erdgeist
697e239027
Moving accesslist stuff to its own object
17 years ago
erdgeist
a8993b8eea
Moving full scrapes to iovec-creating thread safe function
17 years ago
erdgeist
927ac023ce
The expensive DMEM stats is gone. No need for it, too much potential to DOS.
17 years ago
erdgeist
41120f6a0d
Move stats out of opentracker.c, also have an own file handle udp requests.
17 years ago
erdgeist
8900cc0dd9
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.
17 years ago
erdgeist
a859ae78e5
Cleaned up stats return format. Also reflect APPI changes due to MT code in trackerlogic.{c,h}
17 years ago
erdgeist
8acc0ff48f
Introduce configuration dependent number of parameters in a nicer way
17 years ago
erdgeist
5aa41e9d82
The last experiment failed. No more accept deferral
17 years ago
erdgeist
d276addb7a
Trying to accept connections only when data has come in. This is an experiment.
17 years ago
erdgeist
853dc33be2
Allow for more than one admin ip address
17 years ago
erdgeist
dc8e889797
Add more full scrape stats
17 years ago
erdgeist
31eada6168
Reenabled syncing
17 years ago
erdgeist
b19bbd6a85
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.
17 years ago
erdgeist
ab9246618b
some clients chose to accidently send negative numwants
17 years ago
erdgeist
1fce733ea0
Accept clients that forget to spell announce at all...
17 years ago
erdgeist
69c6e2bc99
This uTorrent/1600 client requires some... measures :/ I dont like the workaround.
17 years ago
erdgeist
ac57c3d5d6
Now measure the time a clean_all_torrents takes
17 years ago
erdgeist
6876e16242
assume anything looking like GET /a to be an announce...
17 years ago
erdgeist
3aa286d9b8
Last fix was not enough to get rid off those 1600er buggers.
17 years ago
erdgeist
00972c6c68
Work around the stupid utorrent 1600 bug. That'll be removed after a while.
17 years ago
erdgeist
0de8b53fdd
Now the Holzhammer: GET /an is announce, GET /sc is scrape. Period.
17 years ago
erdgeist
6b2307015f
now accept more classes of announce fuckups
17 years ago
erdgeist
42210450f3
Just one more typo
17 years ago
erdgeist
2cbb480865
Just one more typo
17 years ago
erdgeist
886421d2d5
Work around more typos that might end up in torrent files as announce urls
17 years ago
erdgeist
2b715b89f1
Debug code sometimes copied too much data
17 years ago
erdgeist
ddd85ec7e0
Fixed one leak from struct_http
17 years ago
erdgeist
8098adb48e
Be even more stupidistas-tolerant
17 years ago
erdgeist
84f8738b82
Fix debug output
17 years ago
erdgeist
d768bf6e81
Enable _DEBUG_HTTPERROR to avoid warning...
17 years ago
erdgeist
53bce09735
Include time since startup in full scrape debug output
17 years ago
erdgeist
c59e44ba94
Now limit full scrape to requests sending exactly "GET /scrape HTTP/"
17 years ago
erdgeist
41019c79d8
Introducing multiscrape
17 years ago
erdgeist
d5876ebb6b
Prevent url-decoding for debug by using a temp buffer
17 years ago
erdgeist
271ff156ec
Those stupid "announce ?info_hash" torrents get really annoying. Fix it in announce and scrape.
17 years ago
erdgeist
1b0859411e
be more verbose about what full scrape consumers deliver in their http requests
17 years ago
erdgeist
714ea1b88e
give opentracker the missing header file
17 years ago
erdgeist
f56e40c24e
Must not dealloc mmaped buffers with free()
17 years ago
erdgeist
d25cf9307e
Fix misunderstandig of what libowfat thinks are seconds and what I did...
17 years ago
erdgeist
e1f122f7d1
Since array and io_batch are in a union, and we also have time-outing writable sockets, we need to mark, which of the union is used. Also remove the blessed flag from struct http, since its rarely used and can be done in a seperate memcmp.
17 years ago
erdgeist
6eb441d8a0
Cannot access members after they are freed :/
17 years ago
erdgeist
b86e6382a1
Save a lot of work when skipping through uninteresting http request parameters
17 years ago
erdgeist
32f17413fa
Increase send timeout. Also use an own variable for that.
17 years ago
erdgeist
28de135098
io_timeout is absolute, not relative :/
17 years ago
erdgeist
d1c57d90dd
Free iobatches on timeout
17 years ago
erdgeist
8e78ac7012
HEADS UP: you need to check out the latest libowfat! Huge allocations are now taken from mmap()ed regions, not from heap anymore
17 years ago
erdgeist
1e6b48cbc0
Sockets we write to now do also timeout
17 years ago
erdgeist
28d88e3380
Exclude empty torrents from full scrape, also try to shrink full scrape buffer after it is full
17 years ago
erdgeist
3dc470d672
fixed line breaks in log
17 years ago
erdgeist
65cc45f129
Log all expansive operations to stderr
17 years ago
erdgeist
ffffd80719
Most open trackers do not have the problem of not being called often enough.. so do not introduce an extra timeout when waiting for activity. That also removes one of the last extraneous gettimeofday calls
17 years ago
erdgeist
6c1adb8fc8
since gettimeofday is rather expansive, we do only fetch time once in a 5 second period, when we are delivered a SIGALRM.
17 years ago
erdgeist
4c15dc7a70
make 64bit safe
17 years ago
denis
0d36a8754e
fixed some stat output
17 years ago
erdgeist
c350fa0b3c
New stats for s24s code, this is debug, do not use in real world
17 years ago
erdgeist
33c9c530d0
An announce with event=stopped now returns correct number of leechers and seeders. In TCP and UDP.
17 years ago
erdgeist
3e47339b6c
Added whitelisting to reimplement the WANT_CLOSED_TRACKER functionality
18 years ago
erdgeist
992058383a
Under Linux one gets SIG_ALRMs from time to time when using epoll. Ignore those.
18 years ago
erdgeist
77af0b475f
Sort of addressed a concorrency issue when receiving mutliple SIG_HUPs
18 years ago
erdgeist
afe966b644
some subnet stats... still ugly and untested.
18 years ago
erdgeist
c0b00c0bf5
* 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
18 years ago
erdgeist
3f0a18ba08
Clean up pw-structs
18 years ago
denis
c212b48bec
add comment for udp spec location
18 years ago
denis
78b7c73eab
check for mafic udp id
18 years ago
denis
8f0aaed6b2
fixed an overflow in ot_overall_* variables
18 years ago
erdgeist
77e5241cf8
Clean up help and usage
18 years ago
erdgeist
5c18293c9c
New Makefile, incorporated some patches sent to me by Robin H. Johnson
18 years ago
erdgeist
bd4617a911
Try to figure out nobodys uid und gid using getpwnam
18 years ago
erdgeist
4e9523c489
Fix endianess issue with blessed ips. Document -A feature.
18 years ago
erdgeist
32ba028814
added the -A feature that blesses a certain IP address for administrative use (like sync)
18 years ago
erdgeist
33774078ab
Added inbound part of sync. This also meant to remove the black/whitelisting as we did by now. A more scalable way to blacklist will follow.
18 years ago
erdgeist
b38104b986
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.
18 years ago
denis
bab7c0bbb1
code cleanup and more specific comments. tracker starts now also with udp port 6969 as default
18 years ago
denis
46605bcb2c
some udp scrape style cleanup
18 years ago
denis
8ac7768b96
support for udp scrape
18 years ago
denis
36413e4853
fix the stats a little for not accepting udp scrapes
18 years ago
denis
45d34960b0
udp first connect needs to be counted as a connection too
18 years ago
denis
f07ec869d8
added the udp stats to the main stats
18 years ago
denis
5bb381e20a
new modes for udp4 and tcp4 stats
18 years ago
denis
50fe661aa7
collecting stats for udp connections (2nd try)
18 years ago
denis
f2a0c36721
collecting stats for udp connections
18 years ago
denis
036e874ed4
fixed udp peerport endianess handling
18 years ago
denis
92213c286a
fixed udp remoteport handling
18 years ago
erdgeist
c00c64f9da
some 64bit and endianess issues with udp...
18 years ago
erdgeist
e75c3e25cd
make full scrapes a bit faster, increase checkback time for udp
18 years ago
erdgeist
5b55394cdc
Port is a short, not a long
18 years ago
erdgeist
5d1f07d353
Sanify handling of server sockets
18 years ago
erdgeist
0a90216686
udp is working in theory, return_peers_for_torrent accepts a switch to request an udp answer. Fixed prototypes.
18 years ago
erdgeist
175c37c3e8
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
18 years ago
erdgeist
0d6f508882
zero copy sockets do only work on page aligned buffers for page sized blocks. So remove the code that tried to take advantage of false assuming, it would work on any buffer
18 years ago
denis
222519fdff
fixed the last commit. now it really counts the scrapes and fullscrapes
18 years ago
denis
bbdc36c518
add the scrapes to the valid requests (blue mrtg line)
18 years ago
erdgeist
f9c49e9759
el cheapo vector debugging
18 years ago
erdgeist
62c59983cb
Use multiple output buffers to try to avoid copy on write when unnecessary
18 years ago
erdgeist
10a6586ad1
Try to shake off too stupid clients by sending an Bittorrent error rather than an HTTP 400
18 years ago
erdgeist
6e1ee92de6
Use signed size_t wherever appropriate
18 years ago
erdgeist
11abb03e85
size_t cant get < 0 and thus wont report errors
18 years ago
erdgeist
f0ddf30389
provide mrtg with more sane strings
18 years ago
erdgeist
5a3d6b4029
We now answer as soon as the first \n is seen
...
We do not copy input buffer to socket's input array anymore, if we find \n in first buffer
We simplified parsing of input string
We removed asserts that were needed to debug negative seeder counts
We split static input and output buffers for obvious reasons
We enlarged static output buffer to prepare multi scrape functionality
We removed -lm from Makefile, this lib was only needed, when our random algorithm worked on floats
We close fds now that come with an empty iocookie into write loop, those don't make sense
We documented query string parser in opentracker.c a lot better to help auditing
18 years ago
erdgeist
4c6bcc89ce
complete_D_ ... ECODERTOODUMB
18 years ago