xdp: T2666: initial XDP (generic mode) forwarding support

The CLI command 'set interfaces ethernet <interface> offload-options xdp" enables
the XDP generic mode on the given interface.

vyos@vyos:~$ show interfaces ethernet eth1
eth1: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 xdpgeneric/id:151 qdisc mq state DOWN group default qlen 1000
    link/ether 00:50:56:bf:ef:aa brd ff:ff:ff:ff:ff:ff
    inet6 fe80::250:56ff:febf:efaa/64 scope link tentative
       valid_lft forever preferred_lft forever
    Description: fooa

XDP code is thankfully copied from [1], thank you for this nice tutorial.

NOTE: this is an experimental feature which might break your
forwarding/filtering.

[1]: https://medium.com/swlh/building-a-xdp-express-data-path-based-peering-router-20db4995da66

1 files changed, 202 insertions, 0 deletions

diff --git a/src/ebpf/xdp_router.c b/src/ebpf/xdp_router.c
new file mode 100644
index 000000000..4fb5c7cb1
--- /dev/null
+++ b/src/ebpf/xdp_router.c
@@ -0,0 +1,202 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+// Code thankfully copied from:
+// https://medium.com/swlh/building-a-xdp-express-data-path-based-peering-router-20db4995da66
+
+#include <linux/bpf.h>
+#include <linux/if_ether.h>
+#include <linux/in.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/udp.h>
+#include <linux/socket.h>
+
+#include <bpf/bpf_endian.h>
+#include <bpf/bpf_helpers.h>
+
+#include <stdint.h>
+
+#ifndef XDP_ACTION_MAX
+#define XDP_ACTION_MAX (XDP_REDIRECT + 1)
+#endif
+
+#ifndef memcpy
+#define memcpy(dest, src, n) __builtin_memcpy((dest), (src), (n))
+#endif
+
+#ifndef AF_INET
+#define AF_INET 2
+#endif
+
+#ifndef AF_INET6
+#define AF_INET6 10
+#endif
+
+#ifndef IPV6_FLOWINFO_MASK
+#define IPV6_FLOWINFO_MASK bpf_htonl(0x0FFFFFFF)
+#endif
+
+/* This is the data record stored in the map */
+struct datarec {
+	__u64 rx_packets;
+	__u64 rx_bytes;
+};
+
+/* Keeps stats per (enum) xdp_action */
+struct bpf_map_def SEC("maps") xdp_stats_map = {
+	.type        = BPF_MAP_TYPE_PERCPU_ARRAY,
+	.key_size    = sizeof(__u32),
+	.value_size  = sizeof(struct datarec),
+	.max_entries = XDP_ACTION_MAX,
+};
+
+struct bpf_map_def SEC("maps") tx_port = {
+	.type = BPF_MAP_TYPE_DEVMAP,
+	.key_size = sizeof(int),
+	.value_size = sizeof(int),
+	.max_entries = 256,
+};
+
+/* from include/net/ip.h */
+static __always_inline int ip_decrease_ttl(struct iphdr *iph)
+{
+	__u32 check = iph->check;
+	check += bpf_htons(0x0100);
+	iph->check = (__u16)(check + (check >= 0xFFFF));
+	return --iph->ttl;
+}
+
+static __always_inline
+__u32 xdp_stats_record_action(struct xdp_md *ctx, __u32 action)
+{
+	if (action >= XDP_ACTION_MAX)
+		return XDP_ABORTED;
+
+	/* Lookup in kernel BPF-side return pointer to actual data record */
+	struct datarec *rec = bpf_map_lookup_elem(&xdp_stats_map, &action);
+	if (!rec)
+		return XDP_ABORTED;
+
+	/* BPF_MAP_TYPE_PERCPU_ARRAY returns a data record specific to current
+	 * CPU and XDP hooks runs under Softirq, which makes it safe to update
+	 * without atomic operations.
+	 */
+	rec->rx_packets++;
+	rec->rx_bytes += (ctx->data_end - ctx->data);
+
+	return action;
+}
+
+/* xdp_router is the name of the xdp program */
+SEC("prog")
+int xdp_router_func(struct xdp_md *ctx)
+{
+	/* this is the packet context*/
+	void *data_end = (void *)(long)ctx->data_end;
+	void *data = (void *)(long)ctx->data;
+	struct bpf_fib_lookup fib_params = {};
+	struct ethhdr *eth = data;
+	struct ipv6hdr *ip6h;
+	struct iphdr *iph;
+	__u16 h_proto;
+	__u64 nh_off;
+	int rc;
+	/* default action is to pass */
+	int action = XDP_PASS;
+
+	nh_off = sizeof(*eth);
+	if (data + nh_off > data_end) {
+		action = XDP_DROP;
+		goto out;
+	}
+
+	/* determine if this is IP4 or IPv6 by looking at the Ethernet protocol field */
+	h_proto = eth->h_proto;
+	if (h_proto == bpf_htons(ETH_P_IP)) {
+		/* IPv4 part of the code */
+		iph = data + nh_off;
+
+		if (iph + 1 > data_end) {
+			action = XDP_DROP;
+			goto out;
+		}
+		/* as a real router, we need to check the TTL to prevent never ending loops*/
+		if (iph->ttl <= 1)
+			goto out;
+
+		/* populate the fib_params fields to prepare for the lookup */
+		fib_params.family	= AF_INET;
+		fib_params.tos		= iph->tos;
+		fib_params.l4_protocol	= iph->protocol;
+		fib_params.sport	= 0;
+		fib_params.dport	= 0;
+		fib_params.tot_len	= bpf_ntohs(iph->tot_len);
+		fib_params.ipv4_src	= iph->saddr;
+		fib_params.ipv4_dst	= iph->daddr;
+	} else if (h_proto == bpf_htons(ETH_P_IPV6)) {
+		/* IPv6 part of the code */
+		struct in6_addr *src = (struct in6_addr *) fib_params.ipv6_src;
+		struct in6_addr *dst = (struct in6_addr *) fib_params.ipv6_dst;
+
+		ip6h = data + nh_off;
+		if (ip6h + 1 > data_end) {
+			action = XDP_DROP;
+			goto out;
+		}
+		/* as a real router, we need to check the TTL to prevent never ending loops*/
+		if (ip6h->hop_limit <= 1)
+			goto out;
+
+		/* populate the fib_params fields to prepare for the lookup */
+		fib_params.family	= AF_INET6;
+		fib_params.flowinfo	= *(__be32 *) ip6h & IPV6_FLOWINFO_MASK;
+		fib_params.l4_protocol	= ip6h->nexthdr;
+		fib_params.sport	= 0;
+		fib_params.dport	= 0;
+		fib_params.tot_len	= bpf_ntohs(ip6h->payload_len);
+		*src			= ip6h->saddr;
+		*dst			= ip6h->daddr;
+	} else {
+		goto out;
+	}
+
+	fib_params.ifindex = ctx->ingress_ifindex;
+
+	/* this is where the FIB lookup happens. If the lookup is successful */
+	/* it will populate the fib_params.ifindex with the egress interface index */
+
+	rc = bpf_fib_lookup(ctx, &fib_params, sizeof(fib_params), 0);
+	switch (rc) {
+	case BPF_FIB_LKUP_RET_SUCCESS:         /* lookup successful */
+		   /* we are a router, so we need to decrease the ttl */
+		if (h_proto == bpf_htons(ETH_P_IP))
+			ip_decrease_ttl(iph);
+		else if (h_proto == bpf_htons(ETH_P_IPV6))
+			ip6h->hop_limit--;
+		/* set the correct new source and destionation mac addresses */
+		/* can be found in fib_params.dmac and fib_params.smac */
+		memcpy(eth->h_dest, fib_params.dmac, ETH_ALEN);
+		memcpy(eth->h_source, fib_params.smac, ETH_ALEN);
+		/* and done, now we set the action to bpf_redirect_map with fib_params.ifindex which is the egress port as paramater */
+		action = bpf_redirect_map(&tx_port, fib_params.ifindex, 0);
+		break;
+	case BPF_FIB_LKUP_RET_BLACKHOLE:    /* dest is blackholed; can be dropped */
+	case BPF_FIB_LKUP_RET_UNREACHABLE:  /* dest is unreachable; can be dropped */
+	case BPF_FIB_LKUP_RET_PROHIBIT:     /* dest not allowed; can be dropped */
+		action = XDP_DROP;
+		break;
+	case BPF_FIB_LKUP_RET_NOT_FWDED:    /* packet is not forwarded */
+	case BPF_FIB_LKUP_RET_FWD_DISABLED: /* fwding is not enabled on ingress */
+	case BPF_FIB_LKUP_RET_UNSUPP_LWT:   /* fwd requires encapsulation */
+	case BPF_FIB_LKUP_RET_NO_NEIGH:     /* no neighbor entry for nh */
+	case BPF_FIB_LKUP_RET_FRAG_NEEDED:  /* fragmentation required to fwd */
+		/* PASS */
+		break;
+	}
+
+out:
+	/* and done, update stats and return action */
+	return xdp_stats_record_action(ctx, action);
+}
+
+char _license[] SEC("license") = "GPL";