Re: [PATCH net-next] mlx4: Better use of order-0 pages in RX path

On Mon, 2017-03-13 at 16:40 -0700, Alexei Starovoitov wrote:

that's not how it works. It's a job of submitter to prove
that additional code doesn't cause regressions especially
when there are legitimate concerns.

This test was moved out of the mlx4_en_prepare_rx_desc() section into
the XDP_TX code path.


        if (ring->page_cache.index > 0) {
                /* XDP uses a single page per frame */
                if (!frags->page) {
                        ring->page_cache.index--;
                        frags->page = ring->page_cache.buf[ring->page_cache.index].page;
                        frags->dma  = ring->page_cache.buf[ring->page_cache.index].dma;
                }
                frags->page_offset = XDP_PACKET_HEADROOM;
                rx_desc->data[0].addr = cpu_to_be64(frags->dma +
                                                    XDP_PACKET_HEADROOM);
                return 0;
        }

Can you check again your claim, because I see no additional cost
for XDP_TX.

In fact I removed from the other paths (which are equally important I
believe) a test that had no use, so everybody should be happy.