]> git.ipfire.org Git - thirdparty/tor.git/log
thirdparty/tor.git
15 years agoRemove a needless keep_open_until_flushed
Nick Mathewson [Wed, 8 Sep 2010 14:23:29 +0000 (10:23 -0400)] 
Remove a needless keep_open_until_flushed

15 years agoClose a non-open OR connection *only* after KeepalivePeriod.
Nick Mathewson [Fri, 3 Sep 2010 15:32:35 +0000 (11:32 -0400)] 
Close a non-open OR connection *only* after KeepalivePeriod.

When we introduced the code to close non-open OR connections after
KeepalivePeriod had passed, we replaced some code that said
    if (!connection_is_open(conn)) {
     /* let it keep handshaking forever */
    } else if (do other tests here) {
      ...
with new code that said
    if (!connection_is_open(conn) && past_keepalive) {
     /* let it keep handshaking forever */
    } else if (do other tests here) {
      ...

This was a mistake, since it made all the other tests start applying
to non-open connections, thus causing bug 1840, where non-open
connections get closed way early.

Fixes bug 1840.  Bugfix on 0.2.1.26 (commit 67b38d50).

15 years agoUse -Wno-system-headers on openbsd to resolve 2nd case of bug1848
Nick Mathewson [Tue, 24 Aug 2010 17:43:53 +0000 (13:43 -0400)] 
Use -Wno-system-headers on openbsd to resolve 2nd case of bug1848

15 years agoBackport END_STREAM_REASON_NOROUTE for client use.
Sebastian Hahn [Tue, 3 Aug 2010 14:46:28 +0000 (16:46 +0200)] 
Backport END_STREAM_REASON_NOROUTE for client use.

(Partial backport of 150ed553dfce9161b275028e90, and 4c948ffd6.)

15 years agoMerge branch 'bug1141_v3' into maint-0.2.1
Nick Mathewson [Tue, 17 Aug 2010 15:55:58 +0000 (11:55 -0400)] 
Merge branch 'bug1141_v3' into maint-0.2.1

15 years agoScale CONSENSUS_MIN_SECONDS_BEFORE_CACHING by voting interval
Nick Mathewson [Sat, 31 Jul 2010 17:48:41 +0000 (13:48 -0400)] 
Scale CONSENSUS_MIN_SECONDS_BEFORE_CACHING by voting interval

If the voting interval was short enough, the two-minutes delay
of CONSENSUS_MIN_SECONDS_BEFORE_CACHING would confuse bridges
to the point where they would assert before downloading a consensus.
It it was even shorter (<4 minutes, I think), caches would
assert too.  This patch fixes that by having replacing the
two-minutes value with MIN(2 minutes, interval/16).

Bugfix for 1141; the cache bug could occur since 0.2.0.8-alpha, so
I'm calling this a bugfix on that.  Robert Hogan diagnosed this.
Done as a patch against maint-0.2.1, since it makes it hard to
run some kinds of testing networks.

15 years agoRemove the debian directory from the main git repository
Nick Mathewson [Wed, 4 Aug 2010 16:21:48 +0000 (12:21 -0400)] 
Remove the debian directory from the main git repository

Once upon a time it made sense to keep all the Debian files in the
main Tor distribution, since repeatedly merging them back in was hard.
Now that we're on git, that's no longer so.

Peter's debian repository at debian/tor.git on our git server has the
most recent version of the tor-on-debian packaging stuff, and the versions
in our own repository have gotten out of date.

Resolves bug #1735.

15 years agoUpdate to the August 1 2010 Maxmind GeoLite Country database.
Karsten Loesing [Tue, 3 Aug 2010 11:04:09 +0000 (13:04 +0200)] 
Update to the August 1 2010 Maxmind GeoLite Country database.

15 years agoremove old changes that were folded into 0.2.1.26
Roger Dingledine [Sun, 13 Jun 2010 20:04:32 +0000 (16:04 -0400)] 
remove old changes that were folded into 0.2.1.26

15 years agoAdd maatuska as eighth v3 directory authority.
Karsten Loesing [Thu, 10 Jun 2010 14:56:08 +0000 (16:56 +0200)] 
Add maatuska as eighth v3 directory authority.

15 years agoadd changes file for geoip db update
Nick Mathewson [Mon, 7 Jun 2010 19:09:48 +0000 (15:09 -0400)] 
add changes file for geoip db update

15 years agoUpdate to June 1 2010 Maxmind GeoLite Country database.
Karsten Loesing [Wed, 2 Jun 2010 17:53:19 +0000 (19:53 +0200)] 
Update to June 1 2010 Maxmind GeoLite Country database.

15 years agomove to maxmind geoip db
Roger Dingledine [Thu, 6 May 2010 11:15:22 +0000 (07:15 -0400)] 
move to maxmind geoip db

15 years agorelease notes entry for 0.2.1.26
Roger Dingledine [Wed, 5 May 2010 07:19:41 +0000 (03:19 -0400)] 
release notes entry for 0.2.1.26

15 years agowindows packaging cleanups from phobos
Roger Dingledine [Wed, 21 Apr 2010 00:17:28 +0000 (20:17 -0400)] 
windows packaging cleanups from phobos

15 years agoclose idle tls conns early
Roger Dingledine [Sat, 24 Apr 2010 00:23:00 +0000 (20:23 -0400)] 
close idle tls conns early

15 years agofinally get rid of "clique mode"
Roger Dingledine [Thu, 22 Apr 2010 01:35:18 +0000 (21:35 -0400)] 
finally get rid of "clique mode"

15 years agoclose idle dir-fetch circs early
Roger Dingledine [Fri, 23 Apr 2010 22:35:11 +0000 (18:35 -0400)] 
close idle dir-fetch circs early

15 years agoFix renegotiation on OpenSSL versions that backport RFC5746.
Nick Mathewson [Tue, 13 Apr 2010 18:58:30 +0000 (14:58 -0400)] 
Fix renegotiation on OpenSSL versions that backport RFC5746.

Our code assumed that any version of OpenSSL before 0.9.8l could not
possibly require SSL_OP_ALLOW_UNSAFE_LEGACY_RENEGOTIATION.  This is
so... except that many vendors have backported the flag from later
versions of openssl when they backported the RFC5476 renegotiation
feature.

The new behavior is particularly annoying to detect.  Previously,
leaving SSL_OP_ALLOW_UNSAFE_LEGACY_RENEGOTIATION unset meant that
clients would fail to renegotiate.  People noticed that one fast!
Now, OpenSSL's RFC5476 support means that clients will happily talk to
any servers there are, but servers won't accept renegotiation requests
from unpatched clients unless SSL_OP_ALLOW_etc is set.  More fun:
servers send back a "no renegotiation for you!" error, which unpatched
clients respond to by stalling, and generally producing no useful
error message.

This might not be _the_ cause of bug 1346, but it is quite likely _a_
cause for bug 1346.

15 years agotestsuite: Prevent the main thread from starving the worker threads
Peter Palfrader [Mon, 12 Apr 2010 22:10:56 +0000 (00:10 +0200)] 
testsuite: Prevent the main thread from starving the worker threads

15 years agotestsuite: Only free the main mutex when and if all the worker threads are done
Peter Palfrader [Mon, 12 Apr 2010 20:49:58 +0000 (22:49 +0200)] 
testsuite: Only free the main mutex when and if all the worker threads are done

15 years agofetch relay descriptors from v3 authorities
Roger Dingledine [Sun, 4 Apr 2010 01:53:29 +0000 (21:53 -0400)] 
fetch relay descriptors from v3 authorities

15 years agogive us a blurb; add stanza to the releasenotes tor-0.2.1.25
Roger Dingledine [Tue, 16 Mar 2010 04:44:30 +0000 (00:44 -0400)] 
give us a blurb; add stanza to the releasenotes

15 years agobump to 0.2.1.25
Roger Dingledine [Mon, 15 Mar 2010 22:08:29 +0000 (18:08 -0400)] 
bump to 0.2.1.25

it's perfect, let's ship it

15 years agoclean up the 0.2.1.25 changelog
Roger Dingledine [Sun, 7 Mar 2010 03:39:34 +0000 (22:39 -0500)] 
clean up the 0.2.1.25 changelog

15 years agoApply Roger's bug 1269 fix.
Nick Mathewson [Thu, 4 Mar 2010 23:37:40 +0000 (18:37 -0500)] 
Apply Roger's bug 1269 fix.

From http://archives.seul.org/tor/relays/Mar-2010/msg00006.html :

   As I understand it, the bug should show up on relays that don't set
   Address to an IP address (so they need to resolve their Address
   line or their hostname to guess their IP address), and their
   hostname or Address line fails to resolve -- at that point they'll
   pick a random 4 bytes out of memory and call that their address. At
   the same time, relays that *do* successfully resolve their address
   will ignore the result, and only come up with a useful address if
   their interface address happens to be a public IP address.

15 years agoBackport fix for time-goes-forward test. Fix bug 1267
Nick Mathewson [Tue, 2 Mar 2010 15:40:21 +0000 (10:40 -0500)] 
Backport fix for time-goes-forward test.  Fix bug 1267

15 years agoUpdate Tor Project copyright years
Nick Mathewson [Sat, 27 Feb 2010 22:13:37 +0000 (17:13 -0500)] 
Update Tor Project copyright years

15 years agoProperly handle non-terminated strings
Sebastian Hahn [Thu, 25 Feb 2010 09:31:36 +0000 (10:31 +0100)] 
Properly handle non-terminated strings

Treat strings returned from signed_descriptor_get_body_impl() as not
NUL-terminated. Since the length of the strings is available, this is
not a big problem.

Discovered by rieo.

15 years agoProper NULL checking in circuit_list_path_impl()
Sebastian Hahn [Wed, 24 Feb 2010 08:36:15 +0000 (09:36 +0100)] 
Proper NULL checking in circuit_list_path_impl()

Another dereference-then-NULL-check sequence. No reports of this bug
triggered in the wild. Fixes bugreport 1256.

Thanks to ekir for discovering and reporting this bug.

15 years agoProper NULL checking for hsdesc publication
Sebastian Hahn [Tue, 23 Feb 2010 16:09:02 +0000 (17:09 +0100)] 
Proper NULL checking for hsdesc publication

Fix a dereference-then-NULL-check sequence. This bug wasn't triggered
in the wild, but we should fix it anyways in case it ever happens.
Also make sure users get a note about this being a bug when they
see it in their log.

Thanks to ekir for discovering and reporting this bug.

15 years agoZero a cipher completely before freeing it
Sebastian Hahn [Mon, 22 Feb 2010 10:39:29 +0000 (11:39 +0100)] 
Zero a cipher completely before freeing it

We used to only zero the first ptrsize bytes of the cipher. Since
cipher is large enough, we didn't zero too many bytes. Discovered
and fixed by ekir. Fixes bug 1254.

15 years agoput 0.2.1.24 in release notes too tor-0.2.1.24
Roger Dingledine [Sun, 21 Feb 2010 22:52:52 +0000 (17:52 -0500)] 
put 0.2.1.24 in release notes too

15 years agobump to 0.2.1.24
Roger Dingledine [Sun, 21 Feb 2010 22:27:12 +0000 (17:27 -0500)] 
bump to 0.2.1.24

15 years agoBump version to 0.2.1.23-dev
Nick Mathewson [Thu, 18 Feb 2010 16:57:47 +0000 (11:57 -0500)] 
Bump version to 0.2.1.23-dev

15 years agoAdd changelog for latest openssl fix
Nick Mathewson [Thu, 18 Feb 2010 16:54:26 +0000 (11:54 -0500)] 
Add changelog for latest openssl fix

15 years agoFix compile
Sebastian Hahn [Thu, 18 Feb 2010 12:08:57 +0000 (13:08 +0100)] 
Fix compile

15 years agoEven more conservative option-setting for SSL renegotiation.
Nick Mathewson [Thu, 18 Feb 2010 04:55:03 +0000 (23:55 -0500)] 
Even more conservative option-setting for SSL renegotiation.

This time, set the SSL3_FLAGS_ALLOW_UNSAFE_RENEGOTIATION flag on every
version before OpenSSL 0.9.8l.  I can confirm that the option value (0x0010)
wasn't reused until OpenSSL 1.0.0beta3.

15 years agogive it a blurb, update the date tor-0.2.1.23
Roger Dingledine [Sat, 13 Feb 2010 19:10:57 +0000 (14:10 -0500)] 
give it a blurb, update the date

15 years agonew dannenberg address; make moria2's demise official.
Roger Dingledine [Fri, 12 Feb 2010 19:31:08 +0000 (14:31 -0500)] 
new dannenberg address; make moria2's demise official.

15 years agoprepare for 0.2.1.23
Roger Dingledine [Fri, 12 Feb 2010 17:35:40 +0000 (12:35 -0500)] 
prepare for 0.2.1.23

15 years agoDon't use gethostbyname() in resolve_my_address()
Sebastian Hahn [Mon, 8 Feb 2010 14:35:34 +0000 (15:35 +0100)] 
Don't use gethostbyname() in resolve_my_address()

Tor has tor_lookup_hostname(), which prefers ipv4 addresses automatically.
Bug 1244 occured because gethostbyname() returned an ipv6 address, which
Tor cannot handle currently. Fixes bug 1244; bugfix on 0.0.2pre25.
Reported by Mike Mestnik.

15 years agolookup_last_hid_serv_request() could overflow and leak memory
Sebastian Hahn [Sun, 7 Feb 2010 05:30:55 +0000 (06:30 +0100)] 
lookup_last_hid_serv_request() could overflow and leak memory

The problem was that we didn't allocate enough memory on 32-bit
platforms with 64-bit time_t. The memory leak occured every time
we fetched a hidden service descriptor we've fetched before.

15 years agoLink libssl and libcrypto in the right order.
Nick Mathewson [Tue, 2 Feb 2010 21:12:45 +0000 (16:12 -0500)] 
Link libssl and libcrypto in the right order.

For most linking setups, this doesn't matter.  But for some setups, when
statically linking openssl, it does matter, since you need to link things
with dependencies before you link things they depend on.

Fix for bug 1237.

15 years agoRevise OpenSSL fix to work with OpenSSL 1.0.0beta*
Nick Mathewson [Mon, 1 Feb 2010 03:48:29 +0000 (22:48 -0500)] 
Revise OpenSSL fix to work with OpenSSL 1.0.0beta*

In brief: you mustn't use the SSL3_FLAG solution with anything but 0.9.8l,
and you mustn't use the SSL_OP solution with anything before 0.9.8m, and
you get in _real_ trouble if you try to set the flag in 1.0.0beta, since
they use it for something different.

For the ugly version, see my long comment in tortls.c

15 years agoDecide whether to use SSL flags based on runtime OpenSSL version.
Nick Mathewson [Fri, 29 Jan 2010 22:02:17 +0000 (17:02 -0500)] 
Decide whether to use SSL flags based on runtime OpenSSL version.

We need to do this because Apple doesn't update its dev-tools headers
when it updates its libraries in a security patch.  On the bright
side, this might get us out of shipping a statically linked OpenSSL on
OSX.

May fix bug 1225.

[backported]

15 years agoAdd --enable-static-(openssl|libevent) options
Nick Mathewson [Sun, 24 Jan 2010 19:34:47 +0000 (14:34 -0500)] 
Add --enable-static-(openssl|libevent) options

These options only work when using --with-(openssl|libevent)-dir to
explicitly pick a libevent or openssl location.

15 years agoNew configure option: --enable-gcc-warnings-advisory
Peter Palfrader [Sun, 24 Jan 2010 18:32:41 +0000 (19:32 +0100)] 
New configure option: --enable-gcc-warnings-advisory

the new configure option --enable-gcc-warnings-advisory enables
all the gcc flags that --enable-gcc-warnings does with the
exception of -Werror.

15 years agoFix two rare leaks spotted by rieo.
Nick Mathewson [Sun, 24 Jan 2010 01:46:38 +0000 (20:46 -0500)] 
Fix two rare leaks spotted by rieo.

15 years agoAvoid a possible crash in tls_log_errors.
Nick Mathewson [Fri, 22 Jan 2010 21:32:15 +0000 (16:32 -0500)] 
Avoid a possible crash in tls_log_errors.

We were checking for msg==NULL, but not lib or proc.  This case can
only occur if we have an error whose string we somehow haven't loaded,
but it's worth coding defensively here.

Spotted by rieo on IRC.

15 years agospread guard rotation out throughout the month
Roger Dingledine [Tue, 19 Jan 2010 22:52:52 +0000 (17:52 -0500)] 
spread guard rotation out throughout the month

15 years agoweight guard choice by bandwidth; discard old guards
Roger Dingledine [Tue, 19 Jan 2010 22:30:52 +0000 (17:30 -0500)] 
weight guard choice by bandwidth; discard old guards

15 years agobump to 0.2.1.22, and give it a changelog tor-0.2.1.22
Roger Dingledine [Tue, 19 Jan 2010 19:43:05 +0000 (14:43 -0500)] 
bump to 0.2.1.22, and give it a changelog

15 years agodowngrade a warning
Roger Dingledine [Tue, 19 Jan 2010 19:25:15 +0000 (14:25 -0500)] 
downgrade a warning

this case can now legitimately happen, if you have a cached v2 status
from moria1, and you run with the new list of dirservers that's missing
the old moria1. it's nothing to worry about; the file will die off in
a month or two.

15 years agorotate keys for moria1 and gabelmoo
Roger Dingledine [Tue, 19 Jan 2010 19:12:39 +0000 (14:12 -0500)] 
rotate keys for moria1 and gabelmoo

15 years agostop bridge authorities from leaking their bridge list
Roger Dingledine [Mon, 18 Jan 2010 00:41:22 +0000 (19:41 -0500)] 
stop bridge authorities from leaking their bridge list

15 years agobump to 0.2.1.21 so we can release tor-0.2.1.21
Roger Dingledine [Mon, 21 Dec 2009 08:22:49 +0000 (03:22 -0500)] 
bump to 0.2.1.21 so we can release

15 years agoFix bug 1173: remove an assert(unsigned >= 0).
Nick Mathewson [Tue, 15 Dec 2009 20:51:59 +0000 (15:51 -0500)] 
Fix bug 1173: remove an assert(unsigned >= 0).

16 years agoImproved workaround for disabled OpenSSL renegotiation.
Martin Peck [Fri, 4 Dec 2009 19:25:08 +0000 (14:25 -0500)] 
Improved workaround for disabled OpenSSL renegotiation.

It turns out that OpenSSL 0.9.8m is likely to take a completely
different approach for reenabling renegotiation than OpenSSL 0.9.8l
did, so we need to work with both. :p   Fixes bug 1158.

(patch by coderman; commit message by nickm)

16 years agofix race condition that can cause crashes at client or exit relay
Roger Dingledine [Mon, 23 Nov 2009 15:13:50 +0000 (10:13 -0500)] 
fix race condition that can cause crashes at client or exit relay

Avoid crashing if the client is trying to upload many bytes and the
circuit gets torn down at the same time, or if the flip side
happens on the exit relay. Bugfix on 0.2.0.1-alpha; fixes bug 1150.

16 years agoadd the 0.2.1.20 changelog blurb, plus update the releasenotes
Roger Dingledine [Tue, 17 Nov 2009 20:35:14 +0000 (15:35 -0500)] 
add the 0.2.1.20 changelog blurb, plus update the releasenotes

16 years agoMerge commit 'debian-tor-0.2.1.20-1' into maint-0.2.1
Nick Mathewson [Fri, 13 Nov 2009 20:47:18 +0000 (15:47 -0500)] 
Merge commit 'debian-tor-0.2.1.20-1' into maint-0.2.1

16 years agoMerge branch 'debian-merge' into debian-0.2.1
Peter Palfrader [Fri, 13 Nov 2009 18:58:59 +0000 (19:58 +0100)] 
Merge branch 'debian-merge' into debian-0.2.1

* debian-merge: (37 commits)
  New upstream version
  bump to 0.2.1.20
  Move moria1 and Tonga to alternate IP addresses.
  read the "circwindow" parameter from the consensus
  Code to parse and access network parameters.
  Revert "Teach connection_ap_can_use_exit about Exclude*Nodes"
  Work around a memory leak in openssl 0.9.8g (and maybe others)
  Teach connection_ap_can_use_exit about Exclude*Nodes
  make some bug 1090 warnings go away
  Fix a memory leak when parsing a ns
  Fix obscure 64-bit big-endian hidserv bug
  turns out the packaging changes aren't in 0.2.1.20
  update changelog with bundle details
  Use an _actual_ fix for the byte-reverse warning.
  Use a simpler fix for the byte-reversing warning
  Fix compile warnings on Snow Leopard
  Add getinfo accepted-server-descriptor. Clean spec.
  Reduce log level for bug case that we now know really exists.
  Only send reachability status events on overall success/failure
  update the README instructions and OS X makefiles
  ...

16 years agoNew upstream version
Peter Palfrader [Fri, 13 Nov 2009 18:57:10 +0000 (19:57 +0100)] 
New upstream version

16 years agoMerge commit 'tor-0.2.1.20' into debian-merge
Peter Palfrader [Fri, 13 Nov 2009 18:01:22 +0000 (19:01 +0100)] 
Merge commit 'tor-0.2.1.20' into debian-merge

* commit 'tor-0.2.1.20': (36 commits)
  bump to 0.2.1.20
  Move moria1 and Tonga to alternate IP addresses.
  read the "circwindow" parameter from the consensus
  Code to parse and access network parameters.
  Revert "Teach connection_ap_can_use_exit about Exclude*Nodes"
  Work around a memory leak in openssl 0.9.8g (and maybe others)
  Teach connection_ap_can_use_exit about Exclude*Nodes
  make some bug 1090 warnings go away
  Fix a memory leak when parsing a ns
  Fix obscure 64-bit big-endian hidserv bug
  turns out the packaging changes aren't in 0.2.1.20
  update changelog with bundle details
  Use an _actual_ fix for the byte-reverse warning.
  Use a simpler fix for the byte-reversing warning
  Fix compile warnings on Snow Leopard
  Add getinfo accepted-server-descriptor. Clean spec.
  Reduce log level for bug case that we now know really exists.
  Only send reachability status events on overall success/failure
  update the README instructions and OS X makefiles
  Avoid segfault when accessing hidden service.
  ...

16 years agoMake Tor work with OpenSSL 0.9.8l
Nick Mathewson [Thu, 5 Nov 2009 23:13:08 +0000 (18:13 -0500)] 
Make Tor work with OpenSSL 0.9.8l

To fix a major security problem related to incorrect use of
SSL/TLS renegotiation, OpenSSL has turned off renegotiation by
default.  We are not affected by this security problem, however,
since we do renegotiation right.  (Specifically, we never treat a
renegotiated credential as authenticating previous communication.)
Nevertheless, OpenSSL's new behavior requires us to explicitly
turn renegotiation back on in order to get our protocol working
again.

Amusingly, this is not so simple as "set the flag when you create
the SSL object" , since calling connect or accept seems to clear
the flags.

For belt-and-suspenders purposes, we clear the flag once the Tor
handshake is done.  There's no way to exploit a second handshake
either, but we might as well not allow it.

16 years agoFix an apparently bogus check; fortunately, it seems to be untriggered.
Nick Mathewson [Tue, 27 Oct 2009 03:14:53 +0000 (23:14 -0400)] 
Fix an apparently bogus check; fortunately, it seems to be untriggered.

16 years agoAdd changelog to describe coverity fixes for 0.2.1.21
Nick Mathewson [Tue, 27 Oct 2009 02:39:42 +0000 (22:39 -0400)] 
Add changelog to describe coverity fixes for 0.2.1.21

16 years agoFix two memory leaks found by Coverity (CIDs 417-418)
Nick Mathewson [Tue, 27 Oct 2009 02:12:40 +0000 (22:12 -0400)] 
Fix two memory leaks found by Coverity (CIDs 417-418)

The first happens on an error case when a controller wants an
impossible directory object.  The second happens when we can't write
our fingerprint file.

16 years agoAdd missing break statements for Coverity CIDs #406,407.
Nick Mathewson [Tue, 27 Oct 2009 01:35:26 +0000 (21:35 -0400)] 
Add missing break statements for Coverity CIDs #406,407.

The code for these was super-wrong, but will only break things when we
reset an option on a platform where sizeof(time_t) is different from
sizeof(int).

16 years agoOnly send the if_modified_since header for a v3 consensus.
Nick Mathewson [Tue, 27 Oct 2009 00:12:52 +0000 (20:12 -0400)] 
Only send the if_modified_since header for a v3 consensus.

Spotted by xmux; bugfix on 0.2.0.10-alpha.
(Bug introduced by 20b10859)

16 years agobump to 0.2.1.20 tor-0.2.1.20
Roger Dingledine [Thu, 15 Oct 2009 16:00:35 +0000 (12:00 -0400)] 
bump to 0.2.1.20

16 years agoMove moria1 and Tonga to alternate IP addresses.
Roger Dingledine [Wed, 14 Oct 2009 23:36:08 +0000 (19:36 -0400)] 
Move moria1 and Tonga to alternate IP addresses.

16 years agoMerge commit 'phobos/polipo-updates-maint-0.2.1' into maint-0.2.1
Nick Mathewson [Thu, 15 Oct 2009 01:30:53 +0000 (21:30 -0400)] 
Merge commit 'phobos/polipo-updates-maint-0.2.1' into maint-0.2.1

16 years agoread the "circwindow" parameter from the consensus
Roger Dingledine [Wed, 14 Oct 2009 21:07:32 +0000 (17:07 -0400)] 
read the "circwindow" parameter from the consensus

backport of c43859c5c12361fad505
backport of 0d13e0ed145f4c1b5bd1

16 years agoCode to parse and access network parameters.
Nick Mathewson [Wed, 14 Oct 2009 20:15:41 +0000 (16:15 -0400)] 
Code to parse and access network parameters.

Partial backport of 381766ce4b1145460.
Partial backport of 56c6d78520a98fb64.

16 years agoRevert "Teach connection_ap_can_use_exit about Exclude*Nodes"
Roger Dingledine [Mon, 21 Sep 2009 03:50:48 +0000 (23:50 -0400)] 
Revert "Teach connection_ap_can_use_exit about Exclude*Nodes"

This reverts commit dc3229313b6d2aaff437c6fc7fa55ead4409e93d.

We're going to do this more thoroughly in 0.2.2.x, and not in
maint-0.2.1.

16 years agoWork around a memory leak in openssl 0.9.8g (and maybe others)
Nick Mathewson [Thu, 17 Sep 2009 04:01:20 +0000 (00:01 -0400)] 
Work around a memory leak in openssl 0.9.8g (and maybe others)

16 years agoMerge commit 'karsten/fix-1073' into maint-0.2.1
Nick Mathewson [Thu, 17 Sep 2009 03:36:01 +0000 (23:36 -0400)] 
Merge commit 'karsten/fix-1073' into maint-0.2.1

16 years agoTeach connection_ap_can_use_exit about Exclude*Nodes
Sebastian Hahn [Wed, 16 Sep 2009 00:23:04 +0000 (02:23 +0200)] 
Teach connection_ap_can_use_exit about Exclude*Nodes

To further attempt to fix bug 1090, make sure connection_ap_can_use_exit
always returns 0 when the chosen exit router is excluded. This should fix
bug1090.

16 years agoMerge commit 'sebastian/memleak' into maint-0.2.1
Nick Mathewson [Tue, 15 Sep 2009 23:35:16 +0000 (19:35 -0400)] 
Merge commit 'sebastian/memleak' into maint-0.2.1

16 years agomake some bug 1090 warnings go away
Sebastian Hahn [Fri, 11 Sep 2009 13:40:09 +0000 (15:40 +0200)] 
make some bug 1090 warnings go away

When we excluded some Exits, we were sometimes warning the user that we
were going to use the node regardless. Many of those warnings were in
fact bogus, because the relay in question was not used to connect to
the outside world.

Based on patch by Rotor, thanks!

16 years agoFix a memory leak when parsing a ns
Sebastian Hahn [Sun, 13 Sep 2009 19:47:55 +0000 (21:47 +0200)] 
Fix a memory leak when parsing a ns

Adding the same vote to a networkstatus consensus leads to a memory leak
on the client side. Fix that by only using the first vote from any given
voter, and ignoring the others.

Problem found by Rotor, who also helped writing the patch. Thanks!

16 years agoFix obscure 64-bit big-endian hidserv bug
Roger Dingledine [Thu, 3 Sep 2009 00:36:11 +0000 (20:36 -0400)] 
Fix obscure 64-bit big-endian hidserv bug

Fix an obscure bug where hidden services on 64-bit big-endian
systems might mis-read the timestamp in v3 introduce cells, and
refuse to connect back to the client. Discovered by "rotor".
Bugfix on 0.2.1.6-alpha.

16 years agoturns out the packaging changes aren't in 0.2.1.20
Roger Dingledine [Wed, 2 Sep 2009 02:46:41 +0000 (22:46 -0400)] 
turns out the packaging changes aren't in 0.2.1.20

16 years agoupdate changelog with bundle details
Andrew Lewman [Tue, 1 Sep 2009 19:28:03 +0000 (15:28 -0400)] 
update changelog with bundle details

16 years agoUse an _actual_ fix for the byte-reverse warning.
Nick Mathewson [Tue, 1 Sep 2009 19:51:09 +0000 (15:51 -0400)] 
Use an _actual_ fix for the byte-reverse warning.

(Given that we're pretty much assuming that int is 32 bits, and given that
hex values are always unsigned, taking out the "ul" from 0xff000000 should
be fine.)

16 years agoUse a simpler fix for the byte-reversing warning
Nick Mathewson [Tue, 1 Sep 2009 19:41:38 +0000 (15:41 -0400)] 
Use a simpler fix for the byte-reversing warning

16 years agoFix compile warnings on Snow Leopard
Sebastian Hahn [Tue, 1 Sep 2009 03:23:47 +0000 (05:23 +0200)] 
Fix compile warnings on Snow Leopard

Big thanks to nickm and arma for helping me with this!

16 years agoAdd getinfo accepted-server-descriptor. Clean spec.
Roger Dingledine [Mon, 31 Aug 2009 22:37:25 +0000 (18:37 -0400)] 
Add getinfo accepted-server-descriptor. Clean spec.

Add a "getinfo status/accepted-server-descriptor" controller
command, which is the recommended way for controllers to learn
whether our server descriptor has been successfully received by at
least on directory authority. Un-recommend good-server-descriptor
getinfo and status events until we have a better design for them.

16 years agoReduce log level for bug case that we now know really exists.
Karsten Loesing [Mon, 31 Aug 2009 22:16:33 +0000 (00:16 +0200)] 
Reduce log level for bug case that we now know really exists.

16 years agoOnly send reachability status events on overall success/failure
Roger Dingledine [Mon, 31 Aug 2009 20:14:41 +0000 (16:14 -0400)] 
Only send reachability status events on overall success/failure

We were telling the controller about CHECKING_REACHABILITY and
REACHABILITY_FAILED status events whenever we launch a testing
circuit or notice that one has failed. Instead, only tell the
controller when we want to inform the user of overall success or
overall failure. Bugfix on 0.1.2.6-alpha. Fixes bug 1075. Reported
by SwissTorExit.

16 years agoupdate the README instructions and OS X makefiles
phobos [Mon, 31 Aug 2009 00:40:02 +0000 (20:40 -0400)] 
update the README instructions and OS X makefiles

16 years agoAvoid segfault when accessing hidden service.
Karsten Loesing [Sat, 29 Aug 2009 17:41:08 +0000 (19:41 +0200)] 
Avoid segfault when accessing hidden service.

16 years agoOnly send netinfo clock_skew to controller if an authority told us so
Roger Dingledine [Fri, 28 Aug 2009 07:42:09 +0000 (03:42 -0400)] 
Only send netinfo clock_skew to controller if an authority told us so

We were triggering a CLOCK_SKEW controller status event whenever
we connect via the v2 connection protocol to any relay that has
a wrong clock. Instead, we should only inform the controller when
it's a trusted authority that claims our clock is wrong. Bugfix
on 0.2.0.20-rc; starts to fix bug 1074. Reported by SwissTorExit.

16 years agoextremely infinite? who talks like that?
Roger Dingledine [Fri, 28 Aug 2009 01:54:41 +0000 (21:54 -0400)] 
extremely infinite? who talks like that?

16 years agonobody forward-ported the 0.2.0.35 changelog
Roger Dingledine [Wed, 26 Aug 2009 18:11:53 +0000 (14:11 -0400)] 
nobody forward-ported the 0.2.0.35 changelog

16 years agoMerge commit 'phobos/vidalia-bundle-updates-maint-0.2.1' into maint-0.2.1
Nick Mathewson [Tue, 25 Aug 2009 20:36:39 +0000 (16:36 -0400)] 
Merge commit 'phobos/vidalia-bundle-updates-maint-0.2.1' into maint-0.2.1

16 years agoupdate osx-dmg creation directions for the new methods
phobos [Tue, 25 Aug 2009 19:58:25 +0000 (15:58 -0400)] 
update osx-dmg creation directions for the new methods

16 years agoFix a rare infinite-recursion bug when shutting down.
Nick Mathewson [Thu, 20 Aug 2009 15:51:34 +0000 (11:51 -0400)] 
Fix a rare infinite-recursion bug when shutting down.

Once we had called log_free_all(), anything that tried to log a
message (like a failed tor_assert()) would fail like this:

   1. The logging call eventually invokes the _log() function.
   2. _log() calls tor_mutex_lock(log_mutex).
   3. tor_mutex_lock(m) calls tor_assert(m).
   4. Since we freed the log_mutex, tor_assert() fails, and tries to
      log its failure.
   5. GOTO 1.

Now we allocate the mutex statically, and never destroy it on
shutdown.

Bugfix on 0.2.0.16-alpha, which introduced the log mutex.

This bug was found by Matt Edman.