mirror
/
opentracker
mirror of git://erdgeist.org/opentracker
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
master
erdgeist-fix-tasklists
dynamic-accesslists
OPENTRACKER_WITH_BATCHSYNC
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from '20955311d1'
${ noResults }
opentracker/ot_http.c

510 lines
19 KiB
C
Raw Normal View History Unescape Escape

Move http handling to its own sources
17 years ago
/* This software was written by Dirk Engling <erdgeist@erdgeist.org>
Introduce some kind of versioning
17 years ago
It is considered beerware. Prost. Skol. Cheers or whatever.
Whitespace fixes
16 years ago
Introduce some kind of versioning
17 years ago
$id$ */
Move http handling to its own sources
17 years ago
/* System */
#include <sys/types.h>
Avoid warnings about unused variables without certain defines
17 years ago
#include <arpa/inet.h>
Move http handling to its own sources
17 years ago
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
/* Libowfat */
#include "byte.h"
#include "array.h"
#include "iob.h"
V6
16 years ago
#include "ip6.h"
Move http handling to its own sources
17 years ago
/* Opentracker */
#include "trackerlogic.h"
#include "ot_mutex.h"
#include "ot_http.h"
#include "ot_iovec.h"
#include "scan_urlencoded_query.h"
#include "ot_fullscrape.h"
#include "ot_stats.h"
#include "ot_accesslist.h"
#define OT_MAXMULTISCRAPE_COUNT 64
/ can now be redirected to any URL
17 years ago
extern char *g_redirecturl;
Move http handling to its own sources
17 years ago
Allow /stats to be located anywhere in your trackers path
16 years ago
char *g_stats_path;
ssize_t g_stats_path_len;
Move http handling to its own sources
17 years ago
enum {
SUCCESS_HTTP_HEADER_LENGTH = 80,
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
SUCCESS_HTTP_HEADER_LENGTH_CONTENT_ENCODING = 32,
Move http handling to its own sources
17 years ago
SUCCESS_HTTP_SIZE_OFF = 17 };
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
static void http_senddata( const int64 sock, struct ot_workstruct *ws ) {
struct http_data *cookie = io_getcookie( sock );
Move http handling to its own sources
17 years ago
ssize_t written_size;
/* whoever sends data is not interested in its input-array */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
if( cookie && ( cookie->flag & STRUCT_HTTP_FLAG_ARRAY_USED ) ) {
cookie->flag &= ~STRUCT_HTTP_FLAG_ARRAY_USED;
array_reset( &cookie->data.request );
Move http handling to its own sources
17 years ago
}
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
written_size = write( sock, ws->reply, ws->reply_size );
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( ( written_size < 0 ) || ( written_size == ws->reply_size ) ) {
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
free( cookie ); io_close( sock );
Move http handling to its own sources
17 years ago
} else {
char * outbuf;
tai6464 t;
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
if( !cookie ) return;
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( !( outbuf = malloc( ws->reply_size - written_size ) ) ) {
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
free(cookie); io_close( sock );
Move http handling to its own sources
17 years ago
return;
}
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
iob_reset( &cookie->data.batch );
V6
16 years ago
memcpy( outbuf, ws->reply + written_size, ws->reply_size - written_size );
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
iob_addbuf_free( &cookie->data.batch, outbuf, ws->reply_size - written_size );
cookie->flag |= STRUCT_HTTP_FLAG_IOB_USED;
Move http handling to its own sources
17 years ago
/* writeable short data sockets just have a tcp timeout */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
taia_uint( &t, 0 ); io_timeout( sock, t );
io_dontwantread( sock );
io_wantwrite( sock );
Move http handling to its own sources
17 years ago
}
}
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
#define HTTPERROR_302 return http_issue_error( sock, ws, CODE_HTTPERROR_302 )
#define HTTPERROR_400 return http_issue_error( sock, ws, CODE_HTTPERROR_400 )
#define HTTPERROR_400_PARAM return http_issue_error( sock, ws, CODE_HTTPERROR_400_PARAM )
#define HTTPERROR_400_COMPACT return http_issue_error( sock, ws, CODE_HTTPERROR_400_COMPACT )
#define HTTPERROR_400_DOUBLEHASH return http_issue_error( sock, ws, CODE_HTTPERROR_400_PARAM )
#define HTTPERROR_403_IP return http_issue_error( sock, ws, CODE_HTTPERROR_403_IP )
#define HTTPERROR_404 return http_issue_error( sock, ws, CODE_HTTPERROR_404 )
#define HTTPERROR_500 return http_issue_error( sock, ws, CODE_HTTPERROR_500 )
ssize_t http_issue_error( const int64 sock, struct ot_workstruct *ws, int code ) {
/ can now be redirected to any URL
17 years ago
char *error_code[] = { "302 Found", "400 Invalid Request", "400 Invalid Request", "400 Invalid Request",
Count invalid requests
17 years ago
"403 Access Denied", "404 Not Found", "500 Internal Server Error" };
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
char *title = error_code[code];
/ can now be redirected to any URL
17 years ago
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
ws->reply = ws->outbuf;
/ can now be redirected to any URL
17 years ago
if( code == CODE_HTTPERROR_302 )
V6
16 years ago
ws->reply_size = snprintf( ws->reply, G_OUTBUF_SIZE, "HTTP/1.0 302 Found\r\nContent-Length: 0\r\nLocation: %s\r\n\r\n", g_redirecturl );
/ can now be redirected to any URL
17 years ago
else
V6
16 years ago
ws->reply_size = snprintf( ws->reply, G_OUTBUF_SIZE, "HTTP/1.0 %s\r\nContent-Type: text/html\r\nConnection: close\r\nContent-Length: %zd\r\n\r\n<title>%s</title>\n", title, strlen(title)+16-4,title+4);
Count invalid requests
17 years ago
Move http handling to its own sources
17 years ago
#ifdef _DEBUG_HTTPERROR
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
fprintf( stderr, "DEBUG: invalid request was: %s\n", ws->debugbuf );
Move http handling to its own sources
17 years ago
#endif
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,
16 years ago
stats_issue_event( EVENT_FAILED, FLAG_TCP, code );
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
http_senddata( sock, ws );
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
return ws->reply_size = -2;
Move http handling to its own sources
17 years ago
}
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
ssize_t http_sendiovecdata( const int64 sock, struct ot_workstruct *ws, int iovec_entries, struct iovec *iovector ) {
struct http_data *cookie = io_getcookie( sock );
Move http handling to its own sources
17 years ago
char *header;
int i;
size_t header_size, size = iovec_length( &iovec_entries, &iovector );
tai6464 t;
/* No cookie? Bad socket. Leave. */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
if( !cookie ) {
Move http handling to its own sources
17 years ago
iovec_free( &iovec_entries, &iovector );
HTTPERROR_500;
}
/* If this socket collected request in a buffer,
free it now */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
if( cookie->flag & STRUCT_HTTP_FLAG_ARRAY_USED ) {
cookie->flag &= ~STRUCT_HTTP_FLAG_ARRAY_USED;
array_reset( &cookie->data.request );
Move http handling to its own sources
17 years ago
}
/* If we came here, wait for the answer is over */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
cookie->flag &= ~STRUCT_HTTP_FLAG_WAITINGFORTASK;
Move http handling to its own sources
17 years ago
/* Our answers never are 0 vectors. Return an error. */
if( !iovec_entries ) {
HTTPERROR_500;
}
/* Prepare space for http header */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
header = malloc( SUCCESS_HTTP_HEADER_LENGTH + SUCCESS_HTTP_HEADER_LENGTH_CONTENT_ENCODING );
Move http handling to its own sources
17 years ago
if( !header ) {
iovec_free( &iovec_entries, &iovector );
HTTPERROR_500;
}
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
if( cookie->flag & STRUCT_HTTP_FLAG_GZIP )
Move http handling to its own sources
17 years ago
header_size = sprintf( header, "HTTP/1.0 200 OK\r\nContent-Type: text/plain\r\nContent-Encoding: gzip\r\nContent-Length: %zd\r\n\r\n", size );
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
else if( cookie->flag & STRUCT_HTTP_FLAG_BZIP2 )
Move http handling to its own sources
17 years ago
header_size = sprintf( header, "HTTP/1.0 200 OK\r\nContent-Type: text/plain\r\nContent-Encoding: bzip2\r\nContent-Length: %zd\r\n\r\n", size );
else
header_size = sprintf( header, "HTTP/1.0 200 OK\r\nContent-Type: text/plain\r\nContent-Length: %zd\r\n\r\n", size );
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
iob_reset( &cookie->data.batch );
iob_addbuf_free( &cookie->data.batch, header, header_size );
Move http handling to its own sources
17 years ago
/* Will move to ot_iovec.c */
for( i=0; i<iovec_entries; ++i )
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
iob_addbuf_munmap( &cookie->data.batch, iovector[i].iov_base, iovector[i].iov_len );
Move http handling to its own sources
17 years ago
free( iovector );
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
cookie->flag |= STRUCT_HTTP_FLAG_IOB_USED;
Move http handling to its own sources
17 years ago
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
/* writeable sockets timeout after 10 minutes */
Move http handling to its own sources
17 years ago
taia_now( &t ); taia_addsec( &t, &t, OT_CLIENT_TIMEOUT_SEND );
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
io_timeout( sock, t );
io_dontwantread( sock );
io_wantwrite( sock );
Move http handling to its own sources
17 years ago
return 0;
}
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
static ssize_t http_handle_stats( const int64 sock, struct ot_workstruct *ws, char *read_ptr ) {
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
static const ot_keywords keywords_main[] =
{ { "mode", 1 }, {"format", 2 }, { NULL, -3 } };
static const ot_keywords keywords_mode[] =
Forgot to copy the stats?mode=tcp4 parameter
16 years ago
{ { "peer", TASK_STATS_PEERS }, { "conn", TASK_STATS_CONNS }, { "scrp", TASK_STATS_SCRAPE }, { "udp4", TASK_STATS_UDP }, { "tcp4", TASK_STATS_TCP },
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
{ "busy", TASK_STATS_BUSY_NETWORKS }, { "torr", TASK_STATS_TORRENTS }, { "fscr", TASK_STATS_FULLSCRAPE },
include overall completed count in stats
16 years ago
{ "s24s", TASK_STATS_SLASH24S }, { "tpbs", TASK_STATS_TPB }, { "herr", TASK_STATS_HTTPERRORS }, { "completed", TASK_STATS_COMPLETED },
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
{ "top10", TASK_STATS_TOP10 }, { "renew", TASK_STATS_RENEW }, { "syncs", TASK_STATS_SYNCS }, { "version", TASK_STATS_VERSION },
Fix statedump
16 years ago
{ "everything", TASK_STATS_EVERYTHING }, { "statedump", TASK_FULLSCRAPE_TRACKERSTATE }, { NULL, -3 } };
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
static const ot_keywords keywords_format[] =
{ { "bin", TASK_FULLSCRAPE_TPB_BINARY }, { "ben", TASK_FULLSCRAPE }, { "url", TASK_FULLSCRAPE_TPB_URLENCODED },
{ "txt", TASK_FULLSCRAPE_TPB_ASCII }, { NULL, -3 } };
Move http handling to its own sources
17 years ago
int mode = TASK_STATS_PEERS, scanon = 1, format = 0;
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
Tidy up stats task.
16 years ago
#ifdef WANT_RESTRICT_STATS
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
struct http_data *cookie = io_getcookie( sock );
Tidy up stats task.
16 years ago
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
if( !cookie || !accesslist_isblessed( cookie->ip, OT_PERMISSION_MAY_STAT ) )
Tidy up stats task.
16 years ago
HTTPERROR_403_IP;
#endif
Move http handling to its own sources
17 years ago
while( scanon ) {
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
switch( scan_find_keywords( keywords_main, &read_ptr, SCAN_SEARCHPATH_PARAM ) ) {
Move http handling to its own sources
17 years ago
case -2: scanon = 0; break; /* TERMINATOR */
case -1: HTTPERROR_400_PARAM; /* PARSE ERROR */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
case -3: scan_urlencoded_skipvalue( &read_ptr ); break;
case 1: /* matched "mode" */
if( ( mode = scan_find_keywords( keywords_mode, &read_ptr, SCAN_SEARCHPATH_VALUE ) ) <= 0 ) HTTPERROR_400_PARAM;
Format for full scrapes was not parsed correctly.
16 years ago
break;
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
case 2: /* matched "format" */
if( ( format = scan_find_keywords( keywords_format, &read_ptr, SCAN_SEARCHPATH_VALUE ) ) <= 0 ) HTTPERROR_400_PARAM;
Move http handling to its own sources
17 years ago
break;
}
}
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,
16 years ago
#ifdef WANT_FULLSCRAPE
Move http handling to its own sources
17 years ago
if( mode == TASK_STATS_TPB ) {
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
struct http_data* cookie = io_getcookie( sock );
Move http handling to its own sources
17 years ago
tai6464 t;
#ifdef WANT_COMPRESSION_GZIP
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
ws->request[ws->request_size] = 0;
if( strstr( read_ptr - 1, "gzip" ) ) {
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
cookie->flag |= STRUCT_HTTP_FLAG_GZIP;
Move http handling to its own sources
17 years ago
format |= TASK_FLAG_GZIP;
}
#endif
/* Pass this task to the worker thread */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
cookie->flag |= STRUCT_HTTP_FLAG_WAITINGFORTASK;
Move http handling to its own sources
17 years ago
/* Clients waiting for us should not easily timeout */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
taia_uint( &t, 0 ); io_timeout( sock, t );
fullscrape_deliver( sock, format );
io_dontwantread( sock );
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
return ws->reply_size = -2;
Move http handling to its own sources
17 years ago
}
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,
16 years ago
#endif
Move http handling to its own sources
17 years ago
Avoid warnings about unused variables without certain defines
17 years ago
/* default format for now */
Tidy up stats task.
16 years ago
if( ( mode & TASK_CLASS_MASK ) == TASK_STATS ) {
tai6464 t;
/* Complex stats also include expensive memory debugging tools */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
taia_uint( &t, 0 ); io_timeout( sock, t );
stats_deliver( sock, mode );
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
return ws->reply_size = -2;
Tidy up stats task.
16 years ago
}
Whitespace fixes
16 years ago
Tidy up stats task.
16 years ago
/* Simple stats can be answerred immediately */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( !( ws->reply_size = return_stats_for_tracker( ws->reply, mode, 0 ) ) ) HTTPERROR_500;
Tidy up stats task.
16 years ago
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
return ws->reply_size;
Move http handling to its own sources
17 years ago
}
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,
16 years ago
#ifdef WANT_FULLSCRAPE
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
static ssize_t http_handle_fullscrape( const int64 sock, struct ot_workstruct *ws ) {
struct http_data* cookie = io_getcookie( sock );
Move http handling to its own sources
17 years ago
int format = 0;
tai6464 t;
#ifdef WANT_COMPRESSION_GZIP
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
ws->request[ws->request_size-1] = 0;
if( strstr( ws->request, "gzip" ) ) {
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
cookie->flag |= STRUCT_HTTP_FLAG_GZIP;
Move http handling to its own sources
17 years ago
format = TASK_FLAG_GZIP;
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
stats_issue_event( EVENT_FULLSCRAPE_REQUEST_GZIP, 0, (uintptr_t)cookie->ip );
Move http handling to its own sources
17 years ago
} else
#endif
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
stats_issue_event( EVENT_FULLSCRAPE_REQUEST, 0, (uintptr_t)cookie->ip );
Move http handling to its own sources
17 years ago
#ifdef _DEBUG_HTTPERROR
Fix a pointer arithmetic issue leading to incorrect peer data being copied. Allocate correct size for debugbuffer. Expect exact values on find keywords, not only prefix match.
16 years ago
fprintf( stderr, "%s", ws->debugbuf );
Move http handling to its own sources
17 years ago
#endif
/* Pass this task to the worker thread */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
cookie->flag |= STRUCT_HTTP_FLAG_WAITINGFORTASK;
Move http handling to its own sources
17 years ago
/* Clients waiting for us should not easily timeout */
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
taia_uint( &t, 0 ); io_timeout( sock, t );
fullscrape_deliver( sock, TASK_FULLSCRAPE | format );
io_dontwantread( sock );
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
return ws->reply_size = -2;
Move http handling to its own sources
17 years ago
}
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,
16 years ago
#endif
V6
16 years ago
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
static ssize_t http_handle_scrape( const int64 sock, struct ot_workstruct *ws, char *read_ptr ) {
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
static const ot_keywords keywords_scrape[] = { { "info_hash", 1 }, { NULL, -3 } };
Move http handling to its own sources
17 years ago
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
ot_hash * multiscrape_buf = (ot_hash*)ws->request;
Move http handling to its own sources
17 years ago
int scanon = 1, numwant = 0;
/* This is to hack around stupid clients that send "scrape ?info_hash" */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( read_ptr[-1] != '?' ) {
while( ( *read_ptr != '?' ) && ( *read_ptr != '\n' ) ) ++read_ptr;
if( *read_ptr == '\n' ) HTTPERROR_400_PARAM;
++read_ptr;
Move http handling to its own sources
17 years ago
}
while( scanon ) {
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
switch( scan_find_keywords( keywords_scrape, &read_ptr, SCAN_SEARCHPATH_PARAM ) ) {
Move http handling to its own sources
17 years ago
case -2: scanon = 0; break; /* TERMINATOR */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
default: HTTPERROR_400_PARAM; /* PARSE ERROR */
case -3: scan_urlencoded_skipvalue( &read_ptr ); break;
case 1: /* matched "info_hash" */
Move http handling to its own sources
17 years ago
/* ignore this, when we have less than 20 bytes */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( scan_urlencoded_query( &read_ptr, (char*)(multiscrape_buf + numwant++), SCAN_SEARCHPATH_VALUE ) != (ssize_t)sizeof(ot_hash) )
Move http handling to its own sources
17 years ago
HTTPERROR_400_PARAM;
break;
}
}
/* No info_hash found? Inform user */
if( !numwant ) HTTPERROR_400_PARAM;
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
16 years ago
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
/* Limit number of hashes to process */
if( numwant > OT_MAXMULTISCRAPE_COUNT )
numwant = OT_MAXMULTISCRAPE_COUNT;
Move http handling to its own sources
17 years ago
/* Enough for http header + whole scrape string */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( !( ws->reply_size = return_tcp_scrape_for_torrent( multiscrape_buf, numwant, ws->reply ) ) ) HTTPERROR_500;
stats_issue_event( EVENT_SCRAPE, FLAG_TCP, ws->reply_size );
return ws->reply_size;
Move http handling to its own sources
17 years ago
}
Fix a pointer arithmetic issue leading to incorrect peer data being copied. Allocate correct size for debugbuffer. Expect exact values on find keywords, not only prefix match.
16 years ago
static ot_keywords keywords_announce[] = { { "port", 1 }, { "left", 2 }, { "event", 3 }, { "numwant", 4 }, { "compact", 5 }, { "compact6", 5 }, { "info_hash", 6 },
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
#ifdef WANT_IP_FROM_QUERY_STRING
{ "ip", 7 },
#endif
{ NULL, -3 } };
static ot_keywords keywords_announce_event[] = { { "completed", 1 }, { "stopped", 2 }, { NULL, -3 } };
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
static ssize_t http_handle_announce( const int64 sock, struct ot_workstruct *ws, char *read_ptr ) {
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).
16 years ago
int numwant, tmp, scanon;
ot_peer peer;
ot_hash *hash = NULL;
Don't default the incoming port to 6881 but rather use 0 to indicate a less than valid announce
16 years ago
unsigned short port = 0;
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).
16 years ago
char *write_ptr;
ssize_t len;
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
16 years ago
struct http_data *cookie = io_getcookie( sock );
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).
16 years ago
Move http handling to its own sources
17 years ago
/* This is to hack around stupid clients that send "announce ?info_hash" */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( read_ptr[-1] != '?' ) {
while( ( *read_ptr != '?' ) && ( *read_ptr != '\n' ) ) ++read_ptr;
if( *read_ptr == '\n' ) HTTPERROR_400_PARAM;
++read_ptr;
Move http handling to its own sources
17 years ago
}
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).
16 years ago
#ifdef WANT_IP_FROM_PROXY
if( accesslist_isblessed( cookie->ip, OT_PERMISSION_MAY_PROXY ) ) {
ot_ip6 proxied_ip;
char *fwd, *fwd_new = ws->request;
/* Zero terminate for string routines. Normally we'd only overwrite bollocks */
ws->request[ws->request_size-1] = 0;
/* Find last occurence of the forwarded header */
do {
fwd = fwd_new;
fwd_new = strcasestr( fwd_new, "\nX-Forwarded-For:" );
} while( fwd_new );
/* Skip spaces between : and the ip address */
if( fwd ) {
fwd += 18; /* sizeof( "\nX-Forwarded-For:" ) */
while( *fwd == ' ' ) ++fwd;
}
if( fwd && scan_ip6( fwd, proxied_ip ) )
OT_SETIP( &peer, proxied_ip );
else
OT_SETIP( &peer, cookie->ip );
}
#endif
OT_SETIP( &peer, cookie->ip );
Move http handling to its own sources
17 years ago
OT_SETPORT( &peer, &port );
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
16 years ago
OT_PEERFLAG( &peer ) = 0;
Move http handling to its own sources
17 years ago
numwant = 50;
scanon = 1;
while( scanon ) {
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
switch( scan_find_keywords(keywords_announce, &read_ptr, SCAN_SEARCHPATH_PARAM ) ) {
Move http handling to its own sources
17 years ago
case -2: scanon = 0; break; /* TERMINATOR */
case -1: HTTPERROR_400_PARAM; /* PARSE ERROR */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
case -3: scan_urlencoded_skipvalue( &read_ptr ); break;
case 1: /* matched "port" */
len = scan_urlencoded_query( &read_ptr, write_ptr = read_ptr, SCAN_SEARCHPATH_VALUE );
if( ( len <= 0 ) || scan_fixed_int( write_ptr, len, &tmp ) || ( tmp > 0xffff ) ) HTTPERROR_400_PARAM;
port = htons( tmp ); OT_SETPORT( &peer, &port );
Move http handling to its own sources
17 years ago
break;
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
case 2: /* matched "left" */
if( ( len = scan_urlencoded_query( &read_ptr, write_ptr = read_ptr, SCAN_SEARCHPATH_VALUE ) ) <= 0 ) HTTPERROR_400_PARAM;
if( scan_fixed_int( write_ptr, len, &tmp ) ) tmp = 0;
if( !tmp ) OT_PEERFLAG( &peer ) |= PEER_FLAG_SEEDING;
break;
case 3: /* matched "event" */
switch( scan_find_keywords( keywords_announce_event, &read_ptr, SCAN_SEARCHPATH_VALUE ) ) {
case -1: HTTPERROR_400_PARAM;
case 1: /* matched "completed" */
OT_PEERFLAG( &peer ) |= PEER_FLAG_COMPLETED;
break;
case 2: /* matched "stopped" */
OT_PEERFLAG( &peer ) |= PEER_FLAG_STOPPED;
break;
default:
break;
Move http handling to its own sources
17 years ago
}
break;
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
case 4: /* matched "numwant" */
len = scan_urlencoded_query( &read_ptr, write_ptr = read_ptr, SCAN_SEARCHPATH_VALUE );
if( ( len <= 0 ) || scan_fixed_int( write_ptr, len, &numwant ) ) HTTPERROR_400_PARAM;
if( numwant < 0 ) numwant = 50;
if( numwant > 200 ) numwant = 200;
Move http handling to its own sources
17 years ago
break;
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
case 5: /* matched "compact" */
len = scan_urlencoded_query( &read_ptr, write_ptr = read_ptr, SCAN_SEARCHPATH_VALUE );
if( ( len <= 0 ) || scan_fixed_int( write_ptr, len, &tmp ) ) HTTPERROR_400_PARAM;
if( !tmp ) HTTPERROR_400_COMPACT;
break;
case 6: /* matched "info_hash" */
if( hash ) HTTPERROR_400_DOUBLEHASH;
Move http handling to its own sources
17 years ago
/* ignore this, when we have less than 20 bytes */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( scan_urlencoded_query( &read_ptr, write_ptr = read_ptr, SCAN_SEARCHPATH_VALUE ) != 20 ) HTTPERROR_400_PARAM;
hash = (ot_hash*)write_ptr;
break;
#ifdef WANT_IP_FROM_QUERY_STRING
case 7: /* matched "ip" */
V6
16 years ago
{
char *tmp_buf1 = ws->reply, *tmp_buf2 = ws->reply+16;
len = scan_urlencoded_query( &read_ptr, tmp_buf2, SCAN_SEARCHPATH_VALUE );
tmp_buf2[len] = 0;
Reverse logic... do only fail if scan_ip6 cant parse
16 years ago
if( ( len <= 0 ) || !scan_ip6( tmp_buf2, tmp_buf1 ) ) HTTPERROR_400_PARAM;
V6
16 years ago
OT_SETIP( &peer, tmp_buf1 );
}
Move http handling to its own sources
17 years ago
break;
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
#endif
Move http handling to its own sources
17 years ago
}
}
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
16 years ago
/* XXX DEBUG */
stats_issue_event( EVENT_ACCEPT, FLAG_TCP, (uintptr_t)ws->reply );
Move http handling to its own sources
17 years ago
/* Scanned whole query string */
if( !hash )
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
return ws->reply_size = sprintf( ws->reply, "d14:failure reason80:Your client forgot to send your torrent's info_hash. Please upgrade your client.e" );
Move http handling to its own sources
17 years ago
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
16 years ago
if( OT_PEERFLAG( &peer ) & PEER_FLAG_STOPPED )
V6
16 years ago
ws->reply_size = remove_peer_from_torrent( *hash, &peer, ws->reply, FLAG_TCP );
* 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.
16 years ago
else
V6
16 years ago
ws->reply_size = add_peer_to_torrent_and_return_peers( *hash, &peer, FLAG_TCP, numwant, ws->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.
16 years ago
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( !ws->reply_size ) HTTPERROR_500;
* 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.
16 years ago
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
stats_issue_event( EVENT_ANNOUNCE, FLAG_TCP, ws->reply_size);
return ws->reply_size;
Move http handling to its own sources
17 years ago
}
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
ssize_t http_handle_request( const int64 sock, struct ot_workstruct *ws ) {
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
ssize_t reply_off, len;
char *read_ptr = ws->request, *write_ptr;
Move http handling to its own sources
17 years ago
#ifdef _DEBUG_HTTPERROR
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
reply_off = ws->request_size;
V6
16 years ago
if( ws->request_size >= G_DEBUGBUF_SIZE )
reply_off = G_DEBUGBUF_SIZE - 1;
memcpy( ws->debugbuf, ws->request, reply_off );
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
ws->debugbuf[ reply_off ] = 0;
Move http handling to its own sources
17 years ago
#endif
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
/* Tell subroutines where to put reply data */
ws->reply = ws->outbuf + SUCCESS_HTTP_HEADER_LENGTH;
Move http handling to its own sources
17 years ago
/* This one implicitely tests strlen < 5, too -- remember, it is \n terminated */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( memcmp( read_ptr, "GET /", 5) ) HTTPERROR_400;
Move http handling to its own sources
17 years ago
/* Skip leading '/' */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
for( read_ptr+=4; *read_ptr == '/'; ++read_ptr);
Move http handling to its own sources
17 years ago
/* Try to parse the request.
In reality we abandoned requiring the url to be correct. This now
only decodes url encoded characters, we check for announces and
scrapes by looking for "a*" or "sc" */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
len = scan_urlencoded_query( &read_ptr, write_ptr = read_ptr, SCAN_PATH );
Move http handling to its own sources
17 years ago
white space in comment
16 years ago
/* If parsing returned an error, leave with not found */
/ can now be redirected to any URL
17 years ago
if( g_redirecturl && ( len == -2 ) ) HTTPERROR_302;
Move http handling to its own sources
17 years ago
if( len <= 0 ) HTTPERROR_404;
/* This is the hardcore match for announce*/
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( ( *write_ptr == 'a' ) || ( *write_ptr == '?' ) )
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
http_handle_announce( sock, ws, read_ptr );
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,
16 years ago
#ifdef WANT_FULLSCRAPE
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
else if( !memcmp( write_ptr, "scrape HTTP/", 12 ) )
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
http_handle_fullscrape( sock, ws );
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,
16 years ago
#endif
Move http handling to its own sources
17 years ago
/* This is the hardcore match for scrape */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
else if( !memcmp( write_ptr, "sc", 2 ) )
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
http_handle_scrape( sock, ws, read_ptr );
Move http handling to its own sources
17 years ago
/* All the rest is matched the standard way */
Allow /stats to be located anywhere in your trackers path
16 years ago
else if( len == g_stats_path_len && !memcmp( write_ptr, g_stats_path, len ) )
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
http_handle_stats( sock, ws, read_ptr );
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
else
Move http handling to its own sources
17 years ago
HTTPERROR_404;
/* If routines handled sending themselves, just return */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( ws->reply_size == -2 ) return 0;
Move http handling to its own sources
17 years ago
/* If routine failed, let http error take over */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
if( ws->reply_size == -1 ) HTTPERROR_500;
Move http handling to its own sources
17 years ago
/* This one is rather ugly, so I take you step by step through it.
1. In order to avoid having two buffers, one for header and one for content, we allow all above functions from trackerlogic to
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
write to a fixed location, leaving SUCCESS_HTTP_HEADER_LENGTH bytes in our work buffer, which is enough for the static string
Move http handling to its own sources
17 years ago
plus dynamic space needed to expand our Content-Length value. We reserve SUCCESS_HTTP_SIZE_OFF for its expansion and calculate
the space NOT needed to expand in reply_off
*/
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
reply_off = SUCCESS_HTTP_SIZE_OFF - snprintf( ws->outbuf, 0, "%zd", ws->reply_size );
ws->reply = ws->outbuf + reply_off;
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
16 years ago
Move http handling to its own sources
17 years ago
/* 2. Now we sprintf our header so that sprintf writes its terminating '\0' exactly one byte before content starts. Complete
packet size is increased by size of header plus one byte '\n', we will copy over '\0' in next step */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
ws->reply_size += 1 + sprintf( ws->reply, "HTTP/1.0 200 OK\r\nContent-Type: text/plain\r\nContent-Length: %zd\r\n\r", ws->reply_size );
Move http handling to its own sources
17 years ago
/* 3. Finally we join both blocks neatly */
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
ws->outbuf[ SUCCESS_HTTP_HEADER_LENGTH - 1 ] = '\n';
Fix white spaces Introduce loading tracker states with -l Alter tracker state to a human readable form
16 years ago
Add comments, rename our struct http_data h to cookie, all clientsockets to sock, all size_t from socket_recvs to byte_count. Make signal handler set default handler for the second SIGINT
16 years ago
http_senddata( sock, ws );
* http and udp routines now use thread local buffers passed in workstruct containers. In other words they do not use static_buffer anymore and are considered to be thread safe. * the new workstruct also introduces a well defined buffer and result passing path * a new function scan_find_keywords is a wrapper around scan_urlencoded_query that maps keys in url to values passed in an array of ot_keywords structs * this new function cleans up much of url parameter parsing work, where read_ptr and write_ptr have been introduced rather than the confusing char *c, *data variables * I now use memcmp instead of byte_diff to allow compiler to optimize constant size string compares * got rid of UTORRENT_1600_WORKAROUND * livesync_ticker is now only called from one (currently main) thread to avoid race conditions
16 years ago
return ws->reply_size;
Move http handling to its own sources
17 years ago
}
Introduce some kind of versioning
17 years ago
const char *g_version_http_c = "$Source$: $Revision$\n";