aboutsummaryrefslogtreecommitdiffstats
path: root/target/linux/bcm4908/patches-5.15/072-v6.2-0001-net-broadcom-bcm4908_enet-use-build_skb.patch
blob: 1b4cc9e24ccf041dc6f894f9cd100698e99305b4 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
From 3a1cc23a75abcd9cea585eb84846507363d58397 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rafa=C5=82=20Mi=C5=82ecki?= <rafal@milecki.pl>
Date: Tue, 25 Oct 2022 15:22:45 +0200
Subject: [PATCH] net: broadcom: bcm4908_enet: use build_skb()
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

RX code can be more efficient with the build_skb(). Allocating actual
SKB around eth packet buffer - right before passing it up - results in
a better cache usage.

Without RPS (echo 0 > rps_cpus) BCM4908 NAT masq performance "jumps"
between two speeds: ~900 Mbps and 940 Mbps (it's a 4 CPUs SoC). This
change bumps the lower speed from 905 Mb/s to 918 Mb/s (tested using
single stream iperf 2.0.5 traffic).

There are more optimizations to consider. One obvious to try is GRO
however as BCM4908 doesn't do hw csum is may actually lower performance.
Sometimes. Some early testing:

┌─────────────────────────────────┬─────────────────────┬────────────────────┐
│                                 │ netif_receive_skb() │ napi_gro_receive() │
├─────────────────────────────────┼─────────────────────┼────────────────────┤
│ netdev_alloc_skb()              │            905 Mb/s │           892 Mb/s │
│ napi_alloc_frag() + build_skb() │            918 Mb/s │           917 Mb/s │
└─────────────────────────────────┴─────────────────────┴────────────────────┘

Another ideas:
1. napi_build_skb()
2. skb_copy_from_linear_data() for small packets

Those need proper testing first though. That can be done later.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Link: https://lore.kernel.org/r/20221025132245.22871-1-zajec5@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
 drivers/net/ethernet/broadcom/bcm4908_enet.c | 53 +++++++++++++-------
 1 file changed, 36 insertions(+), 17 deletions(-)

--- a/drivers/net/ethernet/broadcom/bcm4908_enet.c
+++ b/drivers/net/ethernet/broadcom/bcm4908_enet.c
@@ -36,13 +36,24 @@
 #define ENET_MAX_ETH_OVERHEAD			(ETH_HLEN + BRCM_MAX_TAG_LEN + VLAN_HLEN + \
 						 ETH_FCS_LEN + 4) /* 32 */
 
+#define ENET_RX_SKB_BUF_SIZE			(NET_SKB_PAD + NET_IP_ALIGN + \
+						 ETH_HLEN + BRCM_MAX_TAG_LEN + VLAN_HLEN + \
+						 ENET_MTU_MAX + ETH_FCS_LEN + 4)
+#define ENET_RX_SKB_BUF_ALLOC_SIZE		(SKB_DATA_ALIGN(ENET_RX_SKB_BUF_SIZE) + \
+						 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
+#define ENET_RX_BUF_DMA_OFFSET			(NET_SKB_PAD + NET_IP_ALIGN)
+#define ENET_RX_BUF_DMA_SIZE			(ENET_RX_SKB_BUF_SIZE - ENET_RX_BUF_DMA_OFFSET)
+
 struct bcm4908_enet_dma_ring_bd {
 	__le32 ctl;
 	__le32 addr;
 } __packed;
 
 struct bcm4908_enet_dma_ring_slot {
-	struct sk_buff *skb;
+	union {
+		void *buf;			/* RX */
+		struct sk_buff *skb;		/* TX */
+	};
 	unsigned int len;
 	dma_addr_t dma_addr;
 };
@@ -260,22 +271,21 @@ static int bcm4908_enet_dma_alloc_rx_buf
 	u32 tmp;
 	int err;
 
-	slot->len = ENET_MTU_MAX + ENET_MAX_ETH_OVERHEAD;
-
-	slot->skb = netdev_alloc_skb(enet->netdev, slot->len);
-	if (!slot->skb)
+	slot->buf = napi_alloc_frag(ENET_RX_SKB_BUF_ALLOC_SIZE);
+	if (!slot->buf)
 		return -ENOMEM;
 
-	slot->dma_addr = dma_map_single(dev, slot->skb->data, slot->len, DMA_FROM_DEVICE);
+	slot->dma_addr = dma_map_single(dev, slot->buf + ENET_RX_BUF_DMA_OFFSET,
+					ENET_RX_BUF_DMA_SIZE, DMA_FROM_DEVICE);
 	err = dma_mapping_error(dev, slot->dma_addr);
 	if (err) {
 		dev_err(dev, "Failed to map DMA buffer: %d\n", err);
-		kfree_skb(slot->skb);
-		slot->skb = NULL;
+		skb_free_frag(slot->buf);
+		slot->buf = NULL;
 		return err;
 	}
 
-	tmp = slot->len << DMA_CTL_LEN_DESC_BUFLENGTH_SHIFT;
+	tmp = ENET_RX_BUF_DMA_SIZE << DMA_CTL_LEN_DESC_BUFLENGTH_SHIFT;
 	tmp |= DMA_CTL_STATUS_OWN;
 	if (idx == enet->rx_ring.length - 1)
 		tmp |= DMA_CTL_STATUS_WRAP;
@@ -315,11 +325,11 @@ static void bcm4908_enet_dma_uninit(stru
 
 	for (i = rx_ring->length - 1; i >= 0; i--) {
 		slot = &rx_ring->slots[i];
-		if (!slot->skb)
+		if (!slot->buf)
 			continue;
 		dma_unmap_single(dev, slot->dma_addr, slot->len, DMA_FROM_DEVICE);
-		kfree_skb(slot->skb);
-		slot->skb = NULL;
+		skb_free_frag(slot->buf);
+		slot->buf = NULL;
 	}
 }
 
@@ -575,6 +585,7 @@ static int bcm4908_enet_poll_rx(struct n
 	while (handled < weight) {
 		struct bcm4908_enet_dma_ring_bd *buf_desc;
 		struct bcm4908_enet_dma_ring_slot slot;
+		struct sk_buff *skb;
 		u32 ctl;
 		int len;
 		int err;
@@ -598,16 +609,24 @@ static int bcm4908_enet_poll_rx(struct n
 
 		if (len < ETH_ZLEN ||
 		    (ctl & (DMA_CTL_STATUS_SOP | DMA_CTL_STATUS_EOP)) != (DMA_CTL_STATUS_SOP | DMA_CTL_STATUS_EOP)) {
-			kfree_skb(slot.skb);
+			skb_free_frag(slot.buf);
 			enet->netdev->stats.rx_dropped++;
 			break;
 		}
 
-		dma_unmap_single(dev, slot.dma_addr, slot.len, DMA_FROM_DEVICE);
+		dma_unmap_single(dev, slot.dma_addr, ENET_RX_BUF_DMA_SIZE, DMA_FROM_DEVICE);
+
+		skb = build_skb(slot.buf, ENET_RX_SKB_BUF_ALLOC_SIZE);
+		if (unlikely(!skb)) {
+			skb_free_frag(slot.buf);
+			enet->netdev->stats.rx_dropped++;
+			break;
+		}
+		skb_reserve(skb, ENET_RX_BUF_DMA_OFFSET);
+		skb_put(skb, len - ETH_FCS_LEN);
+		skb->protocol = eth_type_trans(skb, enet->netdev);
 
-		skb_put(slot.skb, len - ETH_FCS_LEN);
-		slot.skb->protocol = eth_type_trans(slot.skb, enet->netdev);
-		netif_receive_skb(slot.skb);
+		netif_receive_skb(skb);
 
 		enet->netdev->stats.rx_packets++;
 		enet->netdev->stats.rx_bytes += len;