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.
One-hop dirconn streams all share a session group, and get the
ISO_SESSIONGRP flag: they may share circuits with each other and
nothing else.
Anonymized dirconn streams get a new internal-use-only ISO_STREAM
flag: they may not share circuits with anything, including each other.
The new candidate rule, which arma suggested and I like, is that
the original address as received from the client connection or as
rewritten by the controller is the address that counts.
This is mainly meant as a way to keep clients from accidentally
DOSing themselves by (e.g.) enabling IsolateDestAddr or
IsolateDestPort on a port that they use for HTTP.
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 is the meat of proposal 171: we change circuit_is_acceptable()
to require that the connection is compatible with every connection
that has been linked to the circuit; we update circuit_is_better to
prefer attaching streams to circuits in the way that decreases the
circuits' usefulness the least; and we update link_apconn_to_circ()
to do the appropriate bookkeeping.
The "nym epoch" of a stream is defined as the number of times that
NEWNYM had been called before the stream was opened. All streams
are isolated by nym epoch.
This feature should be redundant with existing signewnym stuff, but
it provides a good belt-and-suspenders way for us to avoid ever
letting any circuit type bypass signewnym.
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.
Proposal 171 gives us a new syntax for parsing client port options.
You can now have as many FooPort options as you want (for Foo in
Socks, Trans, DNS, NATD), and they can have address:port arguments,
and you can specify the level of isolation on those ports.
Additionally, this patch refactors the client port parsing logic to
use a new type, port_cfg_t. Previously, ports to be bound were
half-parsed in config.c, and later re-parsed in connection.c when
we're about to bind them. Now, parsing a port means converting it
into a port_cfg_t, and binding it uses only a port_cfg_t, without
needing to parse the user-provided strings at all.
We should do a related refactoring on other port types. For
control ports, that'll be easy enough. For ORPort and DirPort,
we'll want to do this when we solve proposal 118 (letting servers
bind to and advertise multiple ports).
This implements tickets 3514 and 3515.
Previously we were using router_get_by_id(foo) to test "do we have a
descriptor that will let us make an anonymous circuit to foo". But
that isn't right for microdescs: we should have been using node_t.
Fixes bug 3601; bugfix on 0.2.3.1-alpha.
Instead, use compare_tor_addr_to_node_policy everywhere.
One advantage of this is that compare_tor_addr_to_node_policy can
better distinguish 0.0.0.0 from "unknown", which caused a nasty bug
with microdesc users.
Previously, we had an issue where we'd treat an unknown address as
0, which turned into "0.0.0.0", which looked like a rejected
address. This meant in practice that as soon as we started doing
comparisons of unknown uint32 addresses to short policies, we'd get
'rejected' right away. Because of the circumstances under which
this would be called, it would only happen when we had local DNS
cached entries and we were looking to launch new circuits.
