Revert "router: optimize duplicated PIO comparison"

This reverts commit d354ebb.

As stated by David Härdeman at [1]:
Note that time_t will have an alignment of 4 (unlikely) or 8, meaning
that struct ra_pio will have padding. A straight memcmp() could
therefore generate the wrong result unless both struct ra_pios that are
compared are guaranteed to have been memset to zero.

Avoid the risk of something going wrong here by doing a straight
comparison of the two members of the struct that we want to compare.

[1]: openwrt@d02fce0

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>

Author: Noltari

src/odhcpd.h

@@ -346,13 +346,10 @@ struct dnr_options {
 
 // RA PIO - RFC9096
 struct ra_pio {
-	struct {
-		struct in6_addr prefix;
-		uint8_t length;
-	};
+	struct in6_addr prefix;
+	uint8_t length;
 	time_t lifetime;
 };
-#define ra_pio_cmp_len offsetof(struct ra_pio, lifetime)
 
 
 struct interface {

src/router.c

@@ -539,7 +539,8 @@ static void router_clear_duplicated_ra_pio(struct interface *iface)
 		while (j < iface->pio_cnt) {
 			struct ra_pio *pio_b = &iface->pios[j];
 
-			if (!memcmp(pio_a, pio_b, ra_pio_cmp_len)) {
+			if (pio_a->length == pio_b->length &&
+			    !memcmp(&pio_a->prefix, &pio_b->prefix, sizeof(struct in6_addr))) {
 				warn("rfc9096: %s: clear duplicated %s/%u",
 				     iface->ifname,
 				     inet_ntop(AF_INET6, &pio_a->prefix, ipv6_str, sizeof(ipv6_str)),