On Sat, 12 Oct 2024 21:49:05 +0800
Yi Liu <yi.l.liu@xxxxxxxxx> wrote:
> On 2024/10/1 20:11, Jason Gunthorpe wrote:
> > On Mon, Sep 30, 2024 at 07:55:08AM +0000, Tian, Kevin wrote:
> >
> >>> +struct vfio_device_pasid_attach_iommufd_pt {
> >>> + __u32 argsz;
> >>> + __u32 flags;
> >>> + __u32 pasid;
> >>> + __u32 pt_id;
> >>> +};
> >>> +
> >>> +#define VFIO_DEVICE_PASID_ATTACH_IOMMUFD_PT _IO(VFIO_TYPE,
> >>> VFIO_BASE + 21)
> >>
> >> Not sure whether this was discussed before. Does it make sense
> >> to reuse the existing VFIO_DEVICE_ATTACH_IOMMUFD_PT
> >> by introducing a new pasid field and a new flag bit?
> >
> > Maybe? I don't have a strong feeling either way.
> >
> > There is somewhat less code if you reuse the ioctl at least
>
> I had a rough memory that I was suggested to add a separate ioctl for
> PASID. Let's see Alex's opinion.
I don't recall any previous arguments for separate ioctls, but it seems
to make a lot of sense to me to extend the existing ioctls with a flag
to indicate pasid cscope and id. Thanks,
Alex
