netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Saeed Mahameed <saeed@kernel.org>
To: "David S. Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>
Cc: netdev@vger.kernel.org, Tariq Toukan <tariqt@nvidia.com>,
	Maxim Mikityanskiy <maximmi@nvidia.com>,
	Saeed Mahameed <saeedm@nvidia.com>
Subject: [net-next 04/15] net/mlx5e: RX, Test the XDP program existence out of the handler
Date: Thu, 17 Mar 2022 11:54:13 -0700	[thread overview]
Message-ID: <20220317185424.287982-5-saeed@kernel.org> (raw)
In-Reply-To: <20220317185424.287982-1-saeed@kernel.org>

From: Tariq Toukan <tariqt@nvidia.com>

Instead of early return inside mlx5e_xdp_handle(), let the caller check
if an XDP program is loaded.  This allows saving a few unnecessary
function calls and calculations in case !prog.

Performance test: single core, drop packets in iptables
Before: 3,872,504 pps
After:  3,975,628 pps (+2.66%)

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/en/xdp.c  |  5 +-
 .../net/ethernet/mellanox/mlx5/core/en/xdp.h  |  1 +
 .../ethernet/mellanox/mlx5/core/en/xsk/rx.c   |  9 +++-
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   | 49 ++++++++++++-------
 4 files changed, 39 insertions(+), 25 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
index a7f020399370..fcb84971b138 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
@@ -120,15 +120,12 @@ mlx5e_xmit_xdp_buff(struct mlx5e_xdpsq *sq, struct mlx5e_rq *rq,
 
 /* returns true if packet was consumed by xdp */
 bool mlx5e_xdp_handle(struct mlx5e_rq *rq, struct mlx5e_dma_info *di,
+		      struct bpf_prog *prog,
 		      u32 *len, struct xdp_buff *xdp)
 {
-	struct bpf_prog *prog = rcu_dereference(rq->xdp_prog);
 	u32 act;
 	int err;
 
-	if (!prog)
-		return false;
-
 	act = bpf_prog_run_xdp(prog, xdp);
 	switch (act) {
 	case XDP_PASS:
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
index c62f11d7ef6a..850540e94bb4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
@@ -48,6 +48,7 @@
 struct mlx5e_xsk_param;
 int mlx5e_xdp_max_mtu(struct mlx5e_params *params, struct mlx5e_xsk_param *xsk);
 bool mlx5e_xdp_handle(struct mlx5e_rq *rq, struct mlx5e_dma_info *di,
+		      struct bpf_prog *prog,
 		      u32 *len, struct xdp_buff *xdp);
 void mlx5e_xdp_mpwqe_complete(struct mlx5e_xdpsq *sq);
 bool mlx5e_poll_xdpsq_cq(struct mlx5e_cq *cq);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c
index 8e7b877d8a12..162513594862 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c
@@ -4,6 +4,7 @@
 #include "rx.h"
 #include "en/xdp.h"
 #include <net/xdp_sock_drv.h>
+#include <linux/filter.h>
 
 /* RX data path */
 
@@ -31,6 +32,7 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq,
 {
 	struct xdp_buff *xdp = wi->umr.dma_info[page_idx].xsk;
 	u32 cqe_bcnt32 = cqe_bcnt;
+	struct bpf_prog *prog;
 
 	/* Check packet size. Note LRO doesn't use linear SKB */
 	if (unlikely(cqe_bcnt > rq->hw_mtu)) {
@@ -65,7 +67,8 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq,
 	 * allocated first from the Reuse Ring, so it has enough space.
 	 */
 
-	if (likely(mlx5e_xdp_handle(rq, NULL, &cqe_bcnt32, xdp))) {
+	prog = rcu_dereference(rq->xdp_prog);
+	if (likely(prog && mlx5e_xdp_handle(rq, NULL, prog, &cqe_bcnt32, xdp))) {
 		if (likely(__test_and_clear_bit(MLX5E_RQ_FLAG_XDP_XMIT, rq->flags)))
 			__set_bit(page_idx, wi->xdp_xmit_bitmap); /* non-atomic */
 		return NULL; /* page/packet was consumed by XDP */
@@ -83,6 +86,7 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_linear(struct mlx5e_rq *rq,
 					      u32 cqe_bcnt)
 {
 	struct xdp_buff *xdp = wi->di->xsk;
+	struct bpf_prog *prog;
 
 	/* wi->offset is not used in this function, because xdp->data and the
 	 * DMA address point directly to the necessary place. Furthermore, the
@@ -101,7 +105,8 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_linear(struct mlx5e_rq *rq,
 		return NULL;
 	}
 
-	if (likely(mlx5e_xdp_handle(rq, NULL, &cqe_bcnt, xdp)))
+	prog = rcu_dereference(rq->xdp_prog);
+	if (likely(prog && mlx5e_xdp_handle(rq, NULL, prog, &cqe_bcnt, xdp)))
 		return NULL; /* page/packet was consumed by XDP */
 
 	/* XDP_PASS: copy the data from the UMEM to a new SKB. The frame reuse
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
index b06aac087b2a..60c640fc430c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
@@ -34,6 +34,7 @@
 #include <linux/ipv6.h>
 #include <linux/tcp.h>
 #include <linux/bitmap.h>
+#include <linux/filter.h>
 #include <net/ip6_checksum.h>
 #include <net/page_pool.h>
 #include <net/inet_ecn.h>
@@ -1523,11 +1524,11 @@ mlx5e_skb_from_cqe_linear(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe,
 {
 	struct mlx5e_dma_info *di = wi->di;
 	u16 rx_headroom = rq->buff.headroom;
-	struct xdp_buff xdp;
+	struct bpf_prog *prog;
 	struct sk_buff *skb;
+	u32 metasize = 0;
 	void *va, *data;
 	u32 frag_size;
-	u32 metasize;
 
 	va             = page_address(di->page) + wi->offset;
 	data           = va + rx_headroom;
@@ -1535,16 +1536,21 @@ mlx5e_skb_from_cqe_linear(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe,
 
 	dma_sync_single_range_for_cpu(rq->pdev, di->addr, wi->offset,
 				      frag_size, DMA_FROM_DEVICE);
-	net_prefetchw(va); /* xdp_frame data area */
 	net_prefetch(data);
 
-	mlx5e_fill_xdp_buff(rq, va, rx_headroom, cqe_bcnt, &xdp);
-	if (mlx5e_xdp_handle(rq, di, &cqe_bcnt, &xdp))
-		return NULL; /* page/packet was consumed by XDP */
+	prog = rcu_dereference(rq->xdp_prog);
+	if (prog) {
+		struct xdp_buff xdp;
 
-	rx_headroom = xdp.data - xdp.data_hard_start;
+		net_prefetchw(va); /* xdp_frame data area */
+		mlx5e_fill_xdp_buff(rq, va, rx_headroom, cqe_bcnt, &xdp);
+		if (mlx5e_xdp_handle(rq, di, prog, &cqe_bcnt, &xdp))
+			return NULL; /* page/packet was consumed by XDP */
+
+		rx_headroom = xdp.data - xdp.data_hard_start;
+		metasize = xdp.data - xdp.data_meta;
+	}
 	frag_size = MLX5_SKB_FRAG_SZ(rx_headroom + cqe_bcnt);
-	metasize = xdp.data - xdp.data_meta;
 	skb = mlx5e_build_linear_skb(rq, va, frag_size, rx_headroom, cqe_bcnt, metasize);
 	if (unlikely(!skb))
 		return NULL;
@@ -1842,11 +1848,11 @@ mlx5e_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi,
 	struct mlx5e_dma_info *di = &wi->umr.dma_info[page_idx];
 	u16 rx_headroom = rq->buff.headroom;
 	u32 cqe_bcnt32 = cqe_bcnt;
-	struct xdp_buff xdp;
+	struct bpf_prog *prog;
 	struct sk_buff *skb;
+	u32 metasize = 0;
 	void *va, *data;
 	u32 frag_size;
-	u32 metasize;
 
 	/* Check packet size. Note LRO doesn't use linear SKB */
 	if (unlikely(cqe_bcnt > rq->hw_mtu)) {
@@ -1860,19 +1866,24 @@ mlx5e_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi,
 
 	dma_sync_single_range_for_cpu(rq->pdev, di->addr, head_offset,
 				      frag_size, DMA_FROM_DEVICE);
-	net_prefetchw(va); /* xdp_frame data area */
 	net_prefetch(data);
 
-	mlx5e_fill_xdp_buff(rq, va, rx_headroom, cqe_bcnt32, &xdp);
-	if (mlx5e_xdp_handle(rq, di, &cqe_bcnt32, &xdp)) {
-		if (__test_and_clear_bit(MLX5E_RQ_FLAG_XDP_XMIT, rq->flags))
-			__set_bit(page_idx, wi->xdp_xmit_bitmap); /* non-atomic */
-		return NULL; /* page/packet was consumed by XDP */
-	}
+	prog = rcu_dereference(rq->xdp_prog);
+	if (prog) {
+		struct xdp_buff xdp;
 
-	rx_headroom = xdp.data - xdp.data_hard_start;
+		net_prefetchw(va); /* xdp_frame data area */
+		mlx5e_fill_xdp_buff(rq, va, rx_headroom, cqe_bcnt32, &xdp);
+		if (mlx5e_xdp_handle(rq, di, prog, &cqe_bcnt32, &xdp)) {
+			if (__test_and_clear_bit(MLX5E_RQ_FLAG_XDP_XMIT, rq->flags))
+				__set_bit(page_idx, wi->xdp_xmit_bitmap); /* non-atomic */
+			return NULL; /* page/packet was consumed by XDP */
+		}
+
+		rx_headroom = xdp.data - xdp.data_hard_start;
+		metasize = xdp.data - xdp.data_meta;
+	}
 	frag_size = MLX5_SKB_FRAG_SZ(rx_headroom + cqe_bcnt32);
-	metasize = xdp.data - xdp.data_meta;
 	skb = mlx5e_build_linear_skb(rq, va, frag_size, rx_headroom, cqe_bcnt32, metasize);
 	if (unlikely(!skb))
 		return NULL;
-- 
2.35.1


  parent reply	other threads:[~2022-03-17 18:54 UTC|newest]

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-17 18:54 [pull request][net-next 00/15] mlx5 updates 2022-03-17 Saeed Mahameed
2022-03-17 18:54 ` [net-next 01/15] net/mlx5e: Validate MTU when building non-linear legacy RQ fragments info Saeed Mahameed
2022-03-18 11:00   ` patchwork-bot+netdevbpf
2022-03-17 18:54 ` [net-next 02/15] net/mlx5e: Add headroom only to the first fragment in legacy RQ Saeed Mahameed
2022-03-17 18:54 ` [net-next 03/15] net/mlx5e: Build SKB in place over the first fragment in non-linear " Saeed Mahameed
2022-03-17 18:54 ` Saeed Mahameed [this message]
2022-03-17 18:54 ` [net-next 05/15] net/mlx5e: Drop the len output parameter from mlx5e_xdp_handle Saeed Mahameed
2022-03-17 18:54 ` [net-next 06/15] net/mlx5e: Drop cqe_bcnt32 from mlx5e_skb_from_cqe_mpwrq_linear Saeed Mahameed
2022-03-17 18:54 ` [net-next 07/15] net/mlx5: DR, Adjust structure member to reduce memory hole Saeed Mahameed
2022-03-17 18:54 ` [net-next 08/15] net/mlx5: DR, Remove mr_addr rkey from struct mlx5dr_icm_chunk Saeed Mahameed
2022-03-17 18:54 ` [net-next 09/15] net/mlx5: DR, Remove icm_addr from mlx5dr_icm_chunk to reduce memory Saeed Mahameed
2022-03-17 18:54 ` [net-next 10/15] net/mlx5: DR, Remove num_of_entries byte_size from struct mlx5_dr_icm_chunk Saeed Mahameed
2022-03-17 18:54 ` [net-next 11/15] net/mlx5: DR, Remove 4 members from mlx5dr_ste_htbl to reduce memory Saeed Mahameed
2022-03-17 18:54 ` [net-next 12/15] net/mlx5: DR, Remove hw_ste from mlx5dr_ste " Saeed Mahameed
2022-03-17 18:54 ` [net-next 13/15] net/mlx5: CT: Remove extra rhashtable remove on tuple entries Saeed Mahameed
2022-03-17 18:54 ` [net-next 14/15] net/mlx5: Remove unused exported contiguous coherent buffer allocation API Saeed Mahameed
2022-03-17 18:54 ` [net-next 15/15] net/mlx5: Remove unused fill page array API function Saeed Mahameed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220317185424.287982-5-saeed@kernel.org \
    --to=saeed@kernel.org \
    --cc=davem@davemloft.net \
    --cc=kuba@kernel.org \
    --cc=maximmi@nvidia.com \
    --cc=netdev@vger.kernel.org \
    --cc=saeedm@nvidia.com \
    --cc=tariqt@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).