[Intel-gfx] [PATCH v13 5/7] vfio: ABI for mdev display dma-buf operation
Zhang, Tina
tina.zhang at intel.com
Mon Aug 7 08:11:43 UTC 2017
After going through the previous discussions, here are some summaries may be related to the current discussion:
1. How does user mode figure the device capabilities between region and dma-buf?
VFIO_DEVICE_GET_REGION_INFO could tell if the mdev supports region case.
Otherwise, the mdev supports dma-buf.
2. For dma-buf, how to differentiate unsupported vs not initialized?
For dma-buf, when the mdev doesn't support some arguments, -EINVAL will be returned. And -errno will return when meeting other failures, like -ENOMEM.
If the mdev is not initialized, there won't be any returned err. Just zero all the fields in structure vfio_device_gfx_plane_info.
3. The id field in structure vfio_device_gfx_plane_info
So far we haven't figured out the usage of this field for dma-buf usage. So, this field is changed to "region_index" and only used for region usage.
In previous discussions, we thought this "id" field might be used for both dma-buf and region usages.
So, we proposed some ways, like adding flags field to the structure. Another way to do it was to add this:
enum vfio_device_gfx_type {
VFIO_DEVICE_GFX_NONE,
VFIO_DEVICE_GFX_DMABUF,
VFIO_DEVICE_GFX_REGION,
};
struct vfio_device_gfx_query_caps {
__u32 argsz;
__u32 flags;
enum vfio_device_gfx_type;
};
Obviously, we don't need to consider this again, unless we find the "region_index" means something for dma-buf usage.
Thanks.
BR,
Tina
> -----Original Message-----
> From: Zhang, Tina
> Sent: Monday, August 7, 2017 11:23 AM
> To: 'Alex Williamson' <alex.williamson at redhat.com>
> Cc: Tian, Kevin <kevin.tian at intel.com>; intel-gfx at lists.freedesktop.org; dri-
> devel at lists.freedesktop.org; kwankhede at nvidia.com; kraxel at redhat.com;
> intel-gvt-dev at lists.freedesktop.org; Wang, Zhi A <zhi.a.wang at intel.com>; Lv,
> Zhiyuan <zhiyuan.lv at intel.com>
> Subject: RE: [PATCH v13 5/7] vfio: ABI for mdev display dma-buf operation
>
>
>
> > -----Original Message-----
> > From: dri-devel [mailto:dri-devel-bounces at lists.freedesktop.org] On
> > Behalf Of Alex Williamson
> > Sent: Thursday, August 3, 2017 10:43 PM
> > To: Zhang, Tina <tina.zhang at intel.com>
> > Cc: Tian, Kevin <kevin.tian at intel.com>;
> > intel-gfx at lists.freedesktop.org; dri- devel at lists.freedesktop.org;
> > kwankhede at nvidia.com; kraxel at redhat.com;
> > intel-gvt-dev at lists.freedesktop.org; Wang, Zhi A
> > <zhi.a.wang at intel.com>; Lv, Zhiyuan <zhiyuan.lv at intel.com>
> > Subject: Re: [PATCH v13 5/7] vfio: ABI for mdev display dma-buf
> > operation
> >
> > On Thu, 3 Aug 2017 07:08:15 +0000
> > "Zhang, Tina" <tina.zhang at intel.com> wrote:
> >
> > > > -----Original Message-----
> > > > From: Alex Williamson [mailto:alex.williamson at redhat.com]
> > > > Sent: Thursday, August 3, 2017 11:38 AM
> > > > To: Zhang, Tina <tina.zhang at intel.com>
> > > > Cc: Tian, Kevin <kevin.tian at intel.com>;
> > > > intel-gfx at lists.freedesktop.org; dri- devel at lists.freedesktop.org;
> > > > kwankhede at nvidia.com; kraxel at redhat.com;
> > > > intel-gvt-dev at lists.freedesktop.org; Wang, Zhi A
> > > > <zhi.a.wang at intel.com>; Lv, Zhiyuan <zhiyuan.lv at intel.com>
> > > > Subject: Re: [PATCH v13 5/7] vfio: ABI for mdev display dma-buf
> > > > operation
> > > >
> > > > On Thu, 3 Aug 2017 03:17:09 +0000
> > > > "Zhang, Tina" <tina.zhang at intel.com> wrote:
> > > >
> > > > > > -----Original Message-----
> > > > > > From: dri-devel
> > > > > > [mailto:dri-devel-bounces at lists.freedesktop.org]
> > > > > > On Behalf Of Alex Williamson
> > > > > > Sent: Wednesday, August 2, 2017 5:18 AM
> > > > > > To: Zhang, Tina <tina.zhang at intel.com>
> > > > > > Cc: Tian, Kevin <kevin.tian at intel.com>; kraxel at redhat.com;
> > > > > > intel- gfx at lists.freedesktop.org; Wang, Zhi A
> > > > > > <zhi.a.wang at intel.com>; kwankhede at nvidia.com;
> > > > > > dri-devel at lists.freedesktop.org; intel-gvt-
> > > > > > dev at lists.freedesktop.org; Lv, Zhiyuan <zhiyuan.lv at intel.com>
> > > > > > Subject: Re: [PATCH v13 5/7] vfio: ABI for mdev display
> > > > > > dma-buf operation
> > > > > >
> > > > > > On Tue, 25 Jul 2017 17:28:18 +0800 Tina Zhang
> > > > > > <tina.zhang at intel.com> wrote:
> > > > > >
> > > > > > > Add VFIO_DEVICE_QUERY_GFX_PLANE ioctl command to let user
> > mode
> > > > > > > query and get the plan and its related information.
> > > > > > >
> > > > > > > The dma-buf's life cycle is handled by user mode and tracked by
> kernel.
> > > > > > > The returned fd in struct vfio_device_query_gfx_plane can be
> > > > > > > a new fd or an old fd of a re-exported dma-buf. Host User
> > > > > > > mode can check the value of fd and to see if it needs to
> > > > > > > create new resource according to the new fd or just use the
> > > > > > > existed resource related to the old
> > > > fd.
> > > > > > >
> > > > > > > Signed-off-by: Tina Zhang <tina.zhang at intel.com>
> > > > > > > ---
> > > > > > > include/uapi/linux/vfio.h | 28 ++++++++++++++++++++++++++++
> > > > > > > 1 file changed, 28 insertions(+)
> > > > > > >
> > > > > > > diff --git a/include/uapi/linux/vfio.h
> > > > > > > b/include/uapi/linux/vfio.h index ae46105..827a230 100644
> > > > > > > --- a/include/uapi/linux/vfio.h
> > > > > > > +++ b/include/uapi/linux/vfio.h
> > > > > > > @@ -502,6 +502,34 @@ struct vfio_pci_hot_reset {
> > > > > > >
> > > > > > > #define VFIO_DEVICE_PCI_HOT_RESET _IO(VFIO_TYPE,
> VFIO_BASE +
> > > > > > 13)
> > > > > > >
> > > > > > > +/**
> > > > > > > + * VFIO_DEVICE_QUERY_GFX_PLANE - _IOW(VFIO_TYPE,
> VFIO_BASE
> > +
> > > > 14,
> > > > > > > +struct vfio_device_query_gfx_plane)
> > > > > > > + *
> > > > > > > + * Set the drm_plane_type and retrieve information about
> > > > > > > +the gfx
> > plane.
> > > > > > > + *
> > > > > > > + * Return: 0 on success, -errno on failure.
> > > > > > > + */
> > > > > > > +struct vfio_device_gfx_plane_info {
> > > > > > > + __u32 argsz;
> > > > > > > + __u32 flags;
> > > > > > > + /* in */
> > > > > > > + __u32 drm_plane_type; /* type of plane: DRM_PLANE_TYPE_*
> > */
> > > > > > > + /* out */
> > > > > > > + __u32 drm_format; /* drm format of plane */
> > > > > > > + __u64 drm_format_mod; /* tiled mode */
> > > > > > > + __u32 width; /* width of plane */
> > > > > > > + __u32 height; /* height of plane */
> > > > > > > + __u32 stride; /* stride of plane */
> > > > > > > + __u32 size; /* size of plane in bytes, align on page*/
> > > > > > > + __u32 x_pos; /* horizontal position of cursor plane, upper
> > left corner
> > > > > > in pixels */
> > > > > > > + __u32 y_pos; /* vertical position of cursor plane, upper left
> > corner in
> > > > > > lines*/
> > > > > > > + __u32 region_index;
> > > > > > > + __s32 fd; /* dma-buf fd */
> > > > > >
> > > > > > How do I know which of these is valid, region_index or fd?
> > > > > > Can I ask for one vs the other? What are the errno values to
> > > > > > differentiate unsupported vs not initialized? Is there a "probe"
> > > > > > flag that I can use to determine what the device supports
> > > > > > without allocating
> > > > those resources yet?
> > > > > Dma-buf won't use region_index, which means region_index is zero
> > > > > all the
> > > > time for dma-buf usage.
> > > > > As we discussed, there won't be errno if not initialized, just
> > > > > keep all fields
> > zero.
> > > > > I will add the comments about these in the next version. Thanks
> > > >
> > > > Zero is a valid region index.
> > > If zero is valid, can we find out other invalid value? How about 0xffffffff?
> >
> > Unlikely, but also valid. Isn't this why we have a flags field in the
> > structure, so we don't need to rely on implicit values as invalid.
> > Also, all of the previously discussed usage models needs to be
> > documented, either here in the header or in a Documentation/ file.
> According to the previously discussion, we also have the following propose:
> enum vfio_device_gfx_type {
> VFIO_DEVICE_GFX_NONE,
> VFIO_DEVICE_GFX_DMABUF,
> VFIO_DEVICE_GFX_REGION,
> };
>
> struct vfio_device_gfx_query_caps {
> __u32 argsz;
> __u32 flags;
> enum vfio_device_gfx_type;
> };
> So, we may need to add one more ioctl, e.g. VFIO_DEVICE_QUERY_GFX_CAPS.
> User mode can query this before querying the plan info, and to see which usage
> (dma-buf or region) is supported.
> Does it still make sense?
> Thanks.
So, I think we can rely on VFIO_DEVICE_GET_REGION_INFO to tell user mode whether the region case is using, instead of introducing a new ioctl.
Thanks
Tina
>
> Tina
>
>
> > Thanks,
> >
> > Alex
> > _______________________________________________
> > dri-devel mailing list
> > dri-devel at lists.freedesktop.org
> > https://lists.freedesktop.org/mailman/listinfo/dri-devel
More information about the Intel-gfx
mailing list