[tor-bugs] #32604 [Core Tor/Tor]: Add HiddenServiceExportRendPoint and HiddenServiceExportInstanceID directive

[Tor Bug Tracker & Wiki]

#32604: Add HiddenServiceExportRendPoint and HiddenServiceExportInstanceID
directive
-----------------------------------------+---------------------------------
 Reporter:  moonsikpark                  |          Owner:  (none)
     Type:  enhancement                  |         Status:  needs_revision
 Priority:  Medium                       |      Milestone:  Tor:
                                         |  0.4.3.x-final
Component:  Core Tor/Tor                 |        Version:
 Severity:  Normal                       |     Resolution:
 Keywords:  tor-hs tor-dos extra-review  |  Actual Points:
Parent ID:  #32511                       |         Points:
 Reviewer:  dgoulet, ahf, teor           |        Sponsor:  Sponsor27-can
-----------------------------------------+---------------------------------

Comment (by teor):

 Replying to [comment:21 moonsikpark]:
 > OK, made CI happy by letting `circuit_free_()` free
 `or_circ->build_state->chosen_exit`.

 Thanks!

 > Before extending the exposure range of rp fingerprint, I think there
 might be other metrics we can export that can help onion services.
 Potentially `circ->rend_data->nr_streams` or
 `circ->hs_ident->intro_auth_pk`

 Sure that seems fine.

 > or `circ->hs_ident->rendezvous_cookie`.

 No, I don't think we should export raw cryptographic material. If you
 really want this info, you should export
 `siphash(circ->hs_ident->rendezvous_cookie)`. (The cookie should not be
 common across instances, unless the client is doing a replay attack across
 different instances. If you want to be able to detect relay attacks, you
 should do `H(CONSTANT_PREFIX|circ->hs_ident->rendezvous_cookie)`.)

 > I think `INSTANCE_ID` don't have to be that long, we can change it to 1
 byte and move it to the front(`fdAA` where `AA=INSTANCE_ID`).

 Sure.

 > And I'm curious about `dst_ipv6` fixed to `::1`. `real_addr` of
 `rend_service_port_config_t` is not always `localhost`. Does backends
 ignore the destination address?

 I don't know who is actually using this feature right now. Maybe ahf does?

 > Does it mean we can cram data in `dst_ipv6` too?

 That's 16 bytes, so you could get the whole fingerprint if you wanted it.
 But maybe there are more valuable things.

 I'm going to suggest a way forward:
 1. We decide what to do about the existing broken fields:
 https://trac.torproject.org/projects/tor/ticket/32604?replyto=21#comment:19
 2. We do a design for the final set of fields.
 3. We land this patch with the current fields, in the positions they will
 have in the final design.
 4. We open a new ticket for any new fields.

 This feature is getting complex enough that we might need a proposal, a
 spec, or a very well-designed manual page entry.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/32604#comment:22>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online