Previously, when we ran low on memory, we'd close whichever circuits
had the most queued cells. Now, we close those that have the
*oldest* queued cells, on the theory that those are most responsible
for us running low on memory, and that those are the least likely to
actually drain on their own if we wait a little longer.
Based on analysis from a forthcoming paper by Jansen, Tschorsch,
Johnson, and Scheuermann. Fixes bug 9093.
This implements "algorithm 1" from my discussion of bug #9072: on OOM,
find the circuits with the longest queues, and kill them. It's also a
fix for #9063 -- without the side-effects of bug #9072.
The memory bounds aren't perfect here, and you need to be sure to
allow some slack for the rest of Tor's usage.
This isn't a perfect fix; the rest of the solutions I describe on
codeable.
Apparently some compilers like to eliminate memset() operations on
data that's about to go out-of-scope. I've gone with the safest
possible replacement, which might be a bit slow. I don't think this
is critical path in any way that will affect performance, but if it
is, we can work on that in 0.2.4.
Fixes bug 7352.
Previously, we would treat an intro circuit failure as a timeout iff the
circuit failed due to a mismatch in relay identity keys. (Due to a bug
elsewhere, we only recognize relay identity-key mismatches on the first
hop, so this isn't as bad as it could have been.)
Bugfix on commit eaed37d14c, not yet in any
release.
Back when I added this logic in 20c0581a79, the rule was that whenever
a circuit finished building, we cleared its isolation info. I did that
so that we would still use the circuit even if all the streams that
had previously led us to tentatively set its isolation info had closed.
But there were problems with that approach: We could pretty easily get
into a case where S1 had led us to launch C1 and S2 had led us to
launch C2, but when C1 finished, we cleared its isolation and attached
S2 first. Since C2 was still marked in a way that made S1
unattachable to it, we'd then launch another circuit needlessly.
So instead, we try the following approach now: when a circuit is done
building, we try to attach streams to it. If it remains unused after
we try attaching streams, then we clear its isolation info, and try
again to attach streams.
Thanks to Sebastian for helping me figure this out.
Our old "do we need to launch a circuit for stream S" logic was,
more or less, that if we had a pending circuit that could handle S,
we didn't need to launch a new one.
But now that we have streams isolated from one another, we need
something stronger here: It's possible that some pending C can
handle either S1 or S2, but not both.
This patch reuses the existing isolation logic for a simple
solution: when we decide during circuit launching that some pending
C would satisfy stream S1, we "hypothetically" mark C as though S1
had been connected to it. Now if S2 is incompatible with S1, it
won't be something that can attach to C, and so we'll launch a new
stream.
When the circuit becomes OPEN for the first time (with no streams
attached to it), we reset the circuit's isolation status. I'm not
too sure about this part: I wanted some way to be sure that, if all
streams that would have used a circuit die before the circuit is
done, the circuit can still get used. But I worry that this
approach could also lead to us launching too many circuits. Careful
thought needed here.
This patch adds fields to track how streams should be isolated, and
ensures that those fields are set correctly. It also adds fields to
track what streams can go on a circuit, and adds functions to see
whether a streams can go on a circuit and update the circuit
accordingly. Those functions aren't yet called.
This lets us make a lot of other stuff const, allows the compiler to
generate (slightly) better code, and will make me get slightly fewer
patches from folks who stick mutable stuff into or_options_t.
const: because not every input is an output!
The conflicts were mainly caused by the routerinfo->node transition.
Conflicts:
src/or/circuitbuild.c
src/or/command.c
src/or/connection_edge.c
src/or/directory.c
src/or/dirserv.c
src/or/relay.c
src/or/rendservice.c
src/or/routerlist.c
This patch introduces a few new functions in router.c to produce a
more helpful description of a node than its nickame, and then tweaks
nearly all log messages taking a nickname as an argument to call these
functions instead.
There are a few cases where I left the old log messages alone: in
these cases, the nickname was that of an authority (whose nicknames
are useful and unique), or the message already included an identity
and/or an address. I might have missed a couple more too.
This is a fix for bug 3045.
Conflicts in various places, mainly node-related. Resolved them in
favor of HEAD, with copying of tor_mem* operations from bug3122_memcmp_022.
src/common/Makefile.am
src/or/circuitlist.c
src/or/connection_edge.c
src/or/directory.c
src/or/microdesc.c
src/or/networkstatus.c
src/or/router.c
src/or/routerlist.c
src/test/test_util.c
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.
Resolved conflicts in:
doc/tor.1.txt
src/or/circuitbuild.c
src/or/circuituse.c
src/or/connection_edge.c
src/or/connection_edge.h
src/or/directory.c
src/or/rendclient.c
src/or/routerlist.c
src/or/routerlist.h
These were mostly releated to the routerinfo_t->node_t conversion.
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.
We need to make sure that the worst thing that a weird consensus param
can do to us is to break our Tor (and only if the other Tors are
reliably broken in the same way) so that the majority of directory
authorities can't pull any attacks that are worse than the DoS that
they can trigger by simply shutting down.
One of these worse things was the cbtnummodes parameter, which could
lead to heap corruption on some systems if the value was sufficiently
large.
This commit fixes this particular issue and also introduces sanity
checking for all consensus parameters.
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.
