Our previous algorithm had a nonzero probability of picking no
events to cancel, which is of course incorrect. The new code uses
Vitter's good old reservoir sampling "algorithm R" from 1985.
Fixes bug 26008; bugfix on 0.2.6.3-alpha.
This functionality was covered only accidentally by our voting-test
code, and as such wasn't actually tested at all. The tests that
called it made its coverage nondeterministic, depending on what time
of day you ran the tests.
Closes ticket 26014.
We cleared this value in second_elapsed_callback. But what were we
using it for? For detecting if Libevent returned EINVAL too often!
We already have a way to detect too-frequent events, and that's with
a ratelim_t. Refactor the code to use that instead. Closes ticket
26016.
This test was using the current time to pick the time period number,
and a randomly generated hs key. Therefore, it sometimes picked an
index that would wrap around the example dht, and sometimes would
not.
The fix here is just to fix the time period and the public key.
Fixes bug 25997; bugfix on 0.3.2.1-alpha.
When directory authorities read a zero-byte bandwidth file, they log
a warning with the contents of an uninitialised buffer. Log a warning
about the empty file instead.
Fixes bug 26007; bugfix on 0.2.2.1-alpha.
LibreSSL, despite not having the OpenSSL 1.1 API, does define
OPENSSL_VERSION in crypto.h. Additionally, it apparently annotates
some functions as returning NULL, so that our unit tests need to be
more careful about checking for NULL so they don't get compilation
warnings.
Closes ticket 26006.
This test, in test_client_pick_intro(), will have different coverage
depending on whether it selects a good intro point the first time or
whether it has to try a few times. Since it produces the shorter
coverage with P=1/4, repeat this test 64 times so that it only
provides reduced coverage with P=1/2^128. The performance cost is
negligible.
Closes ticket 25996. This test was introduced in 0.3.2.1-alpha.
I'd prefer not to do this for randomized tests, but as things stand
with this test, it produces nondeterministic test coverage.
Closes ticket 25995; bugfix on 0.2.2.2-alpha when this test was
introduced.
This change should make it impossible for the monotonic time to roll
over from one EWMA tick to the next during this test, and make it so
that this test never invokes scale_active_circuits() (which it
doesn't test).
(Earlier changes during the 0.3.4 series should make this call even
rarer than it was before, since we fixed#25927 and removed
cached_gettimeofday. Because this test didn't update
cached_gettimeofday, the chance of rolling over a 10-second interval
was much higher.)
Closes ticket 25994; bugfix on 0.3.3.1-alpha when this test was
introduced.
Arguably, the conditions under which these events happen should be a
bit different, but the rules are complex enough here that I've tried
to have this commit be pure refactoring.
Closes ticket 25952.
Finally, before this code goes away, take a moment to look at the
amazing way that we used to try to have an event happen
every N seconds:
get_uptime() / N != (get_uptime()+seconds_elapsed) / N
Truly, it is a thing of wonder. I'm glad we didn't start using this
pattern everywhere else.
Implement the ability to set flags per events which influences the set up of
the event.
This commit only adds one flag which is "need network" meaning that the event
is not enabled if tor has disabled the network or if hibernation mode.
Signed-off-by: David Goulet <dgoulet@torproject.org>
Prior to #23100, we were not counting HS circuit build times in our
calculation of the timeout. This could lead to a condition where our timeout
was set too low, based on non HS circuit build times, and then we would
abandon all HS circuits, storing no valid timeouts in the histogram.
This commit avoids the assert.
Previously were using this value to have a cheap highish-resolution
timer. But we were only using it in one place, and current dogma is
to use monotime_coarse_t for this kind of thing.
This part of the code was the only part that used "cached
getttimeofday" feature, which wasn't monotonic, which we updated at
slight expense, and which I'd rather not maintain.
Now that we update our buckets on demand before reading or writing,
we no longer need to update them all every TokenBucketRefillInterval
msec.
When a connection runs out of bandwidth, we do need a way to
reenable it, however. We do this by scheduling a timer to reenable
all blocked connections for TokenBucketRefillInterval msec after a
connection becomes blocked.
(If we were using PerConnBWRate more, it might make sense to have a
per-connection timer, rather than a single timeout. But since
PerConnBWRate is currently (mostly) unused, I'm going to go for the
simpler approach here, since usually whenever one connection has
become blocked on bandwidth, most connections are blocked on
bandwidth.)
Implements ticket 25373.
We used to do this 10x per second in connection_buckets_refill();
instead, we now do it when the bucket becomes empty. This change is
part of the work of making connection_buckets_refill() obsolete.
Closes ticket 25828; bugfix on 0.2.3.5-alpha.
This commit introduces the consensus parameter "circ_max_cell_queue_size"
which controls the maximum number of cells a circuit queue should have.
The default value is currently 50000 cells which is above what should be
expected but keeps us a margin of error for padding cells.
Related to this is #9072. Back in 0.2.4.14-alpha, we've removed that limit due
to a Guard discovery attack. Ticket #25226 details why we are putting back the
limit due to the memory pressure issue on relays.
Fixes#25226
Signed-off-by: David Goulet <dgoulet@torproject.org>
Add to the Denial of Service section of the man page an explanation about the
three different mitigation Tor has.
Fixes#25248.
Signed-off-by: David Goulet <dgoulet@torproject.org>
In d1874b4339, we adjusted this check so that we insist on
using routerinfos for bridges. That's almost correct... but if we
have a bridge that is also a regular relay, then we should use
insist on its routerinfo when connecting to it as a bridge
(directly), and be willing to use its microdescriptor when
connecting to it elsewhere in our circuits.
This bug is a likely cause of some (all?) of the (exit_ei == NULL)
failures we've been seeing.
Fixes bug 25691; bugfix on 0.3.3.4-alpha
This patch changes the algorithm of compute_real_max_mem_in_queues() to
use 0.4 * RAM iff the system has more than or equal to 8 GB of RAM, but
will continue to use the old value of 0.75 * RAM if the system have less
than * GB of RAM available.
This patch also adds tests for compute_real_max_mem_in_queues().
See: https://bugs.torproject.org/24782
This one happens if for some reason you start with DirPort enabled
but server mode turned off entirely.
Fixes a case of bug 23693; bugfix on 0.3.1.1-alpha.
This patch lifts the list of default directory authorities from config.c
into their own auth_dirs.inc file, which is then included in config.c
using the C preprocessor.
Patch by beastr0.
See: https://bugs.torproject.org/24854
Both in geoip_note_client_seen() and options_need_geoip_info(), switch from
accessing the options directly to using the should_record_bridge_info() helper
function.
Fixes#25290
Signed-off-by: David Goulet <dgoulet@torproject.org>
When a relay is collecting internal statistics about how many
create cell requests it has seen of each type, accurately count the
requests from relays that temporarily fall out of the consensus.
(To be extra conservative, we were already ignoring requests from clients
in our counts, and we continue ignoring them here.)
Fixes bug 24910; bugfix on 0.2.4.17-rc.
Directory authorities no longer vote in favor of the Guard flag
for relays that don't advertise directory support.
Starting in Tor 0.3.0.1-alpha, Tor clients have been avoiding using
such relays in the Guard position, leading to increasingly broken load
balancing for the 5%-or-so of Guards that don't advertise directory
support.
Fixes bug 22310; bugfix on 0.3.0.6.
Add a missing lock acquisition around access to queued_control_events
in control_free_all(). Use the reassign-and-unlock strategy as in
queued_events_flush_all(). Fixes bug 25675. Coverity found this bug,
but only after we recently added an access to
flush_queued_event_pending.
The anchor here was originally introduced to make sure that the
manpage would format correctly, but we had mistaken it for a section
anchor, and introduced a bogus section heading in 3563a2c819.
Fixes bug 25582; bugfix on 0.3.3.1-alpha.
Coverity found a null pointer reference in nodelist_add_microdesc().
This is almost certainly impossible assuming that the routerstatus_t
returned by router_get_consensus_status_by_descriptor_digest() always
corresponds to an entry in the nodelist. Fixes bug 25629.
Coverity found a null pointer reference in nodelist_add_microdesc().
This is almost certainly impossible assuming that the routerstatus_t
returned by router_get_consensus_status_by_descriptor_digest() always
corresponds to an entry in the nodelist. Fixes bug 25629.
Coverity found a null pointer reference in nodelist_add_microdesc().
This is almost certainly impossible assuming that the routerstatus_t
returned by router_get_consensus_status_by_descriptor_digest() always
corresponds to an entry in the nodelist. Fixes bug 25629.
If we failed to connect at the TCP level to a relay, note it down and refuse
to connect again for another 60 seconds.
Fixes#24767
Signed-off-by: David Goulet <dgoulet@torproject.org>
This update is needed to make it consistent with the behavior of
node_awaiting_ipv6(), which doesn't believe in the addresses from
routerinfos unless it actually plans to use those routerinfos.
Fixes bug 25213; bugfix on b66b62fb75 in 0.3.3.1-alpha,
which tightened up the definition of node_awaiting_ipv6().
There was a nonfatal assertion in pathbias_should_count that would
trigger if onehop_tunnel was set, but the desired_path_length was
greater than 1. This patch fixes that. Fixes bug 24903; bugfix on
0.2.5.2-alpha.
