bird/proto/rip/packets.c
Ondrej Zajicek (work) 586c1800c4 Nest: Neighbor cache cleanups
Simplify neighbor cache code, fix several minor bugs, and improve
handling of ONLINK flag.
2018-06-27 16:57:07 +02:00

794 lines
20 KiB
C

/*
* BIRD -- Routing Information Protocol (RIP)
*
* (c) 1998--1999 Pavel Machek <pavel@ucw.cz>
* (c) 2004--2013 Ondrej Filip <feela@network.cz>
* (c) 2009--2015 Ondrej Zajicek <santiago@crfreenet.org>
* (c) 2009--2015 CZ.NIC z.s.p.o.
*
* Can be freely distributed and used under the terms of the GNU GPL.
*/
#undef LOCAL_DEBUG
#include "rip.h"
#include "lib/mac.h"
#define RIP_CMD_REQUEST 1 /* want info */
#define RIP_CMD_RESPONSE 2 /* responding to request */
#define RIP_BLOCK_LENGTH 20
#define RIP_PASSWD_LENGTH 16
#define RIP_AF_IPV4 2
#define RIP_AF_AUTH 0xffff
/* RIP packet header */
struct rip_packet
{
u8 command;
u8 version;
u16 unused;
};
/* RTE block for RIPv2 */
struct rip_block_v2
{
u16 family;
u16 tag;
ip4_addr network;
ip4_addr netmask;
ip4_addr next_hop;
u32 metric;
};
/* RTE block for RIPng */
struct rip_block_ng
{
ip6_addr prefix;
u16 tag;
u8 pxlen;
u8 metric;
};
/* Authentication block for RIPv2 */
struct rip_block_auth
{
u16 must_be_ffff;
u16 auth_type;
char password[0];
u16 packet_len;
u8 key_id;
u8 auth_len;
u32 seq_num;
u32 unused1;
u32 unused2;
};
/* Authentication tail, RFC 4822 */
struct rip_auth_tail
{
u16 must_be_ffff;
u16 must_be_0001;
byte auth_data[0];
};
/* Internal representation of RTE block data */
struct rip_block
{
net_addr net;
u32 metric;
u16 tag;
u16 no_af;
ip_addr next_hop;
};
#define DROP(DSC,VAL) do { err_dsc = DSC; err_val = VAL; goto drop; } while(0)
#define DROP1(DSC) do { err_dsc = DSC; goto drop; } while(0)
#define SKIP(DSC) do { err_dsc = DSC; goto skip; } while(0)
#define LOG_PKT(msg, args...) \
log_rl(&p->log_pkt_tbf, L_REMOTE "%s: " msg, p->p.name, args)
#define LOG_PKT_AUTH(msg, args...) \
log_rl(&p->log_pkt_tbf, L_AUTH "%s: " msg, p->p.name, args)
#define LOG_RTE(msg, args...) \
log_rl(&p->log_rte_tbf, L_REMOTE "%s: " msg, p->p.name, args)
static inline void * rip_tx_buffer(struct rip_iface *ifa)
{ return ifa->sk->tbuf; }
static inline uint rip_pkt_hdrlen(struct rip_iface *ifa)
{ return sizeof(struct rip_packet) + (ifa->cf->auth_type ? RIP_BLOCK_LENGTH : 0); }
static inline void
rip_put_block(struct rip_proto *p, byte *pos, struct rip_block *rte)
{
if (rip_is_v2(p))
{
struct rip_block_v2 *block = (void *) pos;
block->family = rte->no_af ? 0 : htons(RIP_AF_IPV4);
block->tag = htons(rte->tag);
block->network = ip4_hton(net4_prefix(&rte->net));
block->netmask = ip4_hton(ip4_mkmask(net4_pxlen(&rte->net)));
block->next_hop = ip4_hton(ipa_to_ip4(rte->next_hop));
block->metric = htonl(rte->metric);
}
else /* RIPng */
{
struct rip_block_ng *block = (void *) pos;
block->prefix = ip6_hton(net6_prefix(&rte->net));
block->tag = htons(rte->tag);
block->pxlen = net6_pxlen(&rte->net);
block->metric = rte->metric;
}
}
static inline void
rip_put_next_hop(struct rip_proto *p UNUSED, byte *pos, struct rip_block *rte)
{
struct rip_block_ng *block = (void *) pos;
block->prefix = ip6_hton(ipa_to_ip6(rte->next_hop));
block->tag = 0;
block->pxlen = 0;
block->metric = 0xff;
}
static inline int
rip_get_block(struct rip_proto *p, byte *pos, struct rip_block *rte)
{
if (rip_is_v2(p))
{
struct rip_block_v2 *block = (void *) pos;
/* Skip blocks with strange AF, including authentication blocks */
if (block->family != (rte->no_af ? 0 : htons(RIP_AF_IPV4)))
return 0;
uint pxlen = ip4_masklen(ip4_ntoh(block->netmask));
net_fill_ip4(&rte->net, ip4_ntoh(block->network), pxlen);
rte->metric = ntohl(block->metric);
rte->tag = ntohs(block->tag);
rte->next_hop = ipa_from_ip4(ip4_ntoh(block->next_hop));
return 1;
}
else /* RIPng */
{
struct rip_block_ng *block = (void *) pos;
/* Handle and skip next hop blocks */
if (block->metric == 0xff)
{
rte->next_hop = ipa_from_ip6(ip6_ntoh(block->prefix));
if (!ipa_is_link_local(rte->next_hop)) rte->next_hop = IPA_NONE;
return 0;
}
uint pxlen = (block->pxlen <= IP6_MAX_PREFIX_LENGTH) ? block->pxlen : 255;
net_fill_ip6(&rte->net, ip6_ntoh(block->prefix), pxlen);
rte->metric = block->metric;
rte->tag = ntohs(block->tag);
/* rte->next_hop is deliberately kept unmodified */;
return 1;
}
}
static inline void
rip_update_csn(struct rip_proto *p UNUSED, struct rip_iface *ifa)
{
/*
* We update crypto sequence numbers at the beginning of update session to
* avoid issues with packet reordering, so packets inside one update session
* have the same CSN. We are using real time, but enforcing monotonicity.
*/
if (ifa->cf->auth_type == RIP_AUTH_CRYPTO)
{
u32 now_real = (u32) (current_real_time() TO_S);
ifa->csn = (ifa->csn < now_real) ? now_real : ifa->csn + 1;
}
}
static void
rip_fill_authentication(struct rip_proto *p, struct rip_iface *ifa, struct rip_packet *pkt, uint *plen)
{
struct rip_block_auth *auth = (void *) (pkt + 1);
struct password_item *pass = password_find(ifa->cf->passwords, 0);
if (!pass)
{
/* FIXME: This should not happen */
log(L_ERR "%s: No suitable password found for authentication", p->p.name);
memset(auth, 0, sizeof(struct rip_block_auth));
return;
}
switch (ifa->cf->auth_type)
{
case RIP_AUTH_PLAIN:
auth->must_be_ffff = htons(0xffff);
auth->auth_type = htons(RIP_AUTH_PLAIN);
strncpy(auth->password, pass->password, RIP_PASSWD_LENGTH);
return;
case RIP_AUTH_CRYPTO:
auth->must_be_ffff = htons(0xffff);
auth->auth_type = htons(RIP_AUTH_CRYPTO);
auth->packet_len = htons(*plen);
auth->key_id = pass->id;
auth->auth_len = mac_type_length(pass->alg);
auth->seq_num = ifa->csn_ready ? htonl(ifa->csn) : 0;
auth->unused1 = 0;
auth->unused2 = 0;
ifa->csn_ready = 1;
if (pass->alg < ALG_HMAC)
auth->auth_len += sizeof(struct rip_auth_tail);
/*
* Note that RFC 4822 is unclear whether auth_len should cover whole
* authentication trailer or just auth_data length.
*
* FIXME: We should use just auth_data length by default. Currently we put
* the whole auth trailer length in keyed hash case to keep old behavior,
* but we put just auth_data length in the new HMAC case. Note that Quagga
* has config option for this.
*
* Crypto sequence numbers are increased by sender in rip_update_csn().
* First CSN should be zero, this is handled by csn_ready.
*/
struct rip_auth_tail *tail = (void *) ((byte *) pkt + *plen);
tail->must_be_ffff = htons(0xffff);
tail->must_be_0001 = htons(0x0001);
uint auth_len = mac_type_length(pass->alg);
*plen += sizeof(struct rip_auth_tail) + auth_len;
/* Append key for keyed hash, append padding for HMAC (RFC 4822 2.5) */
if (pass->alg < ALG_HMAC)
strncpy(tail->auth_data, pass->password, auth_len);
else
memset32(tail->auth_data, HMAC_MAGIC, auth_len / 4);
mac_fill(pass->alg, pass->password, pass->length,
(byte *) pkt, *plen, tail->auth_data);
return;
default:
bug("Unknown authentication type");
}
}
static int
rip_check_authentication(struct rip_proto *p, struct rip_iface *ifa, struct rip_packet *pkt, uint *plen, struct rip_neighbor *n)
{
struct rip_block_auth *auth = (void *) (pkt + 1);
struct password_item *pass = NULL;
const char *err_dsc = NULL;
uint err_val = 0;
uint auth_type = 0;
/* Check for authentication entry */
if ((*plen >= (sizeof(struct rip_packet) + sizeof(struct rip_block_auth))) &&
(auth->must_be_ffff == htons(0xffff)))
auth_type = ntohs(auth->auth_type);
if (auth_type != ifa->cf->auth_type)
DROP("authentication method mismatch", auth_type);
switch (auth_type)
{
case RIP_AUTH_NONE:
return 1;
case RIP_AUTH_PLAIN:
pass = password_find_by_value(ifa->cf->passwords, auth->password, RIP_PASSWD_LENGTH);
if (!pass)
DROP1("wrong password");
return 1;
case RIP_AUTH_CRYPTO:
pass = password_find_by_id(ifa->cf->passwords, auth->key_id);
if (!pass)
DROP("no suitable password found", auth->key_id);
uint data_len = ntohs(auth->packet_len);
uint auth_len = mac_type_length(pass->alg);
uint auth_len2 = sizeof(struct rip_auth_tail) + auth_len;
/*
* Ideally, first check should be check for internal consistency:
* (data_len + sizeof(struct rip_auth_tail) + auth->auth_len) != *plen
*
* Second one should check expected code length:
* auth->auth_len != auth_len
*
* But as auth->auth_len has two interpretations, we simplify this
*/
if (data_len + auth_len2 != *plen)
DROP("packet length mismatch", *plen);
/* Warning: two interpretations of auth_len field */
if ((auth->auth_len != auth_len) && (auth->auth_len != auth_len2))
DROP("wrong authentication length", auth->auth_len);
struct rip_auth_tail *tail = (void *) ((byte *) pkt + data_len);
if ((tail->must_be_ffff != htons(0xffff)) || (tail->must_be_0001 != htons(0x0001)))
DROP1("authentication trailer is missing");
/* Accept higher sequence number, or zero if connectivity is lost */
/* FIXME: sequence number must be password/SA specific */
u32 rcv_csn = ntohl(auth->seq_num);
if ((rcv_csn < n->csn) && (rcv_csn || n->uc))
{
/* We want to report both new and old CSN */
LOG_PKT_AUTH("Authentication failed for %I on %s - "
"lower sequence number (rcv %u, old %u)",
n->nbr->addr, ifa->iface->name, rcv_csn, n->csn);
return 0;
}
byte *auth_data = alloca(auth_len);
memcpy(auth_data, tail->auth_data, auth_len);
/* Append key for keyed hash, append padding for HMAC (RFC 4822 2.5) */
if (pass->alg < ALG_HMAC)
strncpy(tail->auth_data, pass->password, auth_len);
else
memset32(tail->auth_data, HMAC_MAGIC, auth_len / 4);
if (!mac_verify(pass->alg, pass->password, pass->length,
(byte *) pkt, *plen, auth_data))
DROP("wrong authentication code", pass->id);
*plen = data_len;
n->csn = rcv_csn;
return 1;
}
drop:
LOG_PKT_AUTH("Authentication failed for %I on %s - %s (%u)",
n->nbr->addr, ifa->iface->name, err_dsc, err_val);
return 0;
}
static inline int
rip_send_to(struct rip_proto *p, struct rip_iface *ifa, struct rip_packet *pkt, uint plen, ip_addr dst)
{
if (ifa->cf->auth_type)
rip_fill_authentication(p, ifa, pkt, &plen);
return sk_send_to(ifa->sk, plen, dst, 0);
}
void
rip_send_request(struct rip_proto *p, struct rip_iface *ifa)
{
byte *pos = rip_tx_buffer(ifa);
struct rip_packet *pkt = (void *) pos;
pkt->command = RIP_CMD_REQUEST;
pkt->version = ifa->cf->version;
pkt->unused = 0;
pos += rip_pkt_hdrlen(ifa);
struct rip_block b = { .no_af = 1, .metric = p->infinity };
rip_put_block(p, pos, &b);
pos += RIP_BLOCK_LENGTH;
rip_update_csn(p, ifa);
TRACE(D_PACKETS, "Sending request via %s", ifa->iface->name);
rip_send_to(p, ifa, pkt, pos - (byte *) pkt, ifa->addr);
}
static void
rip_receive_request(struct rip_proto *p, struct rip_iface *ifa, struct rip_packet *pkt, uint plen, struct rip_neighbor *from)
{
TRACE(D_PACKETS, "Request received from %I on %s", from->nbr->addr, ifa->iface->name);
byte *pos = (byte *) pkt + rip_pkt_hdrlen(ifa);
/* We expect one regular block */
if (plen != (rip_pkt_hdrlen(ifa) + RIP_BLOCK_LENGTH))
return;
struct rip_block b = { .no_af = 1 };
if (!rip_get_block(p, pos, &b))
return;
/* Special case - infinity metric, for RIPng also zero prefix */
if ((b.metric != p->infinity) ||
(rip_is_ng(p) && !net_zero_ip6((net_addr_ip6 *) &b.net)))
return;
/* We do nothing if TX is already active */
if (ifa->tx_active)
{
TRACE(D_EVENTS, "Skipping request from %I on %s, TX is busy", from->nbr->addr, ifa->iface->name);
return;
}
if (!ifa->cf->passive)
rip_send_table(p, ifa, from->nbr->addr, 0);
}
static int
rip_send_response(struct rip_proto *p, struct rip_iface *ifa)
{
if (! ifa->tx_active)
return 0;
byte *pos = rip_tx_buffer(ifa);
byte *max = rip_tx_buffer(ifa) + ifa->tx_plen -
(rip_is_v2(p) ? RIP_BLOCK_LENGTH : 2*RIP_BLOCK_LENGTH);
ip_addr last_next_hop = IPA_NONE;
btime now_ = current_time();
int send = 0;
struct rip_packet *pkt = (void *) pos;
pkt->command = RIP_CMD_RESPONSE;
pkt->version = ifa->cf->version;
pkt->unused = 0;
pos += rip_pkt_hdrlen(ifa);
FIB_ITERATE_START(&p->rtable, &ifa->tx_fit, struct rip_entry, en)
{
/* Dummy entries */
if (!en->valid)
goto next_entry;
/* Stale entries that should be removed */
if ((en->valid == RIP_ENTRY_STALE) &&
((en->changed + ifa->cf->garbage_time) <= now_))
goto next_entry;
/* Triggered updates */
if (en->changed < ifa->tx_changed)
goto next_entry;
/* Not enough space for current entry */
if (pos > max)
{
FIB_ITERATE_PUT(&ifa->tx_fit);
goto break_loop;
}
struct rip_block rte = {
.metric = en->metric,
.tag = en->tag
};
net_copy(&rte.net, en->n.addr);
if (en->iface == ifa->iface)
rte.next_hop = en->next_hop;
if (rip_is_v2(p) && (ifa->cf->version == RIP_V1))
{
/* Skipping subnets (i.e. not hosts, classful networks or default route) */
if (ip4_masklen(ip4_class_mask(net4_prefix(&rte.net))) != rte.net.pxlen)
goto next_entry;
rte.tag = 0;
rte.net.pxlen = 0;
rte.next_hop = IPA_NONE;
}
/* Split horizon */
if (en->from == ifa->iface && ifa->cf->split_horizon)
{
if (ifa->cf->poison_reverse)
{
rte.metric = p->infinity;
rte.next_hop = IPA_NONE;
}
else
goto next_entry;
}
// TRACE(D_PACKETS, " %N -> %I metric %d", &rte.net, rte.next_hop, rte.metric);
/* RIPng next hop entry */
if (rip_is_ng(p) && !ipa_equal(rte.next_hop, last_next_hop))
{
last_next_hop = rte.next_hop;
rip_put_next_hop(p, pos, &rte);
pos += RIP_BLOCK_LENGTH;
}
rip_put_block(p, pos, &rte);
pos += RIP_BLOCK_LENGTH;
send = 1;
next_entry: ;
}
FIB_ITERATE_END;
ifa->tx_active = 0;
/* Do not send empty packet */
if (!send)
return 0;
break_loop:
TRACE(D_PACKETS, "Sending response via %s", ifa->iface->name);
return rip_send_to(p, ifa, pkt, pos - (byte *) pkt, ifa->tx_addr);
}
/**
* rip_send_table - RIP interface timer hook
* @p: RIP instance
* @ifa: RIP interface
* @addr: destination IP address
* @changed: time limit for triggered updates
*
* The function activates an update session and starts sending routing update
* packets (using rip_send_response()). The session may be finished during the
* call or may continue in rip_tx_hook() until all appropriate routes are
* transmitted. Note that there may be at most one active update session per
* interface, the function will terminate the old active session before
* activating the new one.
*/
void
rip_send_table(struct rip_proto *p, struct rip_iface *ifa, ip_addr addr, btime changed)
{
DBG("RIP: Opening TX session to %I on %s\n", addr, ifa->iface->name);
rip_reset_tx_session(p, ifa);
ifa->tx_active = 1;
ifa->tx_addr = addr;
ifa->tx_changed = changed;
FIB_ITERATE_INIT(&ifa->tx_fit, &p->rtable);
rip_update_csn(p, ifa);
while (rip_send_response(p, ifa) > 0)
;
}
static void
rip_tx_hook(sock *sk)
{
struct rip_iface *ifa = sk->data;
struct rip_proto *p = ifa->rip;
DBG("RIP: TX hook called (iface %s, src %I, dst %I)\n",
sk->iface->name, sk->saddr, sk->daddr);
while (rip_send_response(p, ifa) > 0)
;
}
static void
rip_err_hook(sock *sk, int err)
{
struct rip_iface *ifa = sk->data;
struct rip_proto *p = ifa->rip;
log(L_ERR "%s: Socket error on %s: %M", p->p.name, ifa->iface->name, err);
rip_reset_tx_session(p, ifa);
}
static void
rip_receive_response(struct rip_proto *p, struct rip_iface *ifa, struct rip_packet *pkt, uint plen, struct rip_neighbor *from)
{
struct rip_block rte = {};
const char *err_dsc = NULL;
TRACE(D_PACKETS, "Response received from %I on %s", from->nbr->addr, ifa->iface->name);
byte *pos = (byte *) pkt + sizeof(struct rip_packet);
byte *end = (byte *) pkt + plen;
btime now_ = current_time();
for (; pos < end; pos += RIP_BLOCK_LENGTH)
{
/* Find next regular RTE */
if (!rip_get_block(p, pos, &rte))
continue;
if (rip_is_v2(p) && (pkt->version == RIP_V1))
{
if (ifa->cf->check_zero && (rte.tag || rte.net.pxlen || ipa_nonzero(rte.next_hop)))
SKIP("RIPv1 reserved field is nonzero");
rte.tag = 0;
rte.net.pxlen = ip4_masklen(ip4_class_mask(net4_prefix(&rte.net)));
rte.next_hop = IPA_NONE;
}
if (rte.net.pxlen == 255)
SKIP("invalid prefix length");
net_normalize(&rte.net);
int c = net_classify(&rte.net);
if ((c < 0) || !(c & IADDR_HOST) || ((c & IADDR_SCOPE_MASK) <= SCOPE_LINK))
SKIP("invalid prefix");
if (rte.metric > p->infinity)
SKIP("invalid metric");
if (ipa_nonzero(rte.next_hop))
{
neighbor *nbr = neigh_find(&p->p, rte.next_hop, ifa->iface, 0);
if (!nbr || (nbr->scope <= 0))
rte.next_hop = IPA_NONE;
}
// TRACE(D_PACKETS, " %N -> %I metric %d", &rte.net.n, rte.next_hop, rte.metric);
rte.metric += ifa->cf->metric;
if (rte.metric < p->infinity)
{
struct rip_rte new = {
.from = from,
.next_hop = ipa_nonzero(rte.next_hop) ? rte.next_hop : from->nbr->addr,
.metric = rte.metric,
.tag = rte.tag,
.expires = now_ + ifa->cf->timeout_time
};
rip_update_rte(p, &rte.net, &new);
}
else
rip_withdraw_rte(p, &rte.net, from);
continue;
skip:
LOG_RTE("Ignoring route %N received from %I - %s",
&rte.net, from->nbr->addr, err_dsc);
}
}
static int
rip_rx_hook(sock *sk, uint len)
{
struct rip_iface *ifa = sk->data;
struct rip_proto *p = ifa->rip;
const char *err_dsc = NULL;
uint err_val = 0;
if (sk->lifindex != sk->iface->index)
return 1;
DBG("RIP: RX hook called (iface %s, src %I, dst %I)\n",
sk->iface->name, sk->faddr, sk->laddr);
/* Silently ignore my own packets */
if (ipa_equal(sk->faddr, sk->saddr))
return 1;
if (rip_is_ng(p) && !ipa_is_link_local(sk->faddr))
DROP1("wrong src address");
struct rip_neighbor *n = rip_get_neighbor(p, &sk->faddr, ifa);
if (!n)
DROP1("not from neighbor");
if ((ifa->cf->ttl_security == 1) && (sk->rcv_ttl < 255))
DROP("wrong TTL", sk->rcv_ttl);
if (sk->fport != sk->dport)
DROP("wrong src port", sk->fport);
if (len < sizeof(struct rip_packet))
DROP("too short", len);
if (sk->flags & SKF_TRUNCATED)
DROP("truncated", len);
struct rip_packet *pkt = (struct rip_packet *) sk->rbuf;
uint plen = len;
if (!pkt->version || (ifa->cf->version_only && (pkt->version != ifa->cf->version)))
DROP("wrong version", pkt->version);
/* rip_check_authentication() has its own error logging */
if (rip_is_v2(p) && !rip_check_authentication(p, ifa, pkt, &plen, n))
return 1;
if ((plen - sizeof(struct rip_packet)) % RIP_BLOCK_LENGTH)
DROP("invalid length", plen);
n->last_seen = current_time();
rip_update_bfd(p, n);
switch (pkt->command)
{
case RIP_CMD_REQUEST:
rip_receive_request(p, ifa, pkt, plen, n);
break;
case RIP_CMD_RESPONSE:
rip_receive_response(p, ifa, pkt, plen, n);
break;
default:
DROP("unknown command", pkt->command);
}
return 1;
drop:
LOG_PKT("Bad packet from %I via %s - %s (%u)",
sk->faddr, sk->iface->name, err_dsc, err_val);
return 1;
}
int
rip_open_socket(struct rip_iface *ifa)
{
struct rip_proto *p = ifa->rip;
sock *sk = sk_new(p->p.pool);
sk->type = SK_UDP;
sk->subtype = rip_is_v2(p) ? SK_IPV4 : SK_IPV6;
sk->sport = ifa->cf->port;
sk->dport = ifa->cf->port;
sk->iface = ifa->iface;
sk->saddr = rip_is_v2(p) ? ifa->iface->addr4->ip : ifa->iface->llv6->ip;
sk->vrf = p->p.vrf;
sk->rx_hook = rip_rx_hook;
sk->tx_hook = rip_tx_hook;
sk->err_hook = rip_err_hook;
sk->data = ifa;
sk->tos = ifa->cf->tx_tos;
sk->priority = ifa->cf->tx_priority;
sk->ttl = ifa->cf->ttl_security ? 255 : 1;
sk->flags = SKF_LADDR_RX | ((ifa->cf->ttl_security == 1) ? SKF_TTL_RX : 0);
/* sk->rbsize and sk->tbsize are handled in rip_iface_update_buffers() */
if (sk_open(sk) < 0)
goto err;
if (ifa->cf->mode == RIP_IM_MULTICAST)
{
if (sk_setup_multicast(sk) < 0)
goto err;
if (sk_join_group(sk, ifa->addr) < 0)
goto err;
}
else /* Broadcast */
{
if (sk_setup_broadcast(sk) < 0)
goto err;
if (ipa_zero(ifa->addr))
{
sk->err = "Missing broadcast address";
goto err;
}
}
ifa->sk = sk;
return 1;
err:
sk_log_error(sk, p->p.name);
rfree(sk);
return 0;
}