tor/src/or/connection_or.c

1573 lines
54 KiB
C
Raw Normal View History

2006-02-09 06:46:49 +01:00
/* Copyright (c) 2001 Matej Pfajfar.
* Copyright (c) 2001-2004, Roger Dingledine.
* Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson.
2010-02-27 23:13:37 +01:00
* Copyright (c) 2007-2010, The Tor Project, Inc. */
/* See LICENSE for licensing information */
2002-06-27 00:45:49 +02:00
/**
* \file connection_or.c
* \brief Functions to handle OR connections, TLS handshaking, and
* cells on the network.
**/
2002-06-27 00:45:49 +02:00
#include "or.h"
2010-07-22 00:46:18 +02:00
#include "buffers.h"
2010-07-22 01:21:00 +02:00
#include "circuitbuild.h"
2010-07-22 10:08:32 +02:00
#include "command.h"
2010-07-22 10:22:51 +02:00
#include "config.h"
2010-07-22 10:32:52 +02:00
#include "connection.h"
2010-07-22 10:50:34 +02:00
#include "connection_or.h"
2010-07-22 11:35:09 +02:00
#include "control.h"
2010-07-22 12:09:49 +02:00
#include "dirserv.h"
2010-07-21 14:38:52 +02:00
#include "geoip.h"
2010-07-23 19:58:06 +02:00
#include "main.h"
2010-07-23 20:18:55 +02:00
#include "networkstatus.h"
Initial conversion to use node_t throughout our codebase. A node_t is an abstraction over routerstatus_t, routerinfo_t, and microdesc_t. It should try to present a consistent interface to all of them. There should be a node_t for a server whenever there is * A routerinfo_t for it in the routerlist * A routerstatus_t in the current_consensus. (note that a microdesc_t alone isn't enough to make a node_t exist, since microdescriptors aren't usable on their own.) There are three ways to get a node_t right now: looking it up by ID, looking it up by nickname, and iterating over the whole list of microdescriptors. All (or nearly all) functions that are supposed to return "a router" -- especially those used in building connections and circuits -- should return a node_t, not a routerinfo_t or a routerstatus_t. A node_t should hold all the *mutable* flags about a node. This patch moves the is_foo flags from routerinfo_t into node_t. The flags in routerstatus_t remain, but they get set from the consensus and should not change. Some other highlights of this patch are: * Looking up routerinfo and routerstatus by nickname is now unified and based on the "look up a node by nickname" function. This tries to look only at the values from current consensus, and not get confused by the routerinfo_t->is_named flag, which could get set for other weird reasons. This changes the behavior of how authorities (when acting as clients) deal with nodes that have been listed by nickname. * I tried not to artificially increase the size of the diff here by moving functions around. As a result, some functions that now operate on nodes are now in the wrong file -- they should get moved to nodelist.c once this refactoring settles down. This moving should happen as part of a patch that moves functions AND NOTHING ELSE. * Some old code is now left around inside #if 0/1 blocks, and should get removed once I've verified that I don't want it sitting around to see how we used to do things. There are still some unimplemented functions: these are flagged with "UNIMPLEMENTED_NODELIST()." I'll work on filling in the implementation here, piece by piece. I wish this patch could have been smaller, but there did not seem to be any piece of it that was independent from the rest. Moving flags forces many functions that once returned routerinfo_t * to return node_t *, which forces their friends to change, and so on.
2010-09-29 21:00:41 +02:00
#include "nodelist.h"
2010-07-23 21:08:30 +02:00
#include "reasons.h"
2010-07-23 21:53:11 +02:00
#include "relay.h"
2010-07-23 22:57:20 +02:00
#include "rephist.h"
2010-07-21 16:17:10 +02:00
#include "router.h"
2010-07-21 17:08:11 +02:00
#include "routerlist.h"
#ifdef USE_BUFFEREVENTS
#include <event2/bufferevent_ssl.h>
#endif
static int connection_tls_finish_handshake(or_connection_t *conn);
static int connection_or_process_cells_from_inbuf(or_connection_t *conn);
static int connection_or_send_versions(or_connection_t *conn);
static int connection_init_or_handshake_state(or_connection_t *conn,
int started_here);
static int connection_or_check_valid_tls_handshake(or_connection_t *conn,
int started_here,
char *digest_rcvd_out);
static void connection_or_tls_renegotiated_cb(tor_tls_t *tls, void *_conn);
#ifdef USE_BUFFEREVENTS
static void connection_or_handle_event_cb(struct bufferevent *bufev,
short event, void *arg);
#include <event2/buffer.h>/*XXXX REMOVE */
#endif
2003-09-16 07:41:49 +02:00
/**************************************************************/
/** Map from identity digest of connected OR or desired OR to a connection_t
* with that identity digest. If there is more than one such connection_t,
2006-06-13 07:50:24 +02:00
* they form a linked list, with next_with_same_id as the next pointer. */
static digestmap_t *orconn_identity_map = NULL;
/** If conn is listed in orconn_identity_map, remove it, and clear
* conn->identity_digest. Otherwise do nothing. */
void
connection_or_remove_from_identity_map(or_connection_t *conn)
{
or_connection_t *tmp;
tor_assert(conn);
if (!orconn_identity_map)
return;
tmp = digestmap_get(orconn_identity_map, conn->identity_digest);
if (!tmp) {
if (!tor_digest_is_zero(conn->identity_digest)) {
log_warn(LD_BUG, "Didn't find connection '%s' on identity map when "
"trying to remove it.",
conn->nickname ? conn->nickname : "NULL");
}
return;
}
if (conn == tmp) {
if (conn->next_with_same_id)
digestmap_set(orconn_identity_map, conn->identity_digest,
conn->next_with_same_id);
else
digestmap_remove(orconn_identity_map, conn->identity_digest);
} else {
while (tmp->next_with_same_id) {
if (tmp->next_with_same_id == conn) {
tmp->next_with_same_id = conn->next_with_same_id;
break;
}
tmp = tmp->next_with_same_id;
}
}
memset(conn->identity_digest, 0, DIGEST_LEN);
conn->next_with_same_id = NULL;
}
/** Remove all entries from the identity-to-orconn map, and clear
* all identities in OR conns.*/
void
connection_or_clear_identity_map(void)
{
smartlist_t *conns = get_connection_array();
SMARTLIST_FOREACH(conns, connection_t *, conn,
{
if (conn->type == CONN_TYPE_OR) {
or_connection_t *or_conn = TO_OR_CONN(conn);
memset(or_conn->identity_digest, 0, DIGEST_LEN);
or_conn->next_with_same_id = NULL;
}
});
digestmap_free(orconn_identity_map, NULL);
orconn_identity_map = NULL;
}
/** Change conn->identity_digest to digest, and add conn into
* orconn_digest_map. */
static void
connection_or_set_identity_digest(or_connection_t *conn, const char *digest)
{
or_connection_t *tmp;
tor_assert(conn);
tor_assert(digest);
if (!orconn_identity_map)
orconn_identity_map = digestmap_new();
if (!memcmp(conn->identity_digest, digest, DIGEST_LEN))
return;
/* If the identity was set previously, remove the old mapping. */
if (! tor_digest_is_zero(conn->identity_digest))
connection_or_remove_from_identity_map(conn);
memcpy(conn->identity_digest, digest, DIGEST_LEN);
/* If we're setting the ID to zero, don't add a mapping. */
if (tor_digest_is_zero(digest))
return;
tmp = digestmap_set(orconn_identity_map, digest, conn);
conn->next_with_same_id = tmp;
#if 1
/* Testing code to check for bugs in representation. */
for (; tmp; tmp = tmp->next_with_same_id) {
tor_assert(!memcmp(tmp->identity_digest, digest, DIGEST_LEN));
tor_assert(tmp != conn);
}
#endif
}
/** Pack the cell_t host-order structure <b>src</b> into network-order
* in the buffer <b>dest</b>. See tor-spec.txt for details about the
* wire format.
2007-05-25 21:41:31 +02:00
*
* Note that this function doesn't touch <b>dst</b>-\>next: the caller
* should set it or clear it as appropriate.
*/
void
cell_pack(packed_cell_t *dst, const cell_t *src)
{
char *dest = dst->body;
set_uint16(dest, htons(src->circ_id));
*(uint8_t*)(dest+2) = src->command;
memcpy(dest+3, src->payload, CELL_PAYLOAD_SIZE);
2003-09-16 07:41:49 +02:00
}
/** Unpack the network-order buffer <b>src</b> into a host-order
* cell_t structure <b>dest</b>.
*/
static void
cell_unpack(cell_t *dest, const char *src)
{
dest->circ_id = ntohs(get_uint16(src));
2003-09-16 07:41:49 +02:00
dest->command = *(uint8_t*)(src+2);
memcpy(dest->payload, src+3, CELL_PAYLOAD_SIZE);
2003-09-16 07:41:49 +02:00
}
/** Write the header of <b>cell</b> into the first VAR_CELL_HEADER_SIZE
* bytes of <b>hdr_out</b>. */
void
var_cell_pack_header(const var_cell_t *cell, char *hdr_out)
{
set_uint16(hdr_out, htons(cell->circ_id));
set_uint8(hdr_out+2, cell->command);
set_uint16(hdr_out+3, htons(cell->payload_len));
}
/** Allocate and return a new var_cell_t with <b>payload_len</b> bytes of
* payload space. */
var_cell_t *
var_cell_new(uint16_t payload_len)
{
var_cell_t *cell = tor_malloc(sizeof(var_cell_t)+payload_len-1);
cell->payload_len = payload_len;
cell->command = 0;
cell->circ_id = 0;
return cell;
}
2008-02-09 04:11:10 +01:00
/** Release all space held by <b>cell</b>. */
void
var_cell_free(var_cell_t *cell)
{
tor_free(cell);
}
2008-02-09 04:11:10 +01:00
/** We've received an EOF from <b>conn</b>. Mark it for close and return. */
int
connection_or_reached_eof(or_connection_t *conn)
{
2006-02-13 10:02:35 +01:00
log_info(LD_OR,"OR connection reached EOF. Closing.");
connection_mark_for_close(TO_CONN(conn));
return 0;
}
/** Handle any new bytes that have come in on connection <b>conn</b>.
* If conn is in 'open' state, hand it to
* connection_or_process_cells_from_inbuf()
* (else do nothing).
*/
int
connection_or_process_inbuf(or_connection_t *conn)
{
int ret;
tor_assert(conn);
2002-06-27 00:45:49 +02:00
switch (conn->_base.state) {
case OR_CONN_STATE_PROXY_HANDSHAKING:
ret = connection_read_proxy_handshake(TO_CONN(conn));
/* start TLS after handshake completion, or deal with error */
if (ret == 1) {
tor_assert(TO_CONN(conn)->proxy_state == PROXY_CONNECTED);
if (connection_tls_start_handshake(conn, 0) < 0)
ret = -1;
}
if (ret < 0) {
connection_mark_for_close(TO_CONN(conn));
}
return ret;
case OR_CONN_STATE_TLS_SERVER_RENEGOTIATING:
if (tor_tls_server_got_renegotiate(conn->tls))
connection_or_tls_renegotiated_cb(conn->tls, conn);
if (conn->_base.marked_for_close)
return 0;
/* fall through. */
case OR_CONN_STATE_OPEN:
case OR_CONN_STATE_OR_HANDSHAKING:
return connection_or_process_cells_from_inbuf(conn);
default:
return 0; /* don't do anything */
}
2002-06-27 00:45:49 +02:00
}
/** When adding cells to an OR connection's outbuf, keep adding until the
* outbuf is at least this long, or we run out of cells. */
#define OR_CONN_HIGHWATER (32*1024)
/** Add cells to an OR connection's outbuf whenever the outbuf's data length
* drops below this size. */
#define OR_CONN_LOWWATER (16*1024)
/** Called whenever we have flushed some data on an or_conn: add more data
* from active circuits. */
int
connection_or_flushed_some(or_connection_t *conn)
{
size_t datalen = connection_get_outbuf_len(TO_CONN(conn));
/* If we're under the low water mark, add cells until we're just over the
* high water mark. */
if (datalen < OR_CONN_LOWWATER) {
ssize_t n = CEIL_DIV(OR_CONN_HIGHWATER - datalen, CELL_NETWORK_SIZE);
time_t now = approx_time();
while (conn->active_circuits && n > 0) {
int flushed;
flushed = connection_or_flush_from_first_active_circuit(conn, 1, now);
n -= flushed;
}
}
return 0;
}
/** Connection <b>conn</b> has finished writing and has no bytes left on
* its outbuf.
*
2004-05-10 06:34:48 +02:00
* Otherwise it's in state "open": stop writing and return.
*
* If <b>conn</b> is broken, mark it for close and return -1, else
* return 0.
*/
int
connection_or_finished_flushing(or_connection_t *conn)
{
tor_assert(conn);
assert_connection_ok(TO_CONN(conn),0);
2002-06-27 00:45:49 +02:00
switch (conn->_base.state) {
case OR_CONN_STATE_PROXY_HANDSHAKING:
case OR_CONN_STATE_OPEN:
case OR_CONN_STATE_OR_HANDSHAKING:
break;
default:
log_err(LD_BUG,"Called in unexpected state %d.", conn->_base.state);
tor_fragile_assert();
return -1;
}
return 0;
}
/** Connected handler for OR connections: begin the TLS handshake.
*/
int
connection_or_finished_connecting(or_connection_t *or_conn)
{
int proxy_type;
connection_t *conn;
tor_assert(or_conn);
conn = TO_CONN(or_conn);
tor_assert(conn->state == OR_CONN_STATE_CONNECTING);
log_debug(LD_HANDSHAKE,"OR connect() to router at %s:%u finished.",
2006-02-13 10:02:35 +01:00
conn->address,conn->port);
control_event_bootstrap(BOOTSTRAP_STATUS_HANDSHAKE, 0);
proxy_type = PROXY_NONE;
if (get_options()->HTTPSProxy)
proxy_type = PROXY_CONNECT;
else if (get_options()->Socks4Proxy)
proxy_type = PROXY_SOCKS4;
else if (get_options()->Socks5Proxy)
proxy_type = PROXY_SOCKS5;
if (proxy_type != PROXY_NONE) {
/* start proxy handshake */
if (connection_proxy_connect(conn, proxy_type) < 0) {
connection_mark_for_close(conn);
return -1;
}
connection_start_reading(conn);
conn->state = OR_CONN_STATE_PROXY_HANDSHAKING;
return 0;
}
if (connection_tls_start_handshake(or_conn, 0) < 0) {
/* TLS handshaking error of some kind. */
connection_mark_for_close(conn);
return -1;
2002-06-27 00:45:49 +02:00
}
return 0;
2002-06-27 00:45:49 +02:00
}
/** Return 1 if identity digest <b>id_digest</b> is known to be a
* currently or recently running relay. Otherwise return 0. */
int
connection_or_digest_is_known_relay(const char *id_digest)
{
if (router_get_consensus_status_by_id(id_digest))
return 1; /* It's in the consensus: "yes" */
if (router_get_by_id_digest(id_digest))
return 1; /* Not in the consensus, but we have a descriptor for
* it. Probably it was in a recent consensus. "Yes". */
return 0;
}
/** Set the per-conn read and write limits for <b>conn</b>. If it's a known
* relay, we will rely on the global read and write buckets, so give it
* per-conn limits that are big enough they'll never matter. But if it's
* not a known relay, first check if we set PerConnBwRate/Burst, then
* check if the consensus sets them, else default to 'big enough'.
*/
static void
connection_or_update_token_buckets_helper(or_connection_t *conn, int reset,
or_options_t *options)
{
int rate, burst; /* per-connection rate limiting params */
if (connection_or_digest_is_known_relay(conn->identity_digest)) {
/* It's in the consensus, or we have a descriptor for it meaning it
* was probably in a recent consensus. It's a recognized relay:
* give it full bandwidth. */
rate = (int)options->BandwidthRate;
burst = (int)options->BandwidthBurst;
} else {
/* Not a recognized relay. Squeeze it down based on the suggested
* bandwidth parameters in the consensus, but allow local config
* options to override. */
rate = options->PerConnBWRate ? (int)options->PerConnBWRate :
(int)networkstatus_get_param(NULL, "perconnbwrate",
(int)options->BandwidthRate);
burst = options->PerConnBWBurst ? (int)options->PerConnBWBurst :
(int)networkstatus_get_param(NULL, "perconnbwburst",
(int)options->BandwidthBurst);
}
conn->bandwidthrate = rate;
conn->bandwidthburst = burst;
#ifdef USE_BUFFEREVENTS
{
const struct timeval *tick = tor_libevent_get_one_tick_timeout();
struct ev_token_bucket_cfg *cfg, *old_cfg;
int rate_per_tick = rate / TOR_LIBEVENT_TICKS_PER_SECOND;
cfg = ev_token_bucket_cfg_new(rate_per_tick, burst, rate_per_tick,
burst, tick);
old_cfg = conn->bucket_cfg;
if (conn->_base.bufev)
bufferevent_set_rate_limit(conn->_base.bufev, cfg);
if (old_cfg)
ev_token_bucket_cfg_free(old_cfg);
conn->bucket_cfg = cfg;
(void) reset; /* No way to do this with libevent yet. */
}
#else
if (reset) { /* set up the token buckets to be full */
conn->read_bucket = conn->write_bucket = burst;
return;
}
/* If the new token bucket is smaller, take out the extra tokens.
* (If it's larger, don't -- the buckets can grow to reach the cap.) */
if (conn->read_bucket > burst)
conn->read_bucket = burst;
if (conn->write_bucket > burst)
conn->write_bucket = burst;
#endif
}
/** Either our set of relays or our per-conn rate limits have changed.
* Go through all the OR connections and update their token buckets. */
void
connection_or_update_token_buckets(smartlist_t *conns, or_options_t *options)
{
SMARTLIST_FOREACH(conns, connection_t *, conn,
{
if (connection_speaks_cells(conn))
connection_or_update_token_buckets_helper(TO_OR_CONN(conn), 0, options);
});
}
/** If we don't necessarily know the router we're connecting to, but we
* have an addr/port/id_digest, then fill in as much as we can. Start
* by checking to see if this describes a router we know. */
static void
connection_or_init_conn_from_address(or_connection_t *conn,
const tor_addr_t *addr, uint16_t port,
const char *id_digest,
int started_here)
{
Initial conversion to use node_t throughout our codebase. A node_t is an abstraction over routerstatus_t, routerinfo_t, and microdesc_t. It should try to present a consistent interface to all of them. There should be a node_t for a server whenever there is * A routerinfo_t for it in the routerlist * A routerstatus_t in the current_consensus. (note that a microdesc_t alone isn't enough to make a node_t exist, since microdescriptors aren't usable on their own.) There are three ways to get a node_t right now: looking it up by ID, looking it up by nickname, and iterating over the whole list of microdescriptors. All (or nearly all) functions that are supposed to return "a router" -- especially those used in building connections and circuits -- should return a node_t, not a routerinfo_t or a routerstatus_t. A node_t should hold all the *mutable* flags about a node. This patch moves the is_foo flags from routerinfo_t into node_t. The flags in routerstatus_t remain, but they get set from the consensus and should not change. Some other highlights of this patch are: * Looking up routerinfo and routerstatus by nickname is now unified and based on the "look up a node by nickname" function. This tries to look only at the values from current consensus, and not get confused by the routerinfo_t->is_named flag, which could get set for other weird reasons. This changes the behavior of how authorities (when acting as clients) deal with nodes that have been listed by nickname. * I tried not to artificially increase the size of the diff here by moving functions around. As a result, some functions that now operate on nodes are now in the wrong file -- they should get moved to nodelist.c once this refactoring settles down. This moving should happen as part of a patch that moves functions AND NOTHING ELSE. * Some old code is now left around inside #if 0/1 blocks, and should get removed once I've verified that I don't want it sitting around to see how we used to do things. There are still some unimplemented functions: these are flagged with "UNIMPLEMENTED_NODELIST()." I'll work on filling in the implementation here, piece by piece. I wish this patch could have been smaller, but there did not seem to be any piece of it that was independent from the rest. Moving flags forces many functions that once returned routerinfo_t * to return node_t *, which forces their friends to change, and so on.
2010-09-29 21:00:41 +02:00
const node_t *r = node_get_by_id(id_digest);
connection_or_set_identity_digest(conn, id_digest);
connection_or_update_token_buckets_helper(conn, 1, get_options());
conn->_base.port = port;
tor_addr_copy(&conn->_base.addr, addr);
tor_addr_copy(&conn->real_addr, addr);
if (r) {
Initial conversion to use node_t throughout our codebase. A node_t is an abstraction over routerstatus_t, routerinfo_t, and microdesc_t. It should try to present a consistent interface to all of them. There should be a node_t for a server whenever there is * A routerinfo_t for it in the routerlist * A routerstatus_t in the current_consensus. (note that a microdesc_t alone isn't enough to make a node_t exist, since microdescriptors aren't usable on their own.) There are three ways to get a node_t right now: looking it up by ID, looking it up by nickname, and iterating over the whole list of microdescriptors. All (or nearly all) functions that are supposed to return "a router" -- especially those used in building connections and circuits -- should return a node_t, not a routerinfo_t or a routerstatus_t. A node_t should hold all the *mutable* flags about a node. This patch moves the is_foo flags from routerinfo_t into node_t. The flags in routerstatus_t remain, but they get set from the consensus and should not change. Some other highlights of this patch are: * Looking up routerinfo and routerstatus by nickname is now unified and based on the "look up a node by nickname" function. This tries to look only at the values from current consensus, and not get confused by the routerinfo_t->is_named flag, which could get set for other weird reasons. This changes the behavior of how authorities (when acting as clients) deal with nodes that have been listed by nickname. * I tried not to artificially increase the size of the diff here by moving functions around. As a result, some functions that now operate on nodes are now in the wrong file -- they should get moved to nodelist.c once this refactoring settles down. This moving should happen as part of a patch that moves functions AND NOTHING ELSE. * Some old code is now left around inside #if 0/1 blocks, and should get removed once I've verified that I don't want it sitting around to see how we used to do things. There are still some unimplemented functions: these are flagged with "UNIMPLEMENTED_NODELIST()." I'll work on filling in the implementation here, piece by piece. I wish this patch could have been smaller, but there did not seem to be any piece of it that was independent from the rest. Moving flags forces many functions that once returned routerinfo_t * to return node_t *, which forces their friends to change, and so on.
2010-09-29 21:00:41 +02:00
tor_addr_t node_addr;
node_get_addr(r, &node_addr);
/* XXXX proposal 118 will make this more complex. */
Initial conversion to use node_t throughout our codebase. A node_t is an abstraction over routerstatus_t, routerinfo_t, and microdesc_t. It should try to present a consistent interface to all of them. There should be a node_t for a server whenever there is * A routerinfo_t for it in the routerlist * A routerstatus_t in the current_consensus. (note that a microdesc_t alone isn't enough to make a node_t exist, since microdescriptors aren't usable on their own.) There are three ways to get a node_t right now: looking it up by ID, looking it up by nickname, and iterating over the whole list of microdescriptors. All (or nearly all) functions that are supposed to return "a router" -- especially those used in building connections and circuits -- should return a node_t, not a routerinfo_t or a routerstatus_t. A node_t should hold all the *mutable* flags about a node. This patch moves the is_foo flags from routerinfo_t into node_t. The flags in routerstatus_t remain, but they get set from the consensus and should not change. Some other highlights of this patch are: * Looking up routerinfo and routerstatus by nickname is now unified and based on the "look up a node by nickname" function. This tries to look only at the values from current consensus, and not get confused by the routerinfo_t->is_named flag, which could get set for other weird reasons. This changes the behavior of how authorities (when acting as clients) deal with nodes that have been listed by nickname. * I tried not to artificially increase the size of the diff here by moving functions around. As a result, some functions that now operate on nodes are now in the wrong file -- they should get moved to nodelist.c once this refactoring settles down. This moving should happen as part of a patch that moves functions AND NOTHING ELSE. * Some old code is now left around inside #if 0/1 blocks, and should get removed once I've verified that I don't want it sitting around to see how we used to do things. There are still some unimplemented functions: these are flagged with "UNIMPLEMENTED_NODELIST()." I'll work on filling in the implementation here, piece by piece. I wish this patch could have been smaller, but there did not seem to be any piece of it that was independent from the rest. Moving flags forces many functions that once returned routerinfo_t * to return node_t *, which forces their friends to change, and so on.
2010-09-29 21:00:41 +02:00
if (tor_addr_eq(&conn->_base.addr, &node_addr))
conn->is_canonical = 1;
if (!started_here) {
/* Override the addr/port, so our log messages will make sense.
* This is dangerous, since if we ever try looking up a conn by
* its actual addr/port, we won't remember. Careful! */
/* XXXX arma: this is stupid, and it's the reason we need real_addr
* to track is_canonical properly. What requires it? */
/* XXXX <arma> i believe the reason we did this, originally, is because
* we wanted to log what OR a connection was to, and if we logged the
* right IP address and port 56244, that wouldn't be as helpful. now we
* log the "right" port too, so we know if it's moria1 or moria2.
*/
Initial conversion to use node_t throughout our codebase. A node_t is an abstraction over routerstatus_t, routerinfo_t, and microdesc_t. It should try to present a consistent interface to all of them. There should be a node_t for a server whenever there is * A routerinfo_t for it in the routerlist * A routerstatus_t in the current_consensus. (note that a microdesc_t alone isn't enough to make a node_t exist, since microdescriptors aren't usable on their own.) There are three ways to get a node_t right now: looking it up by ID, looking it up by nickname, and iterating over the whole list of microdescriptors. All (or nearly all) functions that are supposed to return "a router" -- especially those used in building connections and circuits -- should return a node_t, not a routerinfo_t or a routerstatus_t. A node_t should hold all the *mutable* flags about a node. This patch moves the is_foo flags from routerinfo_t into node_t. The flags in routerstatus_t remain, but they get set from the consensus and should not change. Some other highlights of this patch are: * Looking up routerinfo and routerstatus by nickname is now unified and based on the "look up a node by nickname" function. This tries to look only at the values from current consensus, and not get confused by the routerinfo_t->is_named flag, which could get set for other weird reasons. This changes the behavior of how authorities (when acting as clients) deal with nodes that have been listed by nickname. * I tried not to artificially increase the size of the diff here by moving functions around. As a result, some functions that now operate on nodes are now in the wrong file -- they should get moved to nodelist.c once this refactoring settles down. This moving should happen as part of a patch that moves functions AND NOTHING ELSE. * Some old code is now left around inside #if 0/1 blocks, and should get removed once I've verified that I don't want it sitting around to see how we used to do things. There are still some unimplemented functions: these are flagged with "UNIMPLEMENTED_NODELIST()." I'll work on filling in the implementation here, piece by piece. I wish this patch could have been smaller, but there did not seem to be any piece of it that was independent from the rest. Moving flags forces many functions that once returned routerinfo_t * to return node_t *, which forces their friends to change, and so on.
2010-09-29 21:00:41 +02:00
tor_addr_copy(&conn->_base.addr, &node_addr);
conn->_base.port = node_get_orport(r);
}
Initial conversion to use node_t throughout our codebase. A node_t is an abstraction over routerstatus_t, routerinfo_t, and microdesc_t. It should try to present a consistent interface to all of them. There should be a node_t for a server whenever there is * A routerinfo_t for it in the routerlist * A routerstatus_t in the current_consensus. (note that a microdesc_t alone isn't enough to make a node_t exist, since microdescriptors aren't usable on their own.) There are three ways to get a node_t right now: looking it up by ID, looking it up by nickname, and iterating over the whole list of microdescriptors. All (or nearly all) functions that are supposed to return "a router" -- especially those used in building connections and circuits -- should return a node_t, not a routerinfo_t or a routerstatus_t. A node_t should hold all the *mutable* flags about a node. This patch moves the is_foo flags from routerinfo_t into node_t. The flags in routerstatus_t remain, but they get set from the consensus and should not change. Some other highlights of this patch are: * Looking up routerinfo and routerstatus by nickname is now unified and based on the "look up a node by nickname" function. This tries to look only at the values from current consensus, and not get confused by the routerinfo_t->is_named flag, which could get set for other weird reasons. This changes the behavior of how authorities (when acting as clients) deal with nodes that have been listed by nickname. * I tried not to artificially increase the size of the diff here by moving functions around. As a result, some functions that now operate on nodes are now in the wrong file -- they should get moved to nodelist.c once this refactoring settles down. This moving should happen as part of a patch that moves functions AND NOTHING ELSE. * Some old code is now left around inside #if 0/1 blocks, and should get removed once I've verified that I don't want it sitting around to see how we used to do things. There are still some unimplemented functions: these are flagged with "UNIMPLEMENTED_NODELIST()." I'll work on filling in the implementation here, piece by piece. I wish this patch could have been smaller, but there did not seem to be any piece of it that was independent from the rest. Moving flags forces many functions that once returned routerinfo_t * to return node_t *, which forces their friends to change, and so on.
2010-09-29 21:00:41 +02:00
conn->nickname = tor_strdup(node_get_nickname(r));
tor_free(conn->_base.address);
Initial conversion to use node_t throughout our codebase. A node_t is an abstraction over routerstatus_t, routerinfo_t, and microdesc_t. It should try to present a consistent interface to all of them. There should be a node_t for a server whenever there is * A routerinfo_t for it in the routerlist * A routerstatus_t in the current_consensus. (note that a microdesc_t alone isn't enough to make a node_t exist, since microdescriptors aren't usable on their own.) There are three ways to get a node_t right now: looking it up by ID, looking it up by nickname, and iterating over the whole list of microdescriptors. All (or nearly all) functions that are supposed to return "a router" -- especially those used in building connections and circuits -- should return a node_t, not a routerinfo_t or a routerstatus_t. A node_t should hold all the *mutable* flags about a node. This patch moves the is_foo flags from routerinfo_t into node_t. The flags in routerstatus_t remain, but they get set from the consensus and should not change. Some other highlights of this patch are: * Looking up routerinfo and routerstatus by nickname is now unified and based on the "look up a node by nickname" function. This tries to look only at the values from current consensus, and not get confused by the routerinfo_t->is_named flag, which could get set for other weird reasons. This changes the behavior of how authorities (when acting as clients) deal with nodes that have been listed by nickname. * I tried not to artificially increase the size of the diff here by moving functions around. As a result, some functions that now operate on nodes are now in the wrong file -- they should get moved to nodelist.c once this refactoring settles down. This moving should happen as part of a patch that moves functions AND NOTHING ELSE. * Some old code is now left around inside #if 0/1 blocks, and should get removed once I've verified that I don't want it sitting around to see how we used to do things. There are still some unimplemented functions: these are flagged with "UNIMPLEMENTED_NODELIST()." I'll work on filling in the implementation here, piece by piece. I wish this patch could have been smaller, but there did not seem to be any piece of it that was independent from the rest. Moving flags forces many functions that once returned routerinfo_t * to return node_t *, which forces their friends to change, and so on.
2010-09-29 21:00:41 +02:00
conn->_base.address = tor_dup_addr(&node_addr);
} else {
const char *n;
/* If we're an authoritative directory server, we may know a
* nickname for this router. */
n = dirserv_get_nickname_by_digest(id_digest);
if (n) {
conn->nickname = tor_strdup(n);
} else {
conn->nickname = tor_malloc(HEX_DIGEST_LEN+2);
conn->nickname[0] = '$';
base16_encode(conn->nickname+1, HEX_DIGEST_LEN+1,
conn->identity_digest, DIGEST_LEN);
}
tor_free(conn->_base.address);
conn->_base.address = tor_dup_addr(addr);
}
}
/** Return true iff <b>a</b> is "better" than <b>b</b> for new circuits.
*
* A more canonical connection is always better than a less canonical
* connection. That aside, a connection is better if it has circuits and the
* other does not, or if it was created more recently.
*
* Requires that both input connections are open; not is_bad_for_new_circs,
* and not impossibly non-canonical.
*
* If </b>forgive_new_connections</b> is true, then we do not call
* <b>a</b>better than <b>b</b> simply because b has no circuits,
* unless b is also relatively old.
*/
static int
connection_or_is_better(time_t now,
const or_connection_t *a,
const or_connection_t *b,
int forgive_new_connections)
{
int newer;
/** Do not definitively deprecate a new connection with no circuits on it
* until this much time has passed. */
#define NEW_CONN_GRACE_PERIOD (15*60)
if (b->is_canonical && !a->is_canonical)
return 0; /* A canonical connection is better than a non-canonical
* one, no matter how new it is or which has circuits. */
newer = b->_base.timestamp_created < a->_base.timestamp_created;
if (
/* We prefer canonical connections regardless of newness. */
(!b->is_canonical && a->is_canonical) ||
/* If both have circuits we prefer the newer: */
(b->n_circuits && a->n_circuits && newer) ||
/* If neither has circuits we prefer the newer: */
(!b->n_circuits && !a->n_circuits && newer))
return 1;
/* If one has no circuits and the other does... */
if (!b->n_circuits && a->n_circuits) {
/* Then it's bad, unless it's in its grace period and we're forgiving. */
if (forgive_new_connections &&
now < b->_base.timestamp_created + NEW_CONN_GRACE_PERIOD)
return 0;
else
return 1;
}
return 0;
}
/** Return the OR connection we should use to extend a circuit to the router
* whose identity is <b>digest</b>, and whose address we believe (or have been
* told in an extend cell) is <b>target_addr</b>. If there is no good
* connection, set *<b>msg_out</b> to a message describing the connection's
* state and our next action, and set <b>launch_out</b> to a boolean for
* whether we should launch a new connection or not.
*/
or_connection_t *
connection_or_get_for_extend(const char *digest,
const tor_addr_t *target_addr,
const char **msg_out,
int *launch_out)
{
or_connection_t *conn, *best=NULL;
int n_inprogress_goodaddr = 0, n_old = 0, n_noncanonical = 0, n_possible = 0;
time_t now = approx_time();
tor_assert(msg_out);
tor_assert(launch_out);
if (!orconn_identity_map) {
*msg_out = "Router not connected (nothing is). Connecting.";
*launch_out = 1;
return NULL;
}
conn = digestmap_get(orconn_identity_map, digest);
for (; conn; conn = conn->next_with_same_id) {
tor_assert(conn->_base.magic == OR_CONNECTION_MAGIC);
tor_assert(conn->_base.type == CONN_TYPE_OR);
tor_assert(!memcmp(conn->identity_digest, digest, DIGEST_LEN));
if (conn->_base.marked_for_close)
continue;
/* Never return a non-open connection. */
if (conn->_base.state != OR_CONN_STATE_OPEN) {
/* If the address matches, don't launch a new connection for this
* circuit. */
if (!tor_addr_compare(&conn->real_addr, target_addr, CMP_EXACT))
++n_inprogress_goodaddr;
continue;
}
/* Never return a connection that shouldn't be used for circs. */
if (conn->is_bad_for_new_circs) {
++n_old;
continue;
}
/* Never return a non-canonical connection using a recent link protocol
* if the address is not what we wanted.
*
* (For old link protocols, we can't rely on is_canonical getting
* set properly if we're talking to the right address, since we might
* have an out-of-date descriptor, and we will get no NETINFO cell to
* tell us about the right address.) */
if (!conn->is_canonical && conn->link_proto >= 2 &&
tor_addr_compare(&conn->real_addr, target_addr, CMP_EXACT)) {
++n_noncanonical;
continue;
}
++n_possible;
if (!best) {
best = conn; /* If we have no 'best' so far, this one is good enough. */
continue;
}
if (connection_or_is_better(now, conn, best, 0))
best = conn;
}
if (best) {
*msg_out = "Connection is fine; using it.";
*launch_out = 0;
return best;
} else if (n_inprogress_goodaddr) {
*msg_out = "Connection in progress; waiting.";
*launch_out = 0;
return NULL;
} else if (n_old || n_noncanonical) {
*msg_out = "Connections all too old, or too non-canonical. "
" Launching a new one.";
*launch_out = 1;
return NULL;
} else {
*msg_out = "Not connected. Connecting.";
*launch_out = 1;
return NULL;
}
}
/** How old do we let a connection to an OR get before deciding it's
* too old for new circuits? */
#define TIME_BEFORE_OR_CONN_IS_TOO_OLD (60*60*24*7)
/** Given the head of the linked list for all the or_connections with a given
* identity, set elements of that list as is_bad_for_new_circs as
* appropriate. Helper for connection_or_set_bad_connections().
*
* Specifically, we set the is_bad_for_new_circs flag on:
* - all connections if <b>force</b> is true.
* - all connections that are too old.
* - all open non-canonical connections for which a canonical connection
* exists to the same router.
* - all open canonical connections for which a 'better' canonical
* connection exists to the same router.
* - all open non-canonical connections for which a 'better' non-canonical
* connection exists to the same router at the same address.
*
* See connection_or_is_better() for our idea of what makes one OR connection
* better than another.
*/
static void
connection_or_group_set_badness(or_connection_t *head, int force)
{
or_connection_t *or_conn = NULL, *best = NULL;
int n_old = 0, n_inprogress = 0, n_canonical = 0, n_other = 0;
time_t now = time(NULL);
/* Pass 1: expire everything that's old, and see what the status of
* everything else is. */
for (or_conn = head; or_conn; or_conn = or_conn->next_with_same_id) {
if (or_conn->_base.marked_for_close ||
or_conn->is_bad_for_new_circs)
continue;
if (force ||
or_conn->_base.timestamp_created + TIME_BEFORE_OR_CONN_IS_TOO_OLD
< now) {
log_info(LD_OR,
"Marking OR conn to %s:%d as too old for new circuits "
"(fd %d, %d secs old).",
or_conn->_base.address, or_conn->_base.port, or_conn->_base.s,
(int)(now - or_conn->_base.timestamp_created));
or_conn->is_bad_for_new_circs = 1;
}
if (or_conn->is_bad_for_new_circs) {
++n_old;
} else if (or_conn->_base.state != OR_CONN_STATE_OPEN) {
++n_inprogress;
} else if (or_conn->is_canonical) {
++n_canonical;
} else {
++n_other;
}
}
/* Pass 2: We know how about how good the best connection is.
* expire everything that's worse, and find the very best if we can. */
for (or_conn = head; or_conn; or_conn = or_conn->next_with_same_id) {
if (or_conn->_base.marked_for_close ||
or_conn->is_bad_for_new_circs)
continue; /* This one doesn't need to be marked bad. */
if (or_conn->_base.state != OR_CONN_STATE_OPEN)
continue; /* Don't mark anything bad until we have seen what happens
* when the connection finishes. */
if (n_canonical && !or_conn->is_canonical) {
/* We have at least one open canonical connection to this router,
* and this one is open but not canonical. Mark it bad. */
log_info(LD_OR,
"Marking OR conn to %s:%d as unsuitable for new circuits: "
"(fd %d, %d secs old). It is not canonical, and we have "
"another connection to that OR that is.",
or_conn->_base.address, or_conn->_base.port, or_conn->_base.s,
(int)(now - or_conn->_base.timestamp_created));
or_conn->is_bad_for_new_circs = 1;
continue;
}
if (!best || connection_or_is_better(now, or_conn, best, 0))
best = or_conn;
}
if (!best)
return;
/* Pass 3: One connection to OR is best. If it's canonical, mark as bad
* every other open connection. If it's non-canonical, mark as bad
* every other open connection to the same address.
*
* XXXX This isn't optimal; if we have connections to an OR at multiple
* addresses, we'd like to pick the best _for each address_, and mark as
* bad every open connection that isn't best for its address. But this
* can only occur in cases where the other OR is old (so we have no
* canonical connection to it), or where all the connections to the OR are
* at noncanonical addresses and we have no good direct connection (which
* means we aren't at risk of attaching circuits to it anyway). As
* 0.1.2.x dies out, the first case will go away, and the second one is
* "mostly harmless", so a fix can wait until somebody is bored.
*/
for (or_conn = head; or_conn; or_conn = or_conn->next_with_same_id) {
if (or_conn->_base.marked_for_close ||
or_conn->is_bad_for_new_circs ||
or_conn->_base.state != OR_CONN_STATE_OPEN)
continue;
if (or_conn != best && connection_or_is_better(now, best, or_conn, 1)) {
/* This isn't the best conn, _and_ the best conn is better than it,
even when we're being forgiving. */
if (best->is_canonical) {
log_info(LD_OR,
"Marking OR conn to %s:%d as unsuitable for new circuits: "
"(fd %d, %d secs old). We have a better canonical one "
"(fd %d; %d secs old).",
or_conn->_base.address, or_conn->_base.port, or_conn->_base.s,
(int)(now - or_conn->_base.timestamp_created),
best->_base.s, (int)(now - best->_base.timestamp_created));
or_conn->is_bad_for_new_circs = 1;
} else if (!tor_addr_compare(&or_conn->real_addr,
&best->real_addr, CMP_EXACT)) {
log_info(LD_OR,
"Marking OR conn to %s:%d as unsuitable for new circuits: "
"(fd %d, %d secs old). We have a better one with the "
"same address (fd %d; %d secs old).",
or_conn->_base.address, or_conn->_base.port, or_conn->_base.s,
(int)(now - or_conn->_base.timestamp_created),
best->_base.s, (int)(now - best->_base.timestamp_created));
or_conn->is_bad_for_new_circs = 1;
}
}
}
}
/** Go through all the OR connections (or if <b>digest</b> is non-NULL, just
* the OR connections with that digest), and set the is_bad_for_new_circs
* flag based on the rules in connection_or_group_set_badness() (or just
* always set it if <b>force</b> is true).
*/
void
connection_or_set_bad_connections(const char *digest, int force)
{
if (!orconn_identity_map)
return;
DIGESTMAP_FOREACH(orconn_identity_map, identity, or_connection_t *, conn) {
if (!digest || !memcmp(digest, conn->identity_digest, DIGEST_LEN))
connection_or_group_set_badness(conn, force);
} DIGESTMAP_FOREACH_END;
}
/** <b>conn</b> is in the 'connecting' state, and it failed to complete
* a TCP connection. Send notifications appropriately.
*
* <b>reason</b> specifies the or_conn_end_reason for the failure;
* <b>msg</b> specifies the strerror-style error message.
*/
void
connection_or_connect_failed(or_connection_t *conn,
int reason, const char *msg)
{
control_event_or_conn_status(conn, OR_CONN_EVENT_FAILED, reason);
if (!authdir_mode_tests_reachability(get_options()))
control_event_bootstrap_problem(msg, reason);
}
/** Launch a new OR connection to <b>addr</b>:<b>port</b> and expect to
* handshake with an OR with identity digest <b>id_digest</b>.
*
* If <b>id_digest</b> is me, do nothing. If we're already connected to it,
* return that connection. If the connect() is in progress, set the
* new conn's state to 'connecting' and return it. If connect() succeeds,
* call connection_tls_start_handshake() on it.
*
* This function is called from router_retry_connections(), for
* ORs connecting to ORs, and circuit_establish_circuit(), for
* OPs connecting to ORs.
*
* Return the launched conn, or NULL if it failed.
*/
or_connection_t *
connection_or_connect(const tor_addr_t *_addr, uint16_t port,
const char *id_digest)
{
or_connection_t *conn;
or_options_t *options = get_options();
int socket_error = 0;
int using_proxy = 0;
tor_addr_t addr;
2002-06-27 00:45:49 +02:00
tor_assert(_addr);
tor_assert(id_digest);
tor_addr_copy(&addr, _addr);
if (server_mode(options) && router_digest_is_me(id_digest)) {
2006-02-13 10:02:35 +01:00
log_info(LD_PROTOCOL,"Client asked me to connect to myself. Refusing.");
return NULL;
}
conn = or_connection_new(AF_INET);
2002-06-27 00:45:49 +02:00
/* set up conn so it's got all the data we need to remember */
connection_or_init_conn_from_address(conn, &addr, port, id_digest, 1);
conn->_base.state = OR_CONN_STATE_CONNECTING;
control_event_or_conn_status(conn, OR_CONN_EVENT_LAUNCHED, 0);
2002-06-27 00:45:49 +02:00
/* use a proxy server if available */
if (options->HTTPSProxy) {
using_proxy = 1;
tor_addr_copy(&addr, &options->HTTPSProxyAddr);
port = options->HTTPSProxyPort;
} else if (options->Socks4Proxy) {
using_proxy = 1;
tor_addr_copy(&addr, &options->Socks4ProxyAddr);
port = options->Socks4ProxyPort;
} else if (options->Socks5Proxy) {
using_proxy = 1;
tor_addr_copy(&addr, &options->Socks5ProxyAddr);
port = options->Socks5ProxyPort;
}
switch (connection_connect(TO_CONN(conn), conn->_base.address,
&addr, port, &socket_error)) {
case -1:
2005-12-25 00:32:15 +01:00
/* If the connection failed immediately, and we're using
* a proxy, our proxy is down. Don't blame the Tor server. */
if (!using_proxy)
entry_guard_register_connect_status(conn->identity_digest,
0, 1, time(NULL));
connection_or_connect_failed(conn,
errno_to_orconn_end_reason(socket_error),
tor_socket_strerror(socket_error));
connection_free(TO_CONN(conn));
return NULL;
case 0:
connection_watch_events(TO_CONN(conn), READ_EVENT | WRITE_EVENT);
/* writable indicates finish, readable indicates broken link,
error indicates broken link on windows */
return conn;
/* case 1: fall through */
}
if (connection_or_finished_connecting(conn) < 0) {
/* already marked for close */
return NULL;
}
return conn;
}
/** Begin the tls handshake with <b>conn</b>. <b>receiving</b> is 0 if
* we initiated the connection, else it's 1.
*
2005-12-25 00:32:15 +01:00
* Assign a new tls object to conn->tls, begin reading on <b>conn</b>, and
* pass <b>conn</b> to connection_tls_continue_handshake().
*
* Return -1 if <b>conn</b> is broken, else return 0.
*/
int
connection_tls_start_handshake(or_connection_t *conn, int receiving)
{
conn->_base.state = OR_CONN_STATE_TLS_HANDSHAKING;
tor_assert(!conn->tls);
conn->tls = tor_tls_new(conn->_base.s, receiving);
tor_tls_set_logged_address(conn->tls, // XXX client and relay?
escaped_safe_str(conn->_base.address));
if (!conn->tls) {
2006-02-13 10:02:35 +01:00
log_warn(LD_BUG,"tor_tls_new failed. Closing.");
return -1;
}
#ifdef USE_BUFFEREVENTS
if (connection_type_uses_bufferevent(TO_CONN(conn))) {
const int filtering = get_options()->_UseFilteringSSLBufferevents;
struct bufferevent *b =
tor_tls_init_bufferevent(conn->tls, conn->_base.bufev, conn->_base.s,
receiving, filtering);
if (!b) {
log_warn(LD_BUG,"tor_tls_init_bufferevent failed. Closing.");
return -1;
}
conn->_base.bufev = b;
if (conn->bucket_cfg)
bufferevent_set_rate_limit(conn->_base.bufev, conn->bucket_cfg);
connection_enable_rate_limiting(TO_CONN(conn));
connection_configure_bufferevent_callbacks(TO_CONN(conn));
bufferevent_setcb(b,
connection_handle_read_cb,
connection_handle_write_cb,
connection_or_handle_event_cb,/* overriding this one*/
TO_CONN(conn));
}
#endif
connection_start_reading(TO_CONN(conn));
log_debug(LD_HANDSHAKE,"starting TLS handshake on fd %d", conn->_base.s);
note_crypto_pk_op(receiving ? TLS_HANDSHAKE_S : TLS_HANDSHAKE_C);
IF_HAS_BUFFEREVENT(TO_CONN(conn), {
/* ???? */;
}) ELSE_IF_NO_BUFFEREVENT {
if (connection_tls_continue_handshake(conn) < 0)
return -1;
}
return 0;
}
/** Invoked on the server side from inside tor_tls_read() when the server
* gets a successful TLS renegotiation from the client. */
static void
connection_or_tls_renegotiated_cb(tor_tls_t *tls, void *_conn)
{
or_connection_t *conn = _conn;
(void)tls;
/* Don't invoke this again. */
tor_tls_set_renegotiate_callback(tls, NULL, NULL);
tor_tls_block_renegotiation(tls);
if (connection_tls_finish_handshake(conn) < 0) {
/* XXXX_TLS double-check that it's ok to do this from inside read. */
/* XXXX_TLS double-check that this verifies certificates. */
connection_mark_for_close(TO_CONN(conn));
}
}
/** Move forward with the tls handshake. If it finishes, hand
* <b>conn</b> to connection_tls_finish_handshake().
*
* Return -1 if <b>conn</b> is broken, else return 0.
*/
int
connection_tls_continue_handshake(or_connection_t *conn)
{
int result;
check_no_tls_errors();
again:
if (conn->_base.state == OR_CONN_STATE_TLS_CLIENT_RENEGOTIATING) {
// log_notice(LD_OR, "Renegotiate with %p", conn->tls);
result = tor_tls_renegotiate(conn->tls);
// log_notice(LD_OR, "Result: %d", result);
} else {
tor_assert(conn->_base.state == OR_CONN_STATE_TLS_HANDSHAKING);
// log_notice(LD_OR, "Continue handshake with %p", conn->tls);
result = tor_tls_handshake(conn->tls);
// log_notice(LD_OR, "Result: %d", result);
}
switch (result) {
CASE_TOR_TLS_ERROR_ANY:
log_info(LD_OR,"tls error [%s]. breaking connection.",
tor_tls_err_to_string(result));
return -1;
case TOR_TLS_DONE:
if (! tor_tls_used_v1_handshake(conn->tls)) {
if (!tor_tls_is_server(conn->tls)) {
if (conn->_base.state == OR_CONN_STATE_TLS_HANDSHAKING) {
// log_notice(LD_OR,"Done. state was TLS_HANDSHAKING.");
conn->_base.state = OR_CONN_STATE_TLS_CLIENT_RENEGOTIATING;
goto again;
}
// log_notice(LD_OR,"Done. state was %d.", conn->_base.state);
} else {
/* improved handshake, but not a client. */
tor_tls_set_renegotiate_callback(conn->tls,
connection_or_tls_renegotiated_cb,
conn);
conn->_base.state = OR_CONN_STATE_TLS_SERVER_RENEGOTIATING;
connection_stop_writing(TO_CONN(conn));
connection_start_reading(TO_CONN(conn));
return 0;
}
}
return connection_tls_finish_handshake(conn);
case TOR_TLS_WANTWRITE:
connection_start_writing(TO_CONN(conn));
2006-02-13 10:02:35 +01:00
log_debug(LD_OR,"wanted write");
return 0;
case TOR_TLS_WANTREAD: /* handshaking conns are *always* reading */
2006-02-13 10:02:35 +01:00
log_debug(LD_OR,"wanted read");
return 0;
case TOR_TLS_CLOSE:
log_info(LD_OR,"tls closed. breaking connection.");
return -1;
}
return 0;
}
#ifdef USE_BUFFEREVENTS
static void
connection_or_handle_event_cb(struct bufferevent *bufev, short event,
void *arg)
{
struct or_connection_t *conn = TO_OR_CONN(arg);
/* XXXX cut-and-paste code; should become a function. */
if (event & BEV_EVENT_CONNECTED) {
if (conn->_base.state == OR_CONN_STATE_TLS_HANDSHAKING) {
if (tor_tls_finish_handshake(conn->tls) < 0) {
log_warn(LD_OR, "Problem finishing handshake");
connection_mark_for_close(TO_CONN(conn));
return;
}
}
if (! tor_tls_used_v1_handshake(conn->tls)) {
if (!tor_tls_is_server(conn->tls)) {
if (conn->_base.state == OR_CONN_STATE_TLS_HANDSHAKING) {
conn->_base.state = OR_CONN_STATE_TLS_CLIENT_RENEGOTIATING;
tor_tls_unblock_renegotiation(conn->tls);
if (bufferevent_ssl_renegotiate(conn->_base.bufev)<0) {
log_warn(LD_OR, "Start_renegotiating went badly.");
connection_mark_for_close(TO_CONN(conn));
}
tor_tls_unblock_renegotiation(conn->tls);
return; /* ???? */
}
} else if (tor_tls_get_num_server_handshakes(conn->tls) == 1) {
/* improved handshake, as a server. Only got one handshake, so
* wait for the next one. */
tor_tls_set_renegotiate_callback(conn->tls,
connection_or_tls_renegotiated_cb,
conn);
conn->_base.state = OR_CONN_STATE_TLS_SERVER_RENEGOTIATING;
/* return 0; */
return; /* ???? */
} else {
const int handshakes = tor_tls_get_num_server_handshakes(conn->tls);
tor_assert(handshakes >= 2);
if (handshakes == 2) {
/* improved handshake, as a server. Two handshakes happened already,
* so we treat renegotiation as done.
*/
connection_or_tls_renegotiated_cb(conn->tls, conn);
} else {
log_warn(LD_OR, "More than two handshakes done on connection. "
"Closing.");
connection_mark_for_close(TO_CONN(conn));
}
return;
}
}
connection_watch_events(TO_CONN(conn), READ_EVENT|WRITE_EVENT);
if (connection_tls_finish_handshake(conn) < 0)
connection_mark_for_close(TO_CONN(conn)); /* ???? */
return;
}
if (event & BEV_EVENT_ERROR) {
unsigned long err;
while ((err = bufferevent_get_openssl_error(bufev))) {
tor_tls_log_one_error(conn->tls, err, LOG_WARN, LD_OR,
"handshaking (with bufferevent)");
}
}
connection_handle_event_cb(bufev, event, arg);
}
#endif
/** Return 1 if we initiated this connection, or 0 if it started
* out as an incoming connection.
*/
int
connection_or_nonopen_was_started_here(or_connection_t *conn)
{
tor_assert(conn->_base.type == CONN_TYPE_OR);
if (!conn->tls)
return 1; /* it's still in proxy states or something */
if (conn->handshake_state)
return conn->handshake_state->started_here;
return !tor_tls_is_server(conn->tls);
}
/** <b>Conn</b> just completed its handshake. Return 0 if all is well, and
* return -1 if he is lying, broken, or otherwise something is wrong.
*
* If we initiated this connection (<b>started_here</b> is true), make sure
* the other side sent a correctly formed certificate. If I initiated the
* connection, make sure it's the right guy.
*
* Otherwise (if we _didn't_ initiate this connection), it's okay for
* the certificate to be weird or absent.
*
* If we return 0, and the certificate is as expected, write a hash of the
* identity key into <b>digest_rcvd_out</b>, which must have DIGEST_LEN
* space in it.
* If the certificate is invalid or missing on an incoming connection,
* we return 0 and set <b>digest_rcvd_out</b> to DIGEST_LEN NUL bytes.
* (If we return -1, the contents of this buffer are undefined.)
*
* As side effects,
2006-07-04 05:19:59 +02:00
* 1) Set conn->circ_id_type according to tor-spec.txt.
* 2) If we're an authdirserver and we initiated the connection: drop all
* descriptors that claim to be on that IP/port but that aren't
* this guy; and note that this guy is reachable.
* 3) If this is a bridge and we didn't configure its identity
* fingerprint, remember the keyid we just learned.
*/
static int
connection_or_check_valid_tls_handshake(or_connection_t *conn,
int started_here,
char *digest_rcvd_out)
{
crypto_pk_env_t *identity_rcvd=NULL;
or_options_t *options = get_options();
2005-10-17 03:29:28 +02:00
int severity = server_mode(options) ? LOG_PROTOCOL_WARN : LOG_WARN;
const char *safe_address =
started_here ? conn->_base.address :
safe_str_client(conn->_base.address);
const char *conn_type = started_here ? "outgoing" : "incoming";
crypto_pk_env_t *our_identity =
started_here ? get_tlsclient_identity_key() :
get_server_identity_key();
int has_cert = 0, has_identity=0;
check_no_tls_errors();
has_cert = tor_tls_peer_has_cert(conn->tls);
if (started_here && !has_cert) {
log_info(LD_HANDSHAKE,"Tried connecting to router at %s:%d, but it didn't "
"send a cert! Closing.",
safe_address, conn->_base.port);
return -1;
} else if (!has_cert) {
log_debug(LD_HANDSHAKE,"Got incoming connection with no certificate. "
"That's ok.");
}
check_no_tls_errors();
if (has_cert) {
int v = tor_tls_verify(started_here?severity:LOG_INFO,
conn->tls, &identity_rcvd);
if (started_here && v<0) {
log_fn(severity,LD_HANDSHAKE,"Tried connecting to router at %s:%d: It"
" has a cert but it's invalid. Closing.",
safe_address, conn->_base.port);
return -1;
} else if (v<0) {
log_info(LD_HANDSHAKE,"Incoming connection gave us an invalid cert "
"chain; ignoring.");
} else {
log_debug(LD_HANDSHAKE,
"The certificate seems to be valid on %s connection "
"with %s:%d", conn_type, safe_address, conn->_base.port);
}
check_no_tls_errors();
}
if (identity_rcvd) {
has_identity = 1;
crypto_pk_get_digest(identity_rcvd, digest_rcvd_out);
if (crypto_pk_cmp_keys(our_identity, identity_rcvd)<0) {
conn->circ_id_type = CIRC_ID_TYPE_LOWER;
} else {
conn->circ_id_type = CIRC_ID_TYPE_HIGHER;
}
crypto_free_pk_env(identity_rcvd);
} else {
memset(digest_rcvd_out, 0, DIGEST_LEN);
conn->circ_id_type = CIRC_ID_TYPE_NEITHER;
}
if (started_here && tor_digest_is_zero(conn->identity_digest)) {
connection_or_set_identity_digest(conn, digest_rcvd_out);
tor_free(conn->nickname);
conn->nickname = tor_malloc(HEX_DIGEST_LEN+2);
conn->nickname[0] = '$';
base16_encode(conn->nickname+1, HEX_DIGEST_LEN+1,
conn->identity_digest, DIGEST_LEN);
log_info(LD_HANDSHAKE, "Connected to router %s at %s:%d without knowing "
"its key. Hoping for the best.",
conn->nickname, conn->_base.address, conn->_base.port);
/* if it's a bridge and we didn't know its identity fingerprint, now
* we do -- remember it for future attempts. */
learned_router_identity(&conn->_base.addr, conn->_base.port,
digest_rcvd_out);
}
if (started_here) {
int as_advertised = 1;
tor_assert(has_cert);
tor_assert(has_identity);
if (memcmp(digest_rcvd_out, conn->identity_digest, DIGEST_LEN)) {
/* I was aiming for a particular digest. I didn't get it! */
char seen[HEX_DIGEST_LEN+1];
char expected[HEX_DIGEST_LEN+1];
base16_encode(seen, sizeof(seen), digest_rcvd_out, DIGEST_LEN);
base16_encode(expected, sizeof(expected), conn->identity_digest,
DIGEST_LEN);
log_fn(severity, LD_HANDSHAKE,
"Tried connecting to router at %s:%d, but identity key was not "
"as expected: wanted %s but got %s.",
conn->_base.address, conn->_base.port, expected, seen);
entry_guard_register_connect_status(conn->identity_digest, 0, 1,
time(NULL));
control_event_or_conn_status(conn, OR_CONN_EVENT_FAILED,
END_OR_CONN_REASON_OR_IDENTITY);
if (!authdir_mode_tests_reachability(options))
control_event_bootstrap_problem("foo", END_OR_CONN_REASON_OR_IDENTITY);
as_advertised = 0;
}
if (authdir_mode_tests_reachability(options)) {
/* We initiated this connection to address:port. Drop all routers
* with the same address:port and a different key.
*/
dirserv_orconn_tls_done(conn->_base.address, conn->_base.port,
digest_rcvd_out, as_advertised);
}
if (!as_advertised)
return -1;
}
return 0;
}
/** The tls handshake is finished.
*
* Make sure we are happy with the person we just handshaked with.
*
* If he initiated the connection, make sure he's not already connected,
* then initialize conn from the information in router.
*
* If all is successful, call circuit_n_conn_done() to handle events
* that have been pending on the <tls handshake completion. Also set the
* directory to be dirty (only matters if I'm an authdirserver).
*/
static int
connection_tls_finish_handshake(or_connection_t *conn)
{
char digest_rcvd[DIGEST_LEN];
int started_here = connection_or_nonopen_was_started_here(conn);
log_debug(LD_HANDSHAKE,"%s tls handshake on %p with %s done. verifying.",
started_here?"outgoing":"incoming",
conn,
safe_str_client(conn->_base.address));
directory_set_dirty();
if (connection_or_check_valid_tls_handshake(conn, started_here,
digest_rcvd) < 0)
return -1;
circuit_build_times_network_is_live(&circ_times);
if (tor_tls_used_v1_handshake(conn->tls)) {
conn->link_proto = 1;
if (!started_here) {
connection_or_init_conn_from_address(conn, &conn->_base.addr,
conn->_base.port, digest_rcvd, 0);
}
tor_tls_block_renegotiation(conn->tls);
return connection_or_set_state_open(conn);
} else {
conn->_base.state = OR_CONN_STATE_OR_HANDSHAKING;
if (connection_init_or_handshake_state(conn, started_here) < 0)
return -1;
if (!started_here) {
connection_or_init_conn_from_address(conn, &conn->_base.addr,
conn->_base.port, digest_rcvd, 0);
}
return connection_or_send_versions(conn);
}
}
/** Allocate a new connection handshake state for the connection
* <b>conn</b>. Return 0 on success, -1 on failure. */
static int
connection_init_or_handshake_state(or_connection_t *conn, int started_here)
{
or_handshake_state_t *s;
s = conn->handshake_state = tor_malloc_zero(sizeof(or_handshake_state_t));
s->started_here = started_here ? 1 : 0;
return 0;
}
/** Free all storage held by <b>state</b>. */
void
or_handshake_state_free(or_handshake_state_t *state)
{
if (!state)
return;
memset(state, 0xBE, sizeof(or_handshake_state_t));
tor_free(state);
}
/** Set <b>conn</b>'s state to OR_CONN_STATE_OPEN, and tell other subsystems
* as appropriate. Called when we are done with all TLS and OR handshaking.
*/
int
connection_or_set_state_open(or_connection_t *conn)
{
int started_here = connection_or_nonopen_was_started_here(conn);
time_t now = time(NULL);
conn->_base.state = OR_CONN_STATE_OPEN;
control_event_or_conn_status(conn, OR_CONN_EVENT_CONNECTED, 0);
if (started_here) {
circuit_build_times_network_is_live(&circ_times);
rep_hist_note_connect_succeeded(conn->identity_digest, now);
if (entry_guard_register_connect_status(conn->identity_digest,
1, 0, now) < 0) {
/* Close any circuits pending on this conn. We leave it in state
* 'open' though, because it didn't actually *fail* -- we just
* chose not to use it. (Otherwise
* connection_about_to_close_connection() will call a big pile of
* functions to indicate we shouldn't try it again.) */
2008-06-11 01:00:11 +02:00
log_debug(LD_OR, "New entry guard was reachable, but closing this "
"connection so we can retry the earlier entry guards.");
circuit_n_conn_done(conn, 0);
return -1;
}
router_set_status(conn->identity_digest, 1);
} else {
/* only report it to the geoip module if it's not a known router */
if (!router_get_by_id_digest(conn->identity_digest)) {
if (tor_addr_family(&TO_CONN(conn)->addr) == AF_INET) {
/*XXXX IP6 support ipv6 geoip.*/
uint32_t a = tor_addr_to_ipv4h(&TO_CONN(conn)->addr);
geoip_note_client_seen(GEOIP_CLIENT_CONNECT, a, now);
}
}
}
or_handshake_state_free(conn->handshake_state);
conn->handshake_state = NULL;
IF_HAS_BUFFEREVENT(TO_CONN(conn), {
connection_watch_events(TO_CONN(conn), READ_EVENT|WRITE_EVENT);
}) ELSE_IF_NO_BUFFEREVENT {
connection_start_reading(TO_CONN(conn));
}
circuit_n_conn_done(conn, 1); /* send the pending creates, if any. */
return 0;
}
/** Pack <b>cell</b> into wire-format, and write it onto <b>conn</b>'s outbuf.
* For cells that use or affect a circuit, this should only be called by
2008-01-19 21:00:53 +01:00
* connection_or_flush_from_first_active_circuit().
*/
void
connection_or_write_cell_to_buf(const cell_t *cell, or_connection_t *conn)
{
packed_cell_t networkcell;
tor_assert(cell);
tor_assert(conn);
2003-10-09 20:45:14 +02:00
cell_pack(&networkcell, cell);
2003-12-17 22:09:31 +01:00
connection_write_to_buf(networkcell.body, CELL_NETWORK_SIZE, TO_CONN(conn));
if (cell->command != CELL_PADDING)
conn->timestamp_last_added_nonpadding = approx_time();
}
/** Pack a variable-length <b>cell</b> into wire-format, and write it onto
* <b>conn</b>'s outbuf. Right now, this <em>DOES NOT</em> support cells that
* affect a circuit.
*/
void
connection_or_write_var_cell_to_buf(const var_cell_t *cell,
or_connection_t *conn)
{
char hdr[VAR_CELL_HEADER_SIZE];
tor_assert(cell);
tor_assert(conn);
var_cell_pack_header(cell, hdr);
connection_write_to_buf(hdr, sizeof(hdr), TO_CONN(conn));
connection_write_to_buf(cell->payload, cell->payload_len, TO_CONN(conn));
if (cell->command != CELL_PADDING)
conn->timestamp_last_added_nonpadding = approx_time();
}
/** See whether there's a variable-length cell waiting on <b>or_conn</b>'s
* inbuf. Return values as for fetch_var_cell_from_buf(). */
static int
connection_fetch_var_cell_from_buf(or_connection_t *or_conn, var_cell_t **out)
{
connection_t *conn = TO_CONN(or_conn);
IF_HAS_BUFFEREVENT(conn, {
struct evbuffer *input = bufferevent_get_input(conn->bufev);
return fetch_var_cell_from_evbuffer(input, out, or_conn->link_proto);
}) ELSE_IF_NO_BUFFEREVENT {
return fetch_var_cell_from_buf(conn->inbuf, out, or_conn->link_proto);
}
}
/** Process cells from <b>conn</b>'s inbuf.
*
* Loop: while inbuf contains a cell, pull it off the inbuf, unpack it,
* and hand it to command_process_cell().
*
* Always return 0.
*/
static int
connection_or_process_cells_from_inbuf(or_connection_t *conn)
{
var_cell_t *var_cell;
while (1) {
log_debug(LD_OR,
"%d: starting, inbuf_datalen %d (%d pending in tls object).",
conn->_base.s,(int)connection_get_inbuf_len(TO_CONN(conn)),
tor_tls_get_pending_bytes(conn->tls));
if (connection_fetch_var_cell_from_buf(conn, &var_cell)) {
if (!var_cell)
return 0; /* not yet. */
circuit_build_times_network_is_live(&circ_times);
command_process_var_cell(var_cell, conn);
var_cell_free(var_cell);
} else {
char buf[CELL_NETWORK_SIZE];
cell_t cell;
if (connection_get_inbuf_len(TO_CONN(conn))
< CELL_NETWORK_SIZE) /* whole response available? */
return 0; /* not yet */
2003-12-17 22:09:31 +01:00
circuit_build_times_network_is_live(&circ_times);
connection_fetch_from_buf(buf, CELL_NETWORK_SIZE, TO_CONN(conn));
2003-12-17 22:09:31 +01:00
/* retrieve cell info from buf (create the host-order struct from the
* network-order string) */
cell_unpack(&cell, buf);
2003-10-09 20:45:14 +02:00
command_process_cell(&cell, conn);
}
}
}
/** Write a destroy cell with circ ID <b>circ_id</b> and reason <b>reason</b>
* onto OR connection <b>conn</b>. Don't perform range-checking on reason:
* we may want to propagate reasons from other cells.
*
* Return 0.
*/
int
connection_or_send_destroy(circid_t circ_id, or_connection_t *conn, int reason)
{
cell_t cell;
tor_assert(conn);
memset(&cell, 0, sizeof(cell_t));
cell.circ_id = circ_id;
cell.command = CELL_DESTROY;
cell.payload[0] = (uint8_t) reason;
2006-02-13 10:02:35 +01:00
log_debug(LD_OR,"Sending destroy (circID %d).", circ_id);
connection_or_write_cell_to_buf(&cell, conn);
return 0;
}
/** Array of recognized link protocol versions. */
static const uint16_t or_protocol_versions[] = { 1, 2 };
/** Number of versions in <b>or_protocol_versions</b>. */
static const int n_or_protocol_versions =
(int)( sizeof(or_protocol_versions)/sizeof(uint16_t) );
/** Return true iff <b>v</b> is a link protocol version that this Tor
* implementation believes it can support. */
int
is_or_protocol_version_known(uint16_t v)
{
int i;
for (i = 0; i < n_or_protocol_versions; ++i) {
if (or_protocol_versions[i] == v)
return 1;
}
return 0;
}
/** Send a VERSIONS cell on <b>conn</b>, telling the other host about the
* link protocol versions that this Tor can support. */
static int
connection_or_send_versions(or_connection_t *conn)
{
var_cell_t *cell;
int i;
tor_assert(conn->handshake_state &&
!conn->handshake_state->sent_versions_at);
cell = var_cell_new(n_or_protocol_versions * 2);
cell->command = CELL_VERSIONS;
for (i = 0; i < n_or_protocol_versions; ++i) {
uint16_t v = or_protocol_versions[i];
set_uint16(cell->payload+(2*i), htons(v));
}
connection_or_write_var_cell_to_buf(cell, conn);
conn->handshake_state->sent_versions_at = time(NULL);
var_cell_free(cell);
return 0;
}
/** Send a NETINFO cell on <b>conn</b>, telling the other server what we know
* about their address, our address, and the current time. */
int
connection_or_send_netinfo(or_connection_t *conn)
{
cell_t cell;
time_t now = time(NULL);
const routerinfo_t *me;
int len;
char *out;
memset(&cell, 0, sizeof(cell_t));
cell.command = CELL_NETINFO;
/* Timestamp. */
set_uint32(cell.payload, htonl((uint32_t)now));
/* Their address. */
out = cell.payload + 4;
len = append_address_to_payload(out, &conn->_base.addr);
if (len<0)
return -1;
out += len;
/* My address. */
if ((me = router_get_my_routerinfo())) {
tor_addr_t my_addr;
*out++ = 1; /* only one address is supported. */
tor_addr_from_ipv4h(&my_addr, me->addr);
len = append_address_to_payload(out, &my_addr);
if (len < 0)
return -1;
out += len;
} else {
*out++ = 0;
}
connection_or_write_cell_to_buf(&cell, conn);
return 0;
}