The old behavior contributed to unreliability when hidden services and
hsdirs had different consensus versions, and so had different opinions
about who should be cacheing hsdir info.
Bugfix on 0.2.0.10-alpha; based on discussions surrounding bug 2732.
The new behavior is to try to rename the old file if there is one there
that we can't read. In all likelihood, that will fail too, but at least
we tried, and at least it won't crash.
Conflicts throughout. All resolved in favor of taking HEAD and
adding tor_mem* or fast_mem* ops as appropriate.
src/common/Makefile.am
src/or/circuitbuild.c
src/or/directory.c
src/or/dirserv.c
src/or/dirvote.c
src/or/networkstatus.c
src/or/rendclient.c
src/or/rendservice.c
src/or/router.c
src/or/routerlist.c
src/or/routerparse.c
src/or/test.c
Here I looked at the results of the automated conversion and cleaned
them up as follows:
If there was a tor_memcmp or tor_memeq that was in fact "safe"[*] I
changed it to a fast_memcmp or fast_memeq.
Otherwise if there was a tor_memcmp that could turn into a
tor_memneq or tor_memeq, I converted it.
This wants close attention.
[*] I'm erring on the side of caution here, and leaving some things
as tor_memcmp that could in my opinion use the data-dependent
fast_memcmp variant.
Make that explicit by adding an assert and removing a null-check. All of
its callers currently depend on the argument being non-null anyway.
Silences a few clang complaints.
The analyzer assumed that bootstrap_percent could be less than 0 when we
call control_event_bootstrap_problem(), which would mean we're calling
log_fn() with undefined values. The assert makes it clear this can't
happen.
See bug 2850 for rationale: it appears that on some busy exits, the OS
decides that every single port is now unusable because they have been
all used too recently.
Previously we ensured that it would get called periodically by doing
it from inside the code that added microdescriptors. That won't work
though: it would interfere with our code that tried to read microdescs
from disk initially. Instead, we should consider rebuilding the cache
periodically, and on startup.
Previously on 0.2.2, we'd never clean the cache. Now that we can
clean it, we want to add a condition to rebuild it: that should happen
whenever we have dropped enough microdescriptors that we could save a
lot of space.
No changes file, since 0.2.3 doesn't need one and 0.2.2 already has some
changes files for the backport of the microdesc_clean_cahce() function.
Clients and relays haven't used them since early 0.2.0.x. The only
remaining use by authorities learning about new relays ahead of scedule;
see proposal 147 for what we intend to do about that.
We're leaving in an option (FetchV2Networkstatus) to manually fetch v2
networkstatuses, because apparently dnsel and maybe bwauth want them.
This fixes bug 3022.
Previously it would erroneously return true if ListenAddr was set for
a client port, even if that port itself was 0. This would give false
positives, which were not previously harmful... but which were about
to become.
A v0 HS authority stores v0 HS descriptors in the same descriptor
cache that its HS client functionality uses. Thus, if the HS
authority operator clears its client HS descriptor cache, ALL v0
HS descriptors will be lost. That would be bad.
These functions can return NULL for otherwise-valid values of
time_t. Notably, the glibc gmtime manpage says it can return NULL
if the year if greater than INT_MAX, and the windows MSDN gmtime
page says it can return NULL for negative time_t values.
Also, our formatting code is not guaranteed to correctly handle
years after 9999 CE.
This patch tries to correct this by detecting NULL values from
gmtime/localtime_r, and trying to clip them to a reasonable end of
the scale. If they are in the middle of the scale, we call it a
downright error.
Arguably, it's a bug to get out-of-bounds dates like this to begin
with. But we've had bugs of this kind in the past, and warning when
we see a bug is much kinder than doing a NULL-pointer dereference.
Boboper found this one too.
If the user sent a SIGNAL NEWNYM command after we fetched a rendezvous
descriptor, while we were building the introduction-point circuit, we
would give up entirely on trying to connect to the hidden service.
Original patch by rransom slightly edited to go into 0.2.1
Previously, it would remove every trackhostexits-derived mapping
*from* xyz.<exitname>.exit; it was supposed to remove every
trackhostexits-derived mapping *to* xyz.<exitname>.exit.
Bugfix on 0.2.0.20-rc: fixes an XXX020 added while staring at bug-1090
issues.
Now we believe it to be the case that we never build a circuit for our
stream that has an unsuitable exit, so we'll never need to use such
a circuit. The risk is that we have some code that builds the circuit,
but now we refuse to use it, meaning we just build a bazillion circuits
and ignore them all.
This looked at first like another fun way around our node selection
logic: if we had introduction circuits, and we wound up building too
many, we would turn extras into general-purpose circuits. But when we
did so, we wouldn't necessarily check whether the general-purpose
circuits conformed to our node constraints. For example, the last
node could totally be in ExcludedExitNodes and we wouldn't have cared...
...except that the circuit should already be internal, so it won't get user
streams attached to it, so the transition should generally be allowed.
Add an assert to make sure we're right about this, and have it not
check whether ExitNodes is set, since that's irrelevant to internal
circuits.
IOW, if we were using TrackExitHosts, and we added an excluded node or
removed a node from exitnodes, we wouldn't actually remove the mapping
that points us at the new node.
Also, note with an XXX022 comment a place that I think we are looking
at the wrong string.
The routerset_equal function explicitly handles NULL inputs, so
there's no need to check inputs for NULL before calling it.
Also fix a bug in routerset_equal where a non-NULL routerset with no
entries didn't get counted as equal to a NULL routerset. This was
untriggerable, I think, but potentially annoying down the road.
ExcludeExitNodes foo now means that foo.exit doesn't work. If
StrictNodes is set, then ExcludeNodes foo also overrides foo.exit.
foo.exit , however, still works even if foo is not listed in ExitNodes.
This once maybe made sense when ExitNodes meant "Here are 3 exits;
use them all", but now it more typically means "Here are 3
countries; exit from there." Using non-Fast/Stable exits created a
potential partitioning opportunity and an annoying stability
problem.
(Don't worry about the case where all of our ExitNodes are non-Fast
or non-Stable: we handle that later in the function by retrying with
need_capacity and need_uptime set to 0.)
If we're picking a random directory node, never pick an excluded one.
But if we've chosen a specific one (or all), allow it unless strictnodes
is set (in which case warn so the user knows it's their fault).
When warning that we won't connect to a strictly excluded node,
log what it was we were trying to do at that node.
When ExcludeNodes is set but StrictNodes is not set, we only use
non-excluded nodes if we can, but fall back to using excluded nodes
if none of those nodes is usable.
This is a tweak to the bug2917 fix. Basically, if we want to simulate
a signal arriving in the controller, we shouldn't have to pretend that
we're Libevent, or depend on how Tor sets up its Libevent callbacks.
The last entry of the *Maxima values in the state file was inflated by a
factor of NUM_SECS_ROLLING_MEASURE (currently 10). This could lead to
a wrong maximum value propagating through the state file history.
When reading the bw history from the state file, we'd add the 900-second
value as traffic that occured during one second. Fix that by adding the
average value to each second.
This bug was present since 0.2.0.5-alpha, but was hidden until
0.2.23-alpha when we started using the saved values.
Some tor relays would report lines like these in their extrainfo
documents:
dirreq-write-history 2011-03-14 16:46:44 (900 s)
This was confusing to some people who look at the stats. It would happen
whenever a relay first starts up, or when a relay has dirport disabled.
Change this so that lines without actual bw entries are omitted.
Implements ticket 2497.
While doing so, get rid of the now unnecessary function
control_signal_act().
Fixes bug 2917, reported by Robert Ransom. Bugfix on commit
9b4aa8d2ab. This patch is loosely based on
a patch by Robert (Changelog entry).
- Document the structure and variables.
- Make circuits_for_buffer_stats into a static variable.
- Don't die horribly if interval_length is 0.
- Remove the unused local_circ_id field.
- Reorder the fields of circ_buffer_stats_t for cleaner alignment layout.
Instead of answering GETINFO requests about our geoip usage only after
running for 24 hours, this patch makes us answer GETINFO requests
immediately. We still round and quantize as before.
Implements bug2711.
Also, refactor the heck out of the bridge usage formatting code. No
longer should we need to do a generate-parse-and-regenerate cycle to
get the controller string, and that lets us simplify the code a lot.
We've got millisecond timers now, we might as well use them.
This change won't actually make circuits get expiered with microsecond
precision, since we only call the expiry functions once per second.
Still, it should avoid the situation where we have a circuit get
expired too early because of rounding.
A couple of the expiry functions now call tor_gettimeofday: this
should be cheap since we're only doing it once per second. If it gets
to be called more often, though, we should onsider having the current
time be an argument again.
Since svn r1475/git 5b6099e8 in tor-0.0.6, we have responded to an
exhaustion of all 65535 stream IDs on a circuit by marking that
circuit for close. That's not the right response. Instead, we
should mark the circuit as "too dirty for new circuits".
Of course in reality this isn't really right either. If somebody
has managed to cram 65535 streams onto a circuit, the circuit is
probably not going to work well for any of those streams, so maybe
we should be limiting the number of streams on an origin circuit
concurrently.
Also, closing the stream in this case is probably the wrong thing to
do as well, but fixing that can also wait.
We fixed bug 539 (where directories would say "503" but send data
anyway) back in 0.2.0.16-alpha/0.1.2.19. Because most directory
versions were affected, we added workaround to make sure that we
examined the contents of 503-replies to make sure there wasn't any
data for them to find. But now that such routers are nonexistent,
we can remove this code. (Even if somebody fired up an 0.1.2.19
directory cache today, it would still be fine to ignore data in its
erroneous 503 replies.)
The first was genuinely impossible, I think: it could only happen
when the amount we read differed from the amount we wanted to read
by more than INT_MAX.
The second is just very unlikely: it would give incorrect results to
the controller if you somehow wrote or read more than 4GB on one
edge conn in one second. That one is a bugfix on 0.1.2.8-beta.
In afe414 (tor-0.1.0.1-rc~173), when we moved to
connection_edge_end_errno(), we used it in handling errors from
connection_connect(). That's not so good, since by the time
connection_connect() returns, the socket is no longer set, and we're
supposed to be looking at the socket_errno return value from
connection_connect() instead. So do what we should've done, and
look at the socket_errno value that we get from connection_connect().
