On Thu, Jul 13, 2017 at 05:52:06PM +0000, Bart Van Assche wrote:
> On Thu, 2017-07-13 at 11:20 -0600, Jason Gunthorpe wrote:
> > On Thu, Jul 13, 2017 at 12:51:16AM -0700, Christoph Hellwig wrote:
> > > > if (qp->qp_cap_cache & MLX5_RX_CSUM_VALID)
> > > > - wc->wc_flags |= (!!(cqe->hds_ip_ext & MLX5_CQE_L4_OK) &
> > > > - !!(cqe->hds_ip_ext & MLX5_CQE_L3_OK) &
> > > > - (get_cqe_l3_hdr_type(cqe) ==
> > > > - MLX5_CQE_L3_HDR_TYPE_IPV4)) <<
> > > > - IBV_WC_IP_CSUM_OK_SHIFT;
> > > > + wc->wc_flags |=
> > > > + ((bool)(cqe->hds_ip_ext & MLX5_CQE_L4_OK) &
> > > > + (bool)(cqe->hds_ip_ext & MLX5_CQE_L3_OK) &
> > > > + (get_cqe_l3_hdr_type(cqe) ==
> > > > + MLX5_CQE_L3_HDR_TYPE_IPV4))
> > > > + << IBV_WC_IP_CSUM_OK_SHIFT;
> > >
> > > Meh. This code is complete crap. Please factor it out into a little
> > > helper that mere humans can read first. And then replace the odd ^ used
> > > as && with proper if constructs and all should make much more sense.
> >
> > As far as I could make out, this ugly thing is designed like this for
> > performance.
>
> Hello Jason,
>
> How about using an expression like the below to avoid that branches get inserted
> for testing the MLX5_CQE_L4_OK and MLX5_CQE_L3_OK flags?
>
> (cqe->hds_ip_ext & (MLX5_CQE_L4_OK | MLX5_CQE_L3_OK)) == MLX5_CQE_L4_OK | MLX5_CQE_L3_OK
Yes, thanks, that seems a reasonable middle ground:
diff --git a/providers/mlx5/cq.c b/providers/mlx5/cq.c
index b845127de937d0..0a6274e6878d3c 100644
--- a/providers/mlx5/cq.c
+++ b/providers/mlx5/cq.c
@@ -182,6 +182,15 @@ static inline int handle_responder_lazy(struct mlx5_cq *cq, struct mlx5_cqe64 *c
return err;
}
+/* Returns IBV_WC_IP_CSUM_OK or 0 */
+static inline int get_csum_ok(struct mlx5_cqe64 *cqe)
+{
+ return (((cqe->hds_ip_ext & (MLX5_CQE_L4_OK | MLX5_CQE_L3_OK)) ==
+ (MLX5_CQE_L4_OK | MLX5_CQE_L3_OK)) &
+ (get_cqe_l3_hdr_type(cqe) == MLX5_CQE_L3_HDR_TYPE_IPV4))
+ << IBV_WC_IP_CSUM_OK_SHIFT;
+}
+
static inline int handle_responder(struct ibv_wc *wc, struct mlx5_cqe64 *cqe,
struct mlx5_resource *cur_rsc, struct mlx5_srq *srq)
{
@@ -206,12 +215,7 @@ static inline int handle_responder(struct ibv_wc *wc, struct mlx5_cqe64 *cqe,
if (likely(cur_rsc->type == MLX5_RSC_TYPE_QP)) {
wq = &qp->rq;
if (qp->qp_cap_cache & MLX5_RX_CSUM_VALID)
- wc->wc_flags |=
- ((bool)(cqe->hds_ip_ext & MLX5_CQE_L4_OK) &
- (bool)(cqe->hds_ip_ext & MLX5_CQE_L3_OK) &
- (get_cqe_l3_hdr_type(cqe) ==
- MLX5_CQE_L3_HDR_TYPE_IPV4))
- << IBV_WC_IP_CSUM_OK_SHIFT;
+ wc->wc_flags |= get_csum_ok(cqe);
} else {
wq = &(rsc_to_mrwq(cur_rsc)->rq);
}
@@ -1106,11 +1110,7 @@ static inline int mlx5_cq_read_wc_flags(struct ibv_cq_ex *ibcq)
int wc_flags = 0;
if (cq->flags & MLX5_CQ_FLAGS_RX_CSUM_VALID)
- wc_flags = ((bool)(cq->cqe64->hds_ip_ext & MLX5_CQE_L4_OK) &
- (bool)(cq->cqe64->hds_ip_ext & MLX5_CQE_L3_OK) &
- (get_cqe_l3_hdr_type(cq->cqe64) ==
- MLX5_CQE_L3_HDR_TYPE_IPV4))
- << IBV_WC_IP_CSUM_OK_SHIFT;
+ wc_flags = get_csum_ok(cq->cqe64);
switch (mlx5dv_get_cqe_opcode(cq->cqe64)) {
case MLX5_CQE_RESP_WR_IMM:
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
