xdp: make the filter endian independent

src/libknot/xdp/bpf-kernel.c

@@ -24,6 +24,9 @@
 #include "../../contrib/libbpf/include/uapi/linux/bpf.h"
 #include "../../contrib/libbpf/bpf/bpf_helpers.h"
 
+/* Don't fragment flag. */
+#define	IP_DF 0x4000
+
 /* Assume netdev has no more than 128 queues. */
 #define QUEUE_MAX 128
 
@@ -47,7 +50,6 @@ struct ipv6_frag_hdr {
 	unsigned char whatever[7];
 } __attribute__((packed));
 
-/* NOTE: this implementation expects little-endian byte ordering! */
 SEC("xdp_redirect_udp")
 int xdp_redirect_udp_func(struct xdp_md *ctx)
 {
@@ -70,18 +72,19 @@ int xdp_redirect_udp_func(struct xdp_md *ctx)
 
 	/* Parse IPv4 or IPv6 header. */
 	switch (eth->h_proto) {
-	case 0x0008: /* htons(ETH_P_IP) */
+	case __constant_htons(ETH_P_IP):
 		ip4 = data;
 		if ((void *)ip4 + sizeof(*ip4) > data_end) {
 			return XDP_PASS;
 		}
-		if (ip4->frag_off != 0 && ip4->frag_off != 0x0040) { /* htons(IP_DF) */
+		if (ip4->frag_off != 0 &&
+		    ip4->frag_off != __constant_htons(IP_DF)) {
 			fragmented = 1;
 		}
 		ip_proto = ip4->protocol;
 		udp = data + ip4->ihl * 4;
 		break;
-	case 0xDD86: /* htons(ETH_P_IPV6) */
+	case __constant_htons(ETH_P_IPV6):
 		ip6 = data;
 		if ((void *)ip6 + sizeof(*ip6) > data_end) {
 			return XDP_PASS;