We now require that:
- Only actual servers should ever call get_server_identity_key
- If you're being a client or bridge, the client and server keys should
differ.
- If you're being a public relay, the client and server keys
should be the same.
When intro->extend_info is created for an introduction point, it
only starts out with a nickname, not necessarily an identity digest.
Thus, doing router_get_by_digest isn't necessarily safe.
https://trac.torproject.org/projects/tor/ticket/1859
Use router_get_by_digest() instead of router_get_by_hexdigest()
in circuit_discard_optional_exit_enclaves() and
rend_client_get_random_intro(), per Nick's comments.
Using router_get_by_digest() in rend_client_get_random_intro() will
break hidden services published by Tor versions pre 0.1.2.18 and
0.2.07-alpha as they only publish by nickname. This is acceptable
however as these versions only publish to authority tor26 and
don't work for versions in the 0.2.2.x series anyway.
Mainly, this comes from turning two lists that needed to be kept in
synch into a single list of structs. This should save a little RAM,
and make the code simpler.
The old comment was from before I tried a huge pile of crazy stuff to
make the inner loop faster. Short answer: GCC already knows how to
unroll loops pretty well. Other short answer: we should have made the
relay payload size an even multiple of 4, 8, or ideally 16.
There's no reason to keep a time_t and a struct timeval to represent
the same value: highres_created.tv_sec was the same as timestamp_created.
This should save a few bytes per circuit.
The short version is, "where we want to do it, we have nothing real to
chose from and we can't do it easily. Where it's easy to do, we have
no reason to do it yet."
Our old code correctly called bufferevent_flush() on linked
connections to make sure that the other side got an EOF event... but
it didn't call bufferevent_flush() when the connection wasn't
hold_open_until_flushed. Directory connections don't use
hold_open_until_flushed, so the linked exit connection never got an
EOF, so they never sent a RELAY_END cell to the client, and the
client never concluded that data had arrived.
The solution is to make the bufferevent_flush() code apply to _all_
closing linked conns whose partner is not already marked for close.
https://trac.torproject.org/projects/tor/ticket/1859
There are two problems in this bug:
1. When an OP makes a .exit request specifying itself as the exit, and the exit
is not yet listed, Tor gets all the routerinfos needed for the circuit but
discovers in circuit_is_acceptable() that its own routerinfo is not in the
routerdigest list and cannot be used. Tor then gets locked in a cycle of
repeating these two steps. When gathering the routerinfos for a circuit,
specifically when the exit has been chosen by .exit notation, Tor needs to
apply the same rules it uses later on when deciding if it can build a
circuit with those routerinfos.
2. A different bug arises in the above situation when the Tor instance's
routerinfo *is* listed in the routerlist, it shares its nickname with a
number of other Tor nodes, and it does not have 'Named' rights to its
nickname.
So for example, if (i) there are five nodes named Bob in the network, (ii) I
am running one of them but am flagged as 'Unnamed' because someone else
claimed the 'Bob' nickname first, and (iii) I run my Tor as both client
and exit the following can happen to me:
- I go to www.evil.com
- I click on a link www.evil.com.bob.exit
- My request will exit through my own Tor node rather than the 'Named'
node Bob or any of the others.
- www.evil.com now knows I am actually browsing from the same computer
that is running my 'Bob' node
So to solve both issues we need to ensure:
- When fulfilling a .exit request we only choose a routerinfo if it exists in
the routerlist, even when that routerinfo is ours.
- When getting a router by nickname we only return our own router information
if it is not going to be used for building a circuit.
We ensure this by removing the special treatment afforded our own router in
router_get_by_nickname(). This means the function will only return the
routerinfo of our own router if it is in the routerlist built from authority
info and has a unique nickname or is bound to a non-unique nickname.
There are some uses of router_get_by_nickname() where we are looking for the
router by name because of a configuration directive, specifically local
declaration of NodeFamilies and EntryNodes and other routers' declaration of
MyFamily. In these cases it is not at first clear if we need to continue
returning our own routerinfo even if our router is not listed and/or has a
non-unique nickname with the Unnamed flag.
The patch treats each of these cases as follows:
Other Routers' Declaration of MyFamily
This happens in routerlist_add_family(). If another router declares our router
in its family and our router has the Unnamed flag or is not in the routerlist
yet, should we take advantage of the fact that we know our own routerinfo to
add us in anyway? This patch says 'no, treat our own router just like any
other'. This is a safe choice because it ensures our client has the same view
of the network as other clients. We also have no good way of knowing if our
router is Named or not independently of the authorities, so we have to rely on
them in this.
Local declaration of NodeFamilies
Again, we have no way of knowing if the declaration 'NodeFamilies
Bob,Alice,Ringo' refers to our router Bob or the Named router Bob, so we have
to defer to the authorities and treat our own router like any other.
Local declaration of NodeFamilies
Again, same as above. There's also no good reason we would want our client to
choose it's own router as an entry guard if it does not meet the requirements
expected of any other router on the network.
In order to reduce the possibility of error, the patch also replaces two
instances where we were using router_get_by_nickname() with calls to
router_get_by_hexdigest() where the identity digest of the router
is available.
First start of a fix for bug2001, but my test network still isn't
working: the client and the server send each other VERSIONS cells,
but never notice that they got them.
Currently the unit tests test_util_spawn_background_* assume that they
are run from the Tor build directory. This is not the case when running
make distcheck, so the test will fail. This problem is fixed by autoconf
setting BUILDDIR to be the root of the Tor build directory, and this
preprocessor variable being used to specify the absolute path to
test-child. Also, in test-child, do not print out argv[0] because this will
no longer be predictable. Found by Sebastian Hahn.
* Make tor_tls_context_new internal to tortls.c, and return the new
tor_tls_context_t from it.
* Add a public tor_tls_context_init wrapper function to replace it.
- Responsibility of clearing hex_errno is no longer with caller
- More conservative bounds checking
- Length requirement of hex_errno documented
- Output format documented
Some of these functions only work for routerinfo-based nodes, and as
such are only usable for advisory purposes. Fortunately, our uses
of them are compatible with this limitation.
Also, make the NodeFamily option into a list of routersets. This
lets us git rid of router_in_nickname_list (or whatever it was
called) without porting it to work with nodes, and also lets people
specify country codes and IP ranges in NodeFamily
This was the only flag in routerstatus_t that we would previously
change in a routerstatus_t in a consensus. We no longer have reason
to do so -- and probably never did -- as you can now confirm more
easily than you could have done by grepping for is_running before
this patch.
The name change is to emphasize that the routerstatus_t is_running
flag is only there to tell you whether the consensus says it's
running, not whether it *you* think it's running.
A node_t is an abstraction over routerstatus_t, routerinfo_t, and
microdesc_t. It should try to present a consistent interface to all
of them. There should be a node_t for a server whenever there is
* A routerinfo_t for it in the routerlist
* A routerstatus_t in the current_consensus.
(note that a microdesc_t alone isn't enough to make a node_t exist,
since microdescriptors aren't usable on their own.)
There are three ways to get a node_t right now: looking it up by ID,
looking it up by nickname, and iterating over the whole list of
microdescriptors.
All (or nearly all) functions that are supposed to return "a router"
-- especially those used in building connections and circuits --
should return a node_t, not a routerinfo_t or a routerstatus_t.
A node_t should hold all the *mutable* flags about a node. This
patch moves the is_foo flags from routerinfo_t into node_t. The
flags in routerstatus_t remain, but they get set from the consensus
and should not change.
Some other highlights of this patch are:
* Looking up routerinfo and routerstatus by nickname is now
unified and based on the "look up a node by nickname" function.
This tries to look only at the values from current consensus,
and not get confused by the routerinfo_t->is_named flag, which
could get set for other weird reasons. This changes the
behavior of how authorities (when acting as clients) deal with
nodes that have been listed by nickname.
* I tried not to artificially increase the size of the diff here
by moving functions around. As a result, some functions that
now operate on nodes are now in the wrong file -- they should
get moved to nodelist.c once this refactoring settles down.
This moving should happen as part of a patch that moves
functions AND NOTHING ELSE.
* Some old code is now left around inside #if 0/1 blocks, and
should get removed once I've verified that I don't want it
sitting around to see how we used to do things.
There are still some unimplemented functions: these are flagged
with "UNIMPLEMENTED_NODELIST()." I'll work on filling in the
implementation here, piece by piece.
I wish this patch could have been smaller, but there did not seem to
be any piece of it that was independent from the rest. Moving flags
forces many functions that once returned routerinfo_t * to return
node_t *, which forces their friends to change, and so on.
The node_t type is meant to serve two key functions:
1) Abstracting difference between routerinfo_t and microdesc_t
so that clients can use microdesc_t instead of routerinfo_t.
2) Being a central place to hold mutable state about nodes
formerly held in routerstatus_t and routerinfo_t.
This patch implements a nodelist type that holds a node for every
router that we would consider using.
* MINIUPNPC rather than the generic UPNP
* Nick suggested a better abstraction model for tor-fw-helper
* Fix autoconf to build with either natpmp or miniupnpc
* Add AM_PROG_CC_C_O to fix automake complaint
* update spec to address nickm's concern
* refactor nat-pmp to match upnp state
* we prefer tor_snprintf to snprintf
* link properlty for tor_snprintf
* rename test_commandline_options to log_commandline_options
* cast this uint as an int
* detect possible FD_SETSIZE errors
* make note about future enhancements for natpmp
* add upnp enhancement note
* ChangeLog entry
* doxygen and check-spaces cleanup
* create tor-fw-helper.1.txt
tor-fw-helper is a command-line tool to wrap and abstract various
firewall port-forwarding tools.
This commit matches the state of Jacob's tor-fw-helper branch as of
23 September 2010.
(commit msg by Nick)
When picking bridges (or other nodes without a consensus entry (and
thus no bandwidth weights)) we shouldn't just trust the node's
descriptor. So far we believed anything between 0 and 10MB/s, where 0
would mean that a node doesn't get any use from use unless it is our
only one, and 10MB/s would be a quite siginficant weight. To make this
situation better, we now believe weights in the range from 20kB/s to
100kB/s. This should allow new bridges to get use more quickly, and
means that it will be harder for bridges to see almost all our traffic.
In the first 100 circuits, our timeout_ms and close_ms
are the same. So we shouldn't transition circuits to purpose
CIRCUIT_PURPOSE_C_MEASURE_TIMEOUT, since they will just timeout again
next time we check.
We really should ignore any timeouts that have *no* network activity for their
entire measured lifetime, now that we have the 95th percentile measurement
changes. Usually this is up to a minute, even on fast connections.
If we really want all this complexity for these stages here, we need to handle
it better for people with large timeouts. It should probably go away, though.
Rechecking the timeout condition was foolish, because it is checked on the
same codepath. It was also wrong, because we didn't round.
Also, the liveness check itself should be <, and not <=, because we only have
1 second resolution.
Specifically, a circ attempt that we'd launched while the network was
down could timeout after we've marked our entrynodes up, marking them
back down again. The fix is to annotate as bad the OR conns that were
around before we did the retry, so if a circuit that's attached to them
times out we don't do anything about it.
This is needed for IOCP, since telling the IOCP backend about all
your CPUs is a good idea. It'll also come in handy with asn's
multithreaded crypto stuff, and for people who run servers without
reading the manual.
automake 1.6 doesn't like using a conditional += to add stuff to foo_LDADD.
Instead you need to conditionally define a variable, then non-conditionally
put that variable in foo_LDADD.
This requires the latest Git version of Libevent as of 24 March 2010.
In the future, we'll just say it requires Libevent 2.0.5-alpha or
later.
Since Libevent doesn't yet support hierarchical rate limit groups,
there isn't yet support for tracking relayed-bytes separately when
using the bufferevent system. If a future version does add support
for hierarchical buckets, we can add that back in.
We want to make sure that we don't break old torrc files that might have
used something like this made-up example:
ContactInfo UberUser <uber@user.com> # /// Fake email! \\\
Log info file /home/nick.mathewson/projects/tor-info.log
And we also want to support the following style of writing your torrc:
ExcludeNodes \
# Node1337 is run by the Bavarian Illuminati
Node1337, \
# The operator of Node99 looked at me funny
Node99
The code already handles both cases, but the unit test should help prove
it.
Roger correctly pointed out that my code was broken for accounting
periods that shifted forwards, since
start_of_accounting_period_containing(interval_start_time) would not
be equal to interval_start_time, but potentially much earlier.
This function uses GetSystemDirectory() to make sure we load the version
of the library from c:\windows\system32 (or local equivalent) rather than
whatever version lives in the cwd.
The RefuseUnknownExits config option is now a tristate, with "1"
meaning "enable it no matter what the consensus says", "0" meaning
"disable it no matter what the consensus says", and "auto" meaning "do
what the consensus says". If the consensus is silent, we enable
RefuseUnknownExits.
This patch also changes the dirserv logic so that refuseunknownexits
won't make us cache unless we're an exit.
Do not double-report signatures from unrecognized authorities both as
"from unknown authority" and "not present". Fixes bug 1956, bugfix on
0.2.2.16-alpha.
Our attempt to make compilation work on old versions of Windows
again while keeping wince compatibility broke the build for Win2k+.
helix reports this patch fixes the issue for WinXP. Bugfix on
0.2.2.15-alpha; related to bug 1797.
When the CellStatistics option is off, we don't store cell insertion
times. Doing so would also not be very smart, because there seem to
still be some performance issues with this type of statistics. Nothing
harmful happens when we don't have insertion times, so we don't need to
alarm the user.
Previously[*], the function would start with the first stream on the
circuit, and let it package as many cells as it wanted before
proceeding to the next stream in turn. If a circuit had many live
streams that all wanted to package data, the oldest would get
preference, and the newest would get ignored.
Now, we figure out how many cells we're willing to send per stream,
and try to allocate them fairly.
Roger diagnosed this in the comments for bug 1298.
[*] This bug has existed since before the first-ever public release
of Tor. It was added by r152 of Tor on 26 Jan 2003, which was
the first commit to implement streams (then called "topics").
This is not the oldest bug to be fixed in 0.2.2.x: that honor
goes to the windowing bug in r54, which got fixed in e50b7768 by
Roger with diagnosis by Karsten. This is, however, the most
long-lived bug to be fixed in 0.2.2.x: the r54 bug was fixed
2580 days after it was introduced, whereas I am writing this
commit message 2787 days after r152.
I'm going to use this to implement more fairness in
circuit_resume_edge_reading_helper in an attempt to fix bug 1298.
(Updated with fixes from arma and Sebastian)
An authority should never download a consensus if it has a live one,
but when it doesn't, it should admit that it's not going to get one,
and see if anybody else can give it one.
Fixes 1300, fix on 0.2.0.9-alpha
Bridges and other relays not included in the consensus don't
necessarily have a non-zero bandwidth capacity. If all our
configured bridges had a zero bw capacity we would warn the
user. Change that.
Previously, we were also considering the time spent in
soft-hibernation. If this was a long time, we would wind up
underestimating our bandwidth by a lot, and skewing our wakeup time
towards the start of the accounting interval.
This patch also makes us store a few more fields in the state file,
including the time at which we entered soft hibernation.
Fixes bug 1789. Bugfix on 0.0.9pre5.
This will make changes for DST still work, and avoid double-spending
bytes when there are slight changes to configurations.
Fixes bug 1511; the DST issue is a bugfix on 0.0.9pre5.
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).
- Move checks for extra_info to callers
- Change argument name from failed to descs
- Use strlen("fp/") instead of a magic number
- I passed on the suggestion to rename functions from *_failed() to
*_handle_failure(). There are a lot of these so for now just follow
the house style.
It's normal when bootstrapping to have a lot of different certs
missing, so we don't want missing certs to make us warn... unless
the certs we're missing are ones that we've tried to fetch a couple
of times and failed at.
May fix bug 1145.
We frequently add cells to stream-blocked queues for valid reasons
that don't mean we need to block streams. The most obvious reason
is if the cell arrives over a circuit rather than from an edge: we
don't block circuits, no matter how full queues get. The next most
obvious reason is that we allow CONNECTED cells from a newly created
stream to get delivered just fine.
This patch changes the behavior so that we only iterate over the
streams on a circuit when the cell in question came from a stream,
and we only block the stream that generated the cell, so that other
streams can still get their CONNECTEDs in.
This should keep WinCE working (unicode always-on) and get Win98
working again (unicode never-on).
There are two places where we explicitly use ASCII-only APIs, still:
in ntmain.c and in the unit tests.
This patch also fixes a bug in windoes tor_listdir that would cause
the first file to be listed an arbitrary number of times that was
also introduced with WinCE support.
Should fix bug 1797.
Setting CookieAuthFileGroupReadable but without setting CookieAuthFile makes
no sense, because unix directory permissions for the data directory prevent
the group from accessing the file anyways.
When this happens, run through the streams on the circuit and make
sure they're all blocked. If some aren't, that's a bug: block them
all and log it! If they all are, where did the cell come from? Log
it!
(I suspect that this actually happens pretty frequently, so I'm making
these log messages appear at INFO.)
Do not start reading on exit streams when we get a SENDME unless we
have space in the appropriate circuit's cell queue.
Draft fix for bug 1653.
(commit message by nickm)
router_add_to_routerlist() is supposed to be a nice minimal function
that only touches the routerlist structures, but it included a call to
dirserv_single_reachability_test().
We have a function that gets called _after_ adding descriptors
successfully: routerlist_descriptors_added. This patch moves the
responsibility for testing there.
Because the decision of whether to test or not depends on whether
there was an old routerinfo for this router or not, we have to first
detect whether we _will_ want to run the tests if the router is added.
We make this the job of
routers_update_status_from_consensus_networkstatus().
Finally, this patch makes the code notice if a router is going from
hibernating to non-hibernating, and if so causes a reachability test
to get launched.
This solves the problem Roger noted as:
What if the router has a clock that's 5 minutes off, so it
publishes a descriptor for 5 minutes in the future, and we test it
three minutes in. In this edge case, we will continue to advertise
it as Running for the full 45 minute period.
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.
requests to authorities fail due to a network error.
Bug#1138
"When a Tor client starts up using a bridge, and UpdateBridgesFromAuthority
is set, Tor will go to the authority first and look up the bridge by
fingerprint. If the bridge authority is filtered, Tor will never notice that
the bridge authority lookup failed. So it will never fall back."
Add connection_dir_bridge_routerdesc_failed(), a function for unpacking
the bridge information from a failed request, and ensure
connection_dir_request_failed() calls it if the failed request
was for a bridge descriptor.
Test:
1. for ip in `grep -iR 'router ' cached-descriptors|cut -d ' ' -f 3`;
do sudo iptables -A OUTPUT -p tcp -d $ip -j DROP; done
2. remove all files from user tor directory
3. Put the following in torrc:
UseBridges 1
UpdateBridgesFromAuthority 1
Bridge 85.108.88.19:443 7E1B28DB47C175392A0E8E4A287C7CB8686575B7
4. Launch tor - it should fall back to downloading descriptors
directly from the bridge.
Initial patch reviewed and corrected by mingw-san.
Apparently the way we handled cleaning up temporary directories with
atexit() meant that when the child process exited, it would remove the
temporary directory, thus making other tests in the main process fail.
https://trac.torproject.org/projects/tor/ticket/1525
"The codepath taken by the control port "RESOLVE" command to create a
synthetic SOCKS resolve request isn't the same as the path taken by
a real SOCKS request from 'tor-resolve'.
This prevents controllers who set LeaveStreamsUnattached=1 from
being able to attach RESOLVE streams to circuits of their choosing."
Create a new function connection_ap_rewrite_and_attach_if_allowed()
and call that when Tor needs to attach a stream to a circuit but
needs to know if the controller permits it.
No tests added.
Since the rend code doesn't like the port to be 0, we shouldn't generate
the port by declaring crypto_rand_int(65536); instead we should
say crypto_rand_int(65535)+1.
Diagnosed by Matt Edman; fixes bug 1808.
With this patch we stop scheduling when we should write statistics using a
single timestamp in run_scheduled_events(). Instead, we remember when a
statistics interval starts separately for each statistic type in geoip.c
and rephist.c. Every time run_scheduled_events() tries to write stats to
disk, it learns when it should schedule the next such attempt.
This patch also enables all statistics to be stopped and restarted at a
later time.
This patch comes with a few refactorings, some of which were not easily
doable without the patch.
We already had the country code ?? indicating an unknown country, so all we
needed to do to make unknown countries excludable was to make the ?? code
discoverable.
It's okay to get (say) a SocksPort line in the torrc, and then a
SocksPort on the command line to override it, and then a SocksPort via
a controller to override *that*. But if there are two occurrences of
SocksPort in the torrc, or on the command line, or in a single SETCONF
command, then the user is likely confused. Our old code would not
help unconfuse the user, but would instead silently ignore all but
the last occurrence.
This patch changes the behavior so that if the some option is passed
more than once to any torrc, command line, or SETCONF (each of which
coincidentally corresponds to a call to config_assign()), and the
option is not a type that allows multiple occurrences (LINELIST or
LINELIST_X), then we can warn the user.
This closes trac entry 1384.
At best, this patch helps us avoid sending queued relayed cells that
would get ignored during the time between when a destroy cell is
sent and when the circuit is finally freed. At worst, it lets us
release some memory a little earlier than it would otherwise.
Fix for bug #1184. Bugfix on 0.2.0.1-alpha.
The next series of commits begins addressing the issue that we're
currently including the complete or.h file in all of our source files.
To change that, we're splitting function definitions into new header
files (one header file per source file).
Right now it says "552 internal error" because there's no way for
getinfo_helper_*() countries to specify an error message. This
patch changes the getinfo_helper_*() interface, and makes most of the
getinfo helpers give useful error messages in response to failures.
This should prevent recurrences of bug 1699, where a missing GeoIPFile
line in the torrc made GETINFO ip-to-county/* fail in a "not obvious
how to fix" way.
V3 authorities no longer decide not to vote on Guard+Exit. The bandwidth
weights should take care of this now.
Also, lower the max threshold for WFU to 0.98, to allow more nodes to become
guards.
This should make us conflict less with system files named "log.h".
Yes, we shouldn't have been conflicting with those anyway, but some
people's compilers act very oddly.
The actual change was done with one "git mv", by editing
Makefile.am, and running
find . -name '*.[ch]' | xargs perl -i -pe 'if (/^#include.*\Wlog.h/) {s/log.h/torlog.h/; }'
We now record large times as abandoned, to prevent a filter step from
happening and skewing our results.
Also, issue a warn for a rare case that can happen for funky values of Xm or
too many abandoned circuits. Can happen (very rarely) during unit tests, but
should not be possble during live operation, due to network liveness filters
and discard logic.
Many friendly operating systems have 64-bit times, and it's not nice
to pass them to an %ld format.
It's also extremely not-nice to write a time to the log as an
integer. Most people think it's 2010 June 29 23:57 UTC+epsilon, not
1277855805+epsilon.
These timers behave better with non-monotonic clocks than our old
ones, and also try harder to make once-per-second events get called
one second apart, rather than one-plus-epsilon seconds apart.
This fixes bug 943 for everybody using Libevent 2.0 or later.
From the code:
zlib 1.2.4 and 1.2.5 do some "clever" things with macros. Instead of
saying "(defined(FOO) ? FOO : 0)" they like to say "FOO-0", on the theory
that nobody will care if the compile outputs a no-such-identifier warning.
Sorry, but we like -Werror over here, so I guess we need to define these.
I hope that zlib 1.2.6 doesn't break these too.
Possible fix for bug 1526.
This should never happen unless openssl is buggy or some of our
assumptions are deeply wrong, but one of those might have been the
cause of the not-yet-reproducible bug 1209. If it ever happens again,
let's get some info we can use.
We need to ensure that we close timeout measurement circuits. While
we're at it, we should close really old circuits of certain types that
aren't in use, and log really old circuits of other types.
We need to record different statistics at point of timeout, vs the point
of forcible closing.
Also, give some better names to constants and state file variables
to indicate they are not dealing with timeouts, but abandoned circuits.
Having ~/.tor expand into /.tor is, after all, almost certainly not
what the user wanted, and it deserves a warning message.
Also, convert a guess-and-malloc-and-sprintf triple into an asprintf.
In rare cases, we could cannibalize a one-hop circuit, ending up
with a two-hop circuit. This circuit would not be actually used,
but we should prevent its creation in the first place.
Thanks to outofwords and swissknife for helping to analyse this.
Most of the changes here are switches to use APIs available on Windows
CE. The most pervasive change is that Windows CE only provides the
wide-character ("FooW") variants of most of the windows function, and
doesn't support the older ASCII verions at all.
This patch will require use of the wcecompat library to get working
versions of the posix-style fd-based file IO functions.
[commit message by nickm]
Back when we changed the idea of a connection being "too old" for new
circuits into the connection being "bad" for new circuits, we didn't
actually change the info messages. This led to telling the user that
we were labelling connections as "too old" for being worse than
connections that were actually older than them.
Found by Scott on or-talk.
There are now four ways that CBT can be disabled:
1. Network-wide, with the cbtdisabled consensus param.
2. Via config, with "LearnCircuitBuildTimeout 0"
3. Via config, with "AuthoritativeDirectory 1"
4. Via a state file write failure.
This should prevent some asserts and storage of incorrect build times
for the cases where Tor is suspended during a circuit construction, or
just after completing a circuit. The idea is that if the circuit
build time is much greater than we would have cut it off at, we probably
had a suspend event along this codepath, and we should discard the
value.
In case we decide that the timeout rate is now too high due to our
change of the max synthetic quantile value, this consensus parameter
will allow us to restore it to the previous value.
This is for the other issue we saw in Bug 1335. A large number of high
timeouts were causing the timeout calculation to slowly drift upwards,
especially in conditions of load. This fix repeatedly regenerates all of
our synthetic timeouts whenever the timeout changes, to try to prevent
drift.
It also lowers the timeout cap to help for some cases of Bug 1245, where
some timeout values were so large that we ended up allocating a ton of
scratch memory to count the histogram bins.
The downside is that lowering this cap is affecting our timeout rate.
Unfortunately, the buildtimeout quantile is now higher than the actual
completion rate by what appears to be about 7-10%, which probably
represents the skew in the distribution due to lowering this synthetic
cap.
In my state files, I was seeing several peaks, probably due to different
guards having different latency. This change is meant to better capture
this behavior and generate more reasonable timeouts when it happens. It
is improving the timeout values for my collection of state files.
