vDPA/ifcvf: implement management netlink framework for ifcvf
authorZhu Lingshan <lingshan.zhu@intel.com>
Thu, 12 Aug 2021 03:24:54 +0000 (11:24 +0800)
committerMichael S. Tsirkin <mst@redhat.com>
Sun, 5 Sep 2021 20:23:07 +0000 (16:23 -0400)
This commit implements the management netlink framework for ifcvf,
including register and add / remove a device

It works with iproute2:
[root@localhost lszhu]# vdpa mgmtdev show -jp
{
    "mgmtdev": {
        "pci/0000:01:00.5": {
            "supported_classes": [ "net" ]
        },
        "pci/0000:01:00.6": {
            "supported_classes": [ "net" ]
        }
    }
}

[root@localhost lszhu]# vdpa dev add mgmtdev pci/0000:01:00.5 name vdpa0
[root@localhost lszhu]# vdpa dev add mgmtdev pci/0000:01:00.6 name vdpa1

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20210812032454.24486-3-lingshan.zhu@intel.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
drivers/vdpa/ifcvf/ifcvf_base.h
drivers/vdpa/ifcvf/ifcvf_main.c

index 2996db0..1601e87 100644 (file)
@@ -106,6 +106,12 @@ struct ifcvf_lm_cfg {
        struct ifcvf_vring_lm_cfg vring_lm_cfg[IFCVF_MAX_QUEUE_PAIRS];
 };
 
+struct ifcvf_vdpa_mgmt_dev {
+       struct vdpa_mgmt_dev mdev;
+       struct ifcvf_adapter *adapter;
+       struct pci_dev *pdev;
+};
+
 int ifcvf_init_hw(struct ifcvf_hw *hw, struct pci_dev *dev);
 int ifcvf_start_hw(struct ifcvf_hw *hw);
 void ifcvf_stop_hw(struct ifcvf_hw *hw);
index 52465b7..e9a92b3 100644 (file)
@@ -218,7 +218,7 @@ static void ifcvf_vdpa_set_status(struct vdpa_device *vdpa_dev, u8 status)
        int ret;
 
        vf  = vdpa_to_vf(vdpa_dev);
-       adapter = dev_get_drvdata(vdpa_dev->dev.parent);
+       adapter = vdpa_to_adapter(vdpa_dev);
        status_old = ifcvf_get_status(vf);
 
        if (status_old == status)
@@ -458,6 +458,16 @@ static const struct vdpa_config_ops ifc_vdpa_ops = {
        .get_vq_notification = ifcvf_get_vq_notification,
 };
 
+static struct virtio_device_id id_table_net[] = {
+       {VIRTIO_ID_NET, VIRTIO_DEV_ANY_ID},
+       {0},
+};
+
+static struct virtio_device_id id_table_blk[] = {
+       {VIRTIO_ID_BLOCK, VIRTIO_DEV_ANY_ID},
+       {0},
+};
+
 static u32 get_dev_type(struct pci_dev *pdev)
 {
        u32 dev_type;
@@ -478,48 +488,30 @@ static u32 get_dev_type(struct pci_dev *pdev)
        return dev_type;
 }
 
-static int ifcvf_probe(struct pci_dev *pdev, const struct pci_device_id *id)
+static int ifcvf_vdpa_dev_add(struct vdpa_mgmt_dev *mdev, const char *name)
 {
-       struct device *dev = &pdev->dev;
+       struct ifcvf_vdpa_mgmt_dev *ifcvf_mgmt_dev;
        struct ifcvf_adapter *adapter;
+       struct pci_dev *pdev;
        struct ifcvf_hw *vf;
+       struct device *dev;
        int ret, i;
 
-       ret = pcim_enable_device(pdev);
-       if (ret) {
-               IFCVF_ERR(pdev, "Failed to enable device\n");
-               return ret;
-       }
-
-       ret = pcim_iomap_regions(pdev, BIT(0) | BIT(2) | BIT(4),
-                                IFCVF_DRIVER_NAME);
-       if (ret) {
-               IFCVF_ERR(pdev, "Failed to request MMIO region\n");
-               return ret;
-       }
-
-       ret = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
-       if (ret) {
-               IFCVF_ERR(pdev, "No usable DMA configuration\n");
-               return ret;
-       }
-
-       ret = devm_add_action_or_reset(dev, ifcvf_free_irq_vectors, pdev);
-       if (ret) {
-               IFCVF_ERR(pdev,
-                         "Failed for adding devres for freeing irq vectors\n");
-               return ret;
-       }
+       ifcvf_mgmt_dev = container_of(mdev, struct ifcvf_vdpa_mgmt_dev, mdev);
+       if (ifcvf_mgmt_dev->adapter)
+               return -EOPNOTSUPP;
 
+       pdev = ifcvf_mgmt_dev->pdev;
+       dev = &pdev->dev;
        adapter = vdpa_alloc_device(struct ifcvf_adapter, vdpa,
-                                   dev, &ifc_vdpa_ops, NULL);
+                                   dev, &ifc_vdpa_ops, name);
        if (IS_ERR(adapter)) {
                IFCVF_ERR(pdev, "Failed to allocate vDPA structure");
                return PTR_ERR(adapter);
        }
 
-       pci_set_master(pdev);
-       pci_set_drvdata(pdev, adapter);
+       ifcvf_mgmt_dev->adapter = adapter;
+       pci_set_drvdata(pdev, ifcvf_mgmt_dev);
 
        vf = &adapter->vf;
        vf->dev_type = get_dev_type(pdev);
@@ -539,9 +531,10 @@ static int ifcvf_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 
        vf->hw_features = ifcvf_get_hw_features(vf);
 
-       ret = vdpa_register_device(&adapter->vdpa, IFCVF_MAX_QUEUE_PAIRS * 2);
+       adapter->vdpa.mdev = &ifcvf_mgmt_dev->mdev;
+       ret = _vdpa_register_device(&adapter->vdpa, IFCVF_MAX_QUEUE_PAIRS * 2);
        if (ret) {
-               IFCVF_ERR(pdev, "Failed to register ifcvf to vdpa bus");
+               IFCVF_ERR(pdev, "Failed to register to vDPA bus");
                goto err;
        }
 
@@ -552,11 +545,100 @@ err:
        return ret;
 }
 
+static void ifcvf_vdpa_dev_del(struct vdpa_mgmt_dev *mdev, struct vdpa_device *dev)
+{
+       struct ifcvf_vdpa_mgmt_dev *ifcvf_mgmt_dev;
+
+       ifcvf_mgmt_dev = container_of(mdev, struct ifcvf_vdpa_mgmt_dev, mdev);
+       _vdpa_unregister_device(dev);
+       ifcvf_mgmt_dev->adapter = NULL;
+}
+
+static const struct vdpa_mgmtdev_ops ifcvf_vdpa_mgmt_dev_ops = {
+       .dev_add = ifcvf_vdpa_dev_add,
+       .dev_del = ifcvf_vdpa_dev_del
+};
+
+static int ifcvf_probe(struct pci_dev *pdev, const struct pci_device_id *id)
+{
+       struct ifcvf_vdpa_mgmt_dev *ifcvf_mgmt_dev;
+       struct device *dev = &pdev->dev;
+       u32 dev_type;
+       int ret;
+
+       ifcvf_mgmt_dev = kzalloc(sizeof(struct ifcvf_vdpa_mgmt_dev), GFP_KERNEL);
+       if (!ifcvf_mgmt_dev) {
+               IFCVF_ERR(pdev, "Failed to alloc memory for the vDPA management device\n");
+               return -ENOMEM;
+       }
+
+       dev_type = get_dev_type(pdev);
+       switch (dev_type) {
+       case VIRTIO_ID_NET:
+               ifcvf_mgmt_dev->mdev.id_table = id_table_net;
+               break;
+       case VIRTIO_ID_BLOCK:
+               ifcvf_mgmt_dev->mdev.id_table = id_table_blk;
+               break;
+       default:
+               IFCVF_ERR(pdev, "VIRTIO ID %u not supported\n", dev_type);
+               ret = -EOPNOTSUPP;
+               goto err;
+       }
+
+       ifcvf_mgmt_dev->mdev.ops = &ifcvf_vdpa_mgmt_dev_ops;
+       ifcvf_mgmt_dev->mdev.device = dev;
+       ifcvf_mgmt_dev->pdev = pdev;
+
+       ret = pcim_enable_device(pdev);
+       if (ret) {
+               IFCVF_ERR(pdev, "Failed to enable device\n");
+               goto err;
+       }
+
+       ret = pcim_iomap_regions(pdev, BIT(0) | BIT(2) | BIT(4),
+                                IFCVF_DRIVER_NAME);
+       if (ret) {
+               IFCVF_ERR(pdev, "Failed to request MMIO region\n");
+               goto err;
+       }
+
+       ret = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
+       if (ret) {
+               IFCVF_ERR(pdev, "No usable DMA configuration\n");
+               goto err;
+       }
+
+       ret = devm_add_action_or_reset(dev, ifcvf_free_irq_vectors, pdev);
+       if (ret) {
+               IFCVF_ERR(pdev,
+                         "Failed for adding devres for freeing irq vectors\n");
+               goto err;
+       }
+
+       pci_set_master(pdev);
+
+       ret = vdpa_mgmtdev_register(&ifcvf_mgmt_dev->mdev);
+       if (ret) {
+               IFCVF_ERR(pdev,
+                         "Failed to initialize the management interfaces\n");
+               goto err;
+       }
+
+       return 0;
+
+err:
+       kfree(ifcvf_mgmt_dev);
+       return ret;
+}
+
 static void ifcvf_remove(struct pci_dev *pdev)
 {
-       struct ifcvf_adapter *adapter = pci_get_drvdata(pdev);
+       struct ifcvf_vdpa_mgmt_dev *ifcvf_mgmt_dev;
 
-       vdpa_unregister_device(&adapter->vdpa);
+       ifcvf_mgmt_dev = pci_get_drvdata(pdev);
+       vdpa_mgmtdev_unregister(&ifcvf_mgmt_dev->mdev);
+       kfree(ifcvf_mgmt_dev);
 }
 
 static struct pci_device_id ifcvf_pci_ids[] = {