Re: [PATCH v17 Kernel 6/7] vfio iommu: Adds flag to indicate dirty pages tracking capability support

[Alex Williamson <alex.williamson@redhat.com>]

On Mon, 30 Mar 2020 22:20:43 +0530
Kirti Wankhede <kwankhede@nvidia.com> wrote:

> Flag VFIO_IOMMU_INFO_DIRTY_PGS in VFIO_IOMMU_GET_INFO indicates that driver
> support dirty pages tracking.
> 
> Signed-off-by: Kirti Wankhede <kwankhede@nvidia.com>
> Reviewed-by: Neo Jia <cjia@nvidia.com>
> ---
>  drivers/vfio/vfio_iommu_type1.c | 3 ++-
>  include/uapi/linux/vfio.h       | 5 +++--
>  2 files changed, 5 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> index 266550bd7307..9fe12b425976 100644
> --- a/drivers/vfio/vfio_iommu_type1.c
> +++ b/drivers/vfio/vfio_iommu_type1.c
> @@ -2390,7 +2390,8 @@ static long vfio_iommu_type1_ioctl(void *iommu_data,
>  			info.cap_offset = 0; /* output, no-recopy necessary */
>  		}
>  
> -		info.flags = VFIO_IOMMU_INFO_PGSIZES;
> +		info.flags = VFIO_IOMMU_INFO_PGSIZES |
> +			     VFIO_IOMMU_INFO_DIRTY_PGS;
>  
>  		info.iova_pgsizes = vfio_pgsize_bitmap(iommu);
>  
> diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h
> index e3cbf8b78623..0fe7c9a6f211 100644
> --- a/include/uapi/linux/vfio.h
> +++ b/include/uapi/linux/vfio.h
> @@ -985,8 +985,9 @@ struct vfio_device_feature {
>  struct vfio_iommu_type1_info {
>  	__u32	argsz;
>  	__u32	flags;
> -#define VFIO_IOMMU_INFO_PGSIZES (1 << 0)	/* supported page sizes info */
> -#define VFIO_IOMMU_INFO_CAPS	(1 << 1)	/* Info supports caps */
> +#define VFIO_IOMMU_INFO_PGSIZES   (1 << 0) /* supported page sizes info */
> +#define VFIO_IOMMU_INFO_CAPS      (1 << 1) /* Info supports caps */
> +#define VFIO_IOMMU_INFO_DIRTY_PGS (1 << 2) /* supports dirty page tracking */
>  	__u64	iova_pgsizes;	/* Bitmap of supported page sizes */
>  	__u32   cap_offset;	/* Offset within info struct of first cap */
>  };


As I just mentioned in my reply to Yan, I'm wondering if
VFIO_CHECK_EXTENSION would be a better way to expose this.  The
difference is relatively trivial, but currently the only flag
set by VFIO_IOMMU_GET_INFO is to indicate the presence of a field in
the returned structure.  I think this is largely true of other INFO
ioctls within vfio as well and we're already using the
VFIO_CHECK_EXTENSION ioctl to check supported IOMMU models, and IOMMU
cache coherency.  We'd simply need to define a VFIO_DIRTY_PGS_IOMMU
value (9) and return 1 for that case.  Then when we enable support for
dirt pages that can span multiple mappings, we can add a v2 extensions,
or "MULTI" variant of this extension, since it should be backwards
compatible.

The v2/multi version will again require that the user provide a zero'd
bitmap, but I don't think that should be a problem as part of the
definition of that version (we won't know if the user is using v1 or
v2, but a v1 user should only retrieve bitmaps that exactly match
existing mappings, where all bits will be written).  Thanks,

Alex