We were only using it when smartlist_choose_node_by_bandwidth_weights
failed. But that function could only fail in the presence of
buggy/ancient authorities or in the absence of a consensus. Either
way, it's better to use sensible defaults and a nicer algorithm.
Now, if a router ever changes its microdescriptor, but the new
microdescriptor SHA256 hash has the same 160-bit prefix as the old
one, we treat it as a new microdescriptor when deciding whether to
copy status information.
(This function also is used to compare SHA1 digests of router
descriptors, but don't worry: the descriptor_digest field either holds
a SHA256 hash, or a SHA1 hash padded with 0 bytes.)
Silence clang warnings under --enable-expensive-hardening, including:
+ implicit truncation of 64 bit values to 32 bit;
+ const char assignment to self;
+ tautological compare; and
+ additional parentheses around equality tests. (gcc uses these to
silence assignment, so clang warns when they're present in an
equality test. But we need to use extra parentheses in macros to
isolate them from other code).
I concatenated the remaining changes/* files, removed them, made the
headings more uniform, then told format_changelog.py to sort,
collate, and wrap them.
By now, support in the network is widespread and it's time to require
more modern crypto on all Tor instances, whether they're clients or
servers. By doing this early in 0.2.6, we can be sure that at some point
all clients will have reasonable support.
Ensure we securely wipe keys from memory after
crypto_digest_get_digest and init_curve25519_keypair_from_file
have finished using them.
Fixes bug 13477.
Add unit tests for tor_timegm signed overflow,
tor_timegm and parse_rfc1123_time validity checks,
and correct_tm year clamping.
Unit tests (visible) fixes in bug 13476.
Check all date/time values passed to tor_timegm
and parse_rfc1123_time for validity, taking leap
years into account.
Improves HTTP header validation.
Avoid unlikely signed integer overflow in tor_timegm
on systems with 32-bit time_t.
Fixes bug 13476.
Also, refactor the way we handle failed handshakes so that this
warning doesn't propagate itself to "onion_skin_client_handshake
failed" and "circuit_finish_handshake failed" and
"connection_edge_process_relay_cell (at origin) failed."
Resolves warning from 9635.
One pain point in evolving the Tor design and implementing has been
adding code that makes clients reject directory documents that they
previously would have accepted, if those descriptors actually exist.
When this happened, the clients would get the document, reject it,
and then decide to try downloading it again, ad infinitum. This
problem becomes particularly obnoxious with authorities, since if
some authorities accept a descriptor that others don't, the ones
that don't accept it would go crazy trying to re-fetch it over and
over. (See for example ticket #9286.)
This patch tries to solve this problem by tracking, if a descriptor
isn't parseable, what its digest was, and whether it is invalid
because of some flaw that applies to the portion containing the
digest. (This excludes RSA signature problems: RSA signatures
aren't included in the digest. This means that a directory
authority can still put another directory authority into a loop by
mentioning a descriptor, and then serving that descriptor with an
invalid RSA signatures. But that would also make the misbehaving
directory authority get DoSed by the server it's attacking, so it's
not much of an issue.)
We already have a mechanism to mark something undownloadable with
downloadstatus_mark_impossible(); we use that here for
microdescriptors, extrainfos, and router descriptors.
Unit tests to follow in another patch.
Closes ticket #11243.
Fix an instance of integer overflow in format_time_interval() when
taking the absolute value of the supplied signed interval value.
Fixes bug 13393.
Create unit tests for format_time_interval().
Bitwise check for the BRIDGE_DIRINFO flag, rather than checking for
equality.
Fixes a (potential) bug where directories offering BRIDGE_DIRINFO,
and some other flag (i.e. microdescriptors or extrainfo),
would be ignored when looking for bridge directories.
Final fix in series for bug 13163.
Document usage of the NO_DIRINFO and ALL_DIRINFO flags clearly in functions
which take them as arguments. Replace 0 with NO_DIRINFO in a function call
for clarity.
Seeks to prevent future issues like 13163.
Stop using the default authorities in networks which provide both
AlternateDirAuthority and AlternateBridgeAuthority.
This bug occurred due to an ambiguity around the use of NO_DIRINFO.
(Does it mean "any" or "none"?)
Partially fixes bug 13163.
If (GNU) Make 3.81 is running processes in parallel using -j2 (or more),
it waits until all descendent processes have exited before it returns to
the shell.
When a command like "make -j2 test-network" is run, this means that
test-network.sh apparently hangs until it either make is forcibly
terminated, or all the chutney-launched tor processes have exited.
A workaround is to use make without -j, or make -j1 if there is an
existing alias to "make -jn" in the shell.
We resolve this bug in tor by using "chutney stop" after "chutney verify"
in test-network.sh.
Cases that now send errors:
* Malformed IP address (SOCKS5_GENERAL_ERROR)
* CONNECT/RESOLVE request with IP, when SafeSocks is set
(SOCKS5_NOT_ALLOWED)
* RESOLVE_PTR request with FQDN (SOCKS5_ADDRESS_TYPE_NOT_SUPPORTED)
* Malformed FQDN (SOCKS5_GENERAL_ERROR)
* Unknown address type (SOCKS5_ADDRESS_TYPE_NOT_SUPPORTED)
Fixes bug 13314.
Add a --delay option to test-network.sh, which configures the delay before
the chutney network tests for data transmission. The default remains at
18 seconds if the argument isn't specified.
Apparently we should be using bootstrap status for this (eventually).
Partially implements ticket 13161.
The default shell on OS X is bash, which has a builtin echo. When called
in "sh" mode, this echo does not accept "-n". This patch uses "/bin/echo -n"
instead.
Partially fixes issue 13161.
Add the TestingDirAuthVoteExit option, a list of nodes to vote Exit for,
regardless of their uptime, bandwidth, or exit policy.
TestingTorNetwork must be set for this option to have any effect.
Works around an issue where authorities would take up to 35 minutes to
give nodes the Exit flag in a test network, despite short consensus
intervals. Partially implements ticket 13161.
Fixes bug 13295; bugfix on 0.2.5.3-alpha.
The alternative here is to call crypto_global_init() from tor-resolve,
but let's avoid linking openssl into tor-resolve for as long as we
can.
When a spawned process forks, fails, then exits very quickly, (this
typically occurs when exec fails), there is a race condition between the
SIGCHLD handler updating the process_handle's fields, and checking the
process status in those fields. The update can occur before or after the
spawn tests check the process status.
We check whether the process is running or not running (rather than just
checking if it is running) to avoid this issue.
In circuit_build_times_calculate_timeout() in circuitstats.c, avoid dividing
by zero in the pareto calculations.
If either the alpha or p parameters are 0, we would divide by zero, yielding
an infinite result; which would be clamped to INT32_MAX anyway. So rather
than dividing by zero, we just skip the offending calculation(s), and
use INT32_MAX for the result.
Division by zero traps under clang -fsanitize=undefined-trap -fsanitize-undefined-trap-on-error.
Avoid 4 null pointer errors under clang shallow analysis (the default when
building under Xcode) by using tor_assert() to prove that the pointers
aren't null. Resolves issue 13284 via minor code refactoring.
c99 lets us do neat stuff like:
{
int j, k;
foo(&j, &k);
int z = j + k;
}
and also
struct point { int x; int y; };
struct point pt = { .x=5, .y=5 };
This commit makes the configure scripts check to make sure your
compiler implements them. It also disables our longstanding warning
about midblock declarations.
Closes ticket 13233.
When DisableNetwork is set, do not launch pluggable transport plugins,
and if any are running already, terminate the existing instances.
Resolves ticket 13213.
Allow clients to use optimistic data when connecting to a hidden service,
which should cut out the initial round-trip for client-side programs
including Tor Browser.
(Now that Tor 0.2.2.x is obsolete, all hidden services should support
server-side optimistic data.)
See proposal 181 for details. Implements ticket 13211.
Clients are now willing to send optimistic circuit data (before they
receive a 'connected' cell) to relays of any version. We used to
only do it for relays running 0.2.3.1-alpha or later, but now all
relays are new enough.
Resolves ticket 13153.
Return an error when the second or later arguments of the
"setevents" controller command are invalid events. Previously we
would return success while silently skipping invalid events.
Fixes bug 13205; bugfix on 0.2.3.2-alpha. Reported by "fpxnns".
Note that this will likely need to be folded with the changes file for #12751,
as this change is a mere fixup on top of the changes introduced for #12751.
Stop modifying the value of our DirReqStatistics torrc option just
because we're not a bridge or relay. This bug was causing Tor
Browser users to write "DirReqStatistics 0" in their torrc files
as if they had chosen to change the config.
Fixes bug 4244; bugfix on 0.2.3.1-alpha.
Clients now send the correct address for their chosen rendezvous point
when trying to access a hidden service. They used to send the wrong
address, which would still work some of the time because they also
sent the identity digest of the rendezvous point, and if the hidden
service happened to try connecting to the rendezvous point from a relay
that already had a connection open to it, the relay would reuse that
connection. Now connections to hidden services should be more robust
and faster. Also, this bug meant that clients were leaking to the hidden
service whether they were on a little-endian (common) or big-endian (rare)
system, which for some users might have reduced their anonymity.
Fixes bug 13151; bugfix on 0.2.1.5-alpha.
Tor Browser includes several ClientTransportPlugin lines in its
torrc-defaults file, leading every Tor Browser user who looks at her
logs to see these notices and wonder if they're dangerous.
Resolves bug 13124; bugfix on 0.2.5.3-alpha.
Technically, we're not allowed to take the address of a member can't
exist relative to the null pointer. That makes me wonder how any sane
compliant system implements the offsetof macro, but let's let sleeping
balrogs lie.
Fixes 13096; patch on 0.1.1.9-alpha; patch from "teor", who was using
clang -fsanitize=undefined-trap -fsanitize-undefined-trap-on-error -ftrapv
(And replay them once we know our first real logs.)
This is an implementation for issue 6938. It solves the problem of
early log mesages not getting sent to log files, but not the issue of
early log messages not getting sent to controllers.
This fixes bug 13102 (not on any released Tor) where using the
standard SSIZE_MAX name broke mingw64, and we didn't realize.
I did this with
perl -i -pe 's/SIZE_T_MAX/SIZE_MAX/' src/*/*.[ch] src/*/*/*.[ch]
This implements the meat of #12899. This commit should simply remove the
parts of Tor dirauths used to check whether a relay was supposed to be
named or not, it doesn't yet convert to a new mechanism for
reject/invalid/baddir/badexiting relays.
Back in 078d6bcd, we added an event number 0x20, but we didn't make
the event_mask field big enough to compensate.
Patch by "teor". Fixes 13085; bugfix on 0.2.5.1-alpha.
This is in preparation for a big patch series removing the entire Naming
system from Tor. In its wake, the approved-routers file is being
deprecated, and a replacement option to allow only pre-approved routers
is not being implemented.
Otherwise, when we're out of input *and* finalizing, we might report
TOR_ZLIB_OK erroneously and not finalize the buffer.
(I don't believe this can happen in practice, with our code today:
write_to_buf_zlib ensures that we are never trying to write into a
completely empty buffer, and zlib says "Z_OK" if you give it even
one byte to write into.)
Fixes bug 11824; bugfix on 0.1.1.23 (06e09cdd47).
torrc.minimal is now the one that should change as infrequently as
possible. To schedule an change to go into it eventually, make your
change to torrc.minimal.in-sample.
torrc.sample is now the volatile one: we can change it to our hearts'
content.
Closes ticket #11144
This implements a feature from bug 13000. Instead of starting a bwauth
run with this wrong idea about their bw, relays should do the self-test
and then get measured.
When a tor relay starts up and has no historical information about its
bandwidth capability, it uploads a descriptor with a bw estimate of 0.
It then starts its bw selftest, but has to wait 20 minutes to upload the
next descriptor due to the MAX_BANDWIDTH_CHANGE_FREQ delay. This change
should mean that on average, relays start seeing meaningful traffic a
little quicker, since they will have a higher chance to appear in the
consensus with a nonzero bw.
Patch by Roger, changes file and comment by Sebastian.
Most of these are in somewhat non-obvious code where it is probably
a good idea to initialize variables and add extra assertions anyway.
Closes 13036. Patches from "teor".
It's now a protocol-warn, since there's nothing relay operators can
do about a client that sends them a malformed create cell.
Resolves bug 12996; bugfix on 0.0.6rc1.
Any error when acquiring the CryptoAPI context should get treated as
bad. Also, this one can't happen for the arguments we're giving.
Fixes bug 10816; bugfix on 0.0.2pre26.
Previously, we had done this only in the connection_free() case, but
when we called connection_free_() directly from
connections_free_all(), we didn't free the connections.
The fix for bug 4647 accidentally removed our hack from bug 586 that
rewrote HashedControlPassword to __HashedControlSessionPassword when
it appears on the commandline (which allowed the user to set her own
HashedControlPassword in the torrc file while the controller generates
a fresh session password for each run).
Fixes bug 12948; bugfix on 0.2.5.1-alpha.
This way, we don't get locking failures when we hit an assertion in
the unit tests. Also, we might find out about unit test bugs from
folks who can't do gdb.
Two bugs here:
1) We didn't add EXTEND2/EXTENDED2 to relay_command_to_string().
2) relay_command_to_string() didn't log the value of unrecognized
commands.
Both fixed here.
When we merged the cookieauthfile creation logic in 33c3e60a37, we
accidentally took out this feature. Fixes bug 12864, bugfix on
0.2.5.1-alpha.
Also adds an ExtORPortCookieAuthFileGroupReadable, since there's no
reason not to.
We added some AS_VAR_IF-based checks to detect whether we have
managed to compile (but not link) with stack-protector. On autoconf
before 2.63, we don't have AS_VAR_IF, so we just have to let the
user get a compile error rather than a helpful "find libssp" error.
Fixes bug 12693; bugfix on 0.2.5.2-alpha (commit 21ac292820)
When Tor starts with DisabledNetwork set, it would correctly
conclude that it shouldn't try making circuits, but it would
mistakenly cache this conclusion and continue believing it even
when DisableNetwork is set to 0. Fixes the bug introduced by the
fix for bug 11200; bugfix on 0.2.5.4-alpha.
Our current systemd unit uses "Type = simple", so systemd does not expect tor to
fork. If the user has "RunAsDaemon 1" in their torrc, then things won't work as
expected. This is e.g. the case on Debian (and derivatives), since there we pass
"--defaults-torrc /usr/share/tor/tor-service-defaults-torrc" (that contains
"RunAsDaemon 1") by default.
The only solution I could find is to explicitly pass "--RunAsDaemon 0" when
starting tor from the systemd unit file, which this commit does.
Those used to be normal to receive on hidden service circuits due to bug
1038, but the buggy Tor versions are long gone from the network so we
can afford to resume watching for them. Resolves the rest of bug 1038;
bugfix on 0.2.1.19.
This function is supposed to construct a list of all the ciphers in
the "v2 link protocol cipher list" that are supported by Tor's
openssl. It does this by invoking ssl23_get_cipher_by_char on each
two-byte ciphersuite ID to see which ones give a match. But when
ssl23_get_cipher_by_char cannot find a match for a two-byte SSL3/TLS
ciphersuite ID, it checks to see whether it has a match for a
three-byte SSL2 ciphersuite ID. This was causing a read off the end
of the 'cipherid' array.
This was probably harmless in practice, but we shouldn't be having
any uninitialized reads.
(Using ssl23_get_cipher_by_char in this way is a kludge, but then
again the entire existence of the v2 link protocol is kind of a
kludge. Once Tor 0.2.2 clients are all gone, we can drop this code
entirely.)
Found by starlight. Fix on 0.2.4.8-alpha. Fixes bug 12227.
Authorities now assign the Guard flag to the fastest 25% of the
network (it used to be the fastest 50%). Also raise the consensus
weight that guarantees the Guard flag from 250 to 2000. For the
current network, this results in about 1100 guards, down from 2500.
This step paves the way for moving the number of entry guards
down to 1 (proposal 236) while still providing reasonable expected
performance for most users.
Implements ticket 12690.
Found because LibreSSL has OPENSSL_NO_COMP always-on, but this
conflicts with the way that _we_ turn off compression. Patch from
dhill, who attributes it to "OpenBSD". Fixes bug 12602; bugfix on
0.2.1.1-alpha, which introduced this turn-compression-off code.
- Don't try to rm -rf the directory before we start: somebody might
have set it to ~, which would be quite sad.
- Always quote the directory name
- Use 'make reset-gcov' before running tests.
- Use 'make check', not ./src/test/test
Otherwise, it always seems as though our Exclude* options have
changed, since we're comparing modified to unmodified values.
Patch from qwerty1. Fixes bug 9801. Bugfix on 0.2.4.10-alpha, where
GeoIPExcludeUnknown was introduced.
We should only assign a relay the HSDir flag if it is currently
considered valid. We can accomplish this by only considering active
relays, and as a consequence of this we also exclude relays that are
currently hibernating. Fixes#12573
Long ago we supported systems where there was no support for
threads, or where the threading library was broken. We shouldn't
have do that any more: on every OS that matters, threads exist, and
the OS supports running threads across multiple CPUs.
This resolves tickets 9495 and 12439. It's a prerequisite to making
our workqueue code work better, since sensible workqueue
implementations don't split across multiple processes.
I don't know whether we missed these or misclassified them when we
first made the "DIRECTORY AUTHORITY SERVER OPTIONS" section, but they
really belong there.
Check for consistency between the queued destroy cells and the marked
circuit IDs. Check for consistency in the count of queued destroy
cells in several ways. Check to see whether any of the marked circuit
IDs have somehow been marked longer than the channel has existed.
And add a comment about why conditions that would cause us to drop a
cell should get checked before actions that would cause us to send a
destroy cell.
Spotted by 'cypherpunks'.
And note that these issues have been present since 0.0.8pre1 (commit
0da256ef), where we added a "shutting down" state, and started
responding to all create cells with DESTROY when shutting down.
Conflicts:
src/or/channel.c
src/or/circuitlist.c
src/or/connection.c
Conflicts involved removal of next_circ_id and addition of
unusable-circid tracking.
The point of the "idle timeout" for connections is to kill the
connection a while after it has no more circuits. But using "last
added a non-padding cell" as a proxy for that is wrong, since if the
last circuit is closed from the other side of the connection, we
will not have sent anything on that connection since well before the
last circuit closed.
This is part of fixing 6799.
When applied to 0.2.5, it is also a fix for 12023.
Instead of killing an or_connection_t that has had no circuits for
the last 3 minutes, give every or_connection_t a randomized timeout,
so that an observer can't so easily infer from the connection close
time the time at which its last circuit closed.
Also, increase the base timeout for canonical connections from 3
minutes to 15 minutes.
Fix for ticket 6799.
When we find a stranded one-hop circuit, log whether it is dirty,
log information about any streams on it, and log information about
connections they might be linked to.
This function is supposed to construct a list of all the ciphers in
the "v2 link protocol cipher list" that are supported by Tor's
openssl. It does this by invoking ssl23_get_cipher_by_char on each
two-byte ciphersuite ID to see which ones give a match. But when
ssl23_get_cipher_by_char cannot find a match for a two-byte SSL3/TLS
ciphersuite ID, it checks to see whether it has a match for a
three-byte SSL2 ciphersuite ID. This was causing a read off the end
of the 'cipherid' array.
This was probably harmless in practice, but we shouldn't be having
any uninitialized reads.
(Using ssl23_get_cipher_by_char in this way is a kludge, but then
again the entire existence of the v2 link protocol is kind of a
kludge. Once Tor 0.2.2 clients are all gone, we can drop this code
entirely.)
Found by starlight. Fix on 0.2.4.8-alpha. Fixes bug 12227.
This code mis-handled the case where a circuit got the same circuit
ID in both directions. I found three instances of it in the
codebase, by grepping for [pn]_circ_id.
Because of the issue in command_process_relay_cell(), this would
have made roughly one circuit in a million completely nonfunctional.
Fixes bug 12195.
On some profiles of Andrea's from #11332, I found that a great deal
of time can still be attributed to functions called from
update_router_have_minimum_dir_info(). This is making our
digestmap, tor_memeq, and siphash functions take a much bigger
portion of runtime than they really should.
If we're calling update_router_have_minimum_dir_info() too often,
that's because we're calling router_dir_info_changed() too often.
And it looks like most of the callers of router_dir_info_changed()
are coming as tail-calls from router_set_status() as invoked by
channel_do_open_actions().
But we don't need to call router_dir_info_changed() so much! (I'm
not quite sure we need to call it from here at all, but...) Surely
we don't need to call it from router_set_status when the router's
status has not actually changed.
This patch makes us call router_dir_info_changed() from
router_set_status only when we are changing the router's status.
Fix for bug 12170. This is leftover from our fix back in 273ee3e81
in 0.1.2.1-alpha, where we started caching the value of
update_router_have_minimum_dir_info().
tor_memeq has started to show up on profiles, and this is one of the
most frequent callers of that function, appearing as it does on every
cell handled for entry or exit.
59f9097d5c introduced tor_memneq here;
it went into Tor 0.2.1.31. Fixes part of 12169.
I've copied the entries from changes/, labeled the ones that also
appeared in 0.2.4.22, sorted them lightly with a python script
(added to maint), and combined sections with the same name.
I didn't combine sections without a description (e.g. "Minor
bugfixes:"), since we'll probably add a description to those.
Without this fix, when running with bridges, we would try fetching
directory info far too early, and have up to a 60 second delay if we
started with bridge descriptors available.
Fixes bug 11965. Fix on 0.2.3.6-alpha, arma thinks.
The old cache had problems:
* It needed to be manually preloaded. (It didn't remember any
address you didn't tell it to remember)
* It was AF_INET only.
* It looked at its cache even if the sandbox wasn't turned on.
* It couldn't remember errors.
* It had some memory management problems. (You can't use memcpy
to copy an addrinfo safely; it has pointers in.)
This patch fixes those issues, and moves to a hash table.
Fixes bug 11970; bugfix on 0.2.5.1-alpha.
These are needed under some circumstances if we are running with
expensive-hardening and sandbox at the same time.
fixes 11477, bugfix on 0.2.5.4-alpha (where we introduced
expensive-hardening)
None of the things we might exec() can possibly run under the
sanbox, so rather than crash later, we have to refuse to accept the
configuration nice and early.
The longer-term solution is to have an exec() helper, but wow is
that risky.
fixes 12043; bugfix on 0.2.5.1-alpha
When we converted the horrible set of options that previously
controlled "use ORPort or DirPort? Anonymously or Non-anonymouly?" to
a single 'indirection' argument, we missed
directory_post_to_dirservers.
The problematic code was introduced in 5cbeb6080, which went into
0.2.4.3-alpha. This is a fix for bug 11469.
When running with User set, we frequently try to look up our
information in the user database (e.g., /etc/passwd). The seccomp2
sandbox setup doesn't let us open /etc/passwd, and probably
shouldn't.
To fix this, we have a pair of wrappers for getpwnam and getpwuid.
When a real call to getpwnam or getpwuid fails, they fall back to a
cached value, if the uid/gid matches.
(Granting access to /etc/passwd isn't possible with the way we
handle opening files through the sandbox. It's not desirable either.)
On OpenBSD 5.4, time_t is a 32-bit integer. These instances contain
implicit treatment of long and time_t as comparable types, so explicitly
cast to time_t.
Apparently, there exist cross-compiling environments for arm7 where
you can compile a 64x64->128 multiply, but not link it.
Fixes bug 11729; bugfix on 0.2.4.8-alpha. Patch from 'conradev'.
Clients should always believe that v3 directory authorities serve
extra-info documents, regardless of whether their server descriptor
contains a "caches-extra-info" line or not.
Fixes part of #11683.
on #9686, gmorehose reports that the 500 MB lower limit is too high
for raspberry pi users.
This is a backport of 647248729f to 0.2.4.
Note that in 0.2.4, the option is called MaxMemInCellQueues.
When clearing a list of tokens, it's important to do token_clear()
on them first, or else any keys they contain will leak. This didn't
leak memory on any of the successful microdescriptor parsing paths,
but it does leak on some failing paths when the failure happens
during tokenization.
Fixes bug 11618; bugfix on 0.2.2.6-alpha.
The python scripts invoked by 'make check' didn't work on python3
before. That was a problem on systems where 'python' is python3.
Fixes bug 11608; bugfix on 0.2.5.2-alpha.
If we can't detect the physical memory, the new default is 8 GB on
64-bit architectures, and 1 GB on 32-bit architectures.
If we *can* detect the physical memory, the new default is
CLAMP(256 MB, phys_mem * 0.75, MAX_DFLT)
where MAX_DFLT is 8 GB on 64-bit architectures and 2 GB on 32-bit
architectures.
You can still override the default by hand. The logic here is simply
trying to choose a lower default value on systems with less than 12 GB
of physical RAM.
Previously we said "Sandbox is not implemented on this platform" on
Linux boxes without libseccomp. Now we say that you need to build
Tor built with libseccomp. Fixes bug 11543; bugfix on 0.2.5.1-alpha.
Fixes a possible root cause of 11553 by only making 64 attempts at
most to pick a circuitID. Previously, we would test every possible
circuit ID until we found one or ran out.
This algorithm succeeds probabilistically. As the comment says:
This potentially causes us to give up early if our circuit ID
space is nearly full. If we have N circuit IDs in use, then we
will reject a new circuit with probability (N / max_range) ^
MAX_CIRCID_ATTEMPTS. This means that in practice, a few percent
of our circuit ID capacity will go unused.
The alternative here, though, is to do a linear search over the
whole circuit ID space every time we extend a circuit, which is
not so great either.
This makes new vs old clients distinguishable, so we should try to
batch it with other patches that do that, like 11438.
The server cipher list is (thanks to #11513) chosen systematically to
put the best choices for Tor first. The client cipher list is chosen
to resemble a browser. So let's set SSL_OP_CIPHER_SERVER_PREFERENCE
to have the servers pick according to their own preference order.
This isn't on by default; to get it, you need to set "TransProxyType
ipfw". (The original patch had automatic detection for whether
/dev/pf is present and openable, but that seems marginally fragile.)
