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 */
|
|
|
|
/* $Id$ */
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
#include "or.h"
|
|
|
|
|
|
|
|
/********* START VARIABLES **********/
|
|
|
|
|
2002-07-05 08:27:23 +02:00
|
|
|
static circuit_t *global_circuitlist=NULL;
|
2002-06-27 00:45:49 +02:00
|
|
|
|
2002-09-22 00:41:48 +02:00
|
|
|
char *circuit_state_to_string[] = {
|
|
|
|
"receiving the onion", /* 0 */
|
2002-11-27 05:08:20 +01:00
|
|
|
"waiting to process create", /* 1 */
|
|
|
|
"connecting to firsthop", /* 2 */
|
|
|
|
"open" /* 3 */
|
2002-09-22 00:41:48 +02:00
|
|
|
};
|
|
|
|
|
2002-06-27 00:45:49 +02:00
|
|
|
/********* END VARIABLES ************/
|
|
|
|
|
|
|
|
void circuit_add(circuit_t *circ) {
|
|
|
|
|
|
|
|
if(!global_circuitlist) { /* first one */
|
|
|
|
global_circuitlist = circ;
|
|
|
|
circ->next = NULL;
|
|
|
|
} else {
|
|
|
|
circ->next = global_circuitlist;
|
|
|
|
global_circuitlist = circ;
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
void circuit_remove(circuit_t *circ) {
|
|
|
|
circuit_t *tmpcirc;
|
|
|
|
|
2002-07-05 08:27:23 +02:00
|
|
|
assert(circ && global_circuitlist);
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
if(global_circuitlist == circ) {
|
|
|
|
global_circuitlist = global_circuitlist->next;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
for(tmpcirc = global_circuitlist;tmpcirc->next;tmpcirc = tmpcirc->next) {
|
|
|
|
if(tmpcirc->next == circ) {
|
|
|
|
tmpcirc->next = circ->next;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
circuit_t *circuit_new(aci_t p_aci, connection_t *p_conn) {
|
|
|
|
circuit_t *circ;
|
|
|
|
|
|
|
|
circ = (circuit_t *)malloc(sizeof(circuit_t));
|
|
|
|
if(!circ)
|
|
|
|
return NULL;
|
|
|
|
memset(circ,0,sizeof(circuit_t)); /* zero it out */
|
|
|
|
|
|
|
|
circ->p_aci = p_aci;
|
|
|
|
circ->p_conn = p_conn;
|
|
|
|
|
2002-11-27 05:08:20 +01:00
|
|
|
circ->state = CIRCUIT_STATE_ONION_WAIT;
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
/* ACIs */
|
|
|
|
circ->p_aci = p_aci;
|
2002-07-05 08:27:23 +02:00
|
|
|
/* circ->n_aci remains 0 because we haven't identified the next hop yet */
|
2002-06-27 00:45:49 +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
|
|
|
circ->n_receive_circwindow = CIRCWINDOW_START;
|
|
|
|
circ->p_receive_circwindow = CIRCWINDOW_START;
|
2002-07-18 08:37:58 +02:00
|
|
|
|
2002-06-27 00:45:49 +02:00
|
|
|
circuit_add(circ);
|
|
|
|
|
|
|
|
return circ;
|
|
|
|
}
|
|
|
|
|
|
|
|
void circuit_free(circuit_t *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
|
|
|
struct data_queue_t *tmpd;
|
2002-08-22 09:30:03 +02:00
|
|
|
|
|
|
|
if (circ->n_crypto)
|
|
|
|
crypto_free_cipher_env(circ->n_crypto);
|
|
|
|
if (circ->p_crypto)
|
|
|
|
crypto_free_cipher_env(circ->p_crypto);
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
if(circ->onion)
|
|
|
|
free(circ->onion);
|
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
|
|
|
if(circ->cpath)
|
|
|
|
circuit_free_cpath(circ->cpath, circ->cpathlen);
|
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
|
|
|
while(circ->data_queue) {
|
|
|
|
tmpd = circ->data_queue;
|
|
|
|
circ->data_queue = tmpd->next;
|
|
|
|
free(tmpd->cell);
|
|
|
|
free(tmpd);
|
|
|
|
}
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
free(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
|
|
|
}
|
|
|
|
|
2002-08-24 06:59:21 +02:00
|
|
|
void circuit_free_cpath(crypt_path_t **cpath, int cpathlen) {
|
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 i;
|
|
|
|
|
|
|
|
for(i=0;i<cpathlen;i++)
|
|
|
|
free(cpath[i]);
|
2002-06-27 00:45:49 +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
|
|
|
free(cpath);
|
2002-06-27 00:45:49 +02:00
|
|
|
}
|
|
|
|
|
2002-12-30 09:51:41 +01:00
|
|
|
/* return 0 if can't get a unique aci. */
|
2002-06-27 00:45:49 +02:00
|
|
|
aci_t get_unique_aci_by_addr_port(uint32_t addr, uint16_t port, int aci_type) {
|
|
|
|
aci_t test_aci;
|
|
|
|
connection_t *conn;
|
|
|
|
|
2002-12-30 09:51:41 +01:00
|
|
|
try_again:
|
2002-06-27 00:45:49 +02:00
|
|
|
log(LOG_DEBUG,"get_unique_aci_by_addr_port() trying to get a unique aci");
|
|
|
|
|
2002-08-22 09:30:03 +02:00
|
|
|
crypto_pseudo_rand(2, (unsigned char *)&test_aci);
|
2002-06-27 00:45:49 +02:00
|
|
|
|
2002-12-30 09:51:41 +01:00
|
|
|
if(aci_type == ACI_TYPE_LOWER && test_aci >= (2<<15))
|
|
|
|
test_aci -= (2<<15);
|
|
|
|
if(aci_type == ACI_TYPE_HIGHER && test_aci < (2<<15))
|
|
|
|
test_aci += (2<<15);
|
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
|
|
|
/* if aci_type == ACI_BOTH, don't filter any of it */
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
if(test_aci == 0)
|
2002-12-30 09:51:41 +01:00
|
|
|
goto try_again;
|
2002-06-27 00:45:49 +02:00
|
|
|
|
2002-07-08 10:59:15 +02:00
|
|
|
conn = connection_exact_get_by_addr_port(addr,port);
|
2002-06-27 00:45:49 +02:00
|
|
|
if(!conn) /* there can't be a conflict -- no connection of that sort yet */
|
|
|
|
return test_aci;
|
|
|
|
|
|
|
|
if(circuit_get_by_aci_conn(test_aci, conn))
|
2002-12-30 09:51:41 +01:00
|
|
|
goto try_again;
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
return test_aci;
|
|
|
|
}
|
|
|
|
|
|
|
|
int circuit_init(circuit_t *circ, int aci_type) {
|
2002-08-22 09:30:03 +02:00
|
|
|
unsigned char iv[16];
|
2002-06-27 00:45:49 +02:00
|
|
|
unsigned char digest1[20];
|
|
|
|
unsigned char digest2[20];
|
2002-11-27 05:08:20 +01:00
|
|
|
struct timeval start, end;
|
|
|
|
int time_passed;
|
|
|
|
|
2002-12-03 23:18:23 +01:00
|
|
|
assert(circ && circ->onion);
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
log(LOG_DEBUG,"circuit_init(): starting");
|
2002-12-03 23:18:23 +01:00
|
|
|
circ->n_port = ntohs(*(uint16_t *)(circ->onion+2));
|
|
|
|
log(LOG_DEBUG,"circuit_init(): Set port to %u.",circ->n_port);
|
|
|
|
circ->n_addr = ntohl(*(uint32_t *)(circ->onion+4));
|
|
|
|
circ->p_f = *(circ->onion+1) >> 4; /* backf */
|
|
|
|
log(LOG_DEBUG,"circuit_init(): Set BACKF to %u.",circ->p_f);
|
|
|
|
circ->n_f = *(circ->onion+1) & 0x0f; /* forwf */
|
|
|
|
log(LOG_DEBUG,"circuit_init(): Set FORWF to %u.",circ->n_f);
|
2002-06-27 00:45:49 +02:00
|
|
|
circ->state = CIRCUIT_STATE_OPEN;
|
|
|
|
|
|
|
|
log(LOG_DEBUG,"circuit_init(): aci_type = %u.",aci_type);
|
|
|
|
|
2002-11-27 05:08:20 +01:00
|
|
|
|
|
|
|
|
|
|
|
gettimeofday(&start,NULL);
|
|
|
|
|
2002-06-27 00:45:49 +02:00
|
|
|
circ->n_aci = get_unique_aci_by_addr_port(circ->n_addr, circ->n_port, aci_type);
|
2002-12-30 09:51:41 +01:00
|
|
|
if(!circ->n_aci) {
|
|
|
|
log(LOG_ERR,"circuit_init(): failed to get unique aci.");
|
|
|
|
return -1;
|
|
|
|
}
|
2002-06-27 00:45:49 +02:00
|
|
|
|
2002-11-27 05:08:20 +01:00
|
|
|
gettimeofday(&end,NULL);
|
|
|
|
|
|
|
|
if(end.tv_usec < start.tv_usec) {
|
|
|
|
end.tv_sec--;
|
|
|
|
end.tv_usec += 1000000;
|
|
|
|
}
|
|
|
|
time_passed = ((end.tv_sec - start.tv_sec)*1000000) + (end.tv_usec - start.tv_usec);
|
|
|
|
if(time_passed > 1000) { /* more than 1ms */
|
|
|
|
log(LOG_NOTICE,"circuit_init(): get_unique_aci just took %d us!",time_passed);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
2002-06-27 00:45:49 +02:00
|
|
|
log(LOG_DEBUG,"circuit_init(): Chosen ACI %u.",circ->n_aci);
|
|
|
|
|
|
|
|
/* keys */
|
2002-12-03 23:18:23 +01:00
|
|
|
memset(iv, 0, 16);
|
|
|
|
crypto_SHA_digest(circ->onion+12,16,digest1);
|
2002-08-22 09:30:03 +02:00
|
|
|
crypto_SHA_digest(digest1,20,digest2);
|
|
|
|
crypto_SHA_digest(digest2,20,digest1);
|
2002-06-27 00:45:49 +02:00
|
|
|
log(LOG_DEBUG,"circuit_init(): Computed keys.");
|
|
|
|
|
2002-10-02 22:39:51 +02:00
|
|
|
if (!(circ->p_crypto = create_onion_cipher(circ->p_f,digest2,iv,1))) {
|
2002-06-27 00:45:49 +02:00
|
|
|
log(LOG_ERR,"Cipher initialization failed (ACI %u).",circ->n_aci);
|
|
|
|
return -1;
|
|
|
|
}
|
2002-08-22 09:30:03 +02:00
|
|
|
|
2002-10-02 22:39:51 +02:00
|
|
|
if (!(circ->n_crypto = create_onion_cipher(circ->n_f, digest1, iv, 0))) {
|
2002-06-27 00:45:49 +02:00
|
|
|
log(LOG_ERR,"Cipher initialization failed (ACI %u).",circ->n_aci);
|
|
|
|
return -1;
|
|
|
|
}
|
2002-08-22 09:30:03 +02:00
|
|
|
|
2002-06-27 00:45:49 +02:00
|
|
|
log(LOG_DEBUG,"circuit_init(): Cipher initialization complete.");
|
|
|
|
|
2002-12-03 23:18:23 +01:00
|
|
|
circ->expire = ntohl(*(uint32_t *)(circ->onion+8));
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2002-09-24 12:43:57 +02:00
|
|
|
circuit_t *circuit_enumerate_by_naddr_nport(circuit_t *circ, uint32_t naddr, uint16_t nport) {
|
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
|
|
|
|
2002-09-24 12:43:57 +02:00
|
|
|
if(!circ) /* use circ if it's defined, else start from the beginning */
|
|
|
|
circ = global_circuitlist;
|
|
|
|
else
|
|
|
|
circ = circ->next;
|
|
|
|
|
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( ; circ; circ = circ->next) {
|
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
|
|
|
if(circ->n_addr == naddr && circ->n_port == nport)
|
|
|
|
return circ;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2002-06-27 00:45:49 +02:00
|
|
|
circuit_t *circuit_get_by_aci_conn(aci_t aci, connection_t *conn) {
|
|
|
|
circuit_t *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
|
|
|
connection_t *tmpconn;
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
for(circ=global_circuitlist;circ;circ = circ->next) {
|
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(circ->p_aci == aci) {
|
|
|
|
for(tmpconn = circ->p_conn; tmpconn; tmpconn = tmpconn->next_topic) {
|
|
|
|
if(tmpconn == conn)
|
|
|
|
return circ;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if(circ->n_aci == aci) {
|
|
|
|
for(tmpconn = circ->n_conn; tmpconn; tmpconn = tmpconn->next_topic) {
|
|
|
|
if(tmpconn == conn)
|
|
|
|
return circ;
|
|
|
|
}
|
|
|
|
}
|
2002-06-27 00:45:49 +02:00
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
circuit_t *circuit_get_by_conn(connection_t *conn) {
|
|
|
|
circuit_t *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
|
|
|
connection_t *tmpconn;
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
for(circ=global_circuitlist;circ;circ = circ->next) {
|
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(tmpconn = circ->p_conn; tmpconn; tmpconn=tmpconn->next_topic)
|
|
|
|
if(tmpconn == conn)
|
|
|
|
return circ;
|
|
|
|
for(tmpconn = circ->n_conn; tmpconn; tmpconn=tmpconn->next_topic)
|
|
|
|
if(tmpconn == conn)
|
|
|
|
return circ;
|
2002-06-27 00:45:49 +02:00
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2003-02-06 09:00:49 +01:00
|
|
|
circuit_t *circuit_get_by_edge_type(char edge_type) {
|
|
|
|
circuit_t *circ;
|
|
|
|
|
|
|
|
for(circ=global_circuitlist;circ;circ = circ->next) {
|
|
|
|
if(edge_type == EDGE_AP && circ->n_conn && circ->n_conn->type == CONN_TYPE_OR) {
|
|
|
|
log(LOG_DEBUG,"circuit_get_by_edge_type(): Choosing n_aci %d.", circ->n_aci);
|
|
|
|
return circ;
|
|
|
|
}
|
|
|
|
if(edge_type == EDGE_EXIT && circ->p_conn && circ->p_conn->type == CONN_TYPE_OR) {
|
|
|
|
return circ;
|
|
|
|
}
|
|
|
|
log(LOG_DEBUG,"circuit_get_by_edge_type(): Skipping p_aci %d / n_aci %d.", circ->p_aci, circ->n_aci);
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
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 circuit_deliver_data_cell_from_edge(cell_t *cell, circuit_t *circ, char edge_type) {
|
|
|
|
int cell_direction;
|
2003-02-06 09:00:49 +01:00
|
|
|
static int numsent_ap=0, numsent_exit=0;
|
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,"circuit_deliver_data_cell_from_edge(): called, edge_type %d.", edge_type);
|
2002-06-27 00:45:49 +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(edge_type == EDGE_AP) { /* i'm the AP */
|
|
|
|
cell_direction = CELL_DIRECTION_OUT;
|
2003-02-06 09:00:49 +01:00
|
|
|
numsent_ap++;
|
|
|
|
log(LOG_DEBUG,"circuit_deliver_data_cell_from_edge(): now sent %d data cells from ap", numsent_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
|
|
|
if(circ->p_receive_circwindow <= 0) {
|
2003-02-18 02:35:55 +01:00
|
|
|
log(LOG_DEBUG,"circuit_deliver_data_cell_from_edge(): pwindow 0, queueing for later.");
|
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->data_queue = data_queue_add(circ->data_queue, cell);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
circ->p_receive_circwindow--;
|
2003-02-18 02:35:55 +01:00
|
|
|
// log(LOG_INFO,"circuit_deliver_data_cell_from_edge(): p_receive_circwindow now %d.",circ->p_receive_circwindow);
|
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 { /* i'm the exit */
|
|
|
|
cell_direction = CELL_DIRECTION_IN;
|
2003-02-06 09:00:49 +01:00
|
|
|
numsent_exit++;
|
|
|
|
log(LOG_DEBUG,"circuit_deliver_data_cell_from_edge(): now sent %d data cells from exit", numsent_exit);
|
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(circ->n_receive_circwindow <= 0) {
|
2003-02-18 02:35:55 +01:00
|
|
|
log(LOG_DEBUG,"circuit_deliver_data_cell_from_edge(): nwindow 0, queueing for later.");
|
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->data_queue = data_queue_add(circ->data_queue, cell);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
circ->n_receive_circwindow--;
|
|
|
|
}
|
|
|
|
|
|
|
|
if(circuit_deliver_data_cell(cell, circ, cell_direction) < 0) {
|
|
|
|
return -1;
|
|
|
|
}
|
2003-02-18 02:35:55 +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_stop_edge_reading(circ, edge_type); /* has window reached 0? */
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int circuit_deliver_data_cell(cell_t *cell, circuit_t *circ, int cell_direction) {
|
|
|
|
connection_t *conn;
|
|
|
|
|
|
|
|
assert(cell && circ);
|
|
|
|
assert(cell_direction == CELL_DIRECTION_OUT || cell_direction == CELL_DIRECTION_IN);
|
|
|
|
if(cell_direction == CELL_DIRECTION_OUT)
|
|
|
|
conn = circ->n_conn;
|
|
|
|
else
|
|
|
|
conn = circ->p_conn;
|
|
|
|
|
|
|
|
/* first crypt cell->length */
|
|
|
|
if(circuit_crypt(circ, &(cell->length), 1, cell_direction) < 0) {
|
|
|
|
log(LOG_DEBUG,"circuit_deliver_data_cell(): length crypt failed. Dropping connection.");
|
2002-06-27 00:45:49 +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
|
|
|
/* then crypt the payload */
|
|
|
|
if(circuit_crypt(circ, (char *)&(cell->payload), CELL_PAYLOAD_SIZE, cell_direction) < 0) {
|
|
|
|
log(LOG_DEBUG,"circuit_deliver_data_cell(): payload crypt failed. Dropping connection.");
|
2002-06-27 00:45:49 +02:00
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
2003-03-05 21:03:05 +01:00
|
|
|
if(cell_direction == CELL_DIRECTION_OUT && (!conn || conn->type == CONN_TYPE_EXIT)) {
|
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,"circuit_deliver_data_cell(): Sending to exit.");
|
|
|
|
return connection_exit_process_data_cell(cell, circ);
|
2002-06-27 00:45:49 +02:00
|
|
|
}
|
2003-03-05 21:03:05 +01:00
|
|
|
if(cell_direction == CELL_DIRECTION_IN && (!conn || conn->type == CONN_TYPE_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
|
|
|
log(LOG_DEBUG,"circuit_deliver_data_cell(): Sending to AP.");
|
|
|
|
return connection_ap_process_data_cell(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
|
|
|
}
|
|
|
|
/* else send it as a 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
|
|
|
assert(conn);
|
|
|
|
//log(LOG_DEBUG,"circuit_deliver_data_cell(): Sending to connection.");
|
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 connection_write_cell_to_buf(cell, conn);
|
2002-06-27 00:45:49 +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
|
|
|
int circuit_crypt(circuit_t *circ, char *in, int inlen, char cell_direction) {
|
2002-06-27 00:45:49 +02:00
|
|
|
char *out;
|
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 i;
|
|
|
|
crypt_path_t *thishop;
|
2002-06-27 00:45:49 +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
|
|
|
assert(circ && in);
|
2002-06-27 00:45:49 +02:00
|
|
|
|
2002-08-22 09:30:03 +02:00
|
|
|
out = (char *)malloc(inlen);
|
2002-06-27 00:45:49 +02:00
|
|
|
if(!out)
|
|
|
|
return -1;
|
|
|
|
|
2003-03-05 21:03:05 +01:00
|
|
|
if(cell_direction == CELL_DIRECTION_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
|
|
|
if(circ->cpath) { /* we're at the beginning of the circuit. We'll want to do layered crypts. */
|
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 (i=circ->cpathlen-1; i >= 0; i--) /* moving from first to last hop
|
|
|
|
* Remember : cpath is in reverse order, i.e. last hop first
|
|
|
|
*/
|
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
|
|
|
{
|
|
|
|
thishop = circ->cpath[i];
|
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
|
|
|
|
|
|
|
/* decrypt */
|
|
|
|
if(crypto_cipher_decrypt(thishop->b_crypto, in, inlen, out)) {
|
|
|
|
log(LOG_ERR,"Error performing decryption:%s",crypto_perror());
|
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
|
|
|
free(out);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* copy ciphertext back to buf */
|
|
|
|
memcpy(in,out,inlen);
|
|
|
|
}
|
|
|
|
} else { /* we're in the middle. Just one crypt. */
|
2002-08-22 09:30:03 +02:00
|
|
|
if(crypto_cipher_encrypt(circ->p_crypto,in, inlen, out)) {
|
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_ERR,"circuit_encrypt(): Encryption failed for ACI : %u (%s).",
|
2002-08-22 09:30:03 +02:00
|
|
|
circ->p_aci, crypto_perror());
|
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
|
|
|
free(out);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
memcpy(in,out,inlen);
|
2002-06-27 00:45:49 +02:00
|
|
|
}
|
2003-03-05 21:03:05 +01:00
|
|
|
} else if(cell_direction == CELL_DIRECTION_OUT) {
|
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
|
|
|
if(circ->cpath) { /* we're at the beginning of the circuit. We'll want to do layered crypts. */
|
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 (i=0; i < circ->cpathlen; i++) /* moving from last to first hop
|
|
|
|
* Remember : cpath is in reverse order, i.e. last hop first
|
|
|
|
*/
|
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
|
|
|
{
|
|
|
|
thishop = circ->cpath[i];
|
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
|
|
|
|
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
|
|
|
/* encrypt */
|
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(crypto_cipher_encrypt(thishop->f_crypto, in, inlen, (unsigned char *)out)) {
|
|
|
|
log(LOG_ERR,"Error performing encryption:%s",crypto_perror());
|
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
|
|
|
free(out);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* copy ciphertext back to buf */
|
|
|
|
memcpy(in,out,inlen);
|
|
|
|
}
|
|
|
|
} else { /* we're in the middle. Just one crypt. */
|
2002-08-22 09:30:03 +02:00
|
|
|
if(crypto_cipher_decrypt(circ->n_crypto,in, inlen, out)) {
|
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_ERR,"circuit_crypt(): Decryption failed for ACI : %u (%s).",
|
2002-08-22 09:30:03 +02:00
|
|
|
circ->n_aci, crypto_perror());
|
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
|
|
|
free(out);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
memcpy(in,out,inlen);
|
2002-06-27 00:45:49 +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
|
|
|
} else {
|
|
|
|
log(LOG_ERR,"circuit_crypt(): unknown cell direction %d.", cell_direction);
|
|
|
|
assert(0);
|
2002-06-27 00:45:49 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
free(out);
|
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;
|
|
|
|
}
|
|
|
|
|
|
|
|
void circuit_resume_edge_reading(circuit_t *circ, int edge_type) {
|
|
|
|
connection_t *conn;
|
|
|
|
struct data_queue_t *tmpd;
|
|
|
|
|
|
|
|
assert(edge_type == EDGE_EXIT || edge_type == EDGE_AP);
|
|
|
|
|
|
|
|
/* first, send the queue waiting at circ onto the circuit */
|
|
|
|
while(circ->data_queue) {
|
|
|
|
assert(circ->data_queue->cell);
|
2003-02-18 02:35:55 +01:00
|
|
|
if(edge_type == EDGE_EXIT) {
|
|
|
|
circ->n_receive_circwindow--;
|
|
|
|
assert(circ->n_receive_circwindow >= 0);
|
|
|
|
|
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(circuit_deliver_data_cell(circ->data_queue->cell, circ, CELL_DIRECTION_IN) < 0) {
|
|
|
|
circuit_close(circ);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
} else { /* ap */
|
|
|
|
circ->p_receive_circwindow--;
|
|
|
|
assert(circ->p_receive_circwindow >= 0);
|
2003-02-18 02:35:55 +01:00
|
|
|
|
|
|
|
if(circuit_deliver_data_cell(circ->data_queue->cell, circ, CELL_DIRECTION_OUT) < 0) {
|
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;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
tmpd = circ->data_queue;
|
|
|
|
circ->data_queue = tmpd->next;
|
|
|
|
free(tmpd->cell);
|
|
|
|
free(tmpd);
|
|
|
|
|
|
|
|
if(circuit_consider_stop_edge_reading(circ, edge_type))
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2003-02-18 02:35:55 +01:00
|
|
|
if(edge_type == EDGE_EXIT)
|
|
|
|
conn = circ->n_conn;
|
|
|
|
else
|
|
|
|
conn = circ->p_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
|
|
|
for( ; conn; conn=conn->next_topic) {
|
|
|
|
if((edge_type == EDGE_EXIT && conn->n_receive_topicwindow > 0) ||
|
|
|
|
(edge_type == EDGE_AP && conn->p_receive_topicwindow > 0)) {
|
|
|
|
connection_start_reading(conn);
|
|
|
|
connection_package_raw_inbuf(conn); /* handle whatever might still be on the inbuf */
|
|
|
|
}
|
|
|
|
}
|
|
|
|
circuit_consider_stop_edge_reading(circ, edge_type);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* returns 1 if the window is empty, else 0. If it's empty, tell edge conns to stop reading. */
|
|
|
|
int circuit_consider_stop_edge_reading(circuit_t *circ, int edge_type) {
|
|
|
|
connection_t *conn = NULL;
|
|
|
|
|
|
|
|
assert(edge_type == EDGE_EXIT || edge_type == EDGE_AP);
|
|
|
|
|
2003-02-18 02:35:55 +01:00
|
|
|
if(edge_type == EDGE_EXIT && circ->n_receive_circwindow <= 0)
|
|
|
|
conn = circ->n_conn;
|
|
|
|
else if(edge_type == EDGE_AP && circ->p_receive_circwindow <= 0)
|
|
|
|
conn = circ->p_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
|
|
|
else
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
for( ; conn; conn=conn->next_topic)
|
|
|
|
connection_stop_reading(conn);
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
int circuit_consider_sending_sendme(circuit_t *circ, int edge_type) {
|
|
|
|
cell_t sendme;
|
2002-06-27 00:45:49 +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(circ);
|
|
|
|
|
|
|
|
sendme.command = CELL_SENDME;
|
|
|
|
sendme.length = CIRCWINDOW_INCREMENT;
|
|
|
|
|
|
|
|
if(edge_type == EDGE_AP) { /* i'm the AP */
|
2003-02-18 02:35:55 +01:00
|
|
|
while(circ->n_receive_circwindow < CIRCWINDOW_START-CIRCWINDOW_INCREMENT) {
|
|
|
|
log(LOG_INFO,"circuit_consider_sending_sendme(): n_receive_circwindow %d, Queueing sendme forward.", circ->n_receive_circwindow);
|
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->n_receive_circwindow += CIRCWINDOW_INCREMENT;
|
|
|
|
sendme.aci = circ->n_aci;
|
2003-02-18 02:35:55 +01:00
|
|
|
if(connection_write_cell_to_buf(&sendme, circ->n_conn) < 0) {
|
|
|
|
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(edge_type == EDGE_EXIT) { /* i'm the exit */
|
2003-02-18 02:35:55 +01:00
|
|
|
while(circ->p_receive_circwindow < CIRCWINDOW_START-CIRCWINDOW_INCREMENT) {
|
|
|
|
log(LOG_INFO,"circuit_consider_sending_sendme(): p_receive_circwindow %d, Queueing sendme back.", circ->p_receive_circwindow);
|
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->p_receive_circwindow += CIRCWINDOW_INCREMENT;
|
|
|
|
sendme.aci = circ->p_aci;
|
2003-02-18 02:35:55 +01:00
|
|
|
if(connection_write_cell_to_buf(&sendme, circ->p_conn) < 0) {
|
|
|
|
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
|
|
|
}
|
|
|
|
}
|
2002-06-27 00:45:49 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void circuit_close(circuit_t *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
|
|
|
connection_t *conn;
|
|
|
|
|
2003-02-06 09:00:49 +01:00
|
|
|
assert(circ);
|
2002-06-27 00:45:49 +02:00
|
|
|
circuit_remove(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
|
|
|
for(conn=circ->n_conn; conn; conn=conn->next_topic) {
|
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
|
|
|
connection_send_destroy(circ->n_aci, circ->n_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
|
|
|
}
|
|
|
|
for(conn=circ->p_conn; conn; conn=conn->next_topic) {
|
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
|
|
|
connection_send_destroy(circ->p_aci, circ->p_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
|
|
|
}
|
2002-06-27 00:45:49 +02:00
|
|
|
circuit_free(circ);
|
|
|
|
}
|
|
|
|
|
|
|
|
void circuit_about_to_close_connection(connection_t *conn) {
|
|
|
|
/* send destroys for all circuits using conn */
|
|
|
|
/* currently, we assume it's too late to flush conn's buf here.
|
|
|
|
* down the road, maybe we'll consider that eof doesn't mean can't-write
|
|
|
|
*/
|
|
|
|
circuit_t *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
|
|
|
connection_t *prevconn, *tmpconn;
|
|
|
|
cell_t cell;
|
|
|
|
int edge_type;
|
|
|
|
|
|
|
|
if(!connection_speaks_cells(conn)) {
|
|
|
|
/* it's an edge conn. need to remove it from the linked list of
|
|
|
|
* conn's for this circuit. Send an 'end' data topic.
|
|
|
|
* But don't kill the circuit.
|
|
|
|
*/
|
|
|
|
|
|
|
|
circ = circuit_get_by_conn(conn);
|
|
|
|
if(!circ)
|
|
|
|
return;
|
|
|
|
|
|
|
|
memset(&cell, 0, sizeof(cell_t));
|
|
|
|
cell.command = CELL_DATA;
|
|
|
|
cell.length = TOPIC_HEADER_SIZE;
|
2003-02-07 00:48:35 +01:00
|
|
|
*(uint16_t *)(cell.payload+2) = htons(conn->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
|
|
|
*cell.payload = TOPIC_COMMAND_END;
|
2003-02-18 02:35:55 +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
|
|
|
if(conn == circ->p_conn) {
|
|
|
|
circ->p_conn = conn->next_topic;
|
|
|
|
edge_type = EDGE_AP;
|
|
|
|
goto send_end;
|
|
|
|
}
|
|
|
|
if(conn == circ->n_conn) {
|
|
|
|
circ->n_conn = conn->next_topic;
|
|
|
|
edge_type = EDGE_EXIT;
|
|
|
|
goto send_end;
|
|
|
|
}
|
|
|
|
for(prevconn = circ->p_conn; prevconn->next_topic && prevconn->next_topic != conn; prevconn = prevconn->next_topic) ;
|
|
|
|
if(prevconn->next_topic) {
|
|
|
|
prevconn->next_topic = conn->next_topic;
|
|
|
|
edge_type = EDGE_AP;
|
|
|
|
goto send_end;
|
|
|
|
}
|
|
|
|
for(prevconn = circ->n_conn; prevconn->next_topic && prevconn->next_topic != conn; prevconn = prevconn->next_topic) ;
|
|
|
|
if(prevconn->next_topic) {
|
|
|
|
prevconn->next_topic = conn->next_topic;
|
|
|
|
edge_type = EDGE_EXIT;
|
|
|
|
goto send_end;
|
|
|
|
}
|
|
|
|
log(LOG_ERR,"circuit_about_to_close_connection(): edge conn not in circuit's list?");
|
|
|
|
assert(0); /* should never get here */
|
|
|
|
send_end:
|
|
|
|
if(edge_type == EDGE_AP) { /* send to circ->n_conn */
|
|
|
|
log(LOG_INFO,"circuit_about_to_close_connection(): send data end forward (aci %d).",circ->n_aci);
|
|
|
|
cell.aci = circ->n_aci;
|
|
|
|
} else { /* send to circ->p_conn */
|
|
|
|
assert(edge_type == EDGE_EXIT);
|
|
|
|
log(LOG_INFO,"circuit_about_to_close_connection(): send data end backward (aci %d).",circ->p_aci);
|
|
|
|
cell.aci = circ->p_aci;
|
|
|
|
}
|
|
|
|
|
|
|
|
if(circuit_deliver_data_cell_from_edge(&cell, circ, edge_type) < 0) {
|
|
|
|
log(LOG_DEBUG,"circuit_about_to_close_connection(): circuit_deliver_data_cell_from_edge (%d) failed. Closing.", edge_type);
|
|
|
|
circuit_close(circ);
|
|
|
|
}
|
|
|
|
return;
|
|
|
|
}
|
2002-06-27 00:45:49 +02:00
|
|
|
|
|
|
|
while((circ = circuit_get_by_conn(conn))) {
|
|
|
|
circuit_remove(circ);
|
|
|
|
if(circ->n_conn == conn) /* it's closing in front of us */
|
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(tmpconn=circ->p_conn; tmpconn; tmpconn=tmpconn->next_topic) {
|
|
|
|
connection_send_destroy(circ->p_aci, tmpconn);
|
|
|
|
}
|
2002-06-27 00:45:49 +02:00
|
|
|
if(circ->p_conn == conn) /* it's closing behind us */
|
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(tmpconn=circ->n_conn; tmpconn; tmpconn=tmpconn->next_topic) {
|
|
|
|
connection_send_destroy(circ->n_aci, tmpconn);
|
|
|
|
}
|
2002-06-27 00:45:49 +02:00
|
|
|
circuit_free(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
|
|
|
/* FIXME this now leaves some out */
|
2002-09-22 00:41:48 +02:00
|
|
|
void circuit_dump_by_conn(connection_t *conn) {
|
|
|
|
circuit_t *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
|
|
|
connection_t *tmpconn;
|
2002-09-22 00:41:48 +02:00
|
|
|
|
|
|
|
for(circ=global_circuitlist;circ;circ = circ->next) {
|
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(tmpconn=circ->p_conn; tmpconn; tmpconn=tmpconn->next_topic) {
|
|
|
|
if(tmpconn == conn) {
|
|
|
|
printf("Conn %d has App-ward circuit: aci %d (other side %d), state %d (%s)\n",
|
|
|
|
conn->poll_index, circ->p_aci, circ->n_aci, circ->state, circuit_state_to_string[circ->state]);
|
|
|
|
}
|
2002-09-22 00:41:48 +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(tmpconn=circ->n_conn; tmpconn; tmpconn=tmpconn->next_topic) {
|
|
|
|
if(tmpconn == conn) {
|
|
|
|
printf("Conn %d has Exit-ward circuit: aci %d (other side %d), state %d (%s)\n",
|
|
|
|
conn->poll_index, circ->n_aci, circ->p_aci, circ->state, circuit_state_to_string[circ->state]);
|
|
|
|
}
|
2002-09-22 00:41:48 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|