On Thu, Jan 03, 2019 at 10:05:57AM +0200, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@xxxxxxxxxxxx>
>
> Add new general helper to get restrack entry given by ID and
> their respective type.
>
> Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxxxx>
> drivers/infiniband/core/restrack.c | 13 +++++++++++++
> include/rdma/restrack.h | 11 +++++++++++
> 2 files changed, 24 insertions(+)
>
> diff --git a/drivers/infiniband/core/restrack.c b/drivers/infiniband/core/restrack.c
> index 473422da0856..d65758a3a7b6 100644
> +++ b/drivers/infiniband/core/restrack.c
> @@ -249,6 +249,19 @@ int __must_check rdma_restrack_get(struct rdma_restrack_entry *res)
> }
> EXPORT_SYMBOL(rdma_restrack_get);
>
> +struct rdma_restrack_entry *
> +rdma_restrack_get_byid(struct rdma_restrack_root *rt,
> + enum rdma_restrack_type type, u32 id)
> +{
> + struct rdma_restrack_entry *res;
> +
> + res = xa_load(&rt->xa[type], id);
the previous patch did:
+ id = res_to_id(res);
+ ret = xa_insert(&dev->res.xa[res->type], id, res, GFP_KERNEL);
Where res_to_id() returns a pointer casted to an unsigned long - so
the use of 'u32 id' is wrong here.
Also, it seems to expose kernel pointers to user space, which is a
security no-no.
I think this needs actual device-unique IDs for all objects, no use of
pointers.
Probably better to use the xarray in the IDR mode and have it generate
unique ID's for all the cases where a HW ID is not already present.
Jason
