Taylor Yu [Thu, 26 Jul 2018 17:32:34 +0000 (12:32 -0500)]
Don't disable an unsupported compiler warning
Conditionalize the pragma that temporarily disables
-Wunused-const-variable. Some versions of gcc don't support it. We
need to do this because of an apparent bug in some libzstd headers.
Fixes bug 26785; bugfix on 0.3.2.11.
Nick Mathewson [Fri, 20 Jul 2018 15:02:05 +0000 (11:02 -0400)]
Tweak assertion in get_time_period_length() for coverity
This is another attempt to fix 1437668. The assertion here should
be safe, since the rules of networkstatus_get_param() keep the value
it returns in range.
Fix time source bug in sr_state_get_start_time_of_current_protocol_run().
The following bug was causing many issues for this branch in chutney:
In sr_state_get_start_time_of_current_protocol_run() we were using the
consensus valid-after to calculate beginning_of_current_round, but we were
using time(NULL) to calculate the current_round slot. This was causing time
sync issues when the consensus valid-after and time(NULL) were disagreeing on
what the current round is. Our fix is to use the consensus valid-after in both
places.
This also means that we are not using 'now' (aka time(NULL)) anymore in that
function, and hence we can remove that argument from the function (and its
callers). I'll do this in the next commit so that we keep things separated.
The OPE cipher is tied to the current blinded key which is tied to the current
time period. Hence create the OPE cipher structure when we create a new
descriptor (and build its blinded key).
Fix up some unittests by being more careful with the local time.
Now that the rev counter depends on the local time, we need to be more careful
in the unittests. Some unittests were breaking because they were using
consensus values from 1985, but they were not updating the local time
appropriately. That was causing the OPE module to complain that it was trying
to encrypt insanely large values.
Compute the description revision counter using the OPE scheme.
To do so for a given descriptor, we use the "seconds since the SR protocol run"
started, for the SRV that is relevant to this descriptor. This is guaranteed to
be a positive value (since we need an SRV to be able to build a descriptor),
and it's also guaranteed to be a small value (since SRVs stop being listed on a
consensus after 48 hours).
We cannot use the "seconds since the time period started", because for the next
descriptor we use the next time period, so the timestamp would end up negative.
See [SERVICEUPLOAD] from rend-spec-v3.txt for more details.
To do so, we have to introduce a new `is_current` argument to a bunch of
functions, because to use "seconds since the SR protocol run" we need to know
if we are building the current or the next descriptor, since we use a different
SRV for each descriptor.
Nick Mathewson [Thu, 10 May 2018 12:46:36 +0000 (08:46 -0400)]
Implementation for a simple order-preserving encryption scheme.
This is meant for use when encrypting the current time within the
period in order to get a monotonically increasing revision counter
without actually revealing our view of the time.
This scheme is far from the most state-of-the-art: don't use it for
anything else without careful analysis by somebody much smarter than
I am.
See ticket #25552 for some rationale for this logic.
juga0 [Sat, 30 Jun 2018 06:29:11 +0000 (06:29 +0000)]
Move bandwidth file tests to same function
also add tests for bw_file_headers.
Headers are all that is found before a correct relay line or
the terminator.
Tests include:
* a empty bandwidth file
* a bandwidth file with only timestamp
* a bandwidth file with v1.0.0 headers
* a bandwidth file with v1.0.0 headers and relay lines
* a bandwidth file with v1.1.0 headers and v1.0.0 relay lines
* a bandwidth file with v1.0.0 headers, malformed relay lines and
relay lines
* a bandwidth file with v1.0.0 headers, malformed relay lines,
relay lines and malformed relay lines
* a bandwidth file with v1.1.0 headers without terminator
* a bandwidth file with v1.1.0 headers with terminator
* a bandwidth file with v1.1.0 headers without terminator and
relay lines
* a bandwidth file with v1.1.0 headers with terminator and relay
lines
* a bandwidth file with v1.1.0 headers without terminator, bad
relay lines and relay lines
* a bandwidth file with v1.1.0 headers with terminator, bad relay
lines and relay lines
juga0 [Sat, 30 Jun 2018 06:16:17 +0000 (06:16 +0000)]
Replace bwlist by bw_file and terminator condition
If bandwidth file terminator is found, set end of headers flag
and do not store the line.
If it is not, parse a relay line and check whether it is a header
line.
projects / thirdparty / tor.git / log
