Re: [PATCH rdma-next v3 08/17] RDMA/efa: Use ib_umem_get_cq_buf() for user CQ buffer

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, May 06, 2026 at 04:32:29PM +0200, Jiri Pirko wrote:
> Wed, May 06, 2026 at 03:51:39PM +0200, jgg@ziepe.ca wrote:
> >On Mon, May 04, 2026 at 03:57:22PM +0200, Jiri Pirko wrote:
> >
> >> @@ -1172,26 +1174,29 @@ int efa_create_user_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
> >>  	cq->ucontext = ucontext;
> >>  	cq->size = PAGE_ALIGN(cmd.cq_entry_size * entries * cmd.num_sub_cqs);
> >>  
> >> -	if (ibcq->umem) {
> >> -		if (ibcq->umem->length < cq->size) {
> >> -			ibdev_dbg(&dev->ibdev, "External memory too small\n");
> >> -			err = -EINVAL;
> >> -			goto err_out;
> >> -		}
> >> +	umem = ib_umem_get_cq_buf(ibcq->device, udata, cq->size,
> >> +				  IB_ACCESS_LOCAL_WRITE);
> >> +	if (IS_ERR(umem)) {
> >> +		err = PTR_ERR(umem);
> >> +		goto err_out;
> >> +	}
> >> +
> >> +	cq->umem = umem;
> >>  
> >> -		if (!ib_umem_is_contiguous(ibcq->umem)) {
> >> +	if (umem) {
> >> +		if (!ib_umem_is_contiguous(umem)) {
> >
> >This is a little funny, I think umem should not be NULL?
> 
> Yes it is null when there are no CQ umem attrs present. That is
> perfectly fine flow.

Oh, it sets up a kernel created page, how unusual.

> I was thinking about the ib_umem_get() return value scheme and went back
> and forth multiple times, I converged to PTR_ERR in case of something
> went wrong and null in case of attrs missing, which I believe is the
> best option. It is documented in ib_umem_get kdoc.

Hmm, it makes the function tricky to use but as long as it is only
this function and only efa is using it things should be fine

Jason