Commit Graph

3394 Commits

Author SHA1 Message Date
Roger Dingledine
a6e91f8157 The crowning hidden-service patch: prefer non-internal circuits
for normal connections, so we don't spend our internal circs on
other stuff and not have them when we need them.


svn:r3380
2005-01-20 07:02:24 +00:00
Roger Dingledine
38be533c69 Handle unavailable hidden services better. We try each intro point
until none are left, then we try to refetch the descriptor. If it's
the same one we had before, then close streams right then. Whenever
a new stream arrives, even if it's right after, optimistically try
refetching the descriptor, just in case.


svn:r3379
2005-01-19 23:15:59 +00:00
Roger Dingledine
ff48179372 make cannibalizing for rend circs prey on the pleb circs first
if possible, rather than the elite ones.


svn:r3378
2005-01-19 23:13:20 +00:00
Nick Mathewson
f8c14bbf1f Detect windows platform correctly
svn:r3377
2005-01-19 23:10:16 +00:00
Nick Mathewson
2c803bff9a Make detect_compression_method work on platforms with more evil chars.
svn:r3376
2005-01-19 23:07:43 +00:00
Nick Mathewson
1eddb28f82 Add unittests for compression detection. Make all rendezvous descriptors "plausible".
svn:r3375
2005-01-19 22:47:48 +00:00
Nick Mathewson
69fa5be7b6 Workaround for webservers that lie about Content-Encoding: Tor now tries to autodetect compressed directories and compression itself. (resolves bug 65)
svn:r3374
2005-01-19 22:40:33 +00:00
Nick Mathewson
e0bf597665 Give more useful log messages when the tor process closes dns/cpuworker connections. "Error: Success!" makes us seem like crazy people."
svn:r3373
2005-01-19 21:34:42 +00:00
Roger Dingledine
75d83827e0 fix another bug where general exit circs were looking at rendezvous
streams. consolidate the looking into one function so it doesn't
happen again.


svn:r3372
2005-01-19 17:59:37 +00:00
Roger Dingledine
f2a7ff6801 better comments for exit policies in torrc.sample
svn:r3371
2005-01-19 17:26:40 +00:00
Roger Dingledine
cfcc93296c bugfix for cvs: we were needing a rendezvous circ, so we cannibalized
a general circ, and called rend_client_rendcirc_has_opened(), which
called connection_ap_attach_pending(), which was needing a rendezvous
circ, so it cannibalized a general circuit, and called ...


svn:r3370
2005-01-19 17:13:11 +00:00
Roger Dingledine
bd2d740b49 make it clearer why FooPort and FooBindAddress are separate
svn:r3369
2005-01-19 06:06:04 +00:00
Roger Dingledine
e94e7ada43 some early thoughts on the requirements for our network discovery protocol
svn:r3368
2005-01-19 04:36:56 +00:00
Roger Dingledine
d204332c98 make the server setup docs more win32 friendly
svn:r3367
2005-01-19 00:53:25 +00:00
Nick Mathewson
7d4a167c81 Basic windows libevent fixes
svn:r3366
2005-01-18 23:44:55 +00:00
Roger Dingledine
84deb5c690 mention that HiddenServiceNodes and HiddenServiceExcludeNodes
have fine defaults that you can leave alone in normal operation.


svn:r3365
2005-01-18 21:53:11 +00:00
Roger Dingledine
4e369bc9e8 allow bandwidthburst to be smaller, but whine if it's small.
svn:r3364
2005-01-18 19:34:22 +00:00
Roger Dingledine
c8c6af659d write at the top of rend-spec that reading the design paper
first will be helpful.


svn:r3363
2005-01-18 00:10:00 +00:00
Roger Dingledine
ea24dc31bc remove some extra-verbose logs
svn:r3361
2005-01-17 18:49:13 +00:00
Roger Dingledine
d2400a5afd Introduce a notion of 'internal' circs, which are chosen without regard
to the exit policy of the last hop. Intro and rendezvous circs must
be internal circs, to avoid leaking information. Resolve and connect
streams can use internal circs if they want.

New circuit pooling algorithm: make sure to have enough circs around
to satisfy any predicted ports, and also make sure to have 2 internal
circs around if we've required internal circs lately (with high uptime
if we've seen that lately).

Split NewCircuitPeriod config option into NewCircuitPeriod (30 secs),
which describes how often we retry making new circuits if current ones
are dirty, and MaxCircuitDirtiness (10 mins), which describes how long
we're willing to make use of an already-dirty circuit.

Once rendezvous circuits are established, keep using the same circuit as
long as you attach a new stream to it at least every 10 minutes. (So web
browsing doesn't require you to build new rend circs every 30 seconds.)

Cannibalize GENERAL circs to be C_REND, C_INTRO, S_INTRO, and S_REND
circ as necessary, if there are any completed ones lying around when
we try to launch one.

Re-instate the ifdef's to use version-0 style introduce cells, since
there was yet another bug in handling version-1 style. We'll try switching
over again after 0.0.9 is obsolete.

Bugfix: when choosing an exit node for a new non-internal circ, don't take
into account whether it'll be useful for any pending x.onion addresses --
it won't.

Bugfix: we weren't actually publishing the hidden service descriptor when
it became dirty. So we only published it every 20 minutes or so, which
means when you first start your Tor, the hidden service will seem broken.


svn:r3360
2005-01-17 18:13:09 +00:00
Roger Dingledine
9c8c90ec2f Sometime post 0.0.9, we should integrate an http proxy into Tor (maybe
as a third class of worker process), so we can stop shipping with the
beast that is Privoxy.


svn:r3359
2005-01-16 02:49:58 +00:00
Roger Dingledine
372832d8fe make tor-resolve print the Tor version on --version too
svn:r3358
2005-01-16 02:21:14 +00:00
Roger Dingledine
22a48e0a41 our man page still talked about the dirservers file
svn:r3357
2005-01-15 23:18:42 +00:00
Roger Dingledine
c6771e5ce4 reset published uptime when you wake up from hibernation,
since if you hibernate daily a published uptime of >1day
will be misleading.


svn:r3356
2005-01-14 17:49:25 +00:00
Roger Dingledine
dcf8b18215 when the hidden service launches a rendezvous circ, make sure it
provides good uptime if any virtual_port of the service wants that.


svn:r3355
2005-01-14 04:35:53 +00:00
Nick Mathewson
a38bccdca7 Move call to event_init() immediately after start_daemon(); move start_daemon earlier. This should fix using libevent with RunAsDaemon.
svn:r3354
2005-01-13 21:32:08 +00:00
Roger Dingledine
0aff4ff4cb get rid of 0.0.8 backwards compatibility
svn:r3353
2005-01-13 20:22:38 +00:00
Roger Dingledine
22f7934718 start generating version 1 style introduce cells
svn:r3352
2005-01-13 20:21:11 +00:00
Roger Dingledine
0703d01b13 fix some problems in control-spec
svn:r3351
2005-01-13 17:54:18 +00:00
Roger Dingledine
77da824faf point win32 folks at 0.0.9.2, not 0.0.9.1
svn:r3350
2005-01-13 17:54:03 +00:00
Roger Dingledine
49fe4cba74 when deciding if a port is handled and the port demands uptime, don't
consider it handled if there's a circ that fits but isn't high-uptime.


svn:r3349
2005-01-13 07:23:19 +00:00
Peter Palfrader
bc5fffb0b3 Add libevent-dev to build-depends
svn:r3348
2005-01-12 15:44:44 +00:00
Roger Dingledine
abea142665 more cleanup
svn:r3347
2005-01-12 12:22:25 +00:00
Roger Dingledine
bafb3e4a5b fix another integer underflow caused by buckets going negative
svn:r3346
2005-01-12 12:19:00 +00:00
Roger Dingledine
24a97d2c55 fix up some comments
svn:r3345
2005-01-12 12:18:46 +00:00
Roger Dingledine
3809c0ab91 tolerate bandwidtch buckets going negative (i hope)
svn:r3344
2005-01-12 10:00:38 +00:00
Nick Mathewson
ff9618f31e Note that we need to check return codes from more libevent functions
svn:r3343
2005-01-12 07:00:28 +00:00
Roger Dingledine
53dbc66a79 document LongLivedPorts in the man page
svn:r3342
2005-01-12 06:45:00 +00:00
Nick Mathewson
324b192f68 Make Tor use Niels Provos's libevent instead of it's current
poll-but-sometimes-select mess.  This will let us use faster async cores
(like epoll, kpoll, and /dev/poll), and hopefully work better on Windows
too.

There are some fairly nasty changes to main.c here; this will almost
certainly break something.  But hey, that's what alphas are for.


svn:r3341
2005-01-12 06:42:32 +00:00
Roger Dingledine
9b578f2fe2 add more chat and interactive protocols to LongLivedPorts
svn:r3340
2005-01-12 05:18:14 +00:00
Roger Dingledine
84c81e0783 Annotate circuits w/ whether they aim to contain high uptime nodes and/or
high capacity nodes. When building circuits, choose appropriate nodes.

New config option LongLivedPorts to indicate application streams
that will want high uptime circuits.

When attaching a stream to a circuit, pay attention to its requirements.

This means that every single node in an intro rend circuit, not just
the last one, will have a minimum uptime.

Boost the min uptime from an hour to 24 hours.


svn:r3339
2005-01-12 04:58:23 +00:00
Roger Dingledine
51f47d746e clean up smartlist_string_isin so callers don't have to manage bufs
themselves


svn:r3338
2005-01-12 04:53:53 +00:00
Roger Dingledine
70b3208f0a if somebody starts his tor server in jan 2004 and then fixes his
clock, don't make his published uptime be a year.


svn:r3337
2005-01-11 11:54:11 +00:00
Roger Dingledine
45eb8d9660 if you're a server and the auth dirservers are down, go to any other
server. only give up hope on network reachability when no other servers
are reachable either.


svn:r3335
2005-01-10 17:39:41 +00:00
Roger Dingledine
74bc4e5af0 help avoid overflow from len_in + (len_in/n)*len_ins
nick, do you think this does it? if so, we should maybe backport too.


svn:r3334
2005-01-10 16:30:51 +00:00
Roger Dingledine
4b6d55afdf separate time_of_process_start from stats_n_seconds_uptime:
now we reset uptime when a dir fetch entirely fails. this
hopefully has something to do with stability of being on the
network.


svn:r3333
2005-01-10 05:10:22 +00:00
Roger Dingledine
3370f53010 it helps to use the correct name for the struct
svn:r3331
2005-01-10 04:32:59 +00:00
Roger Dingledine
10a81ccf18 when pruning circuit-wait streams when a dir fetch failed,
refer in logs to the destination address, not the address of the
application connection.


svn:r3330
2005-01-10 02:47:50 +00:00
Peter Palfrader
eab35efa67 Forward port patches
svn:r3329
2005-01-09 02:30:28 +00:00
Peter Palfrader
904b321fa7 Add watch file
svn:r3328
2005-01-09 02:30:02 +00:00