On Thu, May 03, 2018 at 04:36:57PM +0300, Leon Romanovsky wrote:
> From: Matan Barak <matanb@xxxxxxxxxxxx>
>
> When CONFIG_INFINIBAND_USER_ACCESS is disabled, we need to declare
> uverbs_idr_class and uverbs_fd_class. This is mandatory, as provider
> drivers could use UVERBS_ATTR_FD and UVERBS_ATTR_IDR macros.
> Exporting them as all zeros structures to allow compiling provider
> drivers without being dependent on ib_uverbs.
>
> Signed-off-by: Matan Barak <matanb@xxxxxxxxxxxx>
> Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxxxx>
> ---
> drivers/infiniband/core/rdma_core.c | 2 ++
> include/rdma/uverbs_types.h | 5 +++++
> 2 files changed, 7 insertions(+)
>
> diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c
> index a6e904973ba8..8035a0a7564c 100644
> --- a/drivers/infiniband/core/rdma_core.c
> +++ b/drivers/infiniband/core/rdma_core.c
> @@ -611,6 +611,7 @@ const struct uverbs_obj_type_class uverbs_idr_class = {
> */
> .needs_kfree_rcu = true,
> };
> +EXPORT_SYMBOL(uverbs_idr_class);
>
> static void _uverbs_close_fd(struct ib_uobject_file *uobj_file)
> {
> @@ -719,6 +720,7 @@ const struct uverbs_obj_type_class uverbs_fd_class = {
> .remove_commit = remove_commit_fd_uobject,
> .needs_kfree_rcu = false,
> };
> +EXPORT_SYMBOL(uverbs_fd_class);
>
> struct ib_uobject *uverbs_get_uobject_from_context(const struct uverbs_obj_type *type_attrs,
> struct ib_ucontext *ucontext,
> diff --git a/include/rdma/uverbs_types.h b/include/rdma/uverbs_types.h
> index cc04ec65588d..7c6d1e99c010 100644
> --- a/include/rdma/uverbs_types.h
> +++ b/include/rdma/uverbs_types.h
> @@ -147,8 +147,13 @@ struct uverbs_obj_fd_type {
> int flags;
> };
>
> +#if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS)
> extern const struct uverbs_obj_type_class uverbs_idr_class;
> extern const struct uverbs_obj_type_class uverbs_fd_class;
> +#else
> +static const struct uverbs_obj_type_class uverbs_idr_class = {};
> +static const struct uverbs_obj_type_class uverbs_fd_class = {};
> +#endif
Hurm.
This doesn't make much sense, if there is no user access then we
shouldn't be building specs at all, or the _class should continue to
be defined?
I don't want to build specs with pointers to bogus data, that seems
like a great way to create runtime bugs.
Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
