2024-02-15 17:48:04

by Jonah Palmer

[permalink] [raw]
Subject: [PATCH RESEND] vdpa/mlx5: Allow CVQ size changes

The MLX driver was not updating its control virtqueue size at set_vq_num
and instead always initialized to MLX5_CVQ_MAX_ENT (16) at
setup_cvq_vring.

Qemu would try to set the size to 64 by default, however, because the
CVQ size always was initialized to 16, an error would be thrown when
sending >16 control messages (as used-ring entry 17 is initialized to 0).
For example, starting a guest with x-svq=on and then executing the
following command would produce the error below:

# for i in {1..20}; do ifconfig eth0 hw ether XX:xx:XX:xx:XX:XX; done

qemu-system-x86_64: Insufficient written data (0)
[ 435.331223] virtio_net virtio0: Failed to set mac address by vq command.
SIOCSIFHWADDR: Invalid argument

Signed-off-by: Jonah Palmer <[email protected]>
---
drivers/vdpa/mlx5/net/mlx5_vnet.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
index 778821bab7d9..c74de1fe6a94 100644
--- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
+++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
@@ -2276,9 +2276,15 @@ static void mlx5_vdpa_set_vq_num(struct vdpa_device *vdev, u16 idx, u32 num)
struct mlx5_vdpa_net *ndev = to_mlx5_vdpa_ndev(mvdev);
struct mlx5_vdpa_virtqueue *mvq;

- if (!is_index_valid(mvdev, idx) || is_ctrl_vq_idx(mvdev, idx))
+ if (!is_index_valid(mvdev, idx))
return;

+ if (is_ctrl_vq_idx(mvdev, idx)) {
+ struct mlx5_control_vq *cvq = &mvdev->cvq;
+ cvq->vring.vring.num = num;
+ return;
+ }
+
mvq = &ndev->vqs[idx];
mvq->num_ent = num;
}
@@ -2963,7 +2969,7 @@ static int setup_cvq_vring(struct mlx5_vdpa_dev *mvdev)
u16 idx = cvq->vring.last_avail_idx;

err = vringh_init_iotlb(&cvq->vring, mvdev->actual_features,
- MLX5_CVQ_MAX_ENT, false,
+ cvq->vring.vring.num, false,
(struct vring_desc *)(uintptr_t)cvq->desc_addr,
(struct vring_avail *)(uintptr_t)cvq->driver_addr,
(struct vring_used *)(uintptr_t)cvq->device_addr);
--
2.39.3



2024-02-15 18:19:05

by Dragos Tatulea

[permalink] [raw]
Subject: Re: [PATCH RESEND] vdpa/mlx5: Allow CVQ size changes

On 02/15, Jonah Palmer wrote:
> The MLX driver was not updating its control virtqueue size at set_vq_num
> and instead always initialized to MLX5_CVQ_MAX_ENT (16) at
> setup_cvq_vring.
>
> Qemu would try to set the size to 64 by default, however, because the
> CVQ size always was initialized to 16, an error would be thrown when
> sending >16 control messages (as used-ring entry 17 is initialized to 0).
> For example, starting a guest with x-svq=on and then executing the
> following command would produce the error below:
>
> # for i in {1..20}; do ifconfig eth0 hw ether XX:xx:XX:xx:XX:XX; done
>
> qemu-system-x86_64: Insufficient written data (0)
> [ 435.331223] virtio_net virtio0: Failed to set mac address by vq command.
> SIOCSIFHWADDR: Invalid argument
>
> Signed-off-by: Jonah Palmer <[email protected]>

Thanks for the patch Jonah!
> ---
> drivers/vdpa/mlx5/net/mlx5_vnet.c | 10 ++++++++--
> 1 file changed, 8 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
> index 778821bab7d9..c74de1fe6a94 100644
> --- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
> +++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
> @@ -2276,9 +2276,15 @@ static void mlx5_vdpa_set_vq_num(struct vdpa_device *vdev, u16 idx, u32 num)
> struct mlx5_vdpa_net *ndev = to_mlx5_vdpa_ndev(mvdev);
> struct mlx5_vdpa_virtqueue *mvq;
>
> - if (!is_index_valid(mvdev, idx) || is_ctrl_vq_idx(mvdev, idx))
> + if (!is_index_valid(mvdev, idx))
> return;
>
> + if (is_ctrl_vq_idx(mvdev, idx)) {
> + struct mlx5_control_vq *cvq = &mvdev->cvq;
Nit: add a space between variable declarations and usage.

> + cvq->vring.vring.num = num;
> + return;
> + }
> +
> mvq = &ndev->vqs[idx];
> mvq->num_ent = num;
> }
> @@ -2963,7 +2969,7 @@ static int setup_cvq_vring(struct mlx5_vdpa_dev *mvdev)
> u16 idx = cvq->vring.last_avail_idx;
>
> err = vringh_init_iotlb(&cvq->vring, mvdev->actual_features,
> - MLX5_CVQ_MAX_ENT, false,
> + cvq->vring.vring.num, false,
> (struct vring_desc *)(uintptr_t)cvq->desc_addr,
> (struct vring_avail *)(uintptr_t)cvq->driver_addr,
> (struct vring_used *)(uintptr_t)cvq->device_addr);

Besides the nit:

Acked-by: Dragos Tatulea <[email protected]>

> --
> 2.39.3
>