net: ipv6: Add support for default gateway discovery.

In IPv6, the default gateway and prefix length are determined by receiving
a router advertisement as defined in -
https://www.rfc-editor.org/rfc/rfc4861.

Add support for sending router solicitation (RS) and processing router
advertisements (RA).

If the RA has prefix info option and following conditions are met, then
gatewayip6 and net_prefix_length of ip6addr env variables are initialized.
These are later consumed by IPv6 code for non-local destination IP.

- "Router Lifetime" != 0
- Prefix is NOT link-local prefix (0xfe80::/10)
- L flag is 1
- "Valid Lifetime" != 0

Timing Parameters:
- MAX_RTR_SOLICITATION_DELAY (0-1s)
- RTR_SOLICITATION_INTERVAL (4s) (min retransmit delay)
- MAX_RTR_SOLICITATIONS (3 RS transmissions)

The functionality is enabled by CONFIG_IPV6_ROUTER_DISCOVERY and invoked
automatically from net_init_loop().

Signed-off-by: Ehsan Mohandesi <emohandesi@linux.microsoft.com>
Tested-by: Viacheslav Mitrofanov <v.v.mitrofanov@yadro.com>Reviewed-by:
Tested-by: Viacheslav Mitrofanov <v.v.mitrofanov@yadro.com>
Reviewed-by: Viacheslav Mitrofanov <v.v.mitrofanov@yadro.com>
Tested-by: Sergei Antonov <saproj@gmail.com>
Reviewed-by: Sergei Antonov <saproj@gmail.com>

cmd/Kconfig

@@ -1942,6 +1942,12 @@ config CMD_NCSI
 	  Normally this happens automatically before other network
 	  operations.
 
+config IPV6_ROUTER_DISCOVERY
+	bool "Do IPv6 router discovery"
+	depends on IPV6
+	help
+	  Will automatically perform router solicitation on first IPv6
+	  network operation
 endif
 
 config CMD_ETHSW

include/ndisc.h

@@ -19,6 +19,20 @@ struct nd_msg {
 	__u8		opt[0];
 };
 
+/* struct rs_msg - ICMPv6 Router Solicitation message format */
+struct rs_msg {
+	struct icmp6hdr	icmph;
+	__u8		opt[0];
+};
+
+/* struct ra_msg - ICMPv6 Router Advertisement message format */
+struct ra_msg {
+	struct icmp6hdr	icmph;
+	__u32	reachable_time;
+	__u32	retransmission_timer;
+	__u8	opt[0];
+};
+
 /* struct echo_msg - ICMPv6 echo request/reply message format */
 struct echo_msg {
 	struct icmp6hdr	icmph;
@@ -57,6 +71,11 @@ extern int net_nd_try;
  */
 void ndisc_init(void);
 
+/*
+ * ip6_send_rs() - Send IPv6 Router Solicitation Message
+ */
+void ip6_send_rs(void);
+
 /**
  * ndisc_receive() - Handle ND packet
  *
@@ -78,6 +97,8 @@ void ndisc_request(void);
  * Return: 0 if no timeout, -1 otherwise
  */
 int ndisc_timeout_check(void);
+bool validate_ra(struct ip6_hdr *ip6);
+int process_ra(struct ip6_hdr *ip6, int len);
 #else
 static inline void ndisc_init(void)
 {
@@ -97,6 +118,20 @@ static inline int ndisc_timeout_check(void)
 {
 	return 0;
 }
+
+static inline void ip6_send_rs(void)
+{
+}
+
+static inline bool validate_ra(struct ip6_hdr *ip6)
+{
+	return true;
+}
+
+static inline int process_ra(struct ip6_hdr *ip6, int len)
+{
+	return 0;
+}
 #endif
 
 #endif /* __NDISC_H__ */

include/net.h

@@ -508,7 +508,7 @@ extern int		net_restart_wrap;	/* Tried all network devices */
 enum proto_t {
 	BOOTP, RARP, ARP, TFTPGET, DHCP, DHCP6, PING, PING6, DNS, NFS, CDP,
 	NETCONS, SNTP, TFTPSRV, TFTPPUT, LINKLOCAL, FASTBOOT_UDP, FASTBOOT_TCP,
-	WOL, UDP, NCSI, WGET
+	WOL, UDP, NCSI, WGET, RS
 };
 
 extern char	net_boot_file_name[1024];/* Boot File name */

include/net6.h

@@ -81,8 +81,17 @@ struct udp_hdr {
 			  0x00, 0x00, 0x00, 0x00, \
 			  0x00, 0x00, 0x00, 0x00, \
 			  0x00, 0x00, 0x00, 0x00 } } }
+/*
+ * All-routers multicast address is the link-local scope address to reach all
+ * routers.
+ */
+#define ALL_ROUTERS_MULT_ADDR { { { 0xFF, 0x02, 0x00, 0x00, \
+				  0x00, 0x00, 0x00, 0x00, \
+				  0x00, 0x00, 0x00, 0x00, \
+				  0x00, 0x00, 0x00, 0x02 } } }
 
 #define IPV6_LINK_LOCAL_PREFIX	0xfe80
+#define IPV6_LINK_LOCAL_MASK	0xffb0 /* The first 10-bit of address mask. */
 
 /* hop limit for neighbour discovery packets */
 #define IPV6_NDISC_HOPLIMIT             255
@@ -166,6 +175,37 @@ struct icmp6hdr {
 #define icmp6_rt_lifetime	icmp6_dataun.u_nd_ra.rt_lifetime
 } __packed;
 
+/*
+ * struct icmp6_ra_prefix_info - Prefix Information option of the ICMPv6 message
+ * The Prefix Information option provides hosts with on-link prefixes and
+ * prefixes for Address Autoconfiguration. Refer to RFC 4861 for more info.
+ */
+struct icmp6_ra_prefix_info {
+	u8	type;		/* Type is 3 for Prefix Information. */
+	u8	len;		/* Len is 4 for Prefix Information. */
+	/* The number of leading bits in the Prefix that are valid. */
+	u8	prefix_len;
+	u8	reserved1:6,	/* MUST be ignored by the receiver. */
+		aac:1,		/* autonomous address-configuration flag */
+	/* Indicates that this prefix can be used for on-link determination. */
+		on_link:1;
+	/*
+	 * The length of time in seconds that the prefix is valid for the
+	 * purpose of on-link determination.
+	 */
+	__be32	valid_lifetime;
+	/* The length of time addresses remain preferred. */
+	__be32	preferred_lifetime;
+	__be32	reserved2;	/* MUST be ignored by the receiver. */
+	/*
+	 * Prefix is an IP address or a prefix of an IP address. The Prefix
+	 * Length field contains the number of valid leading bits in the prefix.
+	 * The bits in the prefix after the prefix length are reserved and MUST
+	 * be initialized to zero by the sender and ignored by the receiver.
+	 */
+	struct in6_addr prefix;
+};
+
 extern struct in6_addr const net_null_addr_ip6;	/* NULL IPv6 address */
 extern struct in6_addr net_gateway6;	/* Our gateways IPv6 address */
 extern struct in6_addr net_ip6;	/* Our IPv6 addr (0 = unknown) */

net/ndisc.c

@@ -13,6 +13,8 @@
 #include <net.h>
 #include <net6.h>
 #include <ndisc.h>
+#include <stdlib.h>
+#include <linux/delay.h>
 
 /* IPv6 destination address of packet waiting for ND */
 struct in6_addr net_nd_sol_packet_ip6 = ZERO_IPV6_ADDR;
@@ -29,31 +31,37 @@ int net_nd_tx_packet_size;
 ulong net_nd_timer_start;
 /* the number of requests we have sent so far */
 int net_nd_try;
+struct in6_addr all_routers = ALL_ROUTERS_MULT_ADDR;
+
+#define MAX_RTR_SOLICITATIONS		3
+/* The maximum time to delay sending the first router solicitation message. */
+#define MAX_SOLICITATION_DELAY		1 // 1 second
+/* The time to wait before sending the next router solicitation message. */
+#define RTR_SOLICITATION_INTERVAL	4000 // 4 seconds
 
 #define IP6_NDISC_OPT_SPACE(len) (((len) + 2 + 7) & ~7)
 
 /**
  * ndisc_insert_option() - Insert an option into a neighbor discovery packet
  *
- * @ndisc:	pointer to ND packet
+ * @opt:	pointer to the option element of the neighbor discovery packet
  * @type:	option type to insert
  * @data:	option data to insert
  * @len:	data length
  * Return: the number of bytes inserted (which may be >= len)
  */
-static int
+static int ndisc_insert_option(__u8 *opt, int type, u8 *data, int len)
-ndisc_insert_option(struct nd_msg *ndisc, int type, u8 *data, int len)
 {
 	int space = IP6_NDISC_OPT_SPACE(len);
 
-	ndisc->opt[0] = type;
+	opt[0] = type;
-	ndisc->opt[1] = space >> 3;
+	opt[1] = space >> 3;
-	memcpy(&ndisc->opt[2], data, len);
+	memcpy(&opt[2], data, len);
 	len += 2;
 
 	/* fill the remainder with 0 */
 	if (space - len > 0)
-		memset(&ndisc->opt[len], '\0', space - len);
+		memset(&opt[len], '\0', space - len);
 
 	return space;
 }
@@ -123,7 +131,7 @@ static void ip6_send_ns(struct in6_addr *neigh_addr)
 
 	/* Set the target address and llsaddr option */
 	net_copy_ip6(&msg->target, neigh_addr);
-	ndisc_insert_option(msg, ND_OPT_SOURCE_LL_ADDR, net_ethaddr,
+	ndisc_insert_option(msg->opt, ND_OPT_SOURCE_LL_ADDR, net_ethaddr,
 			    INETHADDRSZ);
 
 	/* checksum */
@@ -137,6 +145,76 @@ static void ip6_send_ns(struct in6_addr *neigh_addr)
 	net_send_packet(net_tx_packet, (pkt - net_tx_packet));
 }
 
+/*
+ * ip6_send_rs() - Send IPv6 Router Solicitation Message.
+ *
+ * A router solicitation is sent to discover a router. RS message creation is
+ * based on RFC 4861 section 4.1. Router Solicitation Message Format.
+ */
+void ip6_send_rs(void)
+{
+	unsigned char enetaddr[6];
+	struct rs_msg *msg;
+	__u16 icmp_len;
+	uchar *pkt;
+	unsigned short csum;
+	unsigned int pcsum;
+	static unsigned int retry_count;
+
+	if (!ip6_is_unspecified_addr(&net_gateway6) &&
+	    net_prefix_length != 0) {
+		net_set_state(NETLOOP_SUCCESS);
+		return;
+	} else if (retry_count >= MAX_RTR_SOLICITATIONS) {
+		net_set_state(NETLOOP_FAIL);
+		net_set_timeout_handler(0, NULL);
+		retry_count = 0;
+		return;
+	}
+
+	printf("ROUTER SOLICITATION %d\n", retry_count + 1);
+
+	ip6_make_mult_ethdstaddr(enetaddr, &all_routers);
+	/*
+	 * ICMP length is the size of ICMP header (8) + one option (8) = 16.
+	 * The option is 2 bytes of type and length + 6 bytes for MAC.
+	 */
+	icmp_len = sizeof(struct icmp6hdr) + IP6_NDISC_OPT_SPACE(INETHADDRSZ);
+
+	pkt = (uchar *)net_tx_packet;
+	pkt += net_set_ether(pkt, enetaddr, PROT_IP6);
+	pkt += ip6_add_hdr(pkt, &net_link_local_ip6, &all_routers, PROT_ICMPV6,
+			   IPV6_NDISC_HOPLIMIT, icmp_len);
+
+	/* ICMPv6 - RS */
+	msg = (struct rs_msg *)pkt;
+	msg->icmph.icmp6_type = IPV6_NDISC_ROUTER_SOLICITATION;
+	msg->icmph.icmp6_code = 0;
+	memset(&msg->icmph.icmp6_cksum, 0, sizeof(__be16));
+	memset(&msg->icmph.icmp6_unused, 0, sizeof(__be32));
+
+	/* Set the llsaddr option */
+	ndisc_insert_option(msg->opt, ND_OPT_SOURCE_LL_ADDR, net_ethaddr,
+			    INETHADDRSZ);
+
+	/* checksum */
+	pcsum = csum_partial((__u8 *)msg, icmp_len, 0);
+	csum = csum_ipv6_magic(&net_link_local_ip6, &all_routers,
+			       icmp_len, PROT_ICMPV6, pcsum);
+	msg->icmph.icmp6_cksum = csum;
+	pkt += icmp_len;
+
+	/* Wait up to 1 second if it is the first try to get the RA */
+	if (retry_count == 0)
+		udelay(((unsigned int)rand() % 1000000) * MAX_SOLICITATION_DELAY);
+
+	/* send it! */
+	net_send_packet(net_tx_packet, (pkt - net_tx_packet));
+
+	retry_count++;
+	net_set_timeout_handler(RTR_SOLICITATION_INTERVAL, ip6_send_rs);
+}
+
 static void
 ip6_send_na(uchar *eth_dst_addr, struct in6_addr *neigh_addr,
 	    struct in6_addr *target)
@@ -167,7 +245,7 @@ ip6_send_na(uchar *eth_dst_addr, struct in6_addr *neigh_addr,
 	msg->icmph.icmp6_dataun.u_nd_advt.override = 1;
 	/* Set the target address and lltargetaddr option */
 	net_copy_ip6(&msg->target, target);
-	ndisc_insert_option(msg, ND_OPT_TARGET_LL_ADDR, net_ethaddr,
+	ndisc_insert_option(msg->opt, ND_OPT_TARGET_LL_ADDR, net_ethaddr,
 			    INETHADDRSZ);
 
 	/* checksum */
@@ -223,6 +301,10 @@ int ndisc_timeout_check(void)
 	return 1;
 }
 
+/*
+ * ndisc_init() - Make initial steps for ND state machine.
+ * Usually move variables into initial state.
+ */
 void ndisc_init(void)
 {
 	net_nd_packet_mac = NULL;
@@ -234,12 +316,125 @@ void ndisc_init(void)
 	net_nd_tx_packet -= (ulong)net_nd_tx_packet % PKTALIGN;
 }
 
+/*
+ * validate_ra() - Validate the router advertisement message.
+ *
+ * @ip6: Pointer to the router advertisement packet
+ *
+ * Check if the router advertisement message is valid. Conditions are
+ * according to RFC 4861 section 6.1.2. Validation of Router Advertisement
+ * Messages.
+ *
+ * Return: true if the message is valid and false if it is invalid.
+ */
+bool validate_ra(struct ip6_hdr *ip6)
+{
+	struct icmp6hdr *icmp = (struct icmp6hdr *)(ip6 + 1);
+
+	/* ICMP length (derived from the IP length) should be 16 or more octets. */
+	if (ip6->payload_len < 16)
+		return false;
+
+	/* Source IP Address should be a valid link-local address. */
+	if ((ntohs(ip6->saddr.s6_addr16[0]) & IPV6_LINK_LOCAL_MASK) !=
+	    IPV6_LINK_LOCAL_PREFIX)
+		return false;
+
+	/*
+	 * The IP Hop Limit field should have a value of 255, i.e., the packet
+	 * could not possibly have been forwarded by a router.
+	 */
+	if (ip6->hop_limit != 255)
+		return false;
+
+	/* ICMP checksum has already been checked in net_ip6_handler. */
+
+	if (icmp->icmp6_code != 0)
+		return false;
+
+	return true;
+}
+
+/*
+ * process_ra() - Process the router advertisement packet.
+ *
+ * @ip6: Pointer to the router advertisement packet
+ * @len: Length of the router advertisement packet
+ *
+ * Process the received router advertisement message.
+ * Although RFC 4861 requires retaining at least two router addresses, we only
+ * keep one because of the U-Boot limitations and its goal of lightweight code.
+ *
+ * Return: 0 - RA is a default router and contains valid prefix information.
+ * Non-zero - RA options are invalid or do not indicate it is a default router
+ * or do not contain valid prefix information.
+ */
+int process_ra(struct ip6_hdr *ip6, int len)
+{
+	/* Pointer to the ICMP section of the packet */
+	struct icmp6hdr *icmp = (struct icmp6hdr *)(ip6 + 1);
+	struct ra_msg *msg = (struct ra_msg *)icmp;
+	int remaining_option_len = len - IP6_HDR_SIZE - sizeof(struct ra_msg);
+	unsigned short int option_len;	/* Length of each option */
+	/* Pointer to the ICMPv6 message options */
+	unsigned char *option = NULL;
+	/* 8-bit identifier of the type of ICMPv6 option */
+	unsigned char type = 0;
+	struct icmp6_ra_prefix_info *prefix = NULL;
+
+	/* Ignore the packet if router lifetime is 0. */
+	if (!icmp->icmp6_rt_lifetime)
+		return -EOPNOTSUPP;
+
+	/* Processing the options */
+	option = msg->opt;
+	while (remaining_option_len > 0) {
+		/* The 2nd byte of the option is its length. */
+		option_len = option[1];
+		/* All included options should have a positive length. */
+		if (option_len == 0)
+			return -EINVAL;
+
+		type = option[0];
+		/* All option types except Prefix Information are ignored. */
+		switch (type) {
+		case ND_OPT_SOURCE_LL_ADDR:
+		case ND_OPT_TARGET_LL_ADDR:
+		case ND_OPT_REDIRECT_HDR:
+		case ND_OPT_MTU:
+			break;
+		case ND_OPT_PREFIX_INFO:
+			prefix = (struct icmp6_ra_prefix_info *)option;
+			/* The link-local prefix 0xfe80::/10 is ignored. */
+			if ((ntohs(prefix->prefix.s6_addr16[0]) &
+			     IPV6_LINK_LOCAL_MASK) == IPV6_LINK_LOCAL_PREFIX)
+				break;
+			if (prefix->on_link && ntohl(prefix->valid_lifetime)) {
+				net_prefix_length = prefix->prefix_len;
+				net_gateway6 = ip6->saddr;
+				return 0;
+			}
+			break;
+		default:
+			debug("Unknown IPv6 Neighbor Discovery Option 0x%x\n",
+			      type);
+		}
+
+		option_len <<= 3; /* Option length is a multiple of 8. */
+		remaining_option_len -= option_len;
+		option += option_len;
+	}
+
+	return -EADDRNOTAVAIL;
+}
+
 int ndisc_receive(struct ethernet_hdr *et, struct ip6_hdr *ip6, int len)
 {
 	struct icmp6hdr *icmp =
 	    (struct icmp6hdr *)(((uchar *)ip6) + IP6_HDR_SIZE);
 	struct nd_msg *ndisc = (struct nd_msg *)icmp;
 	uchar neigh_eth_addr[6];
+	int err = 0;	// The error code returned calling functions.
 
 	switch (icmp->icmp6_type) {
 	case IPV6_NDISC_NEIGHBOUR_SOLICITATION:
@@ -280,6 +475,36 @@ int ndisc_receive(struct ethernet_hdr *et, struct ip6_hdr *ip6, int len)
 			net_nd_packet_mac = NULL;
 		}
 		break;
+	case IPV6_NDISC_ROUTER_SOLICITATION:
+		break;
+	case IPV6_NDISC_ROUTER_ADVERTISEMENT:
+		debug("Received router advertisement for %pI6c from %pI6c\n",
+		      &ip6->daddr, &ip6->saddr);
+		/*
+		 * If gateway and prefix are set, the RA packet is ignored. The
+		 * reason is that the U-Boot code is supposed to be as compact
+		 * as possible and does not need to take care of multiple
+		 * routers. In addition to that, U-Boot does not want to handle
+		 * scenarios like a router setting its lifetime to zero to
+		 * indicate it is not routing anymore. U-Boot program has a
+		 * short life when the system boots up and does not need such
+		 * sophistication.
+		 */
+		if (!ip6_is_unspecified_addr(&net_gateway6) &&
+		    net_prefix_length != 0) {
+			break;
+		}
+		if (!validate_ra(ip6)) {
+			debug("Invalid router advertisement message.\n");
+			break;
+		}
+		err = process_ra(ip6, len);
+		if (err)
+			debug("Ignored router advertisement. Error: %d\n", err);
+		else
+			printf("Set gatewayip6: %pI6c, prefix_length: %d\n",
+			       &net_gateway6, net_prefix_length);
+		break;
 	default:
 		debug("Unexpected ICMPv6 type 0x%x\n", icmp->icmp6_type);
 		return -1;

net/net.c

@@ -24,7 +24,7 @@
  *			- name of bootfile
  *	Next step:	ARP
  *
- * LINK_LOCAL:
+ * LINKLOCAL:
  *
  *	Prerequisites:	- own ethernet address
  *	We want:	- own IP address
@@ -124,6 +124,7 @@
 #include "wol.h"
 #endif
 #include "dhcpv6.h"
+#include "net_rand.h"
 
 /** BOOTP EXTENTIONS **/
 
@@ -350,6 +351,8 @@ void net_auto_load(void)
 
 static int net_init_loop(void)
 {
+	static bool first_call = true;
+
 	if (eth_get_dev()) {
 		memcpy(net_ethaddr, eth_get_ethaddr(), 6);
 
@@ -369,6 +372,12 @@ static int net_init_loop(void)
 		 */
 		return -ENONET;
 
+	if (IS_ENABLED(CONFIG_IPV6_ROUTER_DISCOVERY))
+		if (first_call && use_ip6) {
+			first_call = false;
+			srand_mac(); /* This is for rand used in ip6_send_rs. */
+			net_loop(RS);
+		}
 	return 0;
 }
 
@@ -587,6 +596,10 @@ restart:
 			ncsi_probe_packages();
 			break;
 #endif
+		case RS:
+			if (IS_ENABLED(CONFIG_IPV6_ROUTER_DISCOVERY))
+				ip6_send_rs();
+			break;
 		default:
 			break;
 		}
@@ -684,7 +697,13 @@ restart:
 			x = time_handler;
 			time_handler = (thand_f *)0;
 			(*x)();
-		}
+		} else if (IS_ENABLED(CONFIG_IPV6_ROUTER_DISCOVERY))
+			if (time_handler && protocol == RS)
+				if (!ip6_is_unspecified_addr(&net_gateway6) &&
+				    net_prefix_length != 0) {
+					net_set_state(NETLOOP_SUCCESS);
+					net_set_timeout_handler(0, NULL);
+				}
 
 		if (net_state == NETLOOP_FAIL)
 			ret = net_start_again();

net/net6.c

@@ -413,6 +413,7 @@ int net_ip6_handler(struct ethernet_hdr *et, struct ip6_hdr *ip6, int len)
 			break;
 		case IPV6_NDISC_NEIGHBOUR_SOLICITATION:
 		case IPV6_NDISC_NEIGHBOUR_ADVERTISEMENT:
+		case IPV6_NDISC_ROUTER_ADVERTISEMENT:
 			ndisc_receive(et, ip6, len);
 			break;
 		default: