Remove restriction to not send relay_early for rend circs

Back in bug #1038 (moved) we added a restriction not to send relay_early cells on rends circuits. The time without relays running versions 0.2.1.3-alpha through 0.2.1.18-alpha has come, so it should be fine to lift the restriction now.

changed milestone to %Tor: 0.2.2.x-final

added component::core tor/tor milestone::Tor: 0.2.2.x-final priority::medium resolution::fixed status::closed type::defect labels

Branch relay_early_rend in my repository for a suggested fix. This is on maint-0.2.2, but it also applies cleanly to master. I think it might want to go into maint.

Trac:
Status: new to needs_review

Looks good to me. Merging to 0.2.2 and master.

Trac:
Resolution: N/A to fixed
Status: needs_review to closed

Trac:
Status: closed to reopened
Resolution: fixed to N/A

Does that mean we should change dirserv_get_status_impl() so it refuses descriptors from relays running these versions?

(Just because nobody's running these versions now doesn't mean nobody will run them in the future.)

We could. If you set them up maliciously, you get better ways to screw with users. I checked that all distributions have earlier/later versions. I didn't think it was necessary due to these conditions

Trac:
Milestone: N/A to Tor: 0.2.2.x-final

See branch bug2081_followup_022 in my public repository.

Trac:
Status: reopened to needs_review

Looks good to me. Maybe we want to incorporate the #2156 (moved) fix here, too?

I wouldn't object. What's the oldest 0.2.0.x that doesn't scare us?

We currently have no active 0.2.0.x relays running anything below 0.2.0.32, so we could safely raise to at least 0.2.0.32. In .33, we fixed Ilja's bug. .34 fixes a remote crash on exits. We have 5 relays running 0.2.0.3[23]. Does either bug scare us enough or do we raise to .32?

I'm not too scared. I would like a second opinion from arma though.

Replying to nickm:

See branch bug2081_followup_022 in my public repository. The source code patch looks ‘obviously correct’ to me, but there are two typos in the changes file:

2s/server/servers/
4s/RELAY_EARY/RELAY_EARLY/

There's also s/verions/versions/

We might also say s/server/relay/g

I should point out that we now have Tor relays in the consensus running these versions:

r Bellum 7j/MKmnxheiRBy8T7pkIzW7ZvqU w2+LIXCt6qZeib6tsLY/BvrxM70 2010-12-06 23:37:14 62.75.223.163 9001 0
s Fast Named Running Stable Valid
v Tor 0.2.1.12-alpha (r18423)
w Bandwidth=16900
p reject 1-65535

r sutor yqIADJ1riF5y3j6ckVdjwM0xmS4 Q/JSbIGYIOmNLDVLOEMq394KnSQ 2010-12-06 13:22:40 62.193.210.68 443 0
s Fast Named Running Valid
v Tor 0.2.1.16-rc
w Bandwidth=30
p reject 1-65535

So a) perhaps these versions aren't really dead yet, and b) once we do deploy sebastian's "relay_early_rend" (looks like we already did), we really will want to lock out Tor relays with these versions.

So is that a "go ahead", or "hang on", or "I want to think more" ?

It's an "in retrospect, maybe we should have not applied the relay_early_rend patch yet".

But that said, we did, so we should apply the next patch too.

Once we do lock these versions out, we should make a stronger effort to monitor the logs on directory authorities to recognize when people are trying to publish, and try to contact them to help them realize they should upgrade.

cleaned up changes file and merged.

Trac:
Status: needs_review to closed
Resolution: N/A to fixed

Trac:
Component: Tor Client to Tor

closed

mentioned in issue #2156 (moved)

moved to tpo/core/tor#2081 (closed)