Merge pull request #1 from netoptimizer/MTU-tests01.public

MTU testing

Simple TC-BPF program for testing different packets sizes that violate the MTU of the interface.

This is part of testing upstream kernel work, for removing the MTU limit in the BPF-helpers that change packet size, and instead add BPF-helper that can check or lookup interface MTU.

MTU-tests/Makefile

@@ -0,0 +1,21 @@
+# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
+
+USER_TARGETS :=
+BPF_TARGETS := tc_bpf_inc_pkt_size
+EXTRA_DEPS := encap.h
+
+LIB_DIR = ../lib
+
+include $(LIB_DIR)/common.mk
+
+# The iproute2 'tc' tool doesn't understand BTF debug info
+# use llvm-strip to remove this debug info from object file
+#
+# *BUT* cannot strip everything as it removes ELF elems needed for
+#  creating maps
+#
+.PHONY: strip_tc_obj
+strip_tc_obj: tc_bpf_inc_pkt_size.o
+	$(Q) llvm-strip --no-strip-all --remove-section .BTF $?
+
+all: strip_tc_obj

MTU-tests/README.org

@@ -0,0 +1,25 @@
+#+OPTIONS: ^:nil
+
+* BPF testing packet size changes
+
+Playing with increasing packet size with TC-BPF. Testing what happens when
+sending packets larger that MTU out an interface.
+
+* Notes loading BPF object
+
+How to manually load TC program:
+
+#+begin_src sh
+export DEV=mlx5p1
+tc qdisc del dev "$DEV" clsact # Also deletes all filters
+tc qdisc add dev "$DEV" clsact
+tc filter add dev "$DEV" pref 1 handle 1 egress bpf da obj tc_bpf_inc_pkt_size.o
+tc filter show dev "$DEV" egress
+#+end_src
+
+Be *VERY* careful with =replace= command, it MUST have same =pref= +=handle=:
+
+#+begin_src sh
+tc filter replace dev "$DEV" pref 1 handle 1 egress bpf da obj tc_bpf_inc_pkt_size.o
+#+end_src
+

MTU-tests/encap.h

@@ -0,0 +1,33 @@
+
+static __always_inline __u16 csum_fold_helper(__u32 csum)
+{
+	__u32 sum;
+	sum = (csum >> 16) + (csum & 0xffff);
+	sum += (sum >> 16);
+	return ~sum;
+}
+
+static void encap_ipv4_ipip(volatile void *data, volatile void *data_end)
+{
+	volatile struct iphdr *iph;
+	size_t len;
+
+	struct iphdr encap_hdr = {
+		.version = 4,
+		.ihl = 5,
+		.protocol = IPPROTO_IPIP,
+		.ttl = 16,
+		.saddr = bpf_htonl(0x0a0b0202),
+		.daddr = bpf_htonl(0x0a0b0201),
+	};
+
+	iph = data + sizeof(struct ethhdr);
+	if (iph + 1 > data_end)
+		return;
+
+	*iph = encap_hdr;
+
+	len = (data_end - data);
+	iph->tot_len = bpf_htons(len - sizeof(struct ethhdr));
+	iph->check = csum_fold_helper(bpf_csum_diff((__be32 *)iph, 0, (__be32 *)iph, sizeof(*iph), 0));
+}

MTU-tests/tc_bpf_inc_pkt_size.c

@@ -0,0 +1,131 @@
+#include <linux/bpf.h>
+#include <linux/in6.h>
+#include <bpf/bpf_helpers.h>
+#include <xdp/parsing_helpers.h>
+#include "encap.h"
+
+/* Cycle through different MTU packet sizes, encoded in BPF-code via switch
+ * statement.  MTU is defined as L3 size (usually 1500 for Ethernet), but
+ * remember TC (and XDP) operate at L2 (adjusted later)
+ */
+static __always_inline __u32 get_pkt_size_l3(__u64 cnt)
+{
+	switch (cnt) {
+	case 0:
+		return 1024;
+	case 1:
+		return 1500;
+	case 2:
+		return 1504;
+	case 3:
+		return 1508;
+	case 4:
+		return 1600;
+	case 5:
+		return 4096 + 128;
+	case 6:
+		return 3520;
+	case 7:
+		return 3528;
+	case 8:
+		return 4096 - 14;
+	case 9:
+		return 4096;
+	case 10:
+		return 8192;
+	case 11:
+		return 16000;
+	default:
+		return 1500;
+	}
+}
+#define CNT_MAX 12
+
+/* The tc tool (iproute2) use another ELF map layout than libbpf, see
+ * struct bpf_elf_map from iproute2, but bpf_map_def from libbpf have
+ * same binary layout until "flags" so use that.
+ */
+struct bpf_map_def  SEC("maps") cnt_map = {
+	.type = BPF_MAP_TYPE_ARRAY,
+	.key_size = sizeof(__u32),
+	.value_size = sizeof(__u64),
+	.max_entries = 1,
+	.map_flags = 0,
+};
+
+/* LLVM maps __sync_fetch_and_add() as a built-in function to the BPF atomic add
+ * instruction (that is BPF_STX | BPF_XADD | BPF_W for word sizes)
+ */
+#ifndef lock_xadd
+#define lock_xadd(ptr, val)     ((void) __sync_fetch_and_add(ptr, val))
+#endif
+
+#define ENCAP_TYPE BPF_F_ADJ_ROOM_ENCAP_L3_IPV4
+
+SEC("classifier") int tc_inc_pkt_sz(struct __sk_buff *skb)
+{
+	volatile void *data, *data_end;
+	int ret = BPF_DROP;
+	struct ethhdr *eth;
+	struct iphdr *iph;
+	int extra_len;
+	int len;
+
+	int key = 0;
+	__u64 *cnt;
+
+	cnt = bpf_map_lookup_elem(&cnt_map, &key);
+	if (!cnt)
+		goto out;
+
+	/* Desired packet size at L2 */
+	int pkt_size_l2 = get_pkt_size_l3(*cnt) + sizeof(*eth) ;
+
+	data     = (void *)(long)skb->data;
+	data_end = (void *)(long)skb->data_end;
+	eth = (struct ethhdr *)data;
+
+	if (data + sizeof(*eth) > data_end)
+		return BPF_DROP;
+
+	/* Keep ARP resolution working */
+	if (eth->h_proto == bpf_htons(ETH_P_ARP)) {
+		ret = BPF_OK;
+		goto out;
+	}
+
+	len = (data_end - data);
+	extra_len = pkt_size_l2 - len;
+	// extra_len= sizeof(*iph); /* Adj that does correct IPIP encap */
+
+	/* Wrapping global counter */
+        lock_xadd(cnt, 1);
+	if (*cnt == CNT_MAX)
+		*cnt = 0;
+
+	if (bpf_skb_adjust_room(skb, extra_len, BPF_ADJ_ROOM_MAC, ENCAP_TYPE)) {
+		/* If adjust fails, then skip this packet length adjustment */
+		ret = BPF_OK;
+		goto out;
+	}
+
+	/* Most re-load after bpf_skb_adjust_room() */
+	data     = (void *)(long)skb->data;
+	data_end = (void *)(long)skb->data_end;
+
+	/* Add IP-header with IPIP */
+	encap_ipv4_ipip(data, data_end);
+
+	eth = (void *)data;
+	iph = (void *)(eth +1);
+	if (iph +1 > data_end)
+		goto out;
+
+	eth->h_proto = bpf_htons(ETH_P_IP);
+
+	ret = BPF_OK;
+out:
+	return ret;
+}
+
+char _license[] SEC("license") = "GPL";