tor/src/or/connection.c

2707 lines
88 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) 2004-2007, Roger Dingledine, Nick Mathewson. */
/* See LICENSE for licensing information */
/* $Id$ */
const char connection_c_id[] =
"$Id$";
2002-06-27 00:45:49 +02:00
/**
* \file connection.c
* \brief General high-level functions to handle reading and writing
* on connections.
**/
2002-06-27 00:45:49 +02:00
#include "or.h"
static connection_t *connection_create_listener(const char *listenaddress,
uint16_t listenport, int type);
static int connection_init_accepted_conn(connection_t *conn,
uint8_t listener_type);
2003-10-09 20:45:14 +02:00
static int connection_handle_listener_read(connection_t *conn, int new_type);
static int connection_read_bucket_should_increase(or_connection_t *conn);
static int connection_finished_flushing(connection_t *conn);
static int connection_flushed_some(connection_t *conn);
static int connection_finished_connecting(connection_t *conn);
static int connection_reached_eof(connection_t *conn);
static int connection_read_to_buf(connection_t *conn, int *max_to_read);
static int connection_process_inbuf(connection_t *conn, int package_partial);
static void client_check_address_changed(int sock);
static uint32_t last_interface_ip = 0;
static smartlist_t *outgoing_addrs = NULL;
/**************************************************************/
/**
* Return the human-readable name for the connection type <b>type</b>
*/
const char *
conn_type_to_string(int type)
{
static char buf[64];
switch (type) {
case CONN_TYPE_OR_LISTENER: return "OR listener";
case CONN_TYPE_OR: return "OR";
case CONN_TYPE_EXIT: return "Exit";
case CONN_TYPE_AP_LISTENER: return "Socks listener";
case CONN_TYPE_AP_TRANS_LISTENER:
return "Transparent pf/netfilter listener";
case CONN_TYPE_AP_NATD_LISTENER: return "Transparent natd listener";
case CONN_TYPE_AP: return "Socks";
case CONN_TYPE_DIR_LISTENER: return "Directory listener";
case CONN_TYPE_DIR: return "Directory";
case CONN_TYPE_CPUWORKER: return "CPU worker";
case CONN_TYPE_CONTROL_LISTENER: return "Control listener";
case CONN_TYPE_CONTROL: return "Control";
default:
log_warn(LD_BUG, "unknown connection type %d", type);
tor_snprintf(buf, sizeof(buf), "unknown [%d]", type);
return buf;
}
}
/**
* Return the human-readable name for the connection state <b>state</b>
* for the connection type <b>type</b>
*/
const char *
conn_state_to_string(int type, int state)
{
static char buf[96];
switch (type) {
case CONN_TYPE_OR_LISTENER:
case CONN_TYPE_AP_LISTENER:
case CONN_TYPE_AP_TRANS_LISTENER:
case CONN_TYPE_AP_NATD_LISTENER:
case CONN_TYPE_DIR_LISTENER:
case CONN_TYPE_CONTROL_LISTENER:
if (state == LISTENER_STATE_READY)
return "ready";
break;
case CONN_TYPE_OR:
switch (state) {
case OR_CONN_STATE_CONNECTING: return "connect()ing";
case OR_CONN_STATE_PROXY_FLUSHING: return "proxy flushing";
case OR_CONN_STATE_PROXY_READING: return "proxy reading";
2006-01-05 13:18:34 +01:00
case OR_CONN_STATE_HANDSHAKING: return "handshaking";
case OR_CONN_STATE_OPEN: return "open";
}
break;
case CONN_TYPE_EXIT:
switch (state) {
case EXIT_CONN_STATE_RESOLVING: return "waiting for dest info";
case EXIT_CONN_STATE_CONNECTING: return "connecting";
case EXIT_CONN_STATE_OPEN: return "open";
case EXIT_CONN_STATE_RESOLVEFAILED: return "resolve failed";
}
break;
case CONN_TYPE_AP:
switch (state) {
case AP_CONN_STATE_SOCKS_WAIT: return "waiting for socks info";
case AP_CONN_STATE_NATD_WAIT: return "waiting for natd dest info";
case AP_CONN_STATE_RENDDESC_WAIT: return "waiting for rendezvous desc";
case AP_CONN_STATE_CONTROLLER_WAIT: return "waiting for controller";
case AP_CONN_STATE_CIRCUIT_WAIT: return "waiting for circuit";
case AP_CONN_STATE_CONNECT_WAIT: return "waiting for connect response";
case AP_CONN_STATE_RESOLVE_WAIT: return "waiting for resolve response";
case AP_CONN_STATE_OPEN: return "open";
}
break;
case CONN_TYPE_DIR:
switch (state) {
case DIR_CONN_STATE_CONNECTING: return "connecting";
case DIR_CONN_STATE_CLIENT_SENDING: return "client sending";
2006-01-07 03:12:38 +01:00
case DIR_CONN_STATE_CLIENT_READING: return "client reading";
case DIR_CONN_STATE_CLIENT_FINISHED: return "client finished";
case DIR_CONN_STATE_SERVER_COMMAND_WAIT: return "waiting for command";
case DIR_CONN_STATE_SERVER_WRITING: return "writing";
}
break;
case CONN_TYPE_CPUWORKER:
switch (state) {
case CPUWORKER_STATE_IDLE: return "idle";
case CPUWORKER_STATE_BUSY_ONION: return "busy with onion";
}
break;
case CONN_TYPE_CONTROL:
switch (state) {
case CONTROL_CONN_STATE_OPEN: return "open (protocol v1)";
case CONTROL_CONN_STATE_NEEDAUTH:
return "waiting for authentication (protocol v1)";
}
break;
}
log_warn(LD_BUG, "unknown connection state %d (type %d)", state, type);
tor_snprintf(buf, sizeof(buf),
"unknown state [%d] on unknown [%s] connection",
state, conn_type_to_string(type));
return buf;
}
/** Allocate space for a new connection_t. This function just initializes
* conn; you must call connection_add() to link it into the main array.
*
* Set conn-\>type to <b>type</b>. Set conn-\>s and conn-\>conn_array_index to
* -1 to signify they are not yet assigned.
*
* If conn is not a listener type, allocate buffers for it. If it's
* an AP type, allocate space to store the socks_request.
*
* Assign a pseudorandom next_circ_id between 0 and 2**15.
*
* Initialize conn's timestamps to now.
*/
connection_t *
connection_new(int type)
{
static uint32_t n_connections_allocated = 1;
2002-06-27 00:45:49 +02:00
connection_t *conn;
time_t now = time(NULL);
size_t length;
uint32_t magic;
2002-06-27 00:45:49 +02:00
switch (type) {
case CONN_TYPE_OR:
length = sizeof(or_connection_t);
magic = OR_CONNECTION_MAGIC;
break;
case CONN_TYPE_EXIT:
case CONN_TYPE_AP:
length = sizeof(edge_connection_t);
magic = EDGE_CONNECTION_MAGIC;
break;
case CONN_TYPE_DIR:
length = sizeof(dir_connection_t);
magic = DIR_CONNECTION_MAGIC;
break;
case CONN_TYPE_CONTROL:
length = sizeof(control_connection_t);
magic = CONTROL_CONNECTION_MAGIC;
break;
default:
length = sizeof(connection_t);
magic = BASE_CONNECTION_MAGIC;
break;
}
conn = tor_malloc_zero(length);
conn->magic = magic;
conn->s = -1; /* give it a default of 'not used' */
conn->conn_array_index = -1; /* also default to 'not used' */
2002-06-27 00:45:49 +02:00
conn->type = type;
if (!connection_is_listener(conn)) { /* listeners never use their buf */
conn->inbuf = buf_new();
conn->outbuf = buf_new();
}
if (type == CONN_TYPE_AP) {
TO_EDGE_CONN(conn)->socks_request =
tor_malloc_zero(sizeof(socks_request_t));
}
if (CONN_IS_EDGE(conn)) {
TO_EDGE_CONN(conn)->global_identifier = n_connections_allocated++;
}
if (type == CONN_TYPE_OR)
TO_OR_CONN(conn)->next_circ_id = crypto_rand_int(1<<15);
conn->timestamp_created = now;
conn->timestamp_lastread = now;
conn->timestamp_lastwritten = now;
2002-06-27 00:45:49 +02:00
return conn;
}
/** Create a link between <b>conn_a</b> and <b>conn_b</b> */
void
connection_link_connections(connection_t *conn_a, connection_t *conn_b)
{
tor_assert(conn_a->s < 0);
tor_assert(conn_b->s < 0);
conn_a->linked = 1;
conn_b->linked = 1;
conn_a->linked_conn = conn_b;
conn_b->linked_conn = conn_a;
}
/** Tell libevent that we don't care about <b>conn</b> any more. */
void
connection_unregister_events(connection_t *conn)
{
if (conn->read_event) {
if (event_del(conn->read_event))
2006-02-13 10:02:35 +01:00
log_warn(LD_BUG, "Error removing read event for %d", conn->s);
tor_free(conn->read_event);
}
if (conn->write_event) {
if (event_del(conn->write_event))
2006-02-13 10:02:35 +01:00
log_warn(LD_BUG, "Error removing write event for %d", conn->s);
tor_free(conn->write_event);
}
}
/** Deallocate memory used by <b>conn</b>. Deallocate its buffers if
* necessary, close its socket if necessary, and mark the directory as dirty
* if <b>conn</b> is an OR or OP connection.
*/
static void
_connection_free(connection_t *conn)
{
void *mem;
switch (conn->type) {
case CONN_TYPE_OR:
tor_assert(conn->magic == OR_CONNECTION_MAGIC);
mem = TO_OR_CONN(conn);
break;
case CONN_TYPE_AP:
case CONN_TYPE_EXIT:
tor_assert(conn->magic == EDGE_CONNECTION_MAGIC);
mem = TO_EDGE_CONN(conn);
break;
case CONN_TYPE_DIR:
tor_assert(conn->magic == DIR_CONNECTION_MAGIC);
mem = TO_DIR_CONN(conn);
break;
case CONN_TYPE_CONTROL:
tor_assert(conn->magic == CONTROL_CONNECTION_MAGIC);
mem = TO_CONTROL_CONN(conn);
break;
default:
tor_assert(conn->magic == BASE_CONNECTION_MAGIC);
mem = conn;
break;
}
2002-06-27 00:45:49 +02:00
if (conn->linked) {
int severity = buf_datalen(conn->inbuf)+buf_datalen(conn->outbuf)
? LOG_NOTICE : LOG_INFO;
log_fn(severity, LD_GENERAL, "Freeing linked %s connection [%s] with %d "
"bytes on inbuf, %d on outbuf.",
conn_type_to_string(conn->type),
conn_state_to_string(conn->type, conn->state),
(int)buf_datalen(conn->inbuf), (int)buf_datalen(conn->outbuf));
// tor_assert(!buf_datalen(conn->outbuf)); /*XXXX020 remove me.*/
}
if (!connection_is_listener(conn)) {
buf_free(conn->inbuf);
buf_free(conn->outbuf);
}
tor_free(conn->address);
2002-06-27 00:45:49 +02:00
if (connection_speaks_cells(conn)) {
or_connection_t *or_conn = TO_OR_CONN(conn);
if (or_conn->tls) {
tor_tls_free(or_conn->tls);
or_conn->tls = NULL;
}
tor_free(or_conn->nickname);
}
if (CONN_IS_EDGE(conn)) {
edge_connection_t *edge_conn = TO_EDGE_CONN(conn);
tor_free(edge_conn->chosen_exit_name);
tor_free(edge_conn->socks_request);
}
if (conn->type == CONN_TYPE_CONTROL) {
control_connection_t *control_conn = TO_CONTROL_CONN(conn);
tor_free(control_conn->incoming_cmd);
2002-06-27 00:45:49 +02:00
}
tor_free(conn->read_event); /* Probably already freed by connection_free. */
tor_free(conn->write_event); /* Probably already freed by connection_free. */
if (conn->type == CONN_TYPE_DIR) {
dir_connection_t *dir_conn = TO_DIR_CONN(conn);
tor_free(dir_conn->requested_resource);
if (dir_conn->zlib_state)
tor_zlib_free(dir_conn->zlib_state);
if (dir_conn->fingerprint_stack) {
SMARTLIST_FOREACH(dir_conn->fingerprint_stack, char *, cp, tor_free(cp));
smartlist_free(dir_conn->fingerprint_stack);
}
if (dir_conn->cached_dir)
cached_dir_decref(dir_conn->cached_dir);
}
if (conn->s >= 0) {
2006-02-13 10:02:35 +01:00
log_debug(LD_NET,"closing fd %d.",conn->s);
tor_close_socket(conn->s);
}
if (conn->type == CONN_TYPE_OR &&
!tor_digest_is_zero(TO_OR_CONN(conn)->identity_digest)) {
2006-02-13 10:02:35 +01:00
log_warn(LD_BUG, "called on OR conn with non-zeroed identity_digest");
connection_or_remove_from_identity_map(TO_OR_CONN(conn));
}
memset(conn, 0xAA, sizeof(connection_t)); /* poison memory */
tor_free(mem);
2002-06-27 00:45:49 +02:00
}
/** Make sure <b>conn</b> isn't in any of the global conn lists; then free it.
*/
void
connection_free(connection_t *conn)
{
tor_assert(conn);
tor_assert(!connection_is_on_closeable_list(conn));
tor_assert(!connection_in_array(conn));
if (conn->linked_conn) {
log_err(LD_BUG, "Called with conn->linked_conn still set.");
tor_fragile_assert();
conn->linked_conn->linked_conn = NULL;
if (! conn->linked_conn->marked_for_close &&
conn->linked_conn->reading_from_linked_conn)
connection_start_reading(conn->linked_conn);
conn->linked_conn = NULL;
}
if (connection_speaks_cells(conn)) {
if (conn->state == OR_CONN_STATE_OPEN)
directory_set_dirty();
if (!tor_digest_is_zero(TO_OR_CONN(conn)->identity_digest)) {
connection_or_remove_from_identity_map(TO_OR_CONN(conn));
}
}
if (conn->type == CONN_TYPE_CONTROL) {
TO_CONTROL_CONN(conn)->event_mask = 0;
control_update_global_event_mask();
}
connection_unregister_events(conn);
_connection_free(conn);
}
/** Call _connection_free() on every connection in our array, and release all
* storage helpd by connection.c. This is used by cpuworkers and dnsworkers
* when they fork, so they don't keep resources held open (especially
* sockets).
*
* Don't do the checks in connection_free(), because they will
* fail.
*/
void
connection_free_all(void)
{
int i, n;
connection_t **carray;
get_connection_array(&carray,&n);
/* We don't want to log any messages to controllers. */
for (i=0;i<n;i++)
if (carray[i]->type == CONN_TYPE_CONTROL)
TO_CONTROL_CONN(carray[i])->event_mask = 0;
control_update_global_event_mask();
/* Unlink everything from the identity map. */
connection_or_clear_identity_map();
for (i=0;i<n;i++)
_connection_free(carray[i]);
if (outgoing_addrs) {
SMARTLIST_FOREACH(outgoing_addrs, void*, addr, tor_free(addr));
smartlist_free(outgoing_addrs);
outgoing_addrs = NULL;
}
}
/** Do any cleanup needed:
* - Directory conns that failed to fetch a rendezvous descriptor
* need to inform pending rendezvous streams.
* - OR conns need to call rep_hist_note_*() to record status.
* - AP conns need to send a socks reject if necessary.
* - Exit conns need to call connection_dns_remove() if necessary.
* - AP and Exit conns need to send an end cell if they can.
* - DNS conns need to fail any resolves that are pending on them.
* - OR and edge connections need to be unlinked from circuits.
*/
void
connection_about_to_close_connection(connection_t *conn)
{
circuit_t *circ;
dir_connection_t *dir_conn;
or_connection_t *or_conn;
edge_connection_t *edge_conn;
time_t now = time(NULL);
assert(conn->marked_for_close);
if (CONN_IS_EDGE(conn)) {
if (!conn->edge_has_sent_end) {
log_warn(LD_BUG, "(Harmless.) Edge connection (marked at %s:%d) "
2006-02-13 10:02:35 +01:00
"hasn't sent end yet?",
conn->marked_for_close_file, conn->marked_for_close);
tor_fragile_assert();
}
}
switch (conn->type) {
case CONN_TYPE_DIR:
dir_conn = TO_DIR_CONN(conn);
if (conn->state < DIR_CONN_STATE_CLIENT_FINISHED) {
/* It's a directory connection and connecting or fetching
* failed: forget about this router, and maybe try again. */
connection_dir_request_failed(dir_conn);
// XXX if it's rend desc we may want to retry -RD
}
if (conn->purpose == DIR_PURPOSE_FETCH_RENDDESC)
rend_client_desc_here(dir_conn->rend_query); /* give it a try */
break;
case CONN_TYPE_OR:
or_conn = TO_OR_CONN(conn);
/* Remember why we're closing this connection. */
if (conn->state != OR_CONN_STATE_OPEN) {
if (connection_or_nonopen_was_started_here(or_conn)) {
rep_hist_note_connect_failed(or_conn->identity_digest, now);
entry_guard_register_connect_status(or_conn->identity_digest,0,now);
router_set_status(or_conn->identity_digest, 0);
control_event_or_conn_status(or_conn, OR_CONN_EVENT_FAILED,
control_tls_error_to_reason(or_conn->tls_error));
}
/* Inform any pending (not attached) circs that they should
* give up. */
circuit_n_conn_done(TO_OR_CONN(conn), 0);
} else if (conn->hold_open_until_flushed) {
/* We only set hold_open_until_flushed when we're intentionally
* closing a connection. */
rep_hist_note_disconnect(or_conn->identity_digest, now);
control_event_or_conn_status(or_conn, OR_CONN_EVENT_CLOSED,
control_tls_error_to_reason(or_conn->tls_error));
} else if (or_conn->identity_digest) {
rep_hist_note_connection_died(or_conn->identity_digest, now);
control_event_or_conn_status(or_conn, OR_CONN_EVENT_CLOSED,
control_tls_error_to_reason(or_conn->tls_error));
}
/* Now close all the attached circuits on it. */
circuit_unlink_all_from_or_conn(TO_OR_CONN(conn),
END_CIRC_REASON_OR_CONN_CLOSED);
break;
case CONN_TYPE_AP:
edge_conn = TO_EDGE_CONN(conn);
if (edge_conn->socks_request->has_finished == 0) {
/* since conn gets removed right after this function finishes,
* there's no point trying to send back a reply at this point. */
log_warn(LD_BUG,"Closing stream (marked at %s:%d) without sending"
2006-02-13 10:02:35 +01:00
" back a socks reply.",
conn->marked_for_close_file, conn->marked_for_close);
}
if (!edge_conn->end_reason) {
log_warn(LD_BUG,"Closing stream (marked at %s:%d) without having"
" set end_reason.",
conn->marked_for_close_file, conn->marked_for_close);
}
control_event_stream_status(edge_conn, STREAM_EVENT_CLOSED,
edge_conn->end_reason);
circ = circuit_get_by_edge_conn(edge_conn);
if (circ)
circuit_detach_stream(circ, edge_conn);
break;
case CONN_TYPE_EXIT:
edge_conn = TO_EDGE_CONN(conn);
circ = circuit_get_by_edge_conn(edge_conn);
if (circ)
circuit_detach_stream(circ, edge_conn);
if (conn->state == EXIT_CONN_STATE_RESOLVING) {
connection_dns_remove(edge_conn);
}
break;
}
}
/** Close the underlying socket for <b>conn</b>, so we don't try to
* flush it. Must be used in conjunction with (right before)
* connection_mark_for_close().
*/
void
connection_close_immediate(connection_t *conn)
{
assert_connection_ok(conn,0);
if (conn->s < 0 && !conn->linked) {
log_err(LD_BUG,"Attempt to close already-closed connection.");
tor_fragile_assert();
return;
}
if (conn->outbuf_flushlen) {
2006-02-13 10:02:35 +01:00
log_info(LD_NET,"fd %d, type %s, state %s, %d bytes on outbuf.",
conn->s, conn_type_to_string(conn->type),
conn_state_to_string(conn->type, conn->state),
(int)conn->outbuf_flushlen);
}
connection_unregister_events(conn);
if (conn->s >= 0)
tor_close_socket(conn->s);
conn->s = -1;
if (!connection_is_listener(conn)) {
buf_clear(conn->outbuf);
conn->outbuf_flushlen = 0;
}
}
/** Mark <b>conn</b> to be closed next time we loop through
* conn_close_if_marked() in main.c. */
void
_connection_mark_for_close(connection_t *conn, int line, const char *file)
{
assert_connection_ok(conn,0);
tor_assert(line);
tor_assert(file);
if (conn->marked_for_close) {
log(LOG_WARN,LD_BUG,"Duplicate call to connection_mark_for_close at %s:%d"
" (first at %s:%d)", file, line, conn->marked_for_close_file,
conn->marked_for_close);
tor_fragile_assert();
return;
}
conn->marked_for_close = line;
conn->marked_for_close_file = file;
add_connection_to_closeable_list(conn);
#if 0
/* XXXX020 Actually, I don't think this is right. */
if (conn->linked_conn && !conn->linked_conn->marked_for_close)
_connection_mark_for_close(conn->linked_conn, line, file);
#endif
/* in case we're going to be held-open-til-flushed, reset
* the number of seconds since last successful write, so
* we get our whole 15 seconds */
conn->timestamp_lastwritten = time(NULL);
}
/** Find each connection that has hold_open_until_flushed set to
* 1 but hasn't written in the past 15 seconds, and set
* hold_open_until_flushed to 0. This means it will get cleaned
* up in the next loop through close_if_marked() in main.c.
*/
void
connection_expire_held_open(void)
{
connection_t **carray, *conn;
int n, i;
time_t now;
now = time(NULL);
get_connection_array(&carray, &n);
for (i = 0; i < n; ++i) {
conn = carray[i];
/* If we've been holding the connection open, but we haven't written
* for 15 seconds...
*/
if (conn->hold_open_until_flushed) {
tor_assert(conn->marked_for_close);
if (now - conn->timestamp_lastwritten >= 15) {
2005-10-17 03:29:28 +02:00
int severity;
if (conn->type == CONN_TYPE_EXIT ||
(conn->type == CONN_TYPE_DIR &&
conn->purpose == DIR_PURPOSE_SERVER))
2005-10-17 03:29:28 +02:00
severity = LOG_INFO;
else
severity = LOG_NOTICE;
log_fn(severity, LD_NET,
"Giving up on marked_for_close conn that's been flushing "
"for 15s (fd %d, type %s, state %s).",
conn->s, conn_type_to_string(conn->type),
conn_state_to_string(conn->type, conn->state));
conn->hold_open_until_flushed = 0;
}
}
}
}
/** Bind a new non-blocking socket listening to
* <b>listenaddress</b>:<b>listenport</b>, and add this new connection
* (of type <b>type</b>) to the connection array.
2004-05-19 22:25:44 +02:00
*
* If <b>listenaddress</b> includes a port, we bind on that port;
* otherwise, we use listenport.
*/
static connection_t *
connection_create_listener(const char *listenaddress, uint16_t listenport,
int type)
{
struct sockaddr_in listenaddr; /* where to bind */
char *address = NULL;
2002-06-27 00:45:49 +02:00
connection_t *conn;
uint16_t usePort;
uint32_t addr;
int s; /* the socket we're going to make */
#ifndef MS_WINDOWS
2002-06-27 00:45:49 +02:00
int one=1;
#endif
2002-06-27 00:45:49 +02:00
memset(&listenaddr,0,sizeof(struct sockaddr_in));
if (parse_addr_port(LOG_WARN, listenaddress, &address, &addr, &usePort)<0) {
2006-02-13 10:02:35 +01:00
log_warn(LD_CONFIG,
"Error parsing/resolving ListenAddress %s", listenaddress);
return NULL;
}
if (usePort==0)
usePort = listenport;
listenaddr.sin_addr.s_addr = htonl(addr);
listenaddr.sin_family = AF_INET;
listenaddr.sin_port = htons((uint16_t) usePort);
2006-02-13 10:02:35 +01:00
log_notice(LD_NET, "Opening %s on %s:%d",
conn_type_to_string(type), address, usePort);
2005-07-15 01:04:31 +02:00
if (get_n_open_sockets() >= get_options()->_ConnLimit-1) {
int n_conns = get_n_open_sockets();
log_warn(LD_NET,"Failing because we have %d connections already. Please "
"raise your ulimit -n.", n_conns);
control_event_general_status(LOG_WARN, "TOO_MANY_CONNECTIONS CURRENT=%d",
n_conns);
return NULL;
}
s = tor_open_socket(PF_INET,SOCK_STREAM,IPPROTO_TCP);
2003-12-17 22:09:31 +01:00
if (s < 0) {
2006-02-13 10:02:35 +01:00
log_warn(LD_NET,"Socket creation failed.");
goto err;
2002-06-27 00:45:49 +02:00
}
#ifndef MS_WINDOWS
/* REUSEADDR on normal places means you can rebind to the port
* right after somebody else has let it go. But REUSEADDR on win32
* means you can bind to the port _even when somebody else
* already has it bound_. So, don't do that on Win32. */
setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (void*) &one, sizeof(one));
#endif
2002-06-27 00:45:49 +02:00
if (bind(s,(struct sockaddr *)&listenaddr,sizeof(listenaddr)) < 0) {
const char *helpfulhint = "";
int e = tor_socket_errno(s);
if (ERRNO_IS_EADDRINUSE(e))
helpfulhint = ". Is Tor already running?";
2006-02-13 10:02:35 +01:00
log_warn(LD_NET, "Could not bind to %s:%u: %s%s", address, usePort,
tor_socket_strerror(e), helpfulhint);
tor_close_socket(s);
goto err;
2002-06-27 00:45:49 +02:00
}
if (listen(s,SOMAXCONN) < 0) {
2006-02-13 10:02:35 +01:00
log_warn(LD_NET, "Could not listen on %s:%u: %s", address, usePort,
tor_socket_strerror(tor_socket_errno(s)));
tor_close_socket(s);
goto err;
2002-06-27 00:45:49 +02:00
}
set_socket_nonblocking(s);
2002-06-27 00:45:49 +02:00
conn = connection_new(type);
conn->s = s;
conn->address = address;
address = NULL;
conn->port = usePort;
2002-06-27 00:45:49 +02:00
if (connection_add(conn) < 0) { /* no space, forget it */
log_warn(LD_NET,"connection_add for listener failed. Giving up.");
2002-06-27 00:45:49 +02:00
connection_free(conn);
goto err;
2002-06-27 00:45:49 +02:00
}
2006-02-13 10:02:35 +01:00
log_debug(LD_NET,"%s listening on port %u.",
conn_type_to_string(type), usePort);
2002-06-27 00:45:49 +02:00
conn->state = LISTENER_STATE_READY;
connection_start_reading(conn);
2002-06-27 00:45:49 +02:00
return conn;
err:
tor_free(address);
return NULL;
2002-06-27 00:45:49 +02:00
}
/** Do basic sanity checking on a newly received socket. Return 0
* if it looks ok, else return -1. */
static int
check_sockaddr_in(struct sockaddr *sa, int len, int level)
{
int ok = 1;
struct sockaddr_in *sin=(struct sockaddr_in*)sa;
if (len != sizeof(struct sockaddr_in)) {
log_fn(level, LD_NET, "Length of address not as expected: %d vs %d",
len,(int)sizeof(struct sockaddr_in));
ok = 0;
}
if (sa->sa_family != AF_INET) {
log_fn(level, LD_NET, "Family of address not as expected: %d vs %d",
sa->sa_family, AF_INET);
ok = 0;
}
if (sin->sin_addr.s_addr == 0 || sin->sin_port == 0) {
log_fn(level, LD_NET,
"Address for new connection has address/port equal to zero.");
ok = 0;
}
return ok ? 0 : -1;
}
/** The listener connection <b>conn</b> told poll() it wanted to read.
2004-05-10 06:34:48 +02:00
* Call accept() on conn-\>s, and add the new connection if necessary.
*/
static int
connection_handle_listener_read(connection_t *conn, int new_type)
{
2002-06-27 00:45:49 +02:00
int news; /* the new socket */
connection_t *newconn;
/* information about the remote peer when connecting to other routers */
struct sockaddr_in remote;
char addrbuf[256];
/* length of the remote address. Must be whatever accept() needs. */
socklen_t remotelen = 256;
char tmpbuf[INET_NTOA_BUF_LEN];
2005-04-08 05:26:44 +02:00
tor_assert((size_t)remotelen >= sizeof(struct sockaddr_in));
memset(addrbuf, 0, sizeof(addrbuf));
2002-06-27 00:45:49 +02:00
news = accept(conn->s,(struct sockaddr *)&addrbuf,&remotelen);
2006-07-21 01:33:11 +02:00
if (news < 0) { /* accept() error */
int e = tor_socket_errno(conn->s);
if (ERRNO_IS_ACCEPT_EAGAIN(e)) {
2002-06-27 00:45:49 +02:00
return 0; /* he hung up before we could accept(). that's fine. */
} else if (ERRNO_IS_ACCEPT_RESOURCE_LIMIT(e)) {
2006-02-13 10:02:35 +01:00
log_notice(LD_NET,"accept failed: %s. Dropping incoming connection.",
tor_socket_strerror(e));
return 0;
}
2002-06-27 00:45:49 +02:00
/* else there was a real error. */
2006-02-13 10:02:35 +01:00
log_warn(LD_NET,"accept() failed: %s. Closing listener.",
tor_socket_strerror(e));
connection_mark_for_close(conn);
2002-06-27 00:45:49 +02:00
return -1;
}
2006-02-13 10:02:35 +01:00
log_debug(LD_NET,
"Connection accepted on socket %d (child of fd %d).",
news,conn->s);
2002-06-27 00:45:49 +02:00
set_socket_nonblocking(news);
if (check_sockaddr_in((struct sockaddr*)addrbuf, remotelen, LOG_INFO)<0) {
2006-02-13 10:02:35 +01:00
log_info(LD_NET,
"accept() returned a strange address; trying getsockname().");
remotelen=256;
memset(addrbuf, 0, sizeof(addrbuf));
if (getsockname(news, (struct sockaddr*)addrbuf, &remotelen)<0) {
int e = tor_socket_errno(news);
log_warn(LD_NET, "getsockname() for new connection failed: %s",
tor_socket_strerror(e));
} else {
if (check_sockaddr_in((struct sockaddr*)addrbuf, remotelen,
LOG_WARN) < 0) {
2006-02-13 10:02:35 +01:00
log_warn(LD_NET,"Something's wrong with this conn. Closing it.");
tor_close_socket(news);
return 0;
}
}
}
memcpy(&remote, addrbuf, sizeof(struct sockaddr_in));
/* process entrance policies here, before we even create the connection */
if (new_type == CONN_TYPE_AP) {
/* check sockspolicy to see if we should accept it */
if (socks_policy_permits_address(ntohl(remote.sin_addr.s_addr)) == 0) {
tor_inet_ntoa(&remote.sin_addr, tmpbuf, sizeof(tmpbuf));
2006-02-13 10:02:35 +01:00
log_notice(LD_APP,"Denying socks connection from untrusted address %s.",
tmpbuf);
tor_close_socket(news);
return 0;
}
}
if (new_type == CONN_TYPE_DIR) {
/* check dirpolicy to see if we should accept it */
if (dir_policy_permits_address(ntohl(remote.sin_addr.s_addr)) == 0) {
tor_inet_ntoa(&remote.sin_addr, tmpbuf, sizeof(tmpbuf));
2006-02-13 10:02:35 +01:00
log_notice(LD_DIRSERV,"Denying dir connection from address %s.",
tmpbuf);
tor_close_socket(news);
return 0;
}
}
2002-06-27 00:45:49 +02:00
newconn = connection_new(new_type);
newconn->s = news;
/* remember the remote address */
newconn->addr = ntohl(remote.sin_addr.s_addr);
newconn->port = ntohs(remote.sin_port);
newconn->address = tor_dup_addr(newconn->addr);
2002-06-27 00:45:49 +02:00
if (connection_add(newconn) < 0) { /* no space, forget it */
2002-06-27 00:45:49 +02:00
connection_free(newconn);
return 0; /* no need to tear down the parent */
2002-06-27 00:45:49 +02:00
}
if (connection_init_accepted_conn(newconn, conn->type) < 0) {
connection_mark_for_close(newconn);
return 0;
}
return 0;
}
/** Initialize states for newly accepted connection <b>conn</b>.
* If conn is an OR, start the tls handshake.
* If conn is a transparent AP, get its original destination
* and place it in circuit_wait.
*/
static int
connection_init_accepted_conn(connection_t *conn, uint8_t listener_type)
{
connection_start_reading(conn);
switch (conn->type) {
case CONN_TYPE_OR:
control_event_or_conn_status(TO_OR_CONN(conn), OR_CONN_EVENT_NEW, 0);
return connection_tls_start_handshake(TO_OR_CONN(conn), 1);
case CONN_TYPE_AP:
switch (listener_type) {
case CONN_TYPE_AP_LISTENER:
conn->state = AP_CONN_STATE_SOCKS_WAIT;
break;
case CONN_TYPE_AP_TRANS_LISTENER:
conn->state = AP_CONN_STATE_CIRCUIT_WAIT;
return connection_ap_process_transparent(TO_EDGE_CONN(conn));
case CONN_TYPE_AP_NATD_LISTENER:
conn->state = AP_CONN_STATE_NATD_WAIT;
break;
}
break;
case CONN_TYPE_DIR:
conn->purpose = DIR_PURPOSE_SERVER;
conn->state = DIR_CONN_STATE_SERVER_COMMAND_WAIT;
break;
case CONN_TYPE_CONTROL:
conn->state = CONTROL_CONN_STATE_NEEDAUTH;
break;
}
return 0;
}
/** Take conn, make a nonblocking socket; try to connect to
* addr:port (they arrive in *host order*). If fail, return -1. Else
* assign s to conn-\>s: if connected return 1, if EAGAIN return 0.
*
* address is used to make the logs useful.
*
2004-05-10 06:34:48 +02:00
* On success, add conn to the list of polled connections.
*/
int
connection_connect(connection_t *conn, const char *address,
uint32_t addr, uint16_t port)
{
int s, inprogress = 0;
struct sockaddr_in dest_addr;
or_options_t *options = get_options();
if (get_n_open_sockets() >= get_options()->_ConnLimit-1) {
int n_conns = get_n_open_sockets();
log_warn(LD_NET,"Failing because we have %d connections already. Please "
"raise your ulimit -n.", n_conns);
control_event_general_status(LOG_WARN, "TOO_MANY_CONNECTIONS CURRENT=%d",
n_conns);
return -1;
}
s = tor_open_socket(PF_INET,SOCK_STREAM,IPPROTO_TCP);
if (s < 0) {
2006-02-13 10:02:35 +01:00
log_warn(LD_NET,"Error creating network socket: %s",
tor_socket_strerror(tor_socket_errno(-1)));
return -1;
}
if (options->OutboundBindAddress) {
struct sockaddr_in ext_addr;
memset(&ext_addr, 0, sizeof(ext_addr));
ext_addr.sin_family = AF_INET;
ext_addr.sin_port = 0;
if (!tor_inet_aton(options->OutboundBindAddress, &ext_addr.sin_addr)) {
2006-02-13 10:02:35 +01:00
log_warn(LD_CONFIG,"Outbound bind address '%s' didn't parse. Ignoring.",
options->OutboundBindAddress);
} else {
if (bind(s, (struct sockaddr*)&ext_addr, sizeof(ext_addr)) < 0) {
2006-02-13 10:02:35 +01:00
log_warn(LD_NET,"Error binding network socket: %s",
tor_socket_strerror(tor_socket_errno(s)));
tor_close_socket(s);
return -1;
}
}
}
set_socket_nonblocking(s);
memset(&dest_addr,0,sizeof(dest_addr));
dest_addr.sin_family = AF_INET;
dest_addr.sin_port = htons(port);
dest_addr.sin_addr.s_addr = htonl(addr);
log_debug(LD_NET,"Connecting to %s:%u.",escaped_safe_str(address),port);
if (connect(s,(struct sockaddr *)&dest_addr,sizeof(dest_addr)) < 0) {
int e = tor_socket_errno(s);
if (!ERRNO_IS_CONN_EINPROGRESS(e)) {
/* yuck. kill it. */
2006-02-13 10:02:35 +01:00
log_info(LD_NET,
"connect() to %s:%u failed: %s",escaped_safe_str(address),
port, tor_socket_strerror(e));
tor_close_socket(s);
return -1;
} else {
inprogress = 1;
}
}
if (!server_mode(options))
client_check_address_changed(s);
/* it succeeded. we're connected. */
log_fn(inprogress?LOG_DEBUG:LOG_INFO, LD_NET,
"Connection to %s:%u %s (sock %d).",escaped_safe_str(address),
port, inprogress?"in progress":"established", s);
conn->s = s;
if (connection_add(conn) < 0) /* no space, forget it */
return -1;
return inprogress ? 0 : 1;
}
/**
* Launch any configured listener connections of type <b>type</b>. (A
* listener is configured if <b>port_option</b> is non-zero. If any
* ListenAddress configuration options are given in <b>cfg</b>, create a
* connection binding to each one. Otherwise, create a single
* connection binding to the address <b>default_addr</b>.)
*
* If <b>force</b> is true, close and re-open all listener connections.
* Otherwise, only relaunch the listeners of this type if the number of
* existing connections is not as configured (e.g., because one died),
* or if the existing connections do not match those configured.
*
* Add all old conns that should be closed to <b>replaced_conns</b>.
* Add all new connections to <b>new_conns</b>.
*/
static int
retry_listeners(int type, config_line_t *cfg,
int port_option, const char *default_addr, int force,
smartlist_t *replaced_conns,
smartlist_t *new_conns,
int never_open_conns)
{
smartlist_t *launch = smartlist_create();
int free_launch_elts = 1;
config_line_t *c;
int n_conn, i;
connection_t *conn;
connection_t **carray;
config_line_t *line;
if (cfg && port_option) {
for (c = cfg; c; c = c->next) {
smartlist_add(launch, c);
}
free_launch_elts = 0;
} else if (port_option) {
line = tor_malloc_zero(sizeof(config_line_t));
line->key = tor_strdup("");
line->value = tor_strdup(default_addr);
smartlist_add(launch, line);
}
/*
SMARTLIST_FOREACH(launch, config_line_t *, l,
log_fn(LOG_NOTICE, "#%s#%s", l->key, l->value));
*/
get_connection_array(&carray,&n_conn);
for (i=0; i < n_conn; ++i) {
conn = carray[i];
if (conn->type != type || conn->marked_for_close)
continue;
if (force) {
/* It's a listener, and we're relaunching all listeners of this
* type. Close this one. */
log_notice(LD_NET, "Force-closing listener %s on %s:%d",
conn_type_to_string(type), conn->address, conn->port);
connection_close_immediate(conn);
connection_mark_for_close(conn);
continue;
}
/* Okay, so this is a listener. Is it configured? */
line = NULL;
SMARTLIST_FOREACH(launch, config_line_t *, wanted,
{
char *address=NULL;
uint16_t port;
if (!parse_addr_port(LOG_WARN, wanted->value, &address, NULL, &port)) {
int addr_matches = !strcasecmp(address, conn->address);
tor_free(address);
if (! port)
port = port_option;
if (port == conn->port && addr_matches) {
line = wanted;
break;
}
}
});
if (! line) {
/* This one isn't configured. Close it. */
log_notice(LD_NET, "Closing no-longer-configured %s on %s:%d",
2006-02-13 10:02:35 +01:00
conn_type_to_string(type), conn->address, conn->port);
if (replaced_conns) {
smartlist_add(replaced_conns, conn);
} else {
connection_close_immediate(conn);
connection_mark_for_close(conn);
}
} else {
/* It's configured; we don't need to launch it. */
2006-02-13 10:02:35 +01:00
// log_debug(LD_NET, "Already have %s on %s:%d",
// conn_type_to_string(type), conn->address, conn->port);
smartlist_remove(launch, line);
2005-09-30 23:31:26 +02:00
if (free_launch_elts)
config_free_lines(line);
}
}
/* Now open all the listeners that are configured but not opened. */
i = 0;
if (!never_open_conns) {
SMARTLIST_FOREACH(launch, config_line_t *, cfg,
{
conn = connection_create_listener(cfg->value, (uint16_t) port_option,
type);
if (!conn) {
i = -1;
} else {
if (new_conns)
smartlist_add(new_conns, conn);
}
});
}
if (free_launch_elts) {
SMARTLIST_FOREACH(launch, config_line_t *, cfg,
config_free_lines(cfg));
}
smartlist_free(launch);
return i;
}
/** (Re)launch listeners for each port you should have open. If
* <b>force</b> is true, close and relaunch all listeners. If <b>force</b>
* is false, then only relaunch listeners when we have the wrong number of
* connections for a given type.
*
* Add all old conns that should be closed to <b>replaced_conns</b>.
* Add all new connections to <b>new_conns</b>.
*/
int
retry_all_listeners(int force, smartlist_t *replaced_conns,
smartlist_t *new_conns)
{
or_options_t *options = get_options();
2002-06-27 00:45:49 +02:00
if (retry_listeners(CONN_TYPE_OR_LISTENER, options->ORListenAddress,
options->ORPort, "0.0.0.0", force,
replaced_conns, new_conns, options->ClientOnly)<0)
return -1;
if (retry_listeners(CONN_TYPE_DIR_LISTENER, options->DirListenAddress,
options->DirPort, "0.0.0.0", force,
replaced_conns, new_conns, 0)<0)
return -1;
if (retry_listeners(CONN_TYPE_AP_LISTENER, options->SocksListenAddress,
options->SocksPort, "127.0.0.1", force,
replaced_conns, new_conns, 0)<0)
return -1;
if (retry_listeners(CONN_TYPE_AP_TRANS_LISTENER, options->TransListenAddress,
options->TransPort, "127.0.0.1", force,
replaced_conns, new_conns, 0)<0)
return -1;
if (retry_listeners(CONN_TYPE_AP_NATD_LISTENER, options->NatdListenAddress,
options->NatdPort, "127.0.0.1", force,
replaced_conns, new_conns, 0)<0)
return -1;
if (retry_listeners(CONN_TYPE_CONTROL_LISTENER,
options->ControlListenAddress,
options->ControlPort, "127.0.0.1", force,
replaced_conns, new_conns, 0)<0)
return -1;
2002-06-27 00:45:49 +02:00
return 0;
}
/** Return 1 if we should apply rate limiting to <b>conn</b>,
* and 0 otherwise. Right now this just checks if it's an internal
* IP address or an internal connection. */
static int
connection_is_rate_limited(connection_t *conn)
{
if (conn->linked || is_internal_IP(conn->addr, 0))
return 0;
else
return 1;
}
extern int global_read_bucket, global_write_bucket;
extern int global_relayed_read_bucket, global_relayed_write_bucket;
/** Did either global write bucket run dry last second? If so,
* we are likely to run dry again this second, so be stingy with the
* tokens we just put in. */
static int write_buckets_empty_last_second = 0;
/** Return 1 if <b>conn</b> should use tokens from the "relayed"
* bandwidth rates, else 0. Currently, only OR conns with bandwidth
* class 1, and directory conns that are serving data out, count.
*/
static int
connection_counts_as_relayed_traffic(connection_t *conn)
{
if (conn->type == CONN_TYPE_OR && !TO_OR_CONN(conn)->client_used)
return 1;
if (conn->type == CONN_TYPE_DIR && DIR_CONN_IS_SERVER(conn))
return 1;
return 0;
}
/** Helper function to decide how many bytes out of <b>global_bucket</b>
* we're willing to use for this transaction. <b>base</b> is the size
* of a cell on the network; <b>priority</b> says whether we should
* write many of them or just a few; and <b>conn_bucket</b> (if
* non-negative) provides an upper limit for our answer. */
static int
connection_bucket_round_robin(int base, int priority,
int global_bucket, int conn_bucket)
{
int at_most;
int num_bytes_high = (priority ? 32 : 16) * base;
int num_bytes_low = (priority ? 4 : 2) * base;
/* Do a rudimentary round-robin so one circuit can't hog a connection.
* Pick at most 32 cells, at least 4 cells if possible, and if we're in
* the middle pick 1/8 of the available bandwidth. */
at_most = global_bucket / 8;
at_most -= (at_most % base); /* round down */
if (at_most > num_bytes_high) /* 16 KB, or 8 KB for low-priority */
at_most = num_bytes_high;
else if (at_most < num_bytes_low) /* 2 KB, or 1 KB for low-priority */
at_most = num_bytes_low;
if (at_most > global_bucket)
at_most = global_bucket;
if (conn_bucket >= 0 && at_most > conn_bucket)
at_most = conn_bucket;
if (at_most < 0)
return 0;
return at_most;
}
/** How many bytes at most can we read onto this connection? */
static int
connection_bucket_read_limit(connection_t *conn)
{
int base = connection_speaks_cells(conn) ?
CELL_NETWORK_SIZE : RELAY_PAYLOAD_SIZE;
int priority = conn->type != CONN_TYPE_DIR;
int conn_bucket = -1;
int global_bucket = global_read_bucket;
if (connection_speaks_cells(conn)) {
or_connection_t *or_conn = TO_OR_CONN(conn);
if (conn->state == OR_CONN_STATE_OPEN)
conn_bucket = or_conn->read_bucket;
}
if (!connection_is_rate_limited(conn)) {
/* be willing to read on local conns even if our buckets are empty */
return conn_bucket>=0 ? conn_bucket : 1<<14;
}
if (connection_counts_as_relayed_traffic(conn) &&
global_relayed_read_bucket <= global_read_bucket)
global_bucket = global_relayed_read_bucket;
return connection_bucket_round_robin(base, priority,
global_bucket, conn_bucket);
}
/** How many bytes at most can we write onto this connection? */
int
connection_bucket_write_limit(connection_t *conn)
{
int base = connection_speaks_cells(conn) ?
CELL_NETWORK_SIZE : RELAY_PAYLOAD_SIZE;
int priority = conn->type != CONN_TYPE_DIR;
int global_bucket = global_write_bucket;
if (!connection_is_rate_limited(conn)) {
/* be willing to write to local conns even if our buckets are empty */
return conn->outbuf_flushlen;
}
if (connection_counts_as_relayed_traffic(conn) &&
global_relayed_write_bucket <= global_write_bucket)
global_bucket = global_relayed_write_bucket;
return connection_bucket_round_robin(base, priority, global_bucket,
conn->outbuf_flushlen);
}
/** Return 1 if the global write buckets are low enough that we
* shouldn't send <b>attempt</b> bytes of low-priority directory stuff
* out to <b>conn</b>. Else return 0.
* Priority is 1 for v1 requests (directories and running-routers),
* and 2 for v2 requests (statuses and descriptors). But see FFFF in
* directory_handle_command_get() for why we don't use priority 2 yet.
*
* There are a lot of parameters we could use here:
* - global_relayed_write_bucket. Low is bad.
* - global_write_bucket. Low is bad.
* - bandwidthrate. Low is bad.
* - bandwidthburst. Not a big factor?
* - attempt. High is bad.
* - total bytes queued on outbufs. High is bad. But I'm wary of
* using this, since a few slow-flushing queues will pump up the
* number without meaning what we meant to mean. What we really
* mean is "total directory bytes added to outbufs recently", but
* that's harder to quantify and harder to keep track of.
*/
2006-02-12 23:27:09 +01:00
int
global_write_bucket_low(connection_t *conn, size_t attempt, int priority)
{
int smaller_bucket = global_write_bucket < global_relayed_write_bucket ?
global_write_bucket : global_relayed_write_bucket;
2007-01-05 07:03:10 +01:00
if (authdir_mode(get_options()) && priority>1)
return 0; /* there's always room to answer v2 if we're an auth dir */
if (!connection_is_rate_limited(conn))
return 0; /* local conns don't get limited */
if (smaller_bucket < (int)attempt)
2007-01-05 07:03:10 +01:00
return 1; /* not enough space no matter the priority */
if (write_buckets_empty_last_second)
return 1; /* we're already hitting our limits, no more please */
if (priority == 1) { /* old-style v1 query */
/* Could we handle *two* of these requests within the next two seconds? */
or_options_t *options = get_options();
int64_t can_write = (int64_t)smaller_bucket
+ 2*(options->RelayBandwidthRate ? options->RelayBandwidthRate :
options->BandwidthRate);
if (can_write < 2*(int64_t)attempt)
return 1;
} else { /* v2 query */
/* no further constraints yet */
}
return 0;
}
/** We just read num_read and wrote num_written onto conn.
* Decrement buckets appropriately. */
static void
connection_buckets_decrement(connection_t *conn, time_t now,
int num_read, int num_written)
{
if (!connection_is_rate_limited(conn))
return; /* local IPs are free */
if (num_read > 0)
rep_hist_note_bytes_read(num_read, now);
if (num_written > 0)
rep_hist_note_bytes_written(num_written, now);
if (connection_counts_as_relayed_traffic(conn)) {
global_relayed_read_bucket -= num_read;
global_relayed_write_bucket -= num_written;
}
global_read_bucket -= num_read;
global_write_bucket -= num_written;
if (connection_speaks_cells(conn) && conn->state == OR_CONN_STATE_OPEN)
TO_OR_CONN(conn)->read_bucket -= num_read;
}
/** If we have exhausted our global buckets, or the buckets for conn,
* stop reading. */
static void
connection_consider_empty_read_buckets(connection_t *conn)
{
const char *reason;
if (global_read_bucket <= 0) {
reason = "global read bucket exhausted. Pausing.";
} else if (connection_counts_as_relayed_traffic(conn) &&
global_relayed_read_bucket <= 0) {
reason = "global relayed read bucket exhausted. Pausing.";
} else if (connection_speaks_cells(conn) &&
conn->state == OR_CONN_STATE_OPEN &&
TO_OR_CONN(conn)->read_bucket <= 0) {
reason = "connection read bucket exhausted. Pausing.";
} else
return; /* all good, no need to stop it */
LOG_FN_CONN(conn, (LOG_DEBUG, LD_NET, "%s", reason));
conn->read_blocked_on_bw = 1;
connection_stop_reading(conn);
}
/** If we have exhausted our global buckets, or the buckets for conn,
* stop writing. */
static void
connection_consider_empty_write_buckets(connection_t *conn)
{
const char *reason;
if (global_write_bucket <= 0) {
reason = "global write bucket exhausted. Pausing.";
} else if (connection_counts_as_relayed_traffic(conn) &&
global_relayed_write_bucket <= 0) {
reason = "global relayed write bucket exhausted. Pausing.";
#if 0
} else if (connection_speaks_cells(conn) &&
conn->state == OR_CONN_STATE_OPEN &&
TO_OR_CONN(conn)->write_bucket <= 0) {
reason = "connection write bucket exhausted. Pausing.";
#endif
} else
return; /* all good, no need to stop it */
LOG_FN_CONN(conn, (LOG_DEBUG, LD_NET, "%s", reason));
conn->write_blocked_on_bw = 1;
connection_stop_writing(conn);
}
/** Initialize the global read bucket to options-\>BandwidthBurst. */
void
connection_bucket_init(void)
{
or_options_t *options = get_options();
/* start it at max traffic */
global_read_bucket = (int)options->BandwidthBurst;
global_write_bucket = (int)options->BandwidthBurst;
if (options->RelayBandwidthRate) {
global_relayed_read_bucket = (int)options->RelayBandwidthBurst;
global_relayed_write_bucket = (int)options->RelayBandwidthBurst;
} else {
global_relayed_read_bucket = (int)options->BandwidthBurst;
global_relayed_write_bucket = (int)options->BandwidthBurst;
}
}
static void
connection_bucket_refill_helper(int *bucket, int rate, int burst,
int seconds_elapsed, const char *name)
{
if (*bucket < burst) {
*bucket += rate*seconds_elapsed;
if (*bucket > burst)
*bucket = burst;
log(LOG_DEBUG, LD_NET,"%s now %d.", name, *bucket);
}
}
/** A second has rolled over; increment buckets appropriately. */
void
connection_bucket_refill(int seconds_elapsed)
{
int i, n;
connection_t *conn;
connection_t **carray;
or_options_t *options = get_options();
int relayrate, relayburst;
if (options->RelayBandwidthRate) {
relayrate = (int)options->RelayBandwidthRate;
relayburst = (int)options->RelayBandwidthBurst;
} else {
relayrate = (int)options->BandwidthRate;
relayburst = (int)options->BandwidthBurst;
}
tor_assert(seconds_elapsed >= 0);
write_buckets_empty_last_second =
global_relayed_write_bucket == 0 || global_write_bucket == 0;
/* refill the global buckets */
connection_bucket_refill_helper(&global_read_bucket,
(int)options->BandwidthRate,
(int)options->BandwidthBurst,
seconds_elapsed, "global_read_bucket");
connection_bucket_refill_helper(&global_write_bucket,
(int)options->BandwidthRate,
(int)options->BandwidthBurst,
seconds_elapsed, "global_write_bucket");
connection_bucket_refill_helper(&global_relayed_read_bucket,
relayrate, relayburst, seconds_elapsed,
"global_relayed_read_bucket");
connection_bucket_refill_helper(&global_relayed_write_bucket,
relayrate, relayburst, seconds_elapsed,
"global_relayed_write_bucket");
/* refill the per-connection buckets */
get_connection_array(&carray,&n);
for (i=0;i<n;i++) {
conn = carray[i];
if (connection_speaks_cells(conn)) {
or_connection_t *or_conn = TO_OR_CONN(conn);
if (connection_read_bucket_should_increase(or_conn)) {
or_conn->read_bucket += or_conn->bandwidthrate*seconds_elapsed;
if (or_conn->read_bucket > or_conn->bandwidthburst)
or_conn->read_bucket = or_conn->bandwidthburst;
//log_fn(LOG_DEBUG,"Receiver bucket %d now %d.", i,
// conn->read_bucket);
}
}
if (conn->read_blocked_on_bw == 1 /* marked to turn reading back on now */
&& global_read_bucket > 0 /* and we're allowed to read */
&& (!connection_counts_as_relayed_traffic(conn) ||
global_relayed_read_bucket > 0) /* even if we're relayed traffic */
&& (!connection_speaks_cells(conn) ||
conn->state != OR_CONN_STATE_OPEN ||
TO_OR_CONN(conn)->read_bucket > 0)) {
/* and either a non-cell conn or a cell conn with non-empty bucket */
LOG_FN_CONN(conn, (LOG_DEBUG,LD_NET,
"waking up conn (fd %d) for read", conn->s));
conn->read_blocked_on_bw = 0;
connection_start_reading(conn);
}
if (conn->write_blocked_on_bw == 1
&& global_write_bucket > 0 /* and we're allowed to write */
&& (!connection_counts_as_relayed_traffic(conn) ||
global_relayed_write_bucket > 0)) {
/* even if we're relayed traffic */
LOG_FN_CONN(conn, (LOG_DEBUG,LD_NET,
"waking up conn (fd %d) for write", conn->s));
conn->write_blocked_on_bw = 0;
connection_start_writing(conn);
}
}
}
/** Is the receiver bucket for connection <b>conn</b> low enough that we
* should add another pile of tokens to it?
*/
static int
connection_read_bucket_should_increase(or_connection_t *conn)
{
tor_assert(conn);
if (conn->_base.state != OR_CONN_STATE_OPEN)
return 0; /* only open connections play the rate limiting game */
if (conn->read_bucket >= conn->bandwidthburst)
return 0;
return 1;
}
/** Read bytes from conn-\>s and process them.
*
* This function gets called from conn_read() in main.c, either
* when poll() has declared that conn wants to read, or (for OR conns)
* when there are pending TLS bytes.
*
* It calls connection_read_to_buf() to bring in any new bytes,
* and then calls connection_process_inbuf() to process them.
*
* Mark the connection and return -1 if you want to close it, else
* return 0.
*/
int
connection_handle_read(connection_t *conn)
{
int max_to_read=-1, try_to_read;
size_t before, n_read = 0;
if (conn->marked_for_close)
return 0; /* do nothing */
conn->timestamp_lastread = time(NULL);
switch (conn->type) {
case CONN_TYPE_OR_LISTENER:
return connection_handle_listener_read(conn, CONN_TYPE_OR);
case CONN_TYPE_AP_LISTENER:
case CONN_TYPE_AP_TRANS_LISTENER:
case CONN_TYPE_AP_NATD_LISTENER:
return connection_handle_listener_read(conn, CONN_TYPE_AP);
case CONN_TYPE_DIR_LISTENER:
return connection_handle_listener_read(conn, CONN_TYPE_DIR);
case CONN_TYPE_CONTROL_LISTENER:
return connection_handle_listener_read(conn, CONN_TYPE_CONTROL);
}
loop_again:
try_to_read = max_to_read;
tor_assert(!conn->marked_for_close);
before = buf_datalen(conn->inbuf);
if (connection_read_to_buf(conn, &max_to_read) < 0) {
/* There's a read error; kill the connection.*/
connection_close_immediate(conn); /* Don't flush; connection is dead. */
if (CONN_IS_EDGE(conn)) {
edge_connection_t *edge_conn = TO_EDGE_CONN(conn);
connection_edge_end_errno(edge_conn);
if (edge_conn->socks_request) /* broken, don't send a socks reply back */
edge_conn->socks_request->has_finished = 1;
}
connection_mark_for_close(conn);
return -1;
}
n_read += buf_datalen(conn->inbuf) - before;
if (CONN_IS_EDGE(conn) && try_to_read != max_to_read) {
/* instruct it not to try to package partial cells. */
if (connection_process_inbuf(conn, 0) < 0) {
return -1;
}
if (!conn->marked_for_close &&
connection_is_reading(conn) &&
!conn->inbuf_reached_eof &&
max_to_read > 0)
goto loop_again; /* try reading again, in case more is here now */
}
/* one last try, packaging partial cells and all. */
if (!conn->marked_for_close &&
connection_process_inbuf(conn, 1) < 0) {
return -1;
}
if (conn->linked_conn) {
/* The other side's handle_write will never actually get called, so
* we need to invoke the appropriate callbacks ourself. */
connection_t *linked = conn->linked_conn;
/* XXXX020 Do we need to ensure that this stuff is called even if
* conn dies in a way that causes us to return -1 earlier? */
if (n_read) {
/* Probably a no-op, but hey. */
connection_buckets_decrement(linked, time(NULL), 0, n_read);
if (connection_flushed_some(linked) < 0)
connection_mark_for_close(linked);
if (!connection_wants_to_flush(linked))
connection_finished_flushing(linked);
}
if (!buf_datalen(linked->outbuf) && conn->active_on_link)
connection_stop_reading_from_linked_conn(conn);
}
/* If we hit the EOF, call connection_reached_eof. */
if (!conn->marked_for_close &&
conn->inbuf_reached_eof &&
connection_reached_eof(conn) < 0) {
return -1;
}
return 0;
}
/** Pull in new bytes from conn-\>s or conn-\>linked_conn onto conn-\>inbuf,
* either directly or via TLS. Reduce the token buckets by the number of bytes
* read.
*
* If *max_to_read is -1, then decide it ourselves, else go with the
* value passed to us. When returning, if it's changed, subtract the
* number of bytes we read from *max_to_read.
*
* Return -1 if we want to break conn, else return 0.
*/
static int
connection_read_to_buf(connection_t *conn, int *max_to_read)
{
int result, at_most = *max_to_read;
size_t bytes_in_buf, more_to_read;
size_t n_read = 0, n_written = 0;
if (at_most == -1) { /* we need to initialize it */
/* how many bytes are we allowed to read? */
at_most = connection_bucket_read_limit(conn);
}
bytes_in_buf = buf_capacity(conn->inbuf) - buf_datalen(conn->inbuf);
again:
if ((size_t)at_most > bytes_in_buf && bytes_in_buf >= 1024) {
more_to_read = at_most - bytes_in_buf;
at_most = bytes_in_buf;
} else {
more_to_read = 0;
}
if (connection_speaks_cells(conn) &&
conn->state > OR_CONN_STATE_PROXY_READING) {
int pending;
or_connection_t *or_conn = TO_OR_CONN(conn);
if (conn->state == OR_CONN_STATE_HANDSHAKING) {
/* continue handshaking even if global token bucket is empty */
return connection_tls_continue_handshake(or_conn);
}
2006-02-13 10:02:35 +01:00
log_debug(LD_NET,
"%d: starting, inbuf_datalen %d (%d pending in tls object)."
" at_most %d.",
conn->s,(int)buf_datalen(conn->inbuf),
tor_tls_get_pending_bytes(or_conn->tls), at_most);
/* else open, or closing */
result = read_to_buf_tls(or_conn->tls, at_most, conn->inbuf);
if (TOR_TLS_IS_ERROR(result) || result == TOR_TLS_CLOSE)
or_conn->tls_error = result;
else
or_conn->tls_error = 0;
switch (result) {
case TOR_TLS_CLOSE:
2006-02-13 10:02:35 +01:00
log_info(LD_NET,"TLS connection closed on read. Closing. "
"(Nickname %s, address %s",
or_conn->nickname ? or_conn->nickname : "not set",
conn->address);
return result;
CASE_TOR_TLS_ERROR_ANY:
2006-02-13 10:02:35 +01:00
log_info(LD_NET,"tls error. breaking (nickname %s, address %s).",
or_conn->nickname ? or_conn->nickname : "not set",
conn->address);
return result;
case TOR_TLS_WANTWRITE:
connection_start_writing(conn);
return 0;
case TOR_TLS_WANTREAD: /* we're already reading */
case TOR_TLS_DONE: /* no data read, so nothing to process */
result = 0;
break; /* so we call bucket_decrement below */
default:
break;
}
pending = tor_tls_get_pending_bytes(or_conn->tls);
if (pending) {
/* If we have any pending bytes, we read them now. This *can*
2005-06-11 08:07:22 +02:00
* take us over our read allotment, but really we shouldn't be
* believing that SSL bytes are the same as TCP bytes anyway. */
int r2 = read_to_buf_tls(or_conn->tls, pending, conn->inbuf);
if (r2<0) {
log_warn(LD_BUG, "apparently, reading pending bytes can fail.");
return -1;
} else {
result += r2;
}
}
tor_tls_get_n_raw_bytes(or_conn->tls, &n_read, &n_written);
log_debug(LD_GENERAL, "After TLS read of %d: %ld read, %ld written",
result, (long)n_read, (long)n_written);
} else if (conn->linked) {
if (conn->linked_conn) {
result = move_buf_to_buf(conn->inbuf, conn->linked_conn->outbuf,
&conn->linked_conn->outbuf_flushlen);
} else {
result = 0;
}
//log_notice(LD_GENERAL, "Moved %d bytes on an internal link!", result);
/* If the other side has disappeared, or if it's been marked for close and
* we flushed its outbuf, then we should set our inbuf_reached_eof. */
if (!conn->linked_conn ||
(conn->linked_conn->marked_for_close &&
buf_datalen(conn->linked_conn->outbuf) == 0))
conn->inbuf_reached_eof = 1;
n_read = (size_t) result;
} else {
/* !connection_speaks_cells, !conn->linked_conn. */
int reached_eof = 0;
CONN_LOG_PROTECT(conn,
result = read_to_buf(conn->s, at_most, conn->inbuf, &reached_eof));
if (reached_eof)
conn->inbuf_reached_eof = 1;
2004-12-22 03:55:26 +01:00
// log_fn(LOG_DEBUG,"read_to_buf returned %d.",read_result);
if (result < 0)
return -1;
n_read = (size_t) result;
}
if (n_read > 0) { /* change *max_to_read */
*max_to_read = at_most - n_read;
}
if (conn->type == CONN_TYPE_AP) {
edge_connection_t *edge_conn = TO_EDGE_CONN(conn);
edge_conn->n_read += n_read;
}
connection_buckets_decrement(conn, time(NULL), n_read, n_written);
if (more_to_read && result == at_most) {
bytes_in_buf = buf_capacity(conn->inbuf) - buf_datalen(conn->inbuf);
tor_assert(bytes_in_buf < 1024);
at_most = more_to_read;
goto again;
}
/* Call even if result is 0, since the global read bucket may
* have reached 0 on a different conn, and this guy needs to
* know to stop reading. */
connection_consider_empty_read_buckets(conn);
2007-01-05 07:03:10 +01:00
if (n_written > 0 && connection_is_writing(conn))
connection_consider_empty_write_buckets(conn);
return 0;
2002-06-27 00:45:49 +02:00
}
/** A pass-through to fetch_from_buf. */
int
connection_fetch_from_buf(char *string, size_t len, connection_t *conn)
{
return fetch_from_buf(string, len, conn->inbuf);
2002-06-27 00:45:49 +02:00
}
2004-05-10 06:34:48 +02:00
/** Return conn-\>outbuf_flushlen: how many bytes conn wants to flush
* from its outbuf. */
int
connection_wants_to_flush(connection_t *conn)
{
return conn->outbuf_flushlen > 0;
}
/** Are there too many bytes on edge connection <b>conn</b>'s outbuf to
* send back a relay-level sendme yet? Return 1 if so, 0 if not. Used by
* connection_edge_consider_sending_sendme().
*/
int
connection_outbuf_too_full(connection_t *conn)
{
return (conn->outbuf_flushlen > 10*CELL_PAYLOAD_SIZE);
}
2004-05-10 06:34:48 +02:00
/** Try to flush more bytes onto conn-\>s.
*
* This function gets called either from conn_write() in main.c
* when poll() has declared that conn wants to write, or below
* from connection_write_to_buf() when an entire TLS record is ready.
*
2004-05-10 06:34:48 +02:00
* Update conn-\>timestamp_lastwritten to now, and call flush_buf
* or flush_buf_tls appropriately. If it succeeds and there are no more
* more bytes on conn->outbuf, then call connection_finished_flushing
* on it too.
*
* If <b>force</b>, then write as many bytes as possible, ignoring bandwidth
* limits. (Used for flushing messages to controller connections on fatal
* errors.)
*
* Mark the connection and return -1 if you want to close it, else
* return 0.
*/
int
connection_handle_write(connection_t *conn, int force)
{
int e;
socklen_t len=sizeof(e);
int result;
int max_to_write;
time_t now = time(NULL);
size_t n_read = 0, n_written = 0;
tor_assert(!connection_is_listener(conn));
if (conn->marked_for_close || conn->s < 0)
return 0; /* do nothing */
conn->timestamp_lastwritten = now;
/* Sometimes, "writable" means "connected". */
if (connection_state_is_connecting(conn)) {
if (getsockopt(conn->s, SOL_SOCKET, SO_ERROR, (void*)&e, &len) < 0) {
2006-02-13 10:02:35 +01:00
log_warn(LD_BUG,
"getsockopt() syscall failed?! Please report to tor-ops.");
if (CONN_IS_EDGE(conn))
connection_edge_end_errno(TO_EDGE_CONN(conn));
connection_mark_for_close(conn);
return -1;
}
if (e) {
/* some sort of error, but maybe just inprogress still */
if (!ERRNO_IS_CONN_EINPROGRESS(e)) {
2006-02-13 10:02:35 +01:00
log_info(LD_NET,"in-progress connect failed. Removing.");
if (CONN_IS_EDGE(conn))
connection_edge_end_errno(TO_EDGE_CONN(conn));
connection_close_immediate(conn);
connection_mark_for_close(conn);
/* it's safe to pass OPs to router_set_status(), since it just
* ignores unrecognized routers
*/
if (conn->type == CONN_TYPE_OR && !get_options()->HttpsProxy)
router_set_status(TO_OR_CONN(conn)->identity_digest, 0);
return -1;
} else {
return 0; /* no change, see if next time is better */
}
}
/* The connection is successful. */
if (connection_finished_connecting(conn)<0)
return -1;
}
max_to_write = force ? (int)conn->outbuf_flushlen
: connection_bucket_write_limit(conn);
if (connection_speaks_cells(conn) &&
conn->state > OR_CONN_STATE_PROXY_READING) {
or_connection_t *or_conn = TO_OR_CONN(conn);
if (conn->state == OR_CONN_STATE_HANDSHAKING) {
connection_stop_writing(conn);
if (connection_tls_continue_handshake(or_conn) < 0) {
/* Don't flush; connection is dead. */
connection_close_immediate(conn);
connection_mark_for_close(conn);
return -1;
}
return 0;
}
/* else open, or closing */
result = flush_buf_tls(or_conn->tls, conn->outbuf,
max_to_write, &conn->outbuf_flushlen);
switch (result) {
CASE_TOR_TLS_ERROR_ANY:
case TOR_TLS_CLOSE:
log_info(LD_NET,result!=TOR_TLS_CLOSE?
2006-02-13 10:02:35 +01:00
"tls error. breaking.":"TLS connection closed on flush");
/* Don't flush; connection is dead. */
connection_close_immediate(conn);
connection_mark_for_close(conn);
return -1;
case TOR_TLS_WANTWRITE:
2006-02-13 10:02:35 +01:00
log_debug(LD_NET,"wanted write.");
/* we're already writing */
return 0;
case TOR_TLS_WANTREAD:
/* Make sure to avoid a loop if the receive buckets are empty. */
2006-02-13 10:02:35 +01:00
log_debug(LD_NET,"wanted read.");
if (!connection_is_reading(conn)) {
connection_stop_writing(conn);
conn->write_blocked_on_bw = 1;
/* we'll start reading again when the next second arrives,
* and then also start writing again.
*/
}
/* else no problem, we're already reading */
return 0;
/* case TOR_TLS_DONE:
* for TOR_TLS_DONE, fall through to check if the flushlen
* is empty, so we can stop writing.
2003-12-17 22:09:31 +01:00
*/
}
tor_tls_get_n_raw_bytes(or_conn->tls, &n_read, &n_written);
log_debug(LD_GENERAL, "After TLS write of %d: %ld read, %ld written",
result, (long)n_read, (long)n_written);
} else {
CONN_LOG_PROTECT(conn,
result = flush_buf(conn->s, conn->outbuf,
max_to_write, &conn->outbuf_flushlen));
if (result < 0) {
if (CONN_IS_EDGE(conn))
connection_edge_end_errno(TO_EDGE_CONN(conn));
connection_close_immediate(conn); /* Don't flush; connection is dead. */
connection_mark_for_close(conn);
return -1;
}
n_written = (size_t) result;
}
if (conn->type == CONN_TYPE_AP) {
edge_connection_t *edge_conn = TO_EDGE_CONN(conn);
edge_conn->n_written += n_written;
}
connection_buckets_decrement(conn, time(NULL), n_read, n_written);
if (result > 0) {
/* If we wrote any bytes from our buffer, then call the appropriate
* functions. */
if (connection_flushed_some(conn) < 0)
connection_mark_for_close(conn);
}
if (!connection_wants_to_flush(conn)) { /* it's done flushing */
if (connection_finished_flushing(conn) < 0) {
/* already marked */
return -1;
}
return 0;
}
/* Call even if result is 0, since the global write bucket may
* have reached 0 on a different conn, and this guy needs to
* know to stop writing. */
connection_consider_empty_write_buckets(conn);
2007-01-05 07:03:10 +01:00
if (n_read > 0 && connection_is_reading(conn))
connection_consider_empty_read_buckets(conn);
return 0;
}
/** Openssl TLS record size is 16383; this is close. The goal here is to
* push data out as soon as we know there's enough for a TLS record, so
* during periods of high load we won't read entire megabytes from
* input before pushing any data out. It also has the feature of not
* growing huge outbufs unless something is slow. */
#define MIN_TLS_FLUSHLEN 15872
/** Append <b>len</b> bytes of <b>string</b> onto <b>conn</b>'s
* outbuf, and ask it to start writing.
*
* If <b>zlib</b> is nonzero, this is a directory connection that should get
* its contents compressed or decompressed as they're written. If zlib is
* negative, this is the last data to be compressed, and the connection's zlib
* state should be flushed.
*
* If it's an OR conn and an entire TLS record is ready, then try to
* flush the record now. Similarly, if it's a local control connection
* and a 64k chunk is ready, try to flush it all, so we don't end up with
* many megabytes of controller info queued at once.
*/
void
_connection_write_to_buf_impl(const char *string, size_t len,
connection_t *conn, int zlib)
{
int r;
size_t old_datalen;
if (!len)
return;
/* if it's marked for close, only allow write if we mean to flush it */
if (conn->marked_for_close && !conn->hold_open_until_flushed)
return;
old_datalen = buf_datalen(conn->outbuf);
if (zlib) {
dir_connection_t *dir_conn = TO_DIR_CONN(conn);
int done = zlib < 0;
CONN_LOG_PROTECT(conn, r = write_to_buf_zlib(conn->outbuf,
dir_conn->zlib_state,
string, len, done));
} else {
CONN_LOG_PROTECT(conn, r = write_to_buf(string, len, conn->outbuf));
}
if (r < 0) {
if (CONN_IS_EDGE(conn)) {
/* if it failed, it means we have our package/delivery windows set
wrong compared to our max outbuf size. close the whole circuit. */
2006-02-13 10:02:35 +01:00
log_warn(LD_NET,
"write_to_buf failed. Closing circuit (fd %d).", conn->s);
circuit_mark_for_close(circuit_get_by_edge_conn(TO_EDGE_CONN(conn)),
END_CIRC_REASON_INTERNAL);
} else {
2006-02-13 10:02:35 +01:00
log_warn(LD_NET,
"write_to_buf failed. Closing connection (fd %d).", conn->s);
connection_mark_for_close(conn);
}
return;
}
connection_start_writing(conn);
if (zlib) {
conn->outbuf_flushlen += buf_datalen(conn->outbuf) - old_datalen;
} else {
int extra = 0;
conn->outbuf_flushlen += len;
if (conn->type == CONN_TYPE_OR &&
conn->outbuf_flushlen-len < MIN_TLS_FLUSHLEN &&
conn->outbuf_flushlen >= MIN_TLS_FLUSHLEN) {
extra = conn->outbuf_flushlen - MIN_TLS_FLUSHLEN;
conn->outbuf_flushlen = MIN_TLS_FLUSHLEN;
} else if (conn->type == CONN_TYPE_CONTROL &&
!connection_is_rate_limited(conn) &&
conn->outbuf_flushlen-len < 1<<16 &&
conn->outbuf_flushlen >= 1<<16) {
/* just try to flush all of it */
} else
return; /* no need to try flushing */
if (connection_handle_write(conn, 0) < 0) {
if (!conn->marked_for_close) {
/* this connection is broken. remove it. */
log_warn(LD_BUG, "unhandled error on write for "
"conn (type %d, fd %d); removing",
conn->type, conn->s);
tor_fragile_assert();
/* do a close-immediate here, so we don't try to flush */
connection_close_immediate(conn);
}
return;
}
if (extra) {
conn->outbuf_flushlen += extra;
connection_start_writing(conn);
}
}
}
/** Return the conn to addr/port that has the most recent
* timestamp_created, or NULL if no such conn exists. */
or_connection_t *
connection_or_exact_get_by_addr_port(uint32_t addr, uint16_t port)
{
int i, n;
connection_t *conn;
or_connection_t *best=NULL;
connection_t **carray;
2003-12-17 22:09:31 +01:00
get_connection_array(&carray,&n);
for (i=0;i<n;i++) {
conn = carray[i];
if (conn->type == CONN_TYPE_OR &&
conn->addr == addr &&
conn->port == port &&
!conn->marked_for_close &&
(!best || best->_base.timestamp_created < conn->timestamp_created))
best = TO_OR_CONN(conn);
}
return best;
}
/** Return a connection with given type, address, port, and purpose;
* or NULL if no such connection exists. */
connection_t *
connection_get_by_type_addr_port_purpose(int type,
uint32_t addr, uint16_t port,
int purpose)
{
int i, n;
connection_t *conn;
connection_t **carray;
get_connection_array(&carray,&n);
for (i=0;i<n;i++) {
conn = carray[i];
if (conn->type == type &&
conn->addr == addr &&
conn->port == port &&
conn->purpose == purpose &&
!conn->marked_for_close)
return conn;
}
return NULL;
}
/** Return the stream with id <b>id</b> if it is not already marked for
* close.
*/
edge_connection_t *
connection_get_by_global_id(uint32_t id)
{
int i, n;
connection_t *conn;
connection_t **carray;
get_connection_array(&carray,&n);
for (i=0;i<n;i++) {
conn = carray[i];
if (CONN_IS_EDGE(conn) && TO_EDGE_CONN(conn)->global_identifier == id) {
if (!conn->marked_for_close)
return TO_EDGE_CONN(conn);
else
return NULL;
}
}
return NULL;
}
/** Return a connection of type <b>type</b> that is not marked for close.
*/
connection_t *
connection_get_by_type(int type)
{
int i, n;
connection_t *conn;
connection_t **carray;
2003-12-17 22:09:31 +01:00
get_connection_array(&carray,&n);
for (i=0;i<n;i++) {
conn = carray[i];
if (conn->type == type && !conn->marked_for_close)
return conn;
}
return NULL;
}
/** Return a connection of type <b>type</b> that is in state <b>state</b>,
* and that is not marked for close.
*/
connection_t *
connection_get_by_type_state(int type, int state)
{
int i, n;
connection_t *conn;
connection_t **carray;
2003-12-17 22:09:31 +01:00
get_connection_array(&carray,&n);
for (i=0;i<n;i++) {
conn = carray[i];
if (conn->type == type && conn->state == state && !conn->marked_for_close)
return conn;
}
return NULL;
}
/** Return the connection of type <b>type</b> that is in state
* <b>state</b>, that was written to least recently, and that is not
* marked for close.
*/
connection_t *
connection_get_by_type_state_lastwritten(int type, int state)
{
int i, n;
connection_t *conn, *best=NULL;
connection_t **carray;
2003-12-17 22:09:31 +01:00
get_connection_array(&carray,&n);
for (i=0;i<n;i++) {
conn = carray[i];
if (conn->type == type && conn->state == state && !conn->marked_for_close)
if (!best || conn->timestamp_lastwritten < best->timestamp_lastwritten)
best = conn;
}
return best;
}
/** Return a connection of type <b>type</b> that has rendquery equal
* to <b>rendquery</b>, and that is not marked for close. If state
* is non-zero, conn must be of that state too.
*/
connection_t *
connection_get_by_type_state_rendquery(int type, int state,
const char *rendquery)
{
int i, n;
connection_t *conn;
connection_t **carray;
tor_assert(type == CONN_TYPE_DIR ||
type == CONN_TYPE_AP || type == CONN_TYPE_EXIT);
get_connection_array(&carray,&n);
for (i=0;i<n;i++) {
conn = carray[i];
if (conn->type == type &&
!conn->marked_for_close &&
(!state || state == conn->state)) {
if (type == CONN_TYPE_DIR &&
!rend_cmp_service_ids(rendquery, TO_DIR_CONN(conn)->rend_query))
return conn;
else if (CONN_IS_EDGE(conn) &&
!rend_cmp_service_ids(rendquery, TO_EDGE_CONN(conn)->rend_query))
return conn;
}
}
return NULL;
}
/** Return an open, non-marked connection of a given type and purpose, or NULL
* if no such connection exists. */
connection_t *
connection_get_by_type_purpose(int type, int purpose)
{
int i, n;
connection_t *conn;
connection_t **carray;
get_connection_array(&carray,&n);
for (i=0;i<n;i++) {
conn = carray[i];
if (conn->type == type &&
!conn->marked_for_close &&
(purpose == conn->purpose))
return conn;
}
return NULL;
}
/** Return 1 if <b>conn</b> is a listener conn, else return 0. */
int
connection_is_listener(connection_t *conn)
{
if (conn->type == CONN_TYPE_OR_LISTENER ||
conn->type == CONN_TYPE_AP_LISTENER ||
conn->type == CONN_TYPE_AP_TRANS_LISTENER ||
conn->type == CONN_TYPE_AP_NATD_LISTENER ||
conn->type == CONN_TYPE_DIR_LISTENER ||
conn->type == CONN_TYPE_CONTROL_LISTENER)
return 1;
return 0;
}
2004-05-10 06:34:48 +02:00
/** Return 1 if <b>conn</b> is in state "open" and is not marked
* for close, else return 0.
*/
int
connection_state_is_open(connection_t *conn)
{
tor_assert(conn);
if (conn->marked_for_close)
return 0;
if ((conn->type == CONN_TYPE_OR && conn->state == OR_CONN_STATE_OPEN) ||
(conn->type == CONN_TYPE_AP && conn->state == AP_CONN_STATE_OPEN) ||
(conn->type == CONN_TYPE_EXIT && conn->state == EXIT_CONN_STATE_OPEN) ||
(conn->type == CONN_TYPE_CONTROL &&
conn->state == CONTROL_CONN_STATE_OPEN))
return 1;
return 0;
}
/** Return 1 if conn is in 'connecting' state, else return 0. */
int
connection_state_is_connecting(connection_t *conn)
{
tor_assert(conn);
if (conn->marked_for_close)
return 0;
switch (conn->type)
{
case CONN_TYPE_OR:
return conn->state == OR_CONN_STATE_CONNECTING;
case CONN_TYPE_EXIT:
return conn->state == EXIT_CONN_STATE_CONNECTING;
case CONN_TYPE_DIR:
return conn->state == DIR_CONN_STATE_CONNECTING;
}
return 0;
}
/** DOCDOC */
int
connection_should_read_from_linked_conn(connection_t *conn)
{
if (conn->linked && conn->reading_from_linked_conn) {
if (! conn->linked_conn ||
(conn->linked_conn->writing_to_linked_conn &&
buf_datalen(conn->linked_conn->outbuf)))
return 1;
}
return 0;
}
2006-12-12 21:19:54 +01:00
/** Allocates a base64'ed authenticator for use in http or https
* auth, based on the input string <b>authenticator</b>. Returns it
* if success, else returns NULL. */
char *
alloc_http_authenticator(const char *authenticator)
{
/* an authenticator in Basic authentication
* is just the string "username:password" */
const int authenticator_length = strlen(authenticator);
/* The base64_encode function needs a minimum buffer length
* of 66 bytes. */
const int base64_authenticator_length = (authenticator_length/48+1)*66;
char *base64_authenticator = tor_malloc(base64_authenticator_length);
if (base64_encode(base64_authenticator, base64_authenticator_length,
authenticator, authenticator_length) < 0) {
tor_free(base64_authenticator); /* free and set to null */
} else {
/* remove extra \n at end of encoding */
base64_authenticator[strlen(base64_authenticator) - 1] = 0;
}
return base64_authenticator;
}
/** Given a socket handle, check whether the local address (sockname) of the
* socket is one that we've connected from before. If so, double-check
* whether our address has changed and we need to generate keys. If we do,
* call init_keys().
*/
static void
client_check_address_changed(int sock)
{
uint32_t iface_ip, ip_out;
struct sockaddr_in out_addr;
socklen_t out_addr_len = sizeof(out_addr);
uint32_t *ip;
if (!last_interface_ip)
get_interface_address(LOG_INFO, &last_interface_ip);
if (!outgoing_addrs)
outgoing_addrs = smartlist_create();
if (getsockname(sock, (struct sockaddr*)&out_addr, &out_addr_len)<0) {
int e = tor_socket_errno(sock);
log_warn(LD_NET, "getsockname() to check for address change failed: %s",
tor_socket_strerror(e));
return;
}
/* Okay. If we've used this address previously, we're okay. */
ip_out = ntohl(out_addr.sin_addr.s_addr);
SMARTLIST_FOREACH(outgoing_addrs, uint32_t*, ip,
if (*ip == ip_out) return;
);
/* Uh-oh. We haven't connected from this address before. Has the interface
* address changed? */
if (get_interface_address(LOG_INFO, &iface_ip)<0)
return;
ip = tor_malloc(sizeof(uint32_t));
*ip = ip_out;
if (iface_ip == last_interface_ip) {
/* Nope, it hasn't changed. Add this address to the list. */
smartlist_add(outgoing_addrs, ip);
} else {
/* The interface changed. We're a client, so we need to regenerate our
* keys. First, reset the state. */
2005-11-11 18:03:35 +01:00
log(LOG_NOTICE, LD_NET, "Our IP has changed. Rotating keys...");
last_interface_ip = iface_ip;
SMARTLIST_FOREACH(outgoing_addrs, void*, ip, tor_free(ip));
smartlist_clear(outgoing_addrs);
smartlist_add(outgoing_addrs, ip);
/* Okay, now change our keys. */
ip_address_changed(1);
}
}
2004-05-10 06:34:48 +02:00
/** Process new bytes that have arrived on conn-\>inbuf.
*
* This function just passes conn to the connection-specific
* connection_*_process_inbuf() function. It also passes in
* package_partial if wanted.
*/
static int
connection_process_inbuf(connection_t *conn, int package_partial)
{
tor_assert(conn);
2002-06-27 00:45:49 +02:00
switch (conn->type) {
2002-06-27 00:45:49 +02:00
case CONN_TYPE_OR:
return connection_or_process_inbuf(TO_OR_CONN(conn));
case CONN_TYPE_EXIT:
case CONN_TYPE_AP:
return connection_edge_process_inbuf(TO_EDGE_CONN(conn),
package_partial);
case CONN_TYPE_DIR:
return connection_dir_process_inbuf(TO_DIR_CONN(conn));
case CONN_TYPE_CPUWORKER:
2003-12-17 22:09:31 +01:00
return connection_cpu_process_inbuf(conn);
case CONN_TYPE_CONTROL:
return connection_control_process_inbuf(TO_CONTROL_CONN(conn));
2002-06-27 00:45:49 +02:00
default:
log_err(LD_BUG,"got unexpected conn type %d.", conn->type);
tor_fragile_assert();
2002-06-27 00:45:49 +02:00
return -1;
}
}
/** Called whenever we've written data on a connection. */
static int
connection_flushed_some(connection_t *conn)
{
if (conn->type == CONN_TYPE_DIR &&
conn->state == DIR_CONN_STATE_SERVER_WRITING)
return connection_dirserv_flushed_some(TO_DIR_CONN(conn));
else if (conn->type == CONN_TYPE_OR)
return connection_or_flushed_some(TO_OR_CONN(conn));
else
return 0;
}
2004-05-10 06:34:48 +02:00
/** We just finished flushing bytes from conn-\>outbuf, and there
* are no more bytes remaining.
*
* This function just passes conn to the connection-specific
* connection_*_finished_flushing() function.
*/
static int
connection_finished_flushing(connection_t *conn)
{
tor_assert(conn);
2002-06-27 00:45:49 +02:00
// log_fn(LOG_DEBUG,"entered. Socket %u.", conn->s);
2002-06-27 00:45:49 +02:00
switch (conn->type) {
2002-06-27 00:45:49 +02:00
case CONN_TYPE_OR:
return connection_or_finished_flushing(TO_OR_CONN(conn));
case CONN_TYPE_AP:
case CONN_TYPE_EXIT:
return connection_edge_finished_flushing(TO_EDGE_CONN(conn));
case CONN_TYPE_DIR:
return connection_dir_finished_flushing(TO_DIR_CONN(conn));
case CONN_TYPE_CPUWORKER:
return connection_cpu_finished_flushing(conn);
case CONN_TYPE_CONTROL:
return connection_control_finished_flushing(TO_CONTROL_CONN(conn));
2002-06-27 00:45:49 +02:00
default:
log_err(LD_BUG,"got unexpected conn type %d.", conn->type);
tor_fragile_assert();
2002-06-27 00:45:49 +02:00
return -1;
}
}
/** Called when our attempt to connect() to another server has just
* succeeded.
*
* This function just passes conn to the connection-specific
* connection_*_finished_connecting() function.
*/
static int
connection_finished_connecting(connection_t *conn)
{
tor_assert(conn);
switch (conn->type)
{
case CONN_TYPE_OR:
return connection_or_finished_connecting(TO_OR_CONN(conn));
case CONN_TYPE_EXIT:
return connection_edge_finished_connecting(TO_EDGE_CONN(conn));
case CONN_TYPE_DIR:
return connection_dir_finished_connecting(TO_DIR_CONN(conn));
default:
log_err(LD_BUG,"got unexpected conn type %d.", conn->type);
tor_fragile_assert();
return -1;
}
}
/** Callback: invoked when a connection reaches an EOF event. */
static int
connection_reached_eof(connection_t *conn)
{
switch (conn->type) {
case CONN_TYPE_OR:
return connection_or_reached_eof(TO_OR_CONN(conn));
case CONN_TYPE_AP:
case CONN_TYPE_EXIT:
return connection_edge_reached_eof(TO_EDGE_CONN(conn));
case CONN_TYPE_DIR:
return connection_dir_reached_eof(TO_DIR_CONN(conn));
case CONN_TYPE_CPUWORKER:
return connection_cpu_reached_eof(conn);
case CONN_TYPE_CONTROL:
return connection_control_reached_eof(TO_CONTROL_CONN(conn));
default:
log_err(LD_BUG,"got unexpected conn type %d.", conn->type);
tor_fragile_assert();
return -1;
}
}
/** Log how many bytes are used by buffers of different kinds and sizes. */
void
connection_dump_buffer_mem_stats(int severity)
{
uint64_t used_by_type[_CONN_TYPE_MAX+1];
uint64_t alloc_by_type[_CONN_TYPE_MAX+1];
int n_conns_by_type[_CONN_TYPE_MAX+1];
uint64_t total_alloc = 0;
uint64_t total_used = 0;
int i, n;
connection_t **carray;
memset(used_by_type, 0, sizeof(used_by_type));
memset(alloc_by_type, 0, sizeof(alloc_by_type));
memset(n_conns_by_type, 0, sizeof(n_conns_by_type));
get_connection_array(&carray,&n);
for (i=0; i<n; ++i) {
connection_t *c = carray[i];
int tp = c->type;
++n_conns_by_type[tp];
if (c->inbuf) {
used_by_type[tp] += buf_datalen(c->inbuf);
alloc_by_type[tp] += buf_capacity(c->inbuf);
}
if (c->outbuf) {
used_by_type[tp] += buf_datalen(c->outbuf);
alloc_by_type[tp] += buf_capacity(c->outbuf);
}
}
for (i=0; i <= _CONN_TYPE_MAX; ++i) {
total_used += used_by_type[i];
total_alloc += alloc_by_type[i];
}
log(severity, LD_GENERAL,
"In buffers for %d connections: "U64_FORMAT" used/"U64_FORMAT" allocated",
n, U64_PRINTF_ARG(total_used), U64_PRINTF_ARG(total_alloc));
for (i=_CONN_TYPE_MIN; i <= _CONN_TYPE_MAX; ++i) {
if (!n_conns_by_type[i])
continue;
log(severity, LD_GENERAL,
" For %d %s connections: "U64_FORMAT" used/"U64_FORMAT" allocated",
n_conns_by_type[i], conn_type_to_string(i),
U64_PRINTF_ARG(used_by_type[i]), U64_PRINTF_ARG(alloc_by_type[i]));
}
}
/** Verify that connection <b>conn</b> has all of its invariants
* correct. Trigger an assert if anything is invalid.
*/
void
assert_connection_ok(connection_t *conn, time_t now)
{
(void) now; /* XXXX unused. */
tor_assert(conn);
tor_assert(conn->type >= _CONN_TYPE_MIN);
tor_assert(conn->type <= _CONN_TYPE_MAX);
switch (conn->type) {
case CONN_TYPE_OR:
tor_assert(conn->magic == OR_CONNECTION_MAGIC);
break;
case CONN_TYPE_AP:
case CONN_TYPE_EXIT:
tor_assert(conn->magic == EDGE_CONNECTION_MAGIC);
break;
case CONN_TYPE_DIR:
tor_assert(conn->magic == DIR_CONNECTION_MAGIC);
break;
case CONN_TYPE_CONTROL:
tor_assert(conn->magic == CONTROL_CONNECTION_MAGIC);
break;
default:
tor_assert(conn->magic == BASE_CONNECTION_MAGIC);
break;
}
if (conn->linked_conn) {
tor_assert(conn->linked_conn->linked_conn == conn);
tor_assert(conn->linked != 0);
}
if (conn->linked)
tor_assert(conn->s < 0);
if (conn->outbuf_flushlen > 0) {
tor_assert(connection_is_writing(conn) || conn->write_blocked_on_bw ||
conn->edge_blocked_on_circ);
}
if (conn->hold_open_until_flushed)
tor_assert(conn->marked_for_close);
/* XXXX check: read_blocked_on_bw, write_blocked_on_bw, s, conn_array_index,
* marked_for_close. */
2003-12-17 22:09:31 +01:00
/* buffers */
if (!connection_is_listener(conn)) {
assert_buf_ok(conn->inbuf);
assert_buf_ok(conn->outbuf);
}
/* XXXX Fix this; no longer so.*/
#if 0
if (conn->type != CONN_TYPE_OR && conn->type != CONN_TYPE_DIR)
tor_assert(!conn->pkey);
/* pkey is set if we're a dir client, or if we're an OR in state OPEN
* connected to another OR.
*/
#endif
2006-10-01 23:42:44 +02:00
if (conn->chosen_exit_optional) {
tor_assert(conn->type == CONN_TYPE_AP);
tor_assert((TO_EDGE_CONN(conn))->chosen_exit_name);
}
if (conn->type == CONN_TYPE_OR) {
or_connection_t *or_conn = TO_OR_CONN(conn);
if (conn->state == OR_CONN_STATE_OPEN) {
/* tor_assert(conn->bandwidth > 0); */
/* the above isn't necessarily true: if we just did a TLS
* handshake but we didn't recognize the other peer, or it
* gave a bad cert/etc, then we won't have assigned bandwidth,
* yet it will be open. -RD
*/
// tor_assert(conn->read_bucket >= 0);
}
// tor_assert(conn->addr && conn->port);
tor_assert(conn->address);
if (conn->state > OR_CONN_STATE_PROXY_READING)
tor_assert(or_conn->tls);
}
2003-12-17 22:09:31 +01:00
if (CONN_IS_EDGE(conn)) {
edge_connection_t *edge_conn = TO_EDGE_CONN(conn);
/* XXX unchecked: package window, deliver window. */
if (conn->type == CONN_TYPE_AP) {
tor_assert(edge_conn->socks_request);
if (conn->state == AP_CONN_STATE_OPEN) {
tor_assert(edge_conn->socks_request->has_finished);
if (!conn->marked_for_close) {
tor_assert(edge_conn->cpath_layer);
assert_cpath_layer_ok(edge_conn->cpath_layer);
}
}
}
if (conn->type == CONN_TYPE_EXIT) {
tor_assert(conn->purpose == EXIT_PURPOSE_CONNECT ||
conn->purpose == EXIT_PURPOSE_RESOLVE);
}
} else if (conn->type == CONN_TYPE_DIR) {
} else {
/* Purpose is only used for dir and exit types currently */
tor_assert(!conn->purpose);
}
switch (conn->type)
{
case CONN_TYPE_OR_LISTENER:
case CONN_TYPE_AP_LISTENER:
case CONN_TYPE_AP_TRANS_LISTENER:
case CONN_TYPE_AP_NATD_LISTENER:
case CONN_TYPE_DIR_LISTENER:
case CONN_TYPE_CONTROL_LISTENER:
tor_assert(conn->state == LISTENER_STATE_READY);
break;
case CONN_TYPE_OR:
tor_assert(conn->state >= _OR_CONN_STATE_MIN);
tor_assert(conn->state <= _OR_CONN_STATE_MAX);
tor_assert(TO_OR_CONN(conn)->n_circuits >= 0);
break;
case CONN_TYPE_EXIT:
tor_assert(conn->state >= _EXIT_CONN_STATE_MIN);
tor_assert(conn->state <= _EXIT_CONN_STATE_MAX);
tor_assert(conn->purpose >= _EXIT_PURPOSE_MIN);
tor_assert(conn->purpose <= _EXIT_PURPOSE_MAX);
break;
case CONN_TYPE_AP:
tor_assert(conn->state >= _AP_CONN_STATE_MIN);
tor_assert(conn->state <= _AP_CONN_STATE_MAX);
tor_assert(TO_EDGE_CONN(conn)->socks_request);
break;
case CONN_TYPE_DIR:
tor_assert(conn->state >= _DIR_CONN_STATE_MIN);
tor_assert(conn->state <= _DIR_CONN_STATE_MAX);
tor_assert(conn->purpose >= _DIR_PURPOSE_MIN);
tor_assert(conn->purpose <= _DIR_PURPOSE_MAX);
break;
case CONN_TYPE_CPUWORKER:
tor_assert(conn->state >= _CPUWORKER_STATE_MIN);
tor_assert(conn->state <= _CPUWORKER_STATE_MAX);
break;
case CONN_TYPE_CONTROL:
tor_assert(conn->state >= _CONTROL_CONN_STATE_MIN);
tor_assert(conn->state <= _CONTROL_CONN_STATE_MAX);
break;
default:
tor_assert(0);
}
}