d4708/opentracker
1
0
Fork 0
mirror of git://erdgeist.org/opentracker synced 2025-01-12 16:00:06 +00:00
Code Issues Projects Releases Wiki Activity
5b98dcf3a3
opentracker/trackerlogic.c

462 lines
15 KiB
C
Raw Normal View History

README\!
2007-01-05 00:00:42 +00:00
/* This software was written by Dirk Engling <erdgeist@erdgeist.org>
Introduce some kind of versioning
2007-12-20 05:59:34 +00:00
It is considered beerware. Prost. Skol. Cheers or whatever.
Whitespace fixes
2008-10-28 01:27:22 +00:00
Introduce some kind of versioning
2007-12-20 05:59:34 +00:00
$id$ */
README\!
2007-01-05 00:00:42 +00:00
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
/* System */
Cleaned up type namespace, mapping now works, introduced anonymous mapping, made init/deinit code, time now works.... Glue code and init code still missing
2006-12-06 18:37:46 +00:00
#include <stdlib.h>
Kickoff
2006-12-05 12:56:56 +00:00
#include <string.h>
#include <stdio.h>
Fixed warnings about newlines at end of file
2007-11-06 12:03:11 +00:00
#include <arpa/inet.h>
Cleaned up type namespace, mapping now works, introduced anonymous mapping, made init/deinit code, time now works.... Glue code and init code still missing
2006-12-06 18:37:46 +00:00
#include <unistd.h>
added live sync code added a config file parser added tracker id changed WANT_CLOSED_TRACKER and WANT_BLACKLIST into WANT_ACCESS_WHITE and WANT_ACCESS_BLACK changed WANT_TRACKER_SYNC to WANT_SYNC_BATCH and added WANT_SYNC_LIVE added an option to switch off fullscrapes cleaned up many internal hardcoded values, like PROTO_FLAG,
2008-10-04 05:40:51 +00:00
#include <errno.h>
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
#include <stdint.h>
Cleaned up type namespace, mapping now works, introduced anonymous mapping, made init/deinit code, time now works.... Glue code and init code still missing
2006-12-06 18:37:46 +00:00
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
/* Libowfat */
This is a complete rewrite... assume nothing works.
2006-12-12 02:37:19 +00:00
#include "byte.h"
Drop ot_{byte,word,dword} and use uint{8,16,32}_t, also simplify includes
2007-12-03 01:07:41 +00:00
#include "io.h"
Allow /stats to be located anywhere in your trackers path
2009-06-17 15:06:31 +00:00
#include "iob.h"
#include "array.h"
This is a complete rewrite... assume nothing works.
2006-12-12 02:37:19 +00:00
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
/* Opentracker */
#include "trackerlogic.h"
#include "ot_mutex.h"
#include "ot_stats.h"
#include "ot_clean.h"
Allow /stats to be located anywhere in your trackers path
2009-06-17 15:06:31 +00:00
#include "ot_http.h"
Moving accesslist stuff to its own object
2007-11-12 04:39:53 +00:00
#include "ot_accesslist.h"
Cleanup now moved to its own thread, too
2007-11-19 21:10:53 +00:00
#include "ot_fullscrape.h"
added live sync code added a config file parser added tracker id changed WANT_CLOSED_TRACKER and WANT_BLACKLIST into WANT_ACCESS_WHITE and WANT_ACCESS_BLACK changed WANT_TRACKER_SYNC to WANT_SYNC_BATCH and added WANT_SYNC_LIVE added an option to switch off fullscrapes cleaned up many internal hardcoded values, like PROTO_FLAG,
2008-10-04 05:40:51 +00:00
#include "ot_livesync.h"
* fixed "nobody " -> "nobody" fuckup when getpwnam-ing * implemented basic blacklisting: ** the file specified with -b <BLACKLIST> is read and added to a blacklist vector ** if an announce hits a torrent in that blacklist vector, add_peer_to_torrent fails ** sending a SIGHUP to the program forces it to reread the blacklists ** the server returns with a 500, which is not exactly nice but does the job for now ** an adaequat "failure reason:" should be delivered... TODO
2007-07-22 00:40:10 +00:00
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
/* Forward declaration */
Turn random() to nrand48 in inner loop where it is not cryptographically relevant to avoid futex storm. Thanks and credits to Bart Smienk
2021-04-20 01:48:50 +00:00
size_t return_peers_for_torrent( struct ot_workstruct * ws, ot_torrent *torrent, size_t amount, char *reply, PROTO_FLAG proto );
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
void free_peerlist( ot_peerlist *peer_list ) {
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
if( peer_list->peers.data ) {
if( OT_PEERLIST_HASBUCKETS( peer_list ) ) {
ot_vector *bucket_list = (ot_vector*)(peer_list->peers.data);
while( peer_list->peers.size-- )
free( bucket_list++->data );
}
free( peer_list->peers.data );
}
This is a complete rewrite... assume nothing works.
2006-12-12 02:37:19 +00:00
free( peer_list );
}
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
2009-03-17 23:57:20 +00:00
void add_torrent_from_saved_state( ot_hash hash, ot_time base, size_t down_count ) {
int exactmatch;
ot_torrent *torrent;
ot_vector *torrents_list = mutex_bucket_lock_by_hash( hash );
if( !accesslist_hashisvalid( hash ) )
return mutex_bucket_unlock_by_hash( hash, 0 );
White space
2021-04-20 01:31:58 +00:00
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
2009-03-17 23:57:20 +00:00
torrent = vector_find_or_insert( torrents_list, (void*)hash, sizeof( ot_torrent ), OT_HASH_COMPARE_SIZE, &exactmatch );
if( !torrent || exactmatch )
return mutex_bucket_unlock_by_hash( hash, 0 );
/* Create a new torrent entry, then */
memcpy( torrent->hash, hash, sizeof(ot_hash) );
White space
2021-04-20 01:31:58 +00:00
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
2009-03-17 23:57:20 +00:00
if( !( torrent->peer_list = malloc( sizeof (ot_peerlist) ) ) ) {
vector_remove_torrent( torrents_list, torrent );
return mutex_bucket_unlock_by_hash( hash, 0 );
}
White space
2021-04-20 01:31:58 +00:00
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
2009-03-17 23:57:20 +00:00
byte_zero( torrent->peer_list, sizeof( ot_peerlist ) );
torrent->peer_list->base = base;
torrent->peer_list->down_count = down_count;
return mutex_bucket_unlock_by_hash( hash, 1 );
}
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
size_t add_peer_to_torrent_and_return_peers( PROTO_FLAG proto, struct ot_workstruct *ws, size_t amount ) {
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
int exactmatch, delta_torrentcount = 0;
Now actually seems to work for the most parts Added scraping Added graceful disconnect
2006-12-14 02:44:50 +00:00
ot_torrent *torrent;
ot_peer *peer_dest;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
ot_vector *torrents_list = mutex_bucket_lock_by_hash( *ws->hash );
Kickoff
2006-12-05 12:56:56 +00:00
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
if( !accesslist_hashisvalid( *ws->hash ) ) {
mutex_bucket_unlock_by_hash( *ws->hash, 0 );
Make attempts to access to unauthorized torrents more verbose.
2009-07-04 20:14:03 +00:00
if( proto == FLAG_TCP ) {
const char invalid_hash[] = "d14:failure reason63:Requested download is not authorized for use with this tracker.e";
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
memcpy( ws->reply, invalid_hash, strlen( invalid_hash ) );
Make attempts to access to unauthorized torrents more verbose.
2009-07-04 20:14:03 +00:00
return strlen( invalid_hash );
}
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
return 0;
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
}
* fixed "nobody " -> "nobody" fuckup when getpwnam-ing * implemented basic blacklisting: ** the file specified with -b <BLACKLIST> is read and added to a blacklist vector ** if an announce hits a torrent in that blacklist vector, add_peer_to_torrent fails ** sending a SIGHUP to the program forces it to reread the blacklists ** the server returns with a 500, which is not exactly nice but does the job for now ** an adaequat "failure reason:" should be delivered... TODO
2007-07-22 00:40:10 +00:00
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
torrent = vector_find_or_insert( torrents_list, (void*)ws->hash, sizeof( ot_torrent ), OT_HASH_COMPARE_SIZE, &exactmatch );
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
if( !torrent ) {
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
mutex_bucket_unlock_by_hash( *ws->hash, 0 );
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
return 0;
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
}
Kickoff
2006-12-05 12:56:56 +00:00
This is a complete rewrite... assume nothing works.
2006-12-12 02:37:19 +00:00
if( !exactmatch ) {
Documentation improved, some reindenting (again), variable types checked, unnecessary defines removed
2007-01-11 01:06:10 +00:00
/* Create a new torrent entry, then */
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
memcpy( torrent->hash, *ws->hash, sizeof(ot_hash) );
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
Add a lot of documentation for static functions in trackerlogic.c. Make vector_remove_peer use hysteresis only, if pool is expected to grow again. Remove too old torrents.
2007-02-04 03:29:29 +00:00
if( !( torrent->peer_list = malloc( sizeof (ot_peerlist) ) ) ) {
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
vector_remove_torrent( torrents_list, torrent );
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
mutex_bucket_unlock_by_hash( *ws->hash, 0 );
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
return 0;
This is a complete rewrite... assume nothing works.
2006-12-12 02:37:19 +00:00
}
Kickoff
2006-12-05 12:56:56 +00:00
Add a lot of documentation for static functions in trackerlogic.c. Make vector_remove_peer use hysteresis only, if pool is expected to grow again. Remove too old torrents.
2007-02-04 03:29:29 +00:00
byte_zero( torrent->peer_list, sizeof( ot_peerlist ) );
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
delta_torrentcount = 1;
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
} else
clean_single_torrent( torrent );
Kickoff
2006-12-05 12:56:56 +00:00
Fix: Did not time stamp a torrent correctly, that once timed out and came back to life
2008-12-11 11:48:34 +00:00
torrent->peer_list->base = g_now_minutes;
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
/* Check for peer in torrent */
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
peer_dest = vector_find_or_insert_peer( &(torrent->peer_list->peers), &ws->peer, &exactmatch );
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
if( !peer_dest ) {
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
mutex_bucket_unlock_by_hash( *ws->hash, delta_torrentcount );
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
return 0;
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
}
/* Tell peer that it's fresh */
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
OT_PEERTIME( &ws->peer ) = 0;
Fixed a bug where we forgot to stamp torrents that are not new.
2007-11-28 22:52:35 +00:00
Can now handle multiple ips/ports, can BIND to udp ports but not yet handle them, now checks if event=completed comes with left=0 and drops it, if it doesnt
2007-03-05 21:14:50 +00:00
/* Sanitize flags: Whoever claims to have completed download, must be a seeder */
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
if( ( OT_PEERFLAG( &ws->peer ) & ( PEER_FLAG_COMPLETED | PEER_FLAG_SEEDING ) ) == PEER_FLAG_COMPLETED )
OT_PEERFLAG( &ws->peer ) ^= PEER_FLAG_COMPLETED;
Can now handle multiple ips/ports, can BIND to udp ports but not yet handle them, now checks if event=completed comes with left=0 and drops it, if it doesnt
2007-03-05 21:14:50 +00:00
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
/* If we hadn't had a match create peer there */
This is a complete rewrite... assume nothing works.
2006-12-12 02:37:19 +00:00
if( !exactmatch ) {
Count completed messages only once per peer
2007-02-16 21:13:44 +00:00
only sync when the peer is actually added to the first pool. this saves ca. 40% syncdata
2008-10-07 02:09:38 +00:00
#ifdef WANT_SYNC_LIVE
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
if( proto == FLAG_MCA )
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
OT_PEERFLAG( &ws->peer ) |= PEER_FLAG_FROM_SYNC;
Reverse logic... do only sync those NOT coming from multi cast
2009-02-03 16:45:44 +00:00
else
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
livesync_tell( ws );
only sync when the peer is actually added to the first pool. this saves ca. 40% syncdata
2008-10-07 02:09:38 +00:00
#endif
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
torrent->peer_list->peer_count++;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
if( OT_PEERFLAG(&ws->peer) & PEER_FLAG_COMPLETED ) {
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
torrent->peer_list->down_count++;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
stats_issue_event( EVENT_COMPLETED, 0, (uintptr_t)ws );
include overall completed count in stats
2009-03-24 16:55:28 +00:00
}
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
if( OT_PEERFLAG(&ws->peer) & PEER_FLAG_SEEDING )
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
torrent->peer_list->seed_count++;
Due to recent issues... assert() seed counts
2007-01-27 16:35:56 +00:00
This is a complete rewrite... assume nothing works.
2006-12-12 02:37:19 +00:00
} else {
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
stats_issue_event( EVENT_RENEW, 0, OT_PEERTIME( peer_dest ) );
Add spotting woodpeckers, thanks to Vasya P. again
2009-09-02 02:18:36 +00:00
#ifdef WANT_SPOT_WOODPECKER
if( ( OT_PEERTIME(peer_dest) > 0 ) && ( OT_PEERTIME(peer_dest) < 20 ) )
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
stats_issue_event( EVENT_WOODPECKER, 0, (uintptr_t)&ws->peer );
Add spotting woodpeckers, thanks to Vasya P. again
2009-09-02 02:18:36 +00:00
#endif
Live Sync for peers that do not come back too early.
2008-12-01 17:56:53 +00:00
#ifdef WANT_SYNC_LIVE
/* Won't live sync peers that come back too fast. Only exception:
fresh "completed" reports */
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
if( proto != FLAG_MCA ) {
Live Sync for peers that do not come back too early.
2008-12-01 17:56:53 +00:00
if( OT_PEERTIME( peer_dest ) > OT_CLIENT_SYNC_RENEW_BOUNDARY ||
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
( !(OT_PEERFLAG(peer_dest) & PEER_FLAG_COMPLETED ) && (OT_PEERFLAG(&ws->peer) & PEER_FLAG_COMPLETED ) ) )
livesync_tell( ws );
Live Sync for peers that do not come back too early.
2008-12-01 17:56:53 +00:00
}
#endif
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
if( (OT_PEERFLAG(peer_dest) & PEER_FLAG_SEEDING ) && !(OT_PEERFLAG(&ws->peer) & PEER_FLAG_SEEDING ) )
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
torrent->peer_list->seed_count--;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
if( !(OT_PEERFLAG(peer_dest) & PEER_FLAG_SEEDING ) && (OT_PEERFLAG(&ws->peer) & PEER_FLAG_SEEDING ) )
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
torrent->peer_list->seed_count++;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
if( !(OT_PEERFLAG(peer_dest) & PEER_FLAG_COMPLETED ) && (OT_PEERFLAG(&ws->peer) & PEER_FLAG_COMPLETED ) ) {
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
torrent->peer_list->down_count++;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
stats_issue_event( EVENT_COMPLETED, 0, (uintptr_t)ws );
include overall completed count in stats
2009-03-24 16:55:28 +00:00
}
Renamed OT_FLAG to OT_PEERFLAG to make code easier to read Introduced READ16/32 and WRITE16/32 makros to abstract loading/storing from unaligned addresses away on cpu's that can actually load/store everywhere Removed all unnecessary memmoves, especially where it only moved 6 bytes in inner loop. I replaced them with WRITE16/32(READ16/32()) makros
2008-12-06 18:46:00 +00:00
if( OT_PEERFLAG(peer_dest) & PEER_FLAG_COMPLETED )
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
OT_PEERFLAG( &ws->peer ) |= PEER_FLAG_COMPLETED;
Kickoff
2006-12-05 12:56:56 +00:00
}
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
memcpy( peer_dest, &ws->peer, sizeof(ot_peer) );
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
#ifdef WANT_SYNC
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
if( proto == FLAG_MCA ) {
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
mutex_bucket_unlock_by_hash( *ws->hash, delta_torrentcount );
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
return 0;
}
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
#endif
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
Turn random() to nrand48 in inner loop where it is not cryptographically relevant to avoid futex storm. Thanks and credits to Bart Smienk
2021-04-20 01:48:50 +00:00
ws->reply_size = return_peers_for_torrent( ws, torrent, amount, ws->reply, proto );
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
mutex_bucket_unlock_by_hash( *ws->hash, delta_torrentcount );
return ws->reply_size;
Kickoff
2006-12-05 12:56:56 +00:00
}
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
static size_t return_peers_all( ot_peerlist *peer_list, char *reply ) {
unsigned int bucket, num_buckets = 1;
Order peers by whether they seed. This way clients can chose if they prefer leechers (at the beginning) or leechers (at the end of the list).
2009-07-14 20:51:14 +00:00
ot_vector * bucket_list = &peer_list->peers;
size_t result = OT_PEER_COMPARE_SIZE * peer_list->peer_count;
char * r_end = reply + result;
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
if( OT_PEERLIST_HASBUCKETS(peer_list) ) {
num_buckets = bucket_list->size;
bucket_list = (ot_vector *)bucket_list->data;
}
for( bucket = 0; bucket<num_buckets; ++bucket ) {
ot_peer * peers = (ot_peer*)bucket_list[bucket].data;
size_t peer_count = bucket_list[bucket].size;
Renamed OT_FLAG to OT_PEERFLAG to make code easier to read Introduced READ16/32 and WRITE16/32 makros to abstract loading/storing from unaligned addresses away on cpu's that can actually load/store everywhere Removed all unnecessary memmoves, especially where it only moved 6 bytes in inner loop. I replaced them with WRITE16/32(READ16/32()) makros
2008-12-06 18:46:00 +00:00
while( peer_count-- ) {
Order peers by whether they seed. This way clients can chose if they prefer leechers (at the beginning) or leechers (at the end of the list).
2009-07-14 20:51:14 +00:00
if( OT_PEERFLAG(peers) & PEER_FLAG_SEEDING ) {
r_end-=OT_PEER_COMPARE_SIZE;
White space
2021-04-20 01:31:58 +00:00
memcpy(r_end,peers++,OT_PEER_COMPARE_SIZE);
Order peers by whether they seed. This way clients can chose if they prefer leechers (at the beginning) or leechers (at the end of the list).
2009-07-14 20:51:14 +00:00
} else {
memcpy(reply,peers++,OT_PEER_COMPARE_SIZE);
reply+=OT_PEER_COMPARE_SIZE;
}
Renamed OT_FLAG to OT_PEERFLAG to make code easier to read Introduced READ16/32 and WRITE16/32 makros to abstract loading/storing from unaligned addresses away on cpu's that can actually load/store everywhere Removed all unnecessary memmoves, especially where it only moved 6 bytes in inner loop. I replaced them with WRITE16/32(READ16/32()) makros
2008-12-06 18:46:00 +00:00
}
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
}
Order peers by whether they seed. This way clients can chose if they prefer leechers (at the beginning) or leechers (at the end of the list).
2009-07-14 20:51:14 +00:00
return result;
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
}
Turn random() to nrand48 in inner loop where it is not cryptographically relevant to avoid futex storm. Thanks and credits to Bart Smienk
2021-04-20 01:48:50 +00:00
static size_t return_peers_selection( struct ot_workstruct *ws, ot_peerlist *peer_list, size_t amount, char *reply ) {
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
unsigned int bucket_offset, bucket_index = 0, num_buckets = 1;
ot_vector * bucket_list = &peer_list->peers;
unsigned int shifted_pc = peer_list->peer_count;
unsigned int shifted_step = 0;
unsigned int shift = 0;
Order peers by whether they seed. This way clients can chose if they prefer leechers (at the beginning) or leechers (at the end of the list).
2009-07-14 20:51:14 +00:00
size_t result = OT_PEER_COMPARE_SIZE * amount;
char * r_end = reply + result;
White space
2021-04-20 01:31:58 +00:00
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
if( OT_PEERLIST_HASBUCKETS(peer_list) ) {
num_buckets = bucket_list->size;
bucket_list = (ot_vector *)bucket_list->data;
}
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
/* Make fixpoint arithmetic as exact as possible */
#define MAXPRECBIT (1<<(8*sizeof(int)-3))
while( !(shifted_pc & MAXPRECBIT ) ) { shifted_pc <<= 1; shift++; }
shifted_step = shifted_pc/amount;
#undef MAXPRECBIT
/* Initialize somewhere in the middle of peers so that
fixpoint's aliasing doesn't alway miss the same peers */
Turn random() to nrand48 in inner loop where it is not cryptographically relevant to avoid futex storm. Thanks and credits to Bart Smienk
2021-04-20 01:48:50 +00:00
bucket_offset = nrand48(ws->rand48_state) % peer_list->peer_count;
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
while( amount-- ) {
Renamed OT_FLAG to OT_PEERFLAG to make code easier to read Introduced READ16/32 and WRITE16/32 makros to abstract loading/storing from unaligned addresses away on cpu's that can actually load/store everywhere Removed all unnecessary memmoves, especially where it only moved 6 bytes in inner loop. I replaced them with WRITE16/32(READ16/32()) makros
2008-12-06 18:46:00 +00:00
ot_peer * peer;
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
/* This is the aliased, non shifted range, next value may fall into */
unsigned int diff = ( ( ( amount + 1 ) * shifted_step ) >> shift ) -
( ( amount * shifted_step ) >> shift );
Turn random() to nrand48 in inner loop where it is not cryptographically relevant to avoid futex storm. Thanks and credits to Bart Smienk
2021-04-20 01:48:50 +00:00
bucket_offset += 1 + nrand48(ws->rand48_state) % diff;
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
while( bucket_offset >= bucket_list[bucket_index].size ) {
bucket_offset -= bucket_list[bucket_index].size;
bucket_index = ( bucket_index + 1 ) % num_buckets;
}
Renamed OT_FLAG to OT_PEERFLAG to make code easier to read Introduced READ16/32 and WRITE16/32 makros to abstract loading/storing from unaligned addresses away on cpu's that can actually load/store everywhere Removed all unnecessary memmoves, especially where it only moved 6 bytes in inner loop. I replaced them with WRITE16/32(READ16/32()) makros
2008-12-06 18:46:00 +00:00
peer = ((ot_peer*)bucket_list[bucket_index].data) + bucket_offset;
Order peers by whether they seed. This way clients can chose if they prefer leechers (at the beginning) or leechers (at the end of the list).
2009-07-14 20:51:14 +00:00
if( OT_PEERFLAG(peer) & PEER_FLAG_SEEDING ) {
r_end-=OT_PEER_COMPARE_SIZE;
White space
2021-04-20 01:31:58 +00:00
memcpy(r_end,peer,OT_PEER_COMPARE_SIZE);
Order peers by whether they seed. This way clients can chose if they prefer leechers (at the beginning) or leechers (at the end of the list).
2009-07-14 20:51:14 +00:00
} else {
memcpy(reply,peer,OT_PEER_COMPARE_SIZE);
reply+=OT_PEER_COMPARE_SIZE;
}
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
}
Order peers by whether they seed. This way clients can chose if they prefer leechers (at the beginning) or leechers (at the end of the list).
2009-07-14 20:51:14 +00:00
return result;
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
}
Documentation improved, some reindenting (again), variable types checked, unnecessary defines removed
2007-01-11 01:06:10 +00:00
/* Compiles a list of random peers for a torrent
Limit ipv6 udp replies to an amount that does not create too large UDP packets. Credits to anonymous donor
2024-03-29 23:34:28 +00:00
* Reply must have enough space to hold:
* 92 + 6 * amount bytes for TCP/IPv4
* 92 + 18 * amount bytes for TCP/IPv6
* 12 + 6 * amount bytes for UDP/IPv4
* 12 + 18 * amount bytes for UDP/IPv6
* Does not yet check not to return self
Documentation improved, some reindenting (again), variable types checked, unnecessary defines removed
2007-01-11 01:06:10 +00:00
*/
Turn random() to nrand48 in inner loop where it is not cryptographically relevant to avoid futex storm. Thanks and credits to Bart Smienk
2021-04-20 01:48:50 +00:00
size_t return_peers_for_torrent( struct ot_workstruct * ws, ot_torrent *torrent, size_t amount, char *reply, PROTO_FLAG proto ) {
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
ot_peerlist *peer_list = torrent->peer_list;
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
char *r = reply;
We do now use the tools for locking buckets in order to prepare MT code
2007-11-06 01:29:57 +00:00
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
if( amount > peer_list->peer_count )
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
amount = peer_list->peer_count;
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
Now enforce that peers do not come back before our interval by giving out a 'min interval' in responses.
2008-12-03 23:25:45 +00:00
if( proto == FLAG_TCP ) {
int erval = OT_CLIENT_REQUEST_INTERVAL_RANDOM;
We sent an incorrect reply dictionary vor v6 announces
2009-01-15 16:20:18 +00:00
r += sprintf( r, "d8:completei%zde10:downloadedi%zde10:incompletei%zde8:intervali%ie12:min intervali%ie" PEERS_BENCODED "%zd:", peer_list->seed_count, peer_list->down_count, peer_list->peer_count-peer_list->seed_count, erval, erval/2, OT_PEER_COMPARE_SIZE*amount );
Now enforce that peers do not come back before our interval by giving out a 'min interval' in responses.
2008-12-03 23:25:45 +00:00
} else {
Drop ot_{byte,word,dword} and use uint{8,16,32}_t, also simplify includes
2007-12-03 01:07:41 +00:00
*(uint32_t*)(r+0) = htonl( OT_CLIENT_REQUEST_INTERVAL_RANDOM );
Fixed an incorrect peer count in udp responses. Credits to Theron Lewis.
2009-06-28 23:11:13 +00:00
*(uint32_t*)(r+4) = htonl( peer_list->peer_count - peer_list->seed_count );
Drop ot_{byte,word,dword} and use uint{8,16,32}_t, also simplify includes
2007-12-03 01:07:41 +00:00
*(uint32_t*)(r+8) = htonl( peer_list->seed_count );
udp is working in theory, return_peers_for_torrent accepts a switch to request an udp answer. Fixed prototypes.
2007-03-06 19:43:47 +00:00
r += 12;
}
New, fixpoint distinct random algorithm for choosing peers from the list... may contain bugs. Feedback welcome
2007-01-13 19:06:39 +00:00
if( amount ) {
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
if( amount == peer_list->peer_count )
r += return_peers_all( peer_list, r );
else
Turn random() to nrand48 in inner loop where it is not cryptographically relevant to avoid futex storm. Thanks and credits to Bart Smienk
2021-04-20 01:48:50 +00:00
r += return_peers_selection( ws, peer_list, amount, r );
Kickoff
2006-12-05 12:56:56 +00:00
}
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
added live sync code added a config file parser added tracker id changed WANT_CLOSED_TRACKER and WANT_BLACKLIST into WANT_ACCESS_WHITE and WANT_ACCESS_BLACK changed WANT_TRACKER_SYNC to WANT_SYNC_BATCH and added WANT_SYNC_LIVE added an option to switch off fullscrapes cleaned up many internal hardcoded values, like PROTO_FLAG,
2008-10-04 05:40:51 +00:00
if( proto == FLAG_TCP )
udp is working in theory, return_peers_for_torrent accepts a switch to request an udp answer. Fixed prototypes.
2007-03-06 19:43:47 +00:00
*r++ = 'e';
announce now returns all mandatory answer dictionary entries
2006-12-14 02:54:08 +00:00
Should work now *cough* *cough*
2006-12-08 21:36:26 +00:00
return r - reply;
Kickoff
2006-12-05 12:56:56 +00:00
}
Documentation improved, some reindenting (again), variable types checked, unnecessary defines removed
2007-01-11 01:06:10 +00:00
/* Fetches scrape info for a specific torrent */
V6
2009-01-13 22:41:17 +00:00
size_t return_udp_scrape_for_torrent( ot_hash hash, char *reply ) {
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
int exactmatch, delta_torrentcount = 0;
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
ot_vector *torrents_list = mutex_bucket_lock_by_hash( hash );
support for udp scrape
2007-03-16 22:46:41 +00:00
ot_torrent *torrent = binary_search( hash, torrents_list->data, torrents_list->size, sizeof( ot_torrent ), OT_HASH_COMPARE_SIZE, &exactmatch );
if( !exactmatch ) {
memset( reply, 0, 12);
endianess fix for udp scrape
2007-03-17 01:24:56 +00:00
} else {
Drop ot_{byte,word,dword} and use uint{8,16,32}_t, also simplify includes
2007-12-03 01:07:41 +00:00
uint32_t *r = (uint32_t*) reply;
support for udp scrape
2007-03-16 22:46:41 +00:00
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
if( clean_single_torrent( torrent ) ) {
vector_remove_torrent( torrents_list, torrent );
memset( reply, 0, 12);
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
delta_torrentcount = -1;
Number of buckets is 1024 now Clean all torrents now only cleans one bucket at a time All torrents that are being worked upon in an announce are being cleaned on demoand torrent's peer lists now keep extra counts for seeds and peers to speed up scrape and announce Sync has gone for now. I will think up a new way to implement. The old one was way to slow.
2007-10-27 14:06:07 +00:00
} else {
r[0] = htonl( torrent->peer_list->seed_count );
r[1] = htonl( torrent->peer_list->down_count );
r[2] = htonl( torrent->peer_list->peer_count-torrent->peer_list->seed_count );
support for udp scrape
2007-03-16 22:46:41 +00:00
}
}
Forgot to notify central counter about torrent removal
2009-02-20 11:07:30 +00:00
mutex_bucket_unlock_by_hash( hash, delta_torrentcount );
support for udp scrape
2007-03-16 22:46:41 +00:00
return 12;
}
/* Fetches scrape info for a specific torrent */
Introducing multiscrape
2007-10-19 21:56:59 +00:00
size_t return_tcp_scrape_for_torrent( ot_hash *hash_list, int amount, char *reply ) {
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
char *r = reply;
int exactmatch, i;
Now actually seems to work for the most parts Added scraping Added graceful disconnect
2006-12-14 02:44:50 +00:00
Introducing multiscrape
2007-10-19 21:56:59 +00:00
r += sprintf( r, "d5:filesd" );
Now actually seems to work for the most parts Added scraping Added graceful disconnect
2006-12-14 02:44:50 +00:00
Introducing multiscrape
2007-10-19 21:56:59 +00:00
for( i=0; i<amount; ++i ) {
Reset delta_torrentcount to zero on each pass
2009-02-20 12:31:55 +00:00
int delta_torrentcount = 0;
Introducing multiscrape
2007-10-19 21:56:59 +00:00
ot_hash *hash = hash_list + i;
V6
2009-01-13 22:41:17 +00:00
ot_vector *torrents_list = mutex_bucket_lock_by_hash( *hash );
Introducing multiscrape
2007-10-19 21:56:59 +00:00
ot_torrent *torrent = binary_search( hash, torrents_list->data, torrents_list->size, sizeof( ot_torrent ), OT_HASH_COMPARE_SIZE, &exactmatch );
Now actually seems to work for the most parts Added scraping Added graceful disconnect
2006-12-14 02:44:50 +00:00
We do now use the tools for locking buckets in order to prepare MT code
2007-11-06 01:29:57 +00:00
if( exactmatch ) {
if( clean_single_torrent( torrent ) ) {
vector_remove_torrent( torrents_list, torrent );
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
delta_torrentcount = -1;
We do now use the tools for locking buckets in order to prepare MT code
2007-11-06 01:29:57 +00:00
} else {
Renamed OT_FLAG to OT_PEERFLAG to make code easier to read Introduced READ16/32 and WRITE16/32 makros to abstract loading/storing from unaligned addresses away on cpu's that can actually load/store everywhere Removed all unnecessary memmoves, especially where it only moved 6 bytes in inner loop. I replaced them with WRITE16/32(READ16/32()) makros
2008-12-06 18:46:00 +00:00
*r++='2';*r++='0';*r++=':';
V6
2009-01-13 22:41:17 +00:00
memcpy( r, hash, sizeof(ot_hash) ); r+=sizeof(ot_hash);
Renamed OT_FLAG to OT_PEERFLAG to make code easier to read Introduced READ16/32 and WRITE16/32 makros to abstract loading/storing from unaligned addresses away on cpu's that can actually load/store everywhere Removed all unnecessary memmoves, especially where it only moved 6 bytes in inner loop. I replaced them with WRITE16/32(READ16/32()) makros
2008-12-06 18:46:00 +00:00
r += sprintf( r, "d8:completei%zde10:downloadedi%zde10:incompletei%zdee",
torrent->peer_list->seed_count, torrent->peer_list->down_count, torrent->peer_list->peer_count-torrent->peer_list->seed_count );
We do now use the tools for locking buckets in order to prepare MT code
2007-11-06 01:29:57 +00:00
}
Introducing multiscrape
2007-10-19 21:56:59 +00:00
}
V6
2009-01-13 22:41:17 +00:00
mutex_bucket_unlock_by_hash( *hash, delta_torrentcount );
Reenabled syncing
2007-10-29 17:22:05 +00:00
}
Added outbound part of sync Proposed format: d4:syncd20:<info_hash>8*N:(xxxxyy)*Nee, therefore had to refactor torrent cleanup now that it will hit all torrents once every OT_POOL_TIMEOUT units.
2007-03-27 12:07:29 +00:00
Reenabled syncing
2007-10-29 17:22:05 +00:00
*r++ = 'e'; *r++ = 'e';
New stats for s24s code, this is debug, do not use in real world
2007-09-26 16:49:13 +00:00
return r - reply;
}
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
static ot_peerlist dummy_list;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
size_t remove_peer_from_torrent( PROTO_FLAG proto, struct ot_workstruct *ws ) {
An announce with event=stopped now returns correct number of leechers and seeders. In TCP and UDP.
2007-09-10 02:43:11 +00:00
int exactmatch;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
ot_vector *torrents_list = mutex_bucket_lock_by_hash( *ws->hash );
ot_torrent *torrent = binary_search( ws->hash, torrents_list->data, torrents_list->size, sizeof( ot_torrent ), OT_HASH_COMPARE_SIZE, &exactmatch );
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
ot_peerlist *peer_list = &dummy_list;
account downloaded before early returns
2007-01-27 16:06:13 +00:00
Whitespace fixes
2008-10-28 01:27:22 +00:00
#ifdef WANT_SYNC_LIVE
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
if( proto != FLAG_MCA ) {
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
OT_PEERFLAG( &ws->peer ) |= PEER_FLAG_STOPPED;
livesync_tell( ws );
An announce with event=stopped now returns correct number of leechers and seeders. In TCP and UDP.
2007-09-10 02:43:11 +00:00
}
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
#endif
An announce with event=stopped now returns correct number of leechers and seeders. In TCP and UDP.
2007-09-10 02:43:11 +00:00
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
if( exactmatch ) {
peer_list = torrent->peer_list;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
switch( vector_remove_peer( &peer_list->peers, &ws->peer ) ) {
fix implicit fallthrough spelling This commit fixes the syntax of the implicit fallthrough comments, in order to be matched by GCC (and probably other compilers as well) with the following regular expression: [ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )? fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)? See: https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#index-Wimplicit-fallthrough=
2021-08-03 11:59:09 +00:00
case 2: peer_list->seed_count--; /* Intentional fallthrough */
case 1: peer_list->peer_count--; /* Intentional fallthrough */
The BIG refactoring [tm]. Too many changes to count them. If it doesn't suite you, revert to last version.
2008-11-28 22:21:10 +00:00
default: break;
Now actually seems to work for the most parts Added scraping Added graceful disconnect
2006-12-14 02:44:50 +00:00
}
An announce with event=stopped now returns correct number of leechers and seeders. In TCP and UDP.
2007-09-10 02:43:11 +00:00
}
Now enforce that peers do not come back before our interval by giving out a 'min interval' in responses.
2008-12-03 23:25:45 +00:00
if( proto == FLAG_TCP ) {
int erval = OT_CLIENT_REQUEST_INTERVAL_RANDOM;
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
ws->reply_size = sprintf( ws->reply, "d8:completei%zde10:incompletei%zde8:intervali%ie12:min intervali%ie" PEERS_BENCODED "0:e", peer_list->seed_count, peer_list->peer_count - peer_list->seed_count, erval, erval / 2 );
Now enforce that peers do not come back before our interval by giving out a 'min interval' in responses.
2008-12-03 23:25:45 +00:00
}
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
added live sync code added a config file parser added tracker id changed WANT_CLOSED_TRACKER and WANT_BLACKLIST into WANT_ACCESS_WHITE and WANT_ACCESS_BLACK changed WANT_TRACKER_SYNC to WANT_SYNC_BATCH and added WANT_SYNC_LIVE added an option to switch off fullscrapes cleaned up many internal hardcoded values, like PROTO_FLAG,
2008-10-04 05:40:51 +00:00
/* Handle UDP reply */
Bugfix: overwrote bencoded data with udp reply data when sending an event=stopped
2008-10-06 19:43:52 +00:00
if( proto == FLAG_UDP ) {
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
((uint32_t*)ws->reply)[2] = htonl( OT_CLIENT_REQUEST_INTERVAL_RANDOM );
((uint32_t*)ws->reply)[3] = htonl( peer_list->peer_count - peer_list->seed_count );
((uint32_t*)ws->reply)[4] = htonl( peer_list->seed_count);
ws->reply_size = 20;
added live sync code added a config file parser added tracker id changed WANT_CLOSED_TRACKER and WANT_BLACKLIST into WANT_ACCESS_WHITE and WANT_ACCESS_BLACK changed WANT_TRACKER_SYNC to WANT_SYNC_BATCH and added WANT_SYNC_LIVE added an option to switch off fullscrapes cleaned up many internal hardcoded values, like PROTO_FLAG,
2008-10-04 05:40:51 +00:00
}
We do now use the tools for locking buckets in order to prepare MT code
2007-11-06 01:29:57 +00:00
** struct ot_workstruct gets ritcher (and will become even ritcher soon). This is where we encapsulate all per-request data from peer to hash to peer_id, so that it is available everywhere without passing hundreds of pointers down the stack. Most functions that do work down the stack now accept an ot_workstruct and some flags. So it can end up in the stats/event-handler where it will be the default parameter in the future. ** peer_id is now being copied by default and moved to ot_workstruct So it is available in stats and subsequent functions. ** sync scrape madness is gone SYNC_SCRAPE was intended to sync tracker state that would normally be lost on restarts i.e. downloaded counts per torrent. The way was to push it in the tracker cloud after finding all neighbouring trackers. This is madness. It never was tested and can be done per tracker by fetching stats/mode=statedump from time to time and starting opentracker with the -l option later. ** livesync thread has its own ot_workstruct now So it can behave like ot_udp and ot_http against trackerlogic.c and get rid of the first half of the embarrassing global variables. The sending half will be fixed soon [tm]. ** stats can log completed events The author recognizes the needs of original content distributors to keep track of the amount of times a work has been downloaded. While not feasible and used on openbittorrent and other open and anonymous tracker installations, a tracker user can now choose to send those events to syslog.
2010-04-22 22:08:42 +00:00
mutex_bucket_unlock_by_hash( *ws->hash, 0 );
return ws->reply_size;
Now actually seems to work for the most parts Added scraping Added graceful disconnect
2006-12-14 02:44:50 +00:00
}
Add a stat option the count bucket stalls Add an iterator for all torrents. It's slower but for stats it's okay. Move some stats collection stuff to the new iterator. More to come. Start a "report all stats" page. Start fixing the code to identify "busy" networks. Add the concept of hosts allowed to proxy. Add a parser for the X-Forwarded-For: HTTP header. Clean up HTTP Header handling code. (Remove some left overs of now vanished sync code).
2009-03-04 14:35:21 +00:00
void iterate_all_torrents( int (*for_each)( ot_torrent* torrent, uintptr_t data ), uintptr_t data ) {
int bucket;
size_t j;
for( bucket=0; bucket<OT_BUCKET_COUNT; ++bucket ) {
ot_vector *torrents_list = mutex_bucket_lock( bucket );
ot_torrent *torrents = (ot_torrent*)(torrents_list->data);
for( j=0; j<torrents_list->size; ++j )
if( for_each( torrents + j, data ) )
break;
mutex_bucket_unlock( bucket, 0 );
if( !g_opentracker_running ) return;
}
}
added live sync code added a config file parser added tracker id changed WANT_CLOSED_TRACKER and WANT_BLACKLIST into WANT_ACCESS_WHITE and WANT_ACCESS_BLACK changed WANT_TRACKER_SYNC to WANT_SYNC_BATCH and added WANT_SYNC_LIVE added an option to switch off fullscrapes cleaned up many internal hardcoded values, like PROTO_FLAG,
2008-10-04 05:40:51 +00:00
void exerr( char * message ) {
fprintf( stderr, "%s\n", message );
exit( 111 );
}
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
void trackerlogic_init( ) {
added live sync code added a config file parser added tracker id changed WANT_CLOSED_TRACKER and WANT_BLACKLIST into WANT_ACCESS_WHITE and WANT_ACCESS_BLACK changed WANT_TRACKER_SYNC to WANT_SYNC_BATCH and added WANT_SYNC_LIVE added an option to switch off fullscrapes cleaned up many internal hardcoded values, like PROTO_FLAG,
2008-10-04 05:40:51 +00:00
g_tracker_id = random();
Whitespace fixes
2008-10-28 01:27:22 +00:00
Allow /stats to be located anywhere in your trackers path
2009-06-17 15:06:31 +00:00
if( !g_stats_path )
g_stats_path = "stats";
g_stats_path_len = strlen( g_stats_path );
Dont ruin the srandomdev initialisation in trackerlogic's init
2012-06-03 00:15:59 +00:00
Cleanup now moved to its own thread, too
2007-11-19 21:10:53 +00:00
/* Initialise background worker threads */
We do now use the tools for locking buckets in order to prepare MT code
2007-11-06 01:29:57 +00:00
mutex_init( );
Cleanup now moved to its own thread, too
2007-11-19 21:10:53 +00:00
clean_init( );
fullscrape_init( );
added live sync code added a config file parser added tracker id changed WANT_CLOSED_TRACKER and WANT_BLACKLIST into WANT_ACCESS_WHITE and WANT_ACCESS_BLACK changed WANT_TRACKER_SYNC to WANT_SYNC_BATCH and added WANT_SYNC_LIVE added an option to switch off fullscrapes cleaned up many internal hardcoded values, like PROTO_FLAG,
2008-10-04 05:40:51 +00:00
accesslist_init( );
livesync_init( );
Drop ot_{byte,word,dword} and use uint{8,16,32}_t, also simplify includes
2007-12-03 01:07:41 +00:00
stats_init( );
Kickoff
2006-12-05 12:56:56 +00:00
}
We do now use the tools for locking buckets in order to prepare MT code
2007-11-06 01:29:57 +00:00
void trackerlogic_deinit( void ) {
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
int bucket, delta_torrentcount = 0;
Introduce another unspec'd command: sync which syncs the most recent peer pool for a torrent out. Do some variable type clean up. Do some code structure clean up.
2007-01-26 16:26:49 +00:00
size_t j;
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
2009-03-17 23:57:20 +00:00
Documentation improved, some reindenting (again), variable types checked, unnecessary defines removed
2007-01-11 01:06:10 +00:00
/* Free all torrents... */
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
for(bucket=0; bucket<OT_BUCKET_COUNT; ++bucket ) {
ot_vector *torrents_list = mutex_bucket_lock( bucket );
if( torrents_list->size ) {
for( j=0; j<torrents_list->size; ++j ) {
ot_torrent *torrent = ((ot_torrent*)(torrents_list->data)) + j;
free_peerlist( torrent->peer_list );
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
delta_torrentcount -= 1;
No one can get access to buckets now without locking them. Also split up the trackerlogic.c-monster in functional sub-units. HEADS UP: this code is untested and not considered stable.
2007-11-06 11:58:32 +00:00
}
free( torrents_list->data );
This is a complete rewrite... assume nothing works.
2006-12-12 02:37:19 +00:00
}
* opentracker now drops permissions in correct order and really chroots() if ran as root * lock passing between add_peer_to_torrent and return_peers_for_torrent is now avoided by providing a more general add_peer_to_torrent_and_return_peers function that can be used with NULL parameters to not return any peers (in sync case) * in order to keep a fast overview how many torrents opentracker maintains, every mutex_bucket_unlock operation expects an additional integer parameter that tells ot_mutex.c how many torrents have been added or removed. A function mutex_get_torrent_count has been introduced.
2009-01-02 08:57:53 +00:00
mutex_bucket_unlock( bucket, delta_torrentcount );
This is a complete rewrite... assume nothing works.
2006-12-12 02:37:19 +00:00
}
Cleanup now moved to its own thread, too
2007-11-19 21:10:53 +00:00
Handle program end more politely
2008-12-07 03:50:51 +00:00
/* Deinitialise background worker threads */
stats_deinit( );
livesync_deinit( );
accesslist_deinit( );
fullscrape_deinit( );
clean_deinit( );
release pools only after all other services have been stopped.
2008-10-17 15:21:47 +00:00
/* Release mutexes */
Cleanup now moved to its own thread, too
2007-11-19 21:10:53 +00:00
mutex_deinit( );
Fixed warnings about newlines at end of file
2007-11-06 12:03:11 +00:00
}
Introduce some kind of versioning
2007-12-20 05:59:34 +00:00
const char *g_version_trackerlogic_c = "$Source$: $Revision$\n";
Reference in New Issue Copy Permalink