diff mbox series

[vhost,22/23] vdpa/mlx5: Don't reset VQs more than necessary

Message ID 20240617-stage-vdpa-vq-precreate-v1-22-8c0483f0ca2a@nvidia.com (mailing list archive)
State Superseded
Headers show
Series vdpa/mlx5: Pre-create HW VQs to reduce LM downtime | expand

Checks

Context Check Description
netdev/tree_selection success Not a local patch

Commit Message

Dragos Tatulea June 17, 2024, 3:07 p.m. UTC 
The vdpa device can be reset many times in sequence without any
significant state changes in between. Previously this was not a problem:
VQs were torn down only on first reset. But after VQ pre-creation was
introduced, each reset will delete and re-create the hardware VQs and
their associated resources.

To solve this problem, avoid resetting hardware VQs if the VQs are still
in a blank state.

Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Cosmin Ratiu <cratiu@nvidia.com>
---
 drivers/vdpa/mlx5/net/mlx5_vnet.c | 30 +++++++++++++++++++++++++++---
 1 file changed, 27 insertions(+), 3 deletions(-)

Comments

Eugenio Perez Martin June 19, 2024, 4:14 p.m. UTC | #1 
On Mon, Jun 17, 2024 at 5:09 PM Dragos Tatulea <dtatulea@nvidia.com> wrote:
>
> The vdpa device can be reset many times in sequence without any
> significant state changes in between. Previously this was not a problem:
> VQs were torn down only on first reset. But after VQ pre-creation was
> introduced, each reset will delete and re-create the hardware VQs and
> their associated resources.
>
> To solve this problem, avoid resetting hardware VQs if the VQs are still
> in a blank state.
>
> Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
> Reviewed-by: Cosmin Ratiu <cratiu@nvidia.com>

Acked-by: Eugenio Pérez <eperezma@redhat.com>

> ---
>  drivers/vdpa/mlx5/net/mlx5_vnet.c | 30 +++++++++++++++++++++++++++---
>  1 file changed, 27 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
> index d80d6b47da61..1a5ee0d2b47f 100644
> --- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
> +++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
> @@ -3134,18 +3134,41 @@ static void init_group_to_asid_map(struct mlx5_vdpa_dev *mvdev)
>                 mvdev->group2asid[i] = 0;
>  }
>
> +static bool needs_vqs_reset(const struct mlx5_vdpa_dev *mvdev)
> +{
> +       struct mlx5_vdpa_net *ndev = to_mlx5_vdpa_ndev(mvdev);
> +       struct mlx5_vdpa_virtqueue *mvq = &ndev->vqs[0];
> +
> +       if (mvdev->status & VIRTIO_CONFIG_S_DRIVER_OK)
> +               return true;
> +
> +       if (mvq->fw_state != MLX5_VIRTIO_NET_Q_OBJECT_STATE_INIT)
> +               return true;
> +
> +       return mvq->modified_fields & (
> +               MLX5_VIRTQ_MODIFY_MASK_STATE |
> +               MLX5_VIRTQ_MODIFY_MASK_VIRTIO_Q_ADDRS |
> +               MLX5_VIRTQ_MODIFY_MASK_VIRTIO_Q_AVAIL_IDX |
> +               MLX5_VIRTQ_MODIFY_MASK_VIRTIO_Q_USED_IDX
> +       );
> +}
> +
>  static int mlx5_vdpa_compat_reset(struct vdpa_device *vdev, u32 flags)
>  {
>         struct mlx5_vdpa_dev *mvdev = to_mvdev(vdev);
>         struct mlx5_vdpa_net *ndev = to_mlx5_vdpa_ndev(mvdev);
> +       bool vq_reset;
>
>         print_status(mvdev, 0, true);
>         mlx5_vdpa_info(mvdev, "performing device reset\n");
>
>         down_write(&ndev->reslock);
>         unregister_link_notifier(ndev);
> -       teardown_vq_resources(ndev);
> -       init_mvqs(ndev);
> +       vq_reset = needs_vqs_reset(mvdev);
> +       if (vq_reset) {
> +               teardown_vq_resources(ndev);
> +               init_mvqs(ndev);
> +       }
>
>         if (flags & VDPA_RESET_F_CLEAN_MAP)
>                 mlx5_vdpa_destroy_mr_resources(&ndev->mvdev);
> @@ -3165,7 +3188,8 @@ static int mlx5_vdpa_compat_reset(struct vdpa_device *vdev, u32 flags)
>                 if (mlx5_vdpa_create_dma_mr(mvdev))
>                         mlx5_vdpa_warn(mvdev, "create MR failed\n");
>         }
> -       setup_vq_resources(ndev, false);
> +       if (vq_reset)
> +               setup_vq_resources(ndev, false);
>         up_write(&ndev->reslock);
>
>         return 0;
>
> --
> 2.45.1
>
diff mbox series

Patch

diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
index d80d6b47da61..1a5ee0d2b47f 100644
--- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
+++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
@@ -3134,18 +3134,41 @@  static void init_group_to_asid_map(struct mlx5_vdpa_dev *mvdev)
 		mvdev->group2asid[i] = 0;
 }
 
+static bool needs_vqs_reset(const struct mlx5_vdpa_dev *mvdev)
+{
+	struct mlx5_vdpa_net *ndev = to_mlx5_vdpa_ndev(mvdev);
+	struct mlx5_vdpa_virtqueue *mvq = &ndev->vqs[0];
+
+	if (mvdev->status & VIRTIO_CONFIG_S_DRIVER_OK)
+		return true;
+
+	if (mvq->fw_state != MLX5_VIRTIO_NET_Q_OBJECT_STATE_INIT)
+		return true;
+
+	return mvq->modified_fields & (
+		MLX5_VIRTQ_MODIFY_MASK_STATE |
+		MLX5_VIRTQ_MODIFY_MASK_VIRTIO_Q_ADDRS |
+		MLX5_VIRTQ_MODIFY_MASK_VIRTIO_Q_AVAIL_IDX |
+		MLX5_VIRTQ_MODIFY_MASK_VIRTIO_Q_USED_IDX
+	);
+}
+
 static int mlx5_vdpa_compat_reset(struct vdpa_device *vdev, u32 flags)
 {
 	struct mlx5_vdpa_dev *mvdev = to_mvdev(vdev);
 	struct mlx5_vdpa_net *ndev = to_mlx5_vdpa_ndev(mvdev);
+	bool vq_reset;
 
 	print_status(mvdev, 0, true);
 	mlx5_vdpa_info(mvdev, "performing device reset\n");
 
 	down_write(&ndev->reslock);
 	unregister_link_notifier(ndev);
-	teardown_vq_resources(ndev);
-	init_mvqs(ndev);
+	vq_reset = needs_vqs_reset(mvdev);
+	if (vq_reset) {
+		teardown_vq_resources(ndev);
+		init_mvqs(ndev);
+	}
 
 	if (flags & VDPA_RESET_F_CLEAN_MAP)
 		mlx5_vdpa_destroy_mr_resources(&ndev->mvdev);
@@ -3165,7 +3188,8 @@  static int mlx5_vdpa_compat_reset(struct vdpa_device *vdev, u32 flags)
 		if (mlx5_vdpa_create_dma_mr(mvdev))
 			mlx5_vdpa_warn(mvdev, "create MR failed\n");
 	}
-	setup_vq_resources(ndev, false);
+	if (vq_reset)
+		setup_vq_resources(ndev, false);
 	up_write(&ndev->reslock);
 
 	return 0;