Implemented link padding and receiver token buckets
Each socket reads at most 'bandwidth' bytes per second sustained, but
can handle bursts of up to 10*bandwidth bytes.
Cells are now sent out at evenly-spaced intervals, with padding sent
out otherwise. Set Linkpadding=0 in the rc file to send cells as soon
as they're available (and to never send padding cells).
Added license/copyrights statements at the top of most files.
router->min and router->max have been merged into a single 'bandwidth'
value. We should make the routerinfo_t reflect this (want to do that,
Mat?)
As the bandwidth increases, and we want to stop sleeping more and more
frequently to send a single cell, cpu usage goes up. At 128kB/s we're
pretty much calling poll with a timeout of 1ms or even 0ms. The current
code takes a timeout of 0-9ms and makes it 10ms. prepare_for_poll()
handles everything that should have happened in the past, so as long as
our buffers don't get too full in that 10ms, we're ok.
Speaking of too full, if you run three servers at 100kB/s with -l debug,
it spends too much time printing debugging messages to be able to keep
up with the cells. The outbuf ultimately fills up and it kills that
connection. If you run with -l err, it works fine up through 500kB/s and
probably beyond. Down the road we'll want to teach it to recognize when
an outbuf is getting full, and back off.
svn:r50
2002-07-16 03:12:15 +02:00
|
|
|
/* Copyright 2001,2002 Roger Dingledine, Matej Pfajfar. */
|
|
|
|
/* See LICENSE for licensing information */
|
|
|
|
/* $Id$ */
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
|
|
|
#include "or.h"
|
|
|
|
|
2003-03-18 02:49:55 +01:00
|
|
|
extern or_options_t options; /* command-line and config-file options */
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
|
|
|
int connection_ap_process_inbuf(connection_t *conn) {
|
|
|
|
|
|
|
|
assert(conn && conn->type == CONN_TYPE_AP);
|
|
|
|
|
|
|
|
if(conn->inbuf_reached_eof) {
|
|
|
|
/* eof reached, kill it. */
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_inbuf(): conn reached eof. Closing.");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
2002-07-22 06:08:37 +02:00
|
|
|
// log(LOG_DEBUG,"connection_ap_process_inbuf(): state %d.",conn->state);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
|
|
|
switch(conn->state) {
|
2002-09-22 13:09:07 +02:00
|
|
|
case AP_CONN_STATE_SOCKS_WAIT:
|
|
|
|
return ap_handshake_process_socks(conn);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
case AP_CONN_STATE_OPEN:
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
if(connection_package_raw_inbuf(conn) < 0)
|
|
|
|
return -1;
|
|
|
|
circuit_consider_stop_edge_reading(circuit_get_by_conn(conn), EDGE_AP);
|
|
|
|
return 0;
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
default:
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_inbuf() called in state where I'm waiting. Ignoring buf for now.");
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
int ap_handshake_process_socks(connection_t *conn) {
|
|
|
|
char c;
|
|
|
|
socks4_t socks4_info;
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
circuit_t *circ;
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
|
|
|
assert(conn);
|
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
log(LOG_DEBUG,"ap_handshake_process_socks() entered.");
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
if(!conn->socks_version) { /* try to pull it in */
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
if(conn->inbuf_datalen < sizeof(socks4_t)) /* basic info available? */
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return 0; /* not yet */
|
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
if(connection_fetch_from_buf((char *)&socks4_info,sizeof(socks4_t),conn) < 0)
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return -1;
|
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
log(LOG_DEBUG,"ap_handshake_process_socks(): Successfully read socks info.");
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
if(socks4_info.version != 4) {
|
|
|
|
log(LOG_NOTICE,"ap_handshake_process_socks(): Unrecognized version %d.",socks4_info.version);
|
|
|
|
ap_handshake_socks_reply(conn, SOCKS4_REQUEST_REJECT);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return -1;
|
|
|
|
}
|
2002-09-22 13:09:07 +02:00
|
|
|
conn->socks_version = socks4_info.version;
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
if(socks4_info.command != 1) { /* not a connect? we don't support it. */
|
|
|
|
log(LOG_NOTICE,"ap_handshake_process_socks(): command %d not '1'.",socks4_info.command);
|
|
|
|
ap_handshake_socks_reply(conn, SOCKS4_REQUEST_REJECT);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
conn->dest_port = ntohs(*(uint16_t*)&socks4_info.destport);
|
|
|
|
if(!conn->dest_port) {
|
|
|
|
log(LOG_NOTICE,"ap_handshake_process_socks(): Port is zero.");
|
|
|
|
ap_handshake_socks_reply(conn, SOCKS4_REQUEST_REJECT);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return -1;
|
|
|
|
}
|
2002-09-22 13:09:07 +02:00
|
|
|
log(LOG_NOTICE,"ap_handshake_process_socks(): Dest port is %d.",conn->dest_port);
|
|
|
|
|
|
|
|
if(socks4_info.destip[0] ||
|
|
|
|
socks4_info.destip[1] ||
|
|
|
|
socks4_info.destip[2] ||
|
2002-09-28 01:16:39 +02:00
|
|
|
!socks4_info.destip[3]) { /* not 0.0.0.x */
|
2002-09-22 13:09:07 +02:00
|
|
|
log(LOG_NOTICE,"ap_handshake_process_socks(): destip not in form 0.0.0.x.");
|
2002-09-28 01:16:39 +02:00
|
|
|
sprintf(conn->dest_tmp, "%d.%d.%d.%d", socks4_info.destip[0],
|
|
|
|
socks4_info.destip[1], socks4_info.destip[2], socks4_info.destip[3]);
|
|
|
|
conn->dest_addr = strdup(conn->dest_tmp);
|
|
|
|
log(LOG_DEBUG,"ap_handshake_process_socks(): Successfully read destip (%s)", conn->dest_addr);
|
2002-09-22 13:09:07 +02:00
|
|
|
}
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
|
|
|
}
|
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
if(!conn->read_username) { /* the socks spec says we've got to read stuff until we get a null */
|
2002-09-23 03:04:59 +02:00
|
|
|
for(;;) {
|
|
|
|
if(!conn->inbuf_datalen)
|
|
|
|
return 0; /* maybe next time */
|
2002-09-22 13:09:07 +02:00
|
|
|
if(connection_fetch_from_buf((char *)&c,1,conn) < 0)
|
|
|
|
return -1;
|
|
|
|
if(!c) {
|
|
|
|
conn->read_username = 1;
|
|
|
|
log(LOG_DEBUG,"ap_handshake_process_socks(): Successfully read username.");
|
|
|
|
break;
|
|
|
|
}
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
if(!conn->dest_addr) { /* no dest_addr found yet */
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
2002-09-23 03:04:59 +02:00
|
|
|
for(;;) {
|
|
|
|
if(!conn->inbuf_datalen)
|
|
|
|
return 0; /* maybe next time */
|
2002-09-22 13:09:07 +02:00
|
|
|
if(connection_fetch_from_buf((char *)&c,1,conn) < 0)
|
|
|
|
return -1;
|
2002-09-23 03:04:59 +02:00
|
|
|
conn->dest_tmp[conn->dest_tmplen++] = c;
|
|
|
|
if(conn->dest_tmplen > 500) {
|
2002-09-22 13:09:07 +02:00
|
|
|
log(LOG_NOTICE,"ap_handshake_process_socks(): dest_addr too long!");
|
|
|
|
ap_handshake_socks_reply(conn, SOCKS4_REQUEST_REJECT);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
if(!c) { /* we found the null; we're done */
|
2002-09-23 03:04:59 +02:00
|
|
|
conn->dest_addr = strdup(conn->dest_tmp);
|
2002-09-22 13:09:07 +02:00
|
|
|
log(LOG_NOTICE,"ap_handshake_process_socks(): successfully read dest addr '%s'",
|
|
|
|
conn->dest_addr);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
}
|
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
/* find the circuit that we should use, if there is one. */
|
2003-02-06 09:00:49 +01:00
|
|
|
circ = circuit_get_by_edge_type(EDGE_AP);
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
|
|
|
|
/* now we're all ready to make an onion or send a begin */
|
|
|
|
|
2003-02-06 09:00:49 +01:00
|
|
|
if(circ && circ->state == CIRCUIT_STATE_OPEN) {
|
2003-02-18 02:35:55 +01:00
|
|
|
/* FIXME if circ not yet open, figure out how to queue this begin? */
|
2003-02-06 09:00:49 +01:00
|
|
|
/* add it into the linked list of topics on this circuit */
|
|
|
|
log(LOG_DEBUG,"ap_handshake_process_socks(): attaching new conn to circ. n_aci %d.", circ->n_aci);
|
|
|
|
conn->next_topic = circ->p_conn;
|
|
|
|
circ->p_conn = conn;
|
|
|
|
|
|
|
|
if(ap_handshake_send_begin(conn, circ) < 0) {
|
|
|
|
circuit_close(circ);
|
|
|
|
return -1;
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if(ap_handshake_create_onion(conn) < 0) {
|
2003-02-06 09:00:49 +01:00
|
|
|
if(circ)
|
|
|
|
circuit_close(circ);
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return 0;
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
int ap_handshake_create_onion(connection_t *conn) {
|
|
|
|
int i;
|
|
|
|
int routelen = 0; /* length of the route */
|
|
|
|
unsigned int *route = NULL; /* hops in the route as an array of indexes into rarray */
|
|
|
|
unsigned char *onion = NULL; /* holds the onion */
|
|
|
|
int onionlen = 0; /* onion length in host order */
|
|
|
|
crypt_path_t **cpath = NULL; /* defines the crypt operations that need to be performed on incoming/outgoing data */
|
|
|
|
|
|
|
|
assert(conn);
|
|
|
|
|
|
|
|
/* choose a route */
|
|
|
|
route = (unsigned int *)router_new_route(&routelen);
|
|
|
|
if (!route) {
|
|
|
|
log(LOG_ERR,"ap_handshake_create_onion(): Error choosing a route through the OR network.");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
log(LOG_DEBUG,"ap_handshake_create_onion(): Chosen a route of length %u : ",routelen);
|
|
|
|
#if 0
|
|
|
|
for (i=routelen-1;i>=0;i--)
|
|
|
|
{
|
|
|
|
log(LOG_DEBUG,"ap_handshake_process_ss() : %u : %s:%u, %u",routelen-i,(routerarray[route[i]])->address,ntohs((routerarray[route[i]])->port),RSA_size((routerarray[route[i]])->pkey));
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
/* allocate memory for the crypt path */
|
|
|
|
cpath = malloc(routelen * sizeof(crypt_path_t *));
|
|
|
|
if (!cpath) {
|
|
|
|
log(LOG_ERR,"ap_handshake_create_onion(): Error allocating memory for cpath.");
|
|
|
|
free(route);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* create an onion and calculate crypto keys */
|
|
|
|
onion = router_create_onion(route,routelen,&onionlen,cpath);
|
|
|
|
if (!onion) {
|
|
|
|
log(LOG_ERR,"ap_handshake_create_onion(): Error creating an onion.");
|
|
|
|
free(route);
|
|
|
|
free(cpath); /* it's got nothing in it, since !onion */
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
log(LOG_DEBUG,"ap_handshake_create_onion(): Created an onion of size %u bytes.",onionlen);
|
|
|
|
log(LOG_DEBUG,"ap_handshake_create_onion(): Crypt path :");
|
|
|
|
for (i=0;i<routelen;i++) {
|
|
|
|
log(LOG_DEBUG,"ap_handshake_create_onion() : %u/%u",(cpath[i])->forwf, (cpath[i])->backf);
|
|
|
|
}
|
|
|
|
|
|
|
|
return ap_handshake_establish_circuit(conn, route, routelen, onion, onionlen, cpath);
|
|
|
|
}
|
|
|
|
|
|
|
|
int ap_handshake_establish_circuit(connection_t *conn, unsigned int *route, int routelen, char *onion,
|
|
|
|
int onionlen, crypt_path_t **cpath) {
|
|
|
|
routerinfo_t *firsthop;
|
|
|
|
connection_t *n_conn;
|
|
|
|
circuit_t *circ;
|
|
|
|
|
|
|
|
/* now see if we're already connected to the first OR in 'route' */
|
|
|
|
firsthop = router_get_first_in_route(route, routelen);
|
|
|
|
assert(firsthop); /* should always be defined */
|
|
|
|
free(route); /* we don't need it anymore */
|
|
|
|
|
|
|
|
circ = circuit_new(0, conn); /* sets circ->p_aci and circ->p_conn */
|
|
|
|
circ->state = CIRCUIT_STATE_OR_WAIT;
|
|
|
|
circ->onion = onion;
|
|
|
|
circ->onionlen = onionlen;
|
|
|
|
circ->cpath = cpath;
|
|
|
|
circ->cpathlen = routelen;
|
|
|
|
|
|
|
|
log(LOG_DEBUG,"ap_handshake_establish_circuit(): Looking for firsthop '%s:%u'",
|
2002-08-24 09:55:49 +02:00
|
|
|
firsthop->address,firsthop->or_port);
|
2002-07-08 10:59:15 +02:00
|
|
|
n_conn = connection_twin_get_by_addr_port(firsthop->addr,firsthop->or_port);
|
2002-09-24 12:43:57 +02:00
|
|
|
if(!n_conn || n_conn->state != OR_CONN_STATE_OPEN) { /* not currently connected */
|
2002-07-19 01:44:57 +02:00
|
|
|
circ->n_addr = firsthop->addr;
|
|
|
|
circ->n_port = firsthop->or_port;
|
2003-03-18 02:49:55 +01:00
|
|
|
if(options.ORPort) { /* we would be connected if he were up. but he's not. */
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
log(LOG_DEBUG,"ap_handshake_establish_circuit(): Route's firsthop isn't connected.");
|
|
|
|
circuit_close(circ);
|
|
|
|
return -1;
|
|
|
|
}
|
2002-09-24 12:43:57 +02:00
|
|
|
|
2002-09-26 07:50:51 +02:00
|
|
|
conn->state = AP_CONN_STATE_OR_WAIT;
|
|
|
|
connection_stop_reading(conn); /* Stop listening for input from the AP! */
|
|
|
|
|
2002-09-24 12:43:57 +02:00
|
|
|
if(!n_conn) { /* launch the connection */
|
2002-10-03 00:54:20 +02:00
|
|
|
n_conn = connection_or_connect_as_op(firsthop);
|
2002-09-24 12:43:57 +02:00
|
|
|
if(!n_conn) { /* connect failed, forget the whole thing */
|
|
|
|
log(LOG_DEBUG,"ap_handshake_establish_circuit(): connect to firsthop failed. Closing.");
|
|
|
|
circuit_close(circ);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
2002-09-26 07:50:51 +02:00
|
|
|
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return 0; /* return success. The onion/circuit/etc will be taken care of automatically
|
|
|
|
* (may already have been) whenever n_conn reaches OR_CONN_STATE_OPEN.
|
|
|
|
*/
|
2002-07-19 01:44:57 +02:00
|
|
|
} else { /* it (or a twin) is already open. use it. */
|
|
|
|
circ->n_addr = n_conn->addr;
|
|
|
|
circ->n_port = n_conn->port;
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return ap_handshake_send_onion(conn, n_conn, circ);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2002-09-24 12:43:57 +02:00
|
|
|
/* find circuits that are waiting on me, if any, and get them to send the onion */
|
|
|
|
void ap_handshake_n_conn_open(connection_t *or_conn) {
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
circuit_t *circ;
|
2002-09-24 12:43:57 +02:00
|
|
|
connection_t *p_conn;
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
|
|
|
log(LOG_DEBUG,"ap_handshake_n_conn_open(): Starting.");
|
2002-09-24 12:43:57 +02:00
|
|
|
circ = circuit_enumerate_by_naddr_nport(NULL, or_conn->addr, or_conn->port);
|
|
|
|
for(;;) {
|
|
|
|
if(!circ)
|
|
|
|
return;
|
|
|
|
|
|
|
|
p_conn = circ->p_conn;
|
|
|
|
if(p_conn->state != AP_CONN_STATE_OR_WAIT) {
|
|
|
|
log(LOG_WARNING,"Bug: ap_handshake_n_conn_open() got an ap_conn not in OR_WAIT state.");
|
|
|
|
}
|
|
|
|
connection_start_reading(p_conn); /* resume listening for reads */
|
|
|
|
log(LOG_DEBUG,"ap_handshake_n_conn_open(): Found circ, sending onion.");
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
if(ap_handshake_send_onion(p_conn, or_conn, circ) < 0) {
|
2002-09-24 12:43:57 +02:00
|
|
|
log(LOG_DEBUG,"ap_handshake_n_conn_open(): circuit marked for closing.");
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
circuit_close(circ);
|
|
|
|
return; /* FIXME will want to try the other circuits too? */
|
2002-09-24 12:43:57 +02:00
|
|
|
}
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
for(p_conn = p_conn->next_topic; p_conn; p_conn = p_conn->next_topic) { /* start up any other pending topics */
|
|
|
|
if(ap_handshake_send_begin(p_conn, circ) < 0) {
|
|
|
|
circuit_close(circ);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
circ = circuit_enumerate_by_naddr_nport(circ, or_conn->addr, or_conn->port);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
int ap_handshake_send_onion(connection_t *ap_conn, connection_t *n_conn, circuit_t *circ) {
|
|
|
|
cell_t cell;
|
|
|
|
int tmpbuflen, dataleft;
|
|
|
|
char *tmpbuf;
|
|
|
|
|
|
|
|
circ->n_aci = get_unique_aci_by_addr_port(circ->n_addr, circ->n_port, ACI_TYPE_BOTH);
|
|
|
|
circ->n_conn = n_conn;
|
2002-08-24 09:55:49 +02:00
|
|
|
log(LOG_DEBUG,"ap_handshake_send_onion(): n_conn is %s:%u",n_conn->address,n_conn->port);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
|
|
|
/* deliver the onion as one or more create cells */
|
|
|
|
cell.command = CELL_CREATE;
|
|
|
|
cell.aci = circ->n_aci;
|
|
|
|
|
|
|
|
tmpbuflen = circ->onionlen+4;
|
|
|
|
tmpbuf = malloc(tmpbuflen);
|
|
|
|
if(!tmpbuf)
|
|
|
|
return -1;
|
2002-08-24 09:55:49 +02:00
|
|
|
*(uint32_t*)tmpbuf = htonl(circ->onionlen);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
memcpy(tmpbuf+4, circ->onion, circ->onionlen);
|
|
|
|
|
|
|
|
dataleft = tmpbuflen;
|
|
|
|
while(dataleft) {
|
|
|
|
cell.command = CELL_CREATE;
|
|
|
|
cell.aci = circ->n_aci;
|
|
|
|
log(LOG_DEBUG,"ap_handshake_send_onion(): Sending a create cell for the onion...");
|
|
|
|
if(dataleft >= CELL_PAYLOAD_SIZE) {
|
|
|
|
cell.length = CELL_PAYLOAD_SIZE;
|
|
|
|
memcpy(cell.payload, tmpbuf + tmpbuflen - dataleft, CELL_PAYLOAD_SIZE);
|
2003-02-18 02:35:55 +01:00
|
|
|
connection_write_cell_to_buf(&cell, n_conn);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
dataleft -= CELL_PAYLOAD_SIZE;
|
|
|
|
} else { /* last cell */
|
|
|
|
cell.length = dataleft;
|
|
|
|
memcpy(cell.payload, tmpbuf + tmpbuflen - dataleft, dataleft);
|
2003-03-11 22:38:38 +01:00
|
|
|
/* fill extra space with 0 bytes */
|
|
|
|
memset(cell.payload + dataleft, 0, CELL_PAYLOAD_SIZE - dataleft);
|
2003-02-18 02:35:55 +01:00
|
|
|
connection_write_cell_to_buf(&cell, n_conn);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
dataleft = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
free(tmpbuf);
|
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
if(ap_handshake_send_begin(ap_conn, circ) < 0) {
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
circ->state = CIRCUIT_STATE_OPEN;
|
|
|
|
/* FIXME should set circ->expire to something here */
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int ap_handshake_send_begin(connection_t *ap_conn, circuit_t *circ) {
|
|
|
|
cell_t cell;
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
memset(&cell, 0, sizeof(cell_t));
|
|
|
|
/* deliver the dest_addr in a data cell */
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
cell.command = CELL_DATA;
|
|
|
|
cell.aci = circ->n_aci;
|
2003-02-07 00:48:35 +01:00
|
|
|
crypto_pseudo_rand(2, cell.payload+2); /* bytes 0-1 is blank, bytes 2-3 are random */
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
/* FIXME check for collisions */
|
2003-02-07 00:48:35 +01:00
|
|
|
ap_conn->topic_id = ntohs(*(uint16_t *)(cell.payload+2));
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
cell.payload[0] = TOPIC_COMMAND_BEGIN;
|
|
|
|
snprintf(cell.payload+4, CELL_PAYLOAD_SIZE-4, "%s,%d", ap_conn->dest_addr, ap_conn->dest_port);
|
|
|
|
cell.length = strlen(cell.payload+TOPIC_HEADER_SIZE)+1+TOPIC_HEADER_SIZE;
|
|
|
|
log(LOG_DEBUG,"ap_handshake_send_begin(): Sending data cell to begin topic %d.", ap_conn->topic_id);
|
|
|
|
if(circuit_deliver_data_cell_from_edge(&cell, circ, EDGE_AP) < 0) {
|
|
|
|
log(LOG_DEBUG,"ap_handshake_send_begin(): failed to deliver begin cell. Closing.");
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return -1;
|
|
|
|
}
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
ap_conn->n_receive_topicwindow = TOPICWINDOW_START;
|
|
|
|
ap_conn->p_receive_topicwindow = TOPICWINDOW_START;
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
ap_conn->state = AP_CONN_STATE_OPEN;
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
log(LOG_INFO,"ap_handshake_send_begin(): Address/port sent, ap socket %d, n_aci %d",ap_conn->s,circ->n_aci);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2002-09-22 13:09:07 +02:00
|
|
|
int ap_handshake_socks_reply(connection_t *conn, char result) {
|
|
|
|
socks4_t socks4_info;
|
2002-09-17 10:14:37 +02:00
|
|
|
|
|
|
|
assert(conn);
|
|
|
|
|
2002-09-28 01:34:51 +02:00
|
|
|
socks4_info.version = 0;
|
2002-09-22 13:09:07 +02:00
|
|
|
socks4_info.command = result;
|
|
|
|
socks4_info.destport[0] = socks4_info.destport[1] = 0;
|
|
|
|
socks4_info.destip[0] = socks4_info.destip[1] = socks4_info.destip[2] = socks4_info.destip[3] = 0;
|
|
|
|
|
2002-10-13 15:17:27 +02:00
|
|
|
if(connection_write_to_buf((char *)&socks4_info, sizeof(socks4_t), conn) < 0)
|
|
|
|
return -1;
|
2002-09-22 13:09:07 +02:00
|
|
|
return connection_flush_buf(conn); /* try to flush it, in case we're about to close the conn */
|
|
|
|
}
|
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
int connection_ap_process_data_cell(cell_t *cell, circuit_t *circ) {
|
|
|
|
connection_t *conn;
|
|
|
|
int topic_command;
|
|
|
|
int topic_id;
|
2003-02-06 09:00:49 +01:00
|
|
|
static int num_seen=0;
|
2002-09-22 13:09:07 +02:00
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
/* an incoming data cell has arrived */
|
2002-09-17 10:14:37 +02:00
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
assert(cell && circ);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
topic_command = *cell->payload;
|
2003-02-07 00:48:35 +01:00
|
|
|
topic_id = ntohs(*(uint16_t *)(cell->payload+2));
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): command %d topic %d", topic_command, topic_id);
|
2003-02-06 09:00:49 +01:00
|
|
|
num_seen++;
|
2003-02-18 02:35:55 +01:00
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): Now seen %d data cells here.", num_seen);
|
2003-02-06 09:00:49 +01:00
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
circuit_consider_sending_sendme(circ, EDGE_AP);
|
|
|
|
|
|
|
|
for(conn = circ->p_conn; conn && conn->topic_id != topic_id; conn = conn->next_topic) ;
|
|
|
|
|
|
|
|
/* now conn is either NULL, in which case we don't recognize the topic_id, or
|
|
|
|
* it is set, in which case cell is talking about this conn.
|
|
|
|
*/
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
if(conn && conn->state != AP_CONN_STATE_OPEN) {
|
2002-07-18 08:37:58 +02:00
|
|
|
/* we should not have gotten this cell */
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): Got a data cell when not in 'open' state. Dropping.");
|
|
|
|
return 0;
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
}
|
|
|
|
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
switch(topic_command) {
|
|
|
|
case TOPIC_COMMAND_BEGIN:
|
|
|
|
log(LOG_INFO,"connection_ap_process_data_cell(): topic begin request unsupported. Dropping.");
|
|
|
|
break;
|
|
|
|
case TOPIC_COMMAND_DATA:
|
|
|
|
if(!conn) {
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): data cell dropped, unknown topic %d.",topic_id);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
if(--conn->n_receive_topicwindow < 0) { /* is it below 0 after decrement? */
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): receive_topicwindow at exit below 0. Killing.");
|
|
|
|
return -1; /* exit node breaking protocol. kill the whole circuit. */
|
|
|
|
}
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): willing to receive %d more cells from circ",conn->n_receive_topicwindow);
|
2002-07-18 08:37:58 +02:00
|
|
|
|
2003-03-17 03:42:45 +01:00
|
|
|
#ifdef USE_ZLIB
|
|
|
|
if(connection_decompress_to_buf(cell->payload + TOPIC_HEADER_SIZE,
|
2003-03-19 21:48:56 +01:00
|
|
|
cell->length - TOPIC_HEADER_SIZE,
|
|
|
|
conn, Z_SYNC_FLUSH) < 0) {
|
2003-03-17 03:42:45 +01:00
|
|
|
log(LOG_INFO,"connection_exit_process_data_cell(): write to buf failed. Marking for close.");
|
|
|
|
conn->marked_for_close = 1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#else
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
if(connection_write_to_buf(cell->payload + TOPIC_HEADER_SIZE,
|
|
|
|
cell->length - TOPIC_HEADER_SIZE, conn) < 0) {
|
|
|
|
conn->marked_for_close = 1;
|
|
|
|
return 0;
|
|
|
|
}
|
2003-03-17 03:42:45 +01:00
|
|
|
#endif
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
if(connection_consider_sending_sendme(conn, EDGE_AP) < 0)
|
|
|
|
conn->marked_for_close = 1;
|
|
|
|
return 0;
|
|
|
|
case TOPIC_COMMAND_END:
|
|
|
|
if(!conn) {
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): end cell dropped, unknown topic %d.",topic_id);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): end cell for topic %d. Removing topic.",topic_id);
|
|
|
|
|
|
|
|
/* go through and identify who points to conn. remove conn from the list. */
|
|
|
|
#if 0
|
|
|
|
if(conn == circ->p_conn) {
|
|
|
|
circ->p_conn = conn->next_topic;
|
|
|
|
}
|
|
|
|
for(prevconn = circ->p_conn; prevconn->next_topic != conn; prevconn = prevconn->next_topic) ;
|
|
|
|
prevconn->next_topic = conn->next_topic;
|
2003-03-17 03:42:45 +01:00
|
|
|
#endif
|
2003-03-19 22:59:07 +01:00
|
|
|
#ifdef HALF_OPEN
|
2003-03-17 03:42:45 +01:00
|
|
|
conn->done_sending = 1;
|
|
|
|
shutdown(conn->s, 1); /* XXX check return; refactor NM */
|
|
|
|
if (conn->done_receiving)
|
2003-03-19 21:48:56 +01:00
|
|
|
conn->marked_for_close = 1;
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
#endif
|
|
|
|
conn->marked_for_close = 1;
|
|
|
|
break;
|
|
|
|
case TOPIC_COMMAND_CONNECTED:
|
|
|
|
if(!conn) {
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): connected cell dropped, unknown topic %d.",topic_id);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): Connected! Notifying application.");
|
|
|
|
if(ap_handshake_socks_reply(conn, SOCKS4_REQUEST_GRANTED) < 0) {
|
|
|
|
conn->marked_for_close = 1;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case TOPIC_COMMAND_SENDME:
|
|
|
|
if(!conn) {
|
2003-02-18 02:35:55 +01:00
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): sendme cell dropped, unknown topic %d.",topic_id);
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
conn->p_receive_topicwindow += TOPICWINDOW_INCREMENT;
|
|
|
|
connection_start_reading(conn);
|
|
|
|
connection_package_raw_inbuf(conn); /* handle whatever might still be on the inbuf */
|
|
|
|
circuit_consider_stop_edge_reading(circ, EDGE_AP);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
log(LOG_DEBUG,"connection_ap_process_data_cell(): unknown topic command %d.",topic_command);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
|
|
|
|
int connection_ap_finished_flushing(connection_t *conn) {
|
|
|
|
|
|
|
|
assert(conn && conn->type == CONN_TYPE_AP);
|
|
|
|
|
|
|
|
switch(conn->state) {
|
|
|
|
case AP_CONN_STATE_OPEN:
|
|
|
|
/* FIXME down the road, we'll clear out circuits that are pending to close */
|
2002-07-18 08:37:58 +02:00
|
|
|
connection_stop_writing(conn);
|
2003-03-19 21:48:56 +01:00
|
|
|
#ifdef USE_ZLIB
|
|
|
|
if (connection_decompress_to_buf(NULL, 0, conn, Z_SYNC_FLUSH) < 0)
|
|
|
|
return 0;
|
|
|
|
#endif
|
major overhaul: dns slave subsystem, topics
on startup, it forks off a master dns handler, which forks off dns
slaves (like the apache model). slaves as spawned as load increases,
and then reused. excess slaves are not ever killed, currently.
implemented topics. each topic has a receive window in each direction
at each edge of the circuit, and sends sendme's at the data level, as
per before. each circuit also has receive windows in each direction at
each hop; an edge sends a circuit-level sendme as soon as enough data
cells have arrived (regardless of whether the data cells were flushed
to the exit conns). removed the 'connected' cell type, since it's now
a topic command within data cells.
at the edge of the circuit, there can be multiple connections associated
with a single circuit. you find them via the linked list conn->next_topic.
currently each new ap connection starts its own circuit, so we ought
to see comparable performance to what we had before. but that's only
because i haven't written the code to reattach to old circuits. please
try to break it as-is, and then i'll make it reuse the same circuit and
we'll try to break that.
svn:r152
2003-01-26 10:02:24 +01:00
|
|
|
return connection_consider_sending_sendme(conn, EDGE_AP);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
default:
|
|
|
|
log(LOG_DEBUG,"Bug: connection_ap_finished_flushing() called in unexpected state.");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2002-10-03 00:54:20 +02:00
|
|
|
int connection_ap_create_listener(struct sockaddr_in *bindaddr) {
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
log(LOG_DEBUG,"connection_create_ap_listener starting");
|
2002-10-03 00:54:20 +02:00
|
|
|
return connection_create_listener(bindaddr, CONN_TYPE_AP_LISTENER);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
int connection_ap_handle_listener_read(connection_t *conn) {
|
2002-09-22 13:09:07 +02:00
|
|
|
log(LOG_NOTICE,"AP: Received a connection request. Waiting for socksinfo.");
|
|
|
|
return connection_handle_listener_read(conn, CONN_TYPE_AP, AP_CONN_STATE_SOCKS_WAIT);
|
Integrated onion proxy into or/
The 'or' process can now be told (by the global_role variable) what
roles this server should play -- connect to all ORs, listen for ORs,
listen for OPs, listen for APs, or any combination.
* everything in /src/op/ is now obsolete.
* connection_ap.c now handles all interactions with application proxies
* "port" is now or_port, op_port, ap_port. But routers are still always
referenced (say, in conn_get_by_addr_port()) by addr / or_port. We
should make routers.c actually read these new ports (currently I've
kludged it so op_port = or_port+10, ap_port=or_port+20)
* circuits currently know if they're at the beginning of the path because
circ->cpath is set. They use this instead for crypts (both ways),
if it's set.
* I still obey the "send a 0 back to the AP when you're ready" protocol,
but I think we should phase it out. I can simply not read from the AP
socket until I'm ready.
I need to do a lot of cleanup work here, but the code appears to work, so
now's a good time for a checkin.
svn:r22
2002-07-02 11:36:58 +02:00
|
|
|
}
|