[PATCH net-next, 3/3] net: mana: Reuse XDP dropped page
From: Haiyang Zhang <haiyangz@microsoft.com>
Date: 2022-01-29 02:04:58
Also in:
linux-hyperv, lkml
Subsystem:
hyper-v/azure core and drivers, networking drivers, the rest · Maintainers:
"K. Y. Srinivasan", Haiyang Zhang, Wei Liu, Dexuan Cui, Long Li, Andrew Lunn, "David S. Miller", Eric Dumazet, Jakub Kicinski, Paolo Abeni, Linus Torvalds
Reuse the dropped page in RX path to save page allocation overhead. Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com> --- drivers/net/ethernet/microsoft/mana/mana.h | 1 + drivers/net/ethernet/microsoft/mana/mana_en.c | 15 +++++++++++++-- 2 files changed, 14 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/microsoft/mana/mana.h b/drivers/net/ethernet/microsoft/mana/mana.h
index 8ead960f898d..d36405af9432 100644
--- a/drivers/net/ethernet/microsoft/mana/mana.h
+++ b/drivers/net/ethernet/microsoft/mana/mana.h@@ -310,6 +310,7 @@ struct mana_rxq { struct bpf_prog __rcu *bpf_prog; struct xdp_rxq_info xdp_rxq; + struct page *xdp_save_page; /* MUST BE THE LAST MEMBER: * Each receive buffer has an associated mana_recv_buf_oob.
diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c
index 12067bf5b7d6..69e791e6abc4 100644
--- a/drivers/net/ethernet/microsoft/mana/mana_en.c
+++ b/drivers/net/ethernet/microsoft/mana/mana_en.c@@ -1059,7 +1059,9 @@ static void mana_rx_skb(void *buf_va, struct mana_rxcomp_oob *cqe, u64_stats_update_end(&rx_stats->syncp); drop: - free_page((unsigned long)buf_va); + WARN_ON_ONCE(rxq->xdp_save_page); + rxq->xdp_save_page = virt_to_page(buf_va); + ++ndev->stats.rx_dropped; return;
@@ -1116,7 +1118,13 @@ static void mana_process_rx_cqe(struct mana_rxq *rxq, struct mana_cq *cq, rxbuf_oob = &rxq->rx_oobs[curr]; WARN_ON_ONCE(rxbuf_oob->wqe_inf.wqe_size_in_bu != 1); - new_page = alloc_page(GFP_ATOMIC); + /* Reuse XDP dropped page if available */ + if (rxq->xdp_save_page) { + new_page = rxq->xdp_save_page; + rxq->xdp_save_page = NULL; + } else { + new_page = alloc_page(GFP_ATOMIC); + } if (new_page) { da = dma_map_page(dev, new_page, XDP_PACKET_HEADROOM, rxq->datasize,
@@ -1403,6 +1411,9 @@ static void mana_destroy_rxq(struct mana_port_context *apc, mana_deinit_cq(apc, &rxq->rx_cq); + if (rxq->xdp_save_page) + __free_page(rxq->xdp_save_page); + for (i = 0; i < rxq->num_rx_buf; i++) { rx_oob = &rxq->rx_oobs[i];
--
2.25.1