Re: [PATCH rdma-next v4 06/16] RDMA/uverbs: Push out CQ buffer umem processing into a helper

[Jason Gunthorpe <jgg@ziepe.ca>]

On Thu, May 07, 2026 at 02:52:21PM +0200, Jiri Pirko wrote:

> +static int uverbs_create_cq_get_buffer_desc(struct uverbs_attr_bundle *attrs,
> +					    struct ib_uverbs_buffer_desc *desc)
> +{
> +	struct ib_device *ib_dev = attrs->context->device;
> +	int ret;
> +
> +	if (uverbs_attr_is_valid(attrs,
> UVERBS_ATTR_CREATE_CQ_BUFFER_VA)) {

I know this is just moving code, but I've always disliked this
function. I learned a trick using a case statement for this recently:

	u32 present_attrs = 0;
	if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_CREATE_CQ_BUFFER_VA))
		present_attrs |= BIT(UVERBS_ATTR_CREATE_CQ_BUFFER_VA);
	if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_CREATE_CQ_BUFFER_LENGTH))
		present_attrs |= BIT(UVERBS_ATTR_CREATE_CQ_BUFFER_LENGTH);
	if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_CREATE_CQ_BUFFER_FD))
		present_attrs |= BIT(UVERBS_ATTR_CREATE_CQ_BUFFER_FD);
	if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_CREATE_CQ_BUFFER_OFFSET))
		present_attrs |= BIT(UVERBS_ATTR_CREATE_CQ_BUFFER_OFFSET);

	switch (present_attrs) {
	case 0:
		return -ENODATA;
	case BIT(UVERBS_ATTR_CREATE_CQ_BUFFER_VA) |
		BIT(UVERBS_ATTR_CREATE_CQ_BUFFER_LENGTH):
[..]
		return 0;
	case BIT(UVERBS_ATTR_CREATE_CQ_BUFFER_FD) |
		BIT(UVERBS_ATTR_CREATE_CQ_BUFFER_OFFSET) |
		BIT(UVERBS_ATTR_CREATE_CQ_BUFFER_LENGTH):
[..]
		return 0;
	default:
		return -EINVAL;
	}

No need to build the complex tests to check in each branch if the
other branch attributes are presented.

Jason