Re: [PATCH v4 01/17] iommu: Add hwpt_type with user_data for domain_alloc_user op

[Jason Gunthorpe <jgg@nvidia.com>]

On Mon, Sep 25, 2023 at 04:01:55PM +0800, Baolu Lu wrote:
> On 2023/9/25 14:22, Yi Liu wrote:
> > On 2023/9/21 20:10, Baolu Lu wrote:
> > > On 2023/9/21 15:51, Yi Liu wrote:
> > > > +/**
> > > > + * iommu_copy_user_data - Copy iommu driver specific user space data
> > > > + * @dst_data: Pointer to an iommu driver specific user data
> > > > that is defined in
> > > > + *            include/uapi/linux/iommufd.h
> > > > + * @src_data: Pointer to a struct iommu_user_data for user
> > > > space data info
> > > > + * @data_len: Length of current user data structure, i.e.
> > > > sizeof(struct _dst)
> > > > + * @min_len: Initial length of user data structure for backward
> > > > compatibility.
> > > > + *           This should be offsetofend using the last member
> > > > in the user data
> > > > + *           struct that was initially added to
> > > > include/uapi/linux/iommufd.h
> > > > + */
> > > > +static inline int iommu_copy_user_data(void *dst_data,
> > > > +                       const struct iommu_user_data *src_data,
> > > > +                       size_t data_len, size_t min_len)
> > > > +{
> > > > +    if (WARN_ON(!dst_data || !src_data))
> > > > +        return -EINVAL;
> > > > +    if (src_data->len < min_len || data_len < src_data->len)
> > > > +        return -EINVAL;
> > > > +    return copy_struct_from_user(dst_data, data_len,
> > > > +                     src_data->uptr, src_data->len);
> > > > +}
> > > 
> > > I am not sure that I understand the purpose of "min_len" correctly. It
> > > seems like it would always be equal to data_len?
> > 
> > no, it will not be equal to data_len once there is extension in the
> > uAPI structure.
> > 
> > > Or, it means the minimal data length that the iommu driver requires?
> > 
> > it is the minimal data length the uAPI requires. min_len is finalized
> > per the upstream of the first version of the uAPI.
> 
> So, it looks like a constant. Perhaps we should document it in the
> uapi/iommuf.h and avoid using it as a parameter of a helper
> function?

It is per-driver, per-struct, so this is the right way to do it

Jason