On Wed, Feb 06, 2019 at 01:43:11PM +0200, Shamir Rabinovitch wrote: > On Mon, Jan 28, 2019 at 12:01:06PM -0700, Jason Gunthorpe wrote: > > On Mon, Jan 28, 2019 at 12:12:23PM +0200, Shamir Rabinovitch wrote: > > > Helper function to get driver's context out of ib_udata wrapped > > > in uverbs_attr_bundle for uer objects or NULL for kernel objects. > > > > > > Signed-off-by: Shamir Rabinovitch <shamir.rabinovitch@xxxxxxxxxx> > > > include/rdma/ib_verbs.h | 14 ++++++++++++++ > > > 1 file changed, 14 insertions(+) > > > > > > diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h > > > index 94b6e1dd4dab..fb4d078d6db3 100644 > > > +++ b/include/rdma/ib_verbs.h > > > @@ -4264,4 +4264,18 @@ static inline struct ib_device *rdma_device_to_ibdev(struct device *device) > > > */ > > > #define rdma_device_to_drv_device(dev, drv_dev_struct, ibdev_member) \ > > > container_of(rdma_device_to_ibdev(dev), drv_dev_struct, ibdev_member) > > > + > > > +/** > > > + * rdma_udata_to_drv_context - Helper macro to get the driver's context out of > > > + * ib_udata which is embadded in uverbs_attr_bundle. > > > + * > > > + * NOTE: This macro return NULL for kernel objects and valid driver context for > > > + * user objects. > > > > Actually at this point I think this just can't fail if udata is set, > > so lets just say that: > > > > /* > > * If udata is not NULL this cannot fail. Otherwise a NULL udata will result > > * in a NULL ucontext pointer. Callers can rely on !NULL ucontext to show the > > * op is being called as part of a user system call. > > */ > > #define rdma_udata_to_drv_context(udata, drv_dev_struct, member) \ > > (udata ? container_of(container_of(udata, struct uverbs_attr_bundle, \ > > driver_udata) \ > > ->context, \ > > drv_dev_struct, member) : \ > > (drv_dev_struct *)NULL) > > > > And get rid of the rdma_get_ucontext helper function, umem can just > > inspect udata->context itself and fail with EIO if NULL. > > > > Jason > > OK. Change accordingly & skip previous mail. Yes, just don't miss the comment from CH to try as much as reasonable without major invasion to not call rdma_udata_to_drv_context if udata is NULL Jason
