MIME-Version: 1.0
References: <20230228094110.37-1-xieyongji@bytedance.com> <20230228094110.37-4-xieyongji@bytedance.com>
In-Reply-To: <20230228094110.37-4-xieyongji@bytedance.com>
From:   Jason Wang <jasowang@redhat.com>
Date:   Thu, 16 Mar 2023 12:02:40 +0800
Message-ID: <CACGkMEvmV7xKc7VnaZT+DGcN2hg64ksGHxRAihW2f=RpXydZoQ@mail.gmail.com>
Subject: Re: [PATCH v3 03/11] vdpa: Add set_irq_affinity callback in vdpa_config_ops
To:     Xie Yongji <xieyongji@bytedance.com>
Cc:     mst@redhat.com, tglx@linutronix.de, hch@lst.de,
        virtualization@lists.linux-foundation.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Tue, Feb 28, 2023 at 5:42=E2=80=AFPM Xie Yongji <xieyongji@bytedance.com=
> wrote:
>
> This introduces set_irq_affinity callback in
> vdpa_config_ops so that vdpa device driver can
> get the interrupt affinity hint from the virtio
> device driver. The interrupt affinity hint would
> be needed by the interrupt affinity spreading
> mechanism.
>
> Signed-off-by: Xie Yongji <xieyongji@bytedance.com>
> ---
>  drivers/virtio/virtio_vdpa.c | 4 ++++
>  include/linux/vdpa.h         | 9 +++++++++
>  2 files changed, 13 insertions(+)
>
> diff --git a/drivers/virtio/virtio_vdpa.c b/drivers/virtio/virtio_vdpa.c
> index f72696b4c1c2..9eee8afabda8 100644
> --- a/drivers/virtio/virtio_vdpa.c
> +++ b/drivers/virtio/virtio_vdpa.c
> @@ -282,9 +282,13 @@ static int virtio_vdpa_find_vqs(struct virtio_device=
 *vdev, unsigned int nvqs,
>         struct virtio_vdpa_device *vd_dev =3D to_virtio_vdpa_device(vdev)=
;
>         struct vdpa_device *vdpa =3D vd_get_vdpa(vdev);
>         const struct vdpa_config_ops *ops =3D vdpa->config;
> +       struct irq_affinity default_affd =3D { 0 };
>         struct vdpa_callback cb;
>         int i, err, queue_idx =3D 0;
>
> +       if (ops->set_irq_affinity)
> +               ops->set_irq_affinity(vdpa, desc ? desc : &default_affd);
> +
>         for (i =3D 0; i < nvqs; ++i) {
>                 if (!names[i]) {
>                         vqs[i] =3D NULL;
> diff --git a/include/linux/vdpa.h b/include/linux/vdpa.h
> index d61f369f9cd6..10bd22387276 100644
> --- a/include/linux/vdpa.h
> +++ b/include/linux/vdpa.h
> @@ -259,6 +259,13 @@ struct vdpa_map_file {
>   *                             @vdev: vdpa device
>   *                             @idx: virtqueue index
>   *                             Returns the irq affinity mask
> + * @set_irq_affinity:          Pass the irq affinity hint (best effort)

Note that this could easily confuse the users. I wonder if we can
unify it with set_irq_affinity. Looking at vduse's implementation, it
should be possible.

(E.g set_vq_affinity implemented by virtio-pci are using irq affinity actua=
lly).

Thanks

> + *                             from the virtio device driver to vdpa
> + *                             driver (optional).
> + *                             Needed by the interrupt affinity spreadin=
g
> + *                             mechanism.
> + *                             @vdev: vdpa device
> + *                             @desc: irq affinity hint
>   * @set_group_asid:            Set address space identifier for a
>   *                             virtqueue group (optional)
>   *                             @vdev: vdpa device
> @@ -353,6 +360,8 @@ struct vdpa_config_ops {
>                                const struct cpumask *cpu_mask);
>         const struct cpumask *(*get_vq_affinity)(struct vdpa_device *vdev=
,
>                                                  u16 idx);
> +       void (*set_irq_affinity)(struct vdpa_device *vdev,
> +                                struct irq_affinity *desc);
>
>         /* DMA ops */
>         int (*set_map)(struct vdpa_device *vdev, unsigned int asid,
> --
> 2.20.1
>