Re: [PATCH rdma-next v11 6/6] RDMA/bnxt_re: Support application specific CQs

[Jason Gunthorpe <jgg@ziepe.ca>]

On Tue, Feb 10, 2026 at 10:29:39PM +0530, Sriharsha Basavapatna wrote:
> diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
> index dc73f0072528..04588b4f79c0 100644
> --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
> +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
> @@ -3372,6 +3372,9 @@ static int bnxt_re_setup_sginfo(struct bnxt_re_dev *rdev,
>  	return 0;
>  }
>  
> +static u64 bnxt_re_cq_cmask_supported = (BNXT_RE_CQ_TOGGLE_PAGE_SUPPORT |
> +					 BNXT_RE_CQ_APP_ALLOC_ENABLE);

Don't mix req and resp flags together!!!

Also don't make a global static variable, just pass the flags directly
to the helper.

>  int bnxt_re_create_cq_umem(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
>  			   struct ib_umem *umem, struct uverbs_attr_bundle *attrs)
>  {
> @@ -3382,6 +3385,7 @@ int bnxt_re_create_cq_umem(struct ib_cq *ibcq, const struct ib_cq_init_attr *att
>  		rdma_udata_to_drv_context(udata, struct bnxt_re_ucontext, ib_uctx);
>  	struct bnxt_qplib_dev_attr *dev_attr = rdev->dev_attr;
>  	struct bnxt_qplib_chip_ctx *cctx;
> +	struct bnxt_re_cq_req req = {};

No need to zero memory being passed to ib_copy_validate_udata_in_cm(),
it always writes to the whole struct.

> diff --git a/include/uapi/rdma/bnxt_re-abi.h b/include/uapi/rdma/bnxt_re-abi.h
> index 1f7685665db1..26eeb78193fa 100644
> --- a/include/uapi/rdma/bnxt_re-abi.h
> +++ b/include/uapi/rdma/bnxt_re-abi.h
> @@ -103,10 +103,12 @@ struct bnxt_re_pd_resp {
>  struct bnxt_re_cq_req {
>  	__aligned_u64 cq_va;
>  	__aligned_u64 cq_handle;
> +	__aligned_u64 comp_mask;
>  };
>  
>  enum bnxt_re_cq_mask {
>  	BNXT_RE_CQ_TOGGLE_PAGE_SUPPORT = 0x1,
> +	BNXT_RE_CQ_APP_ALLOC_ENABLE = 0x2,
>  };

The req/resp comp masks should be kept separate.

The name should be much more specific "BNXT_RE_CQ_EXACT_CQE" or something.

This is all alot better now, thanks

Jhason