Re: [PATCH rdma-next 01/10] RDMA/umem: Add reference counting to ib_umem

[Leon Romanovsky <leon@kernel.org>]

On Wed, Feb 04, 2026 at 01:46:15PM -0400, Jason Gunthorpe wrote:
> On Wed, Feb 04, 2026 at 04:38:22PM +0100, Jiri Pirko wrote:
> 
> > >Generally I would not assign to the driver's umem storage until the
> > >creation is completed to avoid this. ie it stays null until committed.
> > >
> > >But looking at mlx5 that looks like quite a maze there.. Yikes..
> > >
> > >So maybe mlx5 adds some NULL assignments on its error paths and less
> > >convoluted drivers can use a simpler option?
> > 
> > How about we have:
> > 	int (*create_cq_umem)(struct ib_cq *cq,
> > 			      const struct ib_cq_init_attr *attr,
> > 			      struct ib_umem **umem,
> >                                              ^^
> > 
> > 			      struct uverbs_attr_bundle *attrs);
> > 
> > And instead of taking ref in the callee we just do *umem = NULL? :S
> > 
> > This would help to cover the error path vs destroy path differences,
> > Tt could also be used to make sure the op consumed all umems; all
> > should be NULLed on success.
> > 
> > Makes sense?
> 
> I'm ok with that, though never seen the pattern before. If the core
> code fails on !NULL it would be pretty hard to use it wrong.

I'm less excited to see unique coding patterns. We should rather invest
time in effort to layer everything correctly.

Thanks

> 
> Jason
>