core-lightning/daemon/routing.c

235 lines
6.2 KiB
C
Raw Normal View History

#include "lightningd.h"
#include "log.h"
#include "overflows.h"
#include "peer.h"
#include "pseudorand.h"
#include "routing.h"
#include <ccan/array_size/array_size.h>
#include <ccan/crypto/siphash24/siphash24.h>
#include <ccan/htable/htable_type.h>
#include <ccan/structeq/structeq.h>
#include <inttypes.h>
static const secp256k1_pubkey *keyof_node(const struct node *n)
{
return &n->id.pubkey;
}
static size_t hash_key(const secp256k1_pubkey *key)
{
return siphash24(siphash_seed(), key, sizeof(key));
}
static bool node_eq(const struct node *n, const secp256k1_pubkey *key)
{
return structeq(&n->id.pubkey, key);
}
HTABLE_DEFINE_TYPE(struct node, keyof_node, hash_key, node_eq, node_map);
struct node_map *empty_node_map(struct lightningd_state *dstate)
{
struct node_map *map = tal(dstate, struct node_map);
node_map_init(map);
return map;
}
struct node *get_node(struct lightningd_state *dstate,
const struct pubkey *id)
{
return node_map_get(dstate->nodes, &id->pubkey);
}
struct node *new_node(struct lightningd_state *dstate,
const struct pubkey *id)
{
struct node *n;
assert(!get_node(dstate, id));
n = tal(dstate, struct node);
n->id = *id;
n->conns = tal_arr(n, struct node_connection, 0);
node_map_add(dstate->nodes, n);
return n;
}
static struct node_connection *
get_or_make_connection(struct lightningd_state *dstate,
struct node *from, struct node *to)
{
size_t i, n = tal_count(to->conns);
for (i = 0; i < n; i++) {
if (to->conns[i].src == from) {
log_debug_struct(dstate->base_log,
"Updating existing route from %s",
struct pubkey, &from->id);
log_add_struct(dstate->base_log, " to %s",
struct pubkey, &to->id);
return &to->conns[i];
}
}
log_debug_struct(dstate->base_log, "Creating new route from %s",
struct pubkey, &from->id);
log_add_struct(dstate->base_log, " to %s", struct pubkey, &to->id);
tal_resize(&to->conns, i+1);
to->conns[i].src = from;
to->conns[i].dst = to;
return &to->conns[i];
}
/* Updates existing route if required. */
struct node_connection *add_connection(struct lightningd_state *dstate,
struct node *from, struct node *to,
u32 base_fee, s32 proportional_fee,
u32 delay, u32 min_blocks)
{
struct node_connection *c = get_or_make_connection(dstate, from, to);
c->base_fee = base_fee;
c->proportional_fee = proportional_fee;
c->delay = delay;
c->min_blocks = min_blocks;
return c;
}
/* Too big to reach, but don't overflow if added. */
#define INFINITE 0x3FFFFFFFFFFFFFFFULL
static void clear_bfg(struct node_map *nodes)
{
struct node *n;
struct node_map_iter it;
for (n = node_map_first(nodes, &it); n; n = node_map_next(nodes, &it)) {
size_t i;
for (i = 0; i < ARRAY_SIZE(n->bfg); i++)
n->bfg[i].total = INFINITE;
}
}
s64 connection_fee(const struct node_connection *c, u64 msatoshi)
{
s64 fee;
if (mul_overflows_s64(c->proportional_fee, msatoshi))
return INFINITE;
fee = (c->proportional_fee * msatoshi) / 1000000;
/* This can't overflow: c->base_fee is a u32 */
return c->base_fee + fee;
}
/* We track totals, rather than costs. That's because the fee depends
* on the current amount passing through. */
static void bfg_one_edge(struct node *node, size_t edgenum)
{
struct node_connection *c = &node->conns[edgenum];
size_t h;
assert(c->dst == node);
for (h = 0; h < ROUTING_MAX_HOPS; h++) {
/* FIXME: Bias towards smaller expiry values. */
/* FIXME: Bias against smaller channels. */
s64 fee = connection_fee(c, node->bfg[h].total);
if (node->bfg[h].total + fee < c->src->bfg[h+1].total) {
c->src->bfg[h+1].total = node->bfg[h].total + fee;
c->src->bfg[h+1].prev = c;
}
}
}
struct peer *find_route(struct lightningd_state *dstate,
const struct pubkey *to,
u64 msatoshi, s64 *fee,
struct node_connection ***route)
{
struct node *n, *src, *dst;
struct node_map_iter it;
struct peer *first;
int runs, i, best;
/* Note: we map backwards, since we know the amount of satoshi we want
* at the end, and need to derive how much we need to send. */
dst = get_node(dstate, &dstate->id);
src = get_node(dstate, to);
if (!src) {
log_info_struct(dstate->base_log, "find_route: cannot find %s",
struct pubkey, to);
return NULL;
}
/* Reset all the information. */
clear_bfg(dstate->nodes);
/* Bellman-Ford-Gibson: like Bellman-Ford, but keep values for
* every path length. */
src->bfg[0].total = msatoshi;
for (runs = 0; runs < ROUTING_MAX_HOPS; runs++) {
/* Run through every edge. */
for (n = node_map_first(dstate->nodes, &it);
n;
n = node_map_next(dstate->nodes, &it)) {
size_t num_edges = tal_count(n->conns);
for (i = 0; i < num_edges; i++) {
bfg_one_edge(n, i);
}
}
}
best = 0;
for (i = 1; i <= ROUTING_MAX_HOPS; i++) {
if (dst->bfg[i].total < dst->bfg[best].total)
best = i;
}
/* No route? */
if (dst->bfg[best].total >= INFINITE) {
log_info_struct(dstate->base_log, "find_route: No route to %s",
struct pubkey, to);
return NULL;
}
/* Save route from *next* hop (we return first hop as peer).
* Note that we take our own fees into account for routing, even
* though we don't pay them: it presumably effects preference. */
dst = dst->bfg[best].prev->dst;
best--;
*fee = dst->bfg[best].total - msatoshi;
*route = tal_arr(dstate, struct node_connection *, best);
for (i = 0, n = dst;
i < best;
n = n->bfg[best-i].prev->dst, i++) {
(*route)[i] = n->bfg[best-i].prev;
}
assert(n == src);
/* We should only add routes if we have a peer. */
first = find_peer(dstate, &(*route)[0]->src->id);
if (!first) {
log_broken_struct(dstate->base_log, "No peer %s?",
struct pubkey, &(*route)[0]->src->id);
return NULL;
}
msatoshi += *fee;
log_info(dstate->base_log, "find_route:");
log_add_struct(dstate->base_log, "via %s", struct pubkey, &first->id);
for (i = 0; i < best; i++) {
log_add_struct(dstate->base_log, " %s",
struct pubkey, &(*route)[i]->dst->id);
log_add(dstate->base_log, "(%i+%i=%"PRIu64")",
(*route)[i]->base_fee,
(*route)[i]->proportional_fee,
connection_fee((*route)[i], msatoshi));
msatoshi -= connection_fee((*route)[i], msatoshi);
}
log_add(dstate->base_log, "=%"PRIi64"(%+"PRIi64")",
(*route)[best-1]->dst->bfg[best-1].total, *fee);
return first;
}