umem_3_size, mvq->umem1.size);
+	MLX5_SET(virtio_q, vq_ctx, umem_3_size, mvq->umem3.size);
 	MLX5_SET(virtio_q, vq_ctx, pd, ndev->mvdev.res.pdn);
 	if (MLX5_CAP_DEV_VDPA_EMULATION(ndev->mvdev.mdev, eth_frame_offload_type))
 		MLX5_SET(virtio_q, vq_ctx, virtio_version_1_0, 1);
@@ -1428,8 +1423,8 @@ static int mlx5_vdpa_set_vq_state(struct vdpa_device *vdev, u16 idx,
 		return -EINVAL;
 	}
 
-	mvq->used_idx = state->avail_index;
-	mvq->avail_idx = state->avail_index;
+	mvq->used_idx = state->split.avail_index;
+	mvq->avail_idx = state->split.avail_index;
 	return 0;
 }
 
@@ -1450,7 +1445,7 @@ static int mlx5_vdpa_get_vq_state(struct vdpa_device *vdev, u16 idx, struct vdpa
 		 * Since both values should be identical, we take the value of
 		 * used_idx which is reported correctly.
 		 */
-		state->avail_index = mvq->used_idx;
+		state->split.avail_index = mvq->used_idx;
 		return 0;
 	}
 
@@ -1459,7 +1454,7 @@ static int mlx5_vdpa_get_vq_state(struct vdpa_device *vdev, u16 idx, struct vdpa
 		mlx5_vdpa_warn(mvdev, "failed to query virtqueue\n");
 		return err;
 	}
-	state->avail_index = attr.used_index;
+	state->split.avail_index = attr.used_index;
 	return 0;
 }
 
@@ -1772,6 +1767,14 @@ out:
 	mutex_unlock(&ndev->reslock);
 }
 
+static void clear_vqs_ready(struct mlx5_vdpa_net *ndev)
+{
+	int i;
+
+	for (i = 0; i < ndev->mvdev.max_vqs; i++)
+		ndev->vqs[i].ready = false;
+}
+
 static void mlx5_vdpa_set_status(struct vdpa_device *vdev, u8 status)
 {
 	struct mlx5_vdpa_dev *mvdev = to_mvdev(vdev);
@@ -1782,10 +1785,15 @@ static void mlx5_vdpa_set_status(struct vdpa_device *vdev, u8 status)
 	if (!status) {
 		mlx5_vdpa_info(mvdev, "performing device reset\n");
 		teardown_driver(ndev);
+		clear_vqs_ready(ndev);
 		mlx5_vdpa_destroy_mr(&ndev->mvdev);
 		ndev->mvdev.status = 0;
 		ndev->mvdev.mlx_features = 0;
 		++mvdev->generation;
+		if (MLX5_CAP_GEN(mvdev->mdev, umem_uid_0)) {
+			if (mlx5_vdpa_create_mr(mvdev, NULL))
+				mlx5_vdpa_warn(mvdev, "create MR failed\n");
+		}
 		return;
 	}
 
@@ -1866,6 +1874,7 @@ static void mlx5_vdpa_free(struct vdpa_device *vdev)
 	ndev = to_mlx5_vdpa_ndev(mvdev);
 
 	free_resources(ndev);
+	mlx5_vdpa_destroy_mr(mvdev);
 	if (!is_zero_ether_addr(ndev->config.mac)) {
 		pfmdev = pci_get_drvdata(pci_physfn(mvdev->mdev->pdev));
 		mlx5_mpfs_del_mac(pfmdev, ndev->config.mac);
@@ -1876,8 +1885,22 @@ static void mlx5_vdpa_free(struct vdpa_device *vdev)
 
 static struct vdpa_notification_area mlx5_get_vq_notification(struct vdpa_device *vdev, u16 idx)
 {
+	struct mlx5_vdpa_dev *mvdev = to_mvdev(vdev);
 	struct vdpa_notification_area ret = {};
+	struct mlx5_vdpa_net *ndev;
+	phys_addr_t addr;
+
+	/* If SF BAR size is smaller than PAGE_SIZE, do not use direct
+	 * notification to avoid the risk of mapping pages that contain BAR of more
+	 * than one SF
+	 */
+	if (MLX5_CAP_GEN(mvdev->mdev, log_min_sf_size) + 12 < PAGE_SHIFT)
+		return ret;
 
+	ndev = to_mlx5_vdpa_ndev(mvdev);
+	addr = (phys_addr_t)ndev->mvdev.res.phys_kick_addr;
+	ret.addr = addr;
+	ret.size = PAGE_SIZE;
 	return ret;
 }
 
@@ -2037,14 +2060,20 @@ static int mlx5_vdpa_dev_add(struct vdpa_mgmt_dev *v_mdev, const char *name)
 			goto err_mtu;
 	}
 
-	mvdev->vdev.dma_dev = mdev->device;
+	mvdev->vdev.dma_dev = &mdev->pdev->dev;
 	err = mlx5_vdpa_alloc_resources(&ndev->mvdev);
 	if (err)
 		goto err_mpfs;
 
+	if (MLX5_CAP_GEN(mvdev->mdev, umem_uid_0)) {
+		err = mlx5_vdpa_create_mr(mvdev, NULL);
+		if (err)
+			goto err_res;
+	}
+
 	err = alloc_resources(ndev);
 	if (err)
-		goto err_res;
+		goto err_mr;
 
 	mvdev->vdev.mdev = &mgtdev->mgtdev;
 	err = _vdpa_register_device(&mvdev->vdev, 2 * mlx5_vdpa_max_qps(max_vqs));
@@ -2056,6 +2085,8 @@ static int mlx5_vdpa_dev_add(struct vdpa_mgmt_dev *v_mdev, const char *name)
 
 err_reg:
 	free_resources(ndev);
+err_mr:
+	mlx5_vdpa_destroy_mr(mvdev);
 err_res:
 	mlx5_vdpa_free_resources(&ndev->mvdev);
 err_mpfs: