MIME-Version: 1.0
References: <20230302113421.174582-1-sgarzare@redhat.com> <20230302113421.174582-2-sgarzare@redhat.com>
In-Reply-To: <20230302113421.174582-2-sgarzare@redhat.com>
From:   Jason Wang <jasowang@redhat.com>
Date:   Tue, 14 Mar 2023 11:39:42 +0800
Message-ID: <CACGkMEv24Zw-OUbBBSne21pF7=4XCZ6JGj7Y_cC7cMFYTjbF1Q@mail.gmail.com>
Subject: Re: [PATCH v2 1/8] vdpa: add bind_mm/unbind_mm callbacks
To:     Stefano Garzarella <sgarzare@redhat.com>
Cc:     virtualization@lists.linux-foundation.org,
        Andrey Zhadchenko <andrey.zhadchenko@virtuozzo.com>,
        eperezma@redhat.com, netdev@vger.kernel.org, stefanha@redhat.com,
        linux-kernel@vger.kernel.org,
        "Michael S. Tsirkin" <mst@redhat.com>, kvm@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Mar 2, 2023 at 7:34=E2=80=AFPM Stefano Garzarella <sgarzare@redhat.=
com> wrote:
>
> These new optional callbacks is used to bind/unbind the device to
> a specific address space so the vDPA framework can use VA when
> these callbacks are implemented.
>
> Suggested-by: Jason Wang <jasowang@redhat.com>
> Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
> ---

One thing that came into my mind is that after this commit:

commit 5ce995f313ce56c0c62425c3ddc37c5c50fc33db
Author: Jason Wang <jasowang@redhat.com>
Date:   Fri May 29 16:02:59 2020 +0800

    vhost: use mmgrab() instead of mmget() for non worker device

    For the device that doesn't use vhost worker and use_mm(), mmget() is
    too heavy weight and it may brings troubles for implementing mmap()
    support for vDPA device.

We don't hold the address space after this commit, so the userspace
mapping could be invalid if the owner exits?

Thanks

>
> Notes:
>     v2:
>     - removed `struct task_struct *owner` param (unused for now, maybe
>       useful to support cgroups) [Jason]
>     - add unbind_mm callback [Jason]
>
>  include/linux/vdpa.h | 10 ++++++++++
>  1 file changed, 10 insertions(+)
>
> diff --git a/include/linux/vdpa.h b/include/linux/vdpa.h
> index 43f59ef10cc9..369c21394284 100644
> --- a/include/linux/vdpa.h
> +++ b/include/linux/vdpa.h
> @@ -290,6 +290,14 @@ struct vdpa_map_file {
>   *                             @vdev: vdpa device
>   *                             @idx: virtqueue index
>   *                             Returns pointer to structure device or er=
ror (NULL)
> + * @bind_mm:                   Bind the device to a specific address spa=
ce
> + *                             so the vDPA framework can use VA when thi=
s
> + *                             callback is implemented. (optional)
> + *                             @vdev: vdpa device
> + *                             @mm: address space to bind
> + * @unbind_mm:                 Unbind the device from the address space
> + *                             bound using the bind_mm callback. (option=
al)
> + *                             @vdev: vdpa device
>   * @free:                      Free resources that belongs to vDPA (opti=
onal)
>   *                             @vdev: vdpa device
>   */
> @@ -351,6 +359,8 @@ struct vdpa_config_ops {
>         int (*set_group_asid)(struct vdpa_device *vdev, unsigned int grou=
p,
>                               unsigned int asid);
>         struct device *(*get_vq_dma_dev)(struct vdpa_device *vdev, u16 id=
x);
> +       int (*bind_mm)(struct vdpa_device *vdev, struct mm_struct *mm);
> +       void (*unbind_mm)(struct vdpa_device *vdev);
>
>         /* Free device resources */
>         void (*free)(struct vdpa_device *vdev);
> --
> 2.39.2
>