linux-kernel - Re: [RFC PATCH net-next 11/11] net: enetc: add TX support for zero-copy XDP sockets

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20230206101921.zmslsxfzdm6ovxf5@skbuf>
Date:   Mon, 6 Feb 2023 12:19:21 +0200
From:   Vladimir Oltean <vladimir.oltean@....com>
To:     netdev@...r.kernel.org
Cc:     "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>,
        Claudiu Manoil <claudiu.manoil@....com>,
        Björn Töpel <bjorn@...nel.org>,
        Magnus Karlsson <magnus.karlsson@...el.com>,
        Maciej Fijalkowski <maciej.fijalkowski@...el.com>,
        Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Jesper Dangaard Brouer <hawk@...nel.org>,
        John Fastabend <john.fastabend@...il.com>, bpf@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH net-next 11/11] net: enetc: add TX support for
 zero-copy XDP sockets

Yikes, ran git format-patch one patch too soon. There's also a hidden,
bonus patch "12/11" below. Doesn't affect first 11 patches in any way,
though. Here it is.

>From f7f10232622309d66a7a1ae1932d0c081936d546 Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <vladimir.oltean@....com>
Date: Tue, 25 Oct 2022 02:32:24 +0300
Subject: [RFC PATCH net-next 12/12] net: enetc: add support for XDP_TX
 with zero-copy XDP sockets

Add support for the case when the BPF program attached to a ring with an
XSK pool returns the XDP_TX verdict. The frame needs to go back on the
interface it came from.

No dma_map or dma_sync_for_device is necessary, just a small impedance
matching logic with the XDP_TX procedure we have in place for non-XSK,
since the data structures are different (xdp_buff vs xdp_frame; cannot
have multi-buffer with XSK).

In the TX confirmation routine, just release the RX buffer (as opposed
to non-XSK XDP_TX). Recycling might be possible, but I haven't
experimented with it.

Signed-off-by: Vladimir Oltean <vladimir.oltean@....com>
---
 drivers/net/ethernet/freescale/enetc/enetc.c | 35 ++++++++++++++++++--
 drivers/net/ethernet/freescale/enetc/enetc.h |  1 +
 2 files changed, 34 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/freescale/enetc/enetc.c b/drivers/net/ethernet/freescale/enetc/enetc.c
index bc0db788afc7..06aaf0334dc3 100644
--- a/drivers/net/ethernet/freescale/enetc/enetc.c
+++ b/drivers/net/ethernet/freescale/enetc/enetc.c
@@ -855,7 +855,9 @@ static bool enetc_clean_tx_ring(struct enetc_bdr *tx_ring, int napi_budget,
 				tx_win_drop++;
 		}
 
-		if (tx_swbd->is_xsk)
+		if (tx_swbd->is_xsk && tx_swbd->is_xdp_tx)
+			xsk_buff_free(tx_swbd->xsk_buff);
+		else if (tx_swbd->is_xsk)
 			(*xsk_confirmed)++;
 		else if (tx_swbd->is_xdp_tx)
 			enetc_recycle_xdp_tx_buff(tx_ring, tx_swbd);
@@ -1661,6 +1663,21 @@ static int enetc_rx_swbd_to_xdp_tx_swbd(struct enetc_tx_swbd *xdp_tx_arr,
 	return n;
 }
 
+static bool enetc_xsk_xdp_tx(struct enetc_bdr *tx_ring,
+			     struct xdp_buff *xsk_buff)
+{
+	struct enetc_tx_swbd tx_swbd = {
+		.dma = xsk_buff_xdp_get_dma(xsk_buff),
+		.len = xdp_get_buff_len(xsk_buff),
+		.is_xdp_tx = true,
+		.is_xsk = true,
+		.is_eof = true,
+		.xsk_buff = xsk_buff,
+	};
+
+	return enetc_xdp_tx(tx_ring, &tx_swbd, 1);
+}
+
 static void enetc_xdp_drop(struct enetc_bdr *rx_ring, int rx_ring_first,
 			   int rx_ring_last)
 {
@@ -1839,10 +1856,12 @@ static int enetc_clean_rx_ring_xsk(struct enetc_bdr *rx_ring,
 				   struct bpf_prog *prog,
 				   struct xsk_buff_pool *pool)
 {
+	struct enetc_ndev_priv *priv = netdev_priv(rx_ring->ndev);
+	struct enetc_bdr *tx_ring = priv->xdp_tx_ring[rx_ring->index];
+	int xdp_redirect_frm_cnt = 0, xdp_tx_frm_cnt = 0;
 	struct net_device *ndev = rx_ring->ndev;
 	union enetc_rx_bd *rxbd, *orig_rxbd;
 	int rx_frm_cnt = 0, rx_byte_cnt = 0;
-	int xdp_redirect_frm_cnt = 0;
 	struct xdp_buff *xsk_buff;
 	int buffs_missing, err, i;
 	bool wakeup_xsk = false;
@@ -1895,6 +1914,15 @@ static int enetc_clean_rx_ring_xsk(struct enetc_bdr *rx_ring,
 			enetc_xsk_buff_to_skb(xsk_buff, rx_ring, orig_rxbd,
 					      napi);
 			break;
+		case XDP_TX:
+			if (enetc_xsk_xdp_tx(tx_ring, xsk_buff)) {
+				xdp_tx_frm_cnt++;
+				tx_ring->stats.xdp_tx++;
+			} else {
+				xsk_buff_free(xsk_buff);
+				tx_ring->stats.xdp_tx_drops++;
+			}
+			break;
 		case XDP_REDIRECT:
 			err = xdp_do_redirect(ndev, xsk_buff, prog);
 			if (unlikely(err)) {
@@ -1919,6 +1947,9 @@ static int enetc_clean_rx_ring_xsk(struct enetc_bdr *rx_ring,
 	if (xdp_redirect_frm_cnt)
 		xdp_do_flush_map();
 
+	if (xdp_tx_frm_cnt)
+		enetc_update_tx_ring_tail(tx_ring);
+
 	if (xsk_uses_need_wakeup(pool)) {
 		if (wakeup_xsk)
 			xsk_set_rx_need_wakeup(pool);
diff --git a/drivers/net/ethernet/freescale/enetc/enetc.h b/drivers/net/ethernet/freescale/enetc/enetc.h
index 403f40473b52..58df6c32cfb3 100644
--- a/drivers/net/ethernet/freescale/enetc/enetc.h
+++ b/drivers/net/ethernet/freescale/enetc/enetc.h
@@ -24,6 +24,7 @@ struct enetc_tx_swbd {
 	union {
 		struct sk_buff *skb;
 		struct xdp_frame *xdp_frame;
+		struct xdp_buff *xsk_buff;
 	};
 	dma_addr_t dma;
 	struct page *page;	/* valid only if is_xdp_tx */
-- 
2.34.1