Exiting data over tunnels is only allowed after opt-in via settings #1325

rjruigrok · 2015-03-12T14:08:08Z

tribler-ci · 2015-03-12T14:36:11Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3678/
Test FAILed.

tribler-ci · 2015-03-12T14:53:31Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester-isolated_devel/287/
Test FAILed.

rjruigrok · 2015-03-12T15:23:45Z

retest this please

tribler-ci · 2015-03-12T15:42:21Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3680/
Test FAILed.

tribler-ci · 2015-03-12T15:44:58Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3681/
Test FAILed.

tribler-ci · 2015-03-12T15:45:33Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester-isolated_devel/289/
Test FAILed.

tribler-ci · 2015-03-12T15:45:39Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester-isolated_devel/290/
Test FAILed.

rjruigrok · 2015-03-12T15:49:21Z

retest this please

tribler-ci · 2015-03-12T16:18:42Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3682/
Test FAILed.

…setting

tribler-ci · 2015-03-12T16:25:21Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester-isolated_devel/291/
Test FAILed.

tribler-ci · 2015-03-12T16:51:06Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3683/
Test FAILed.

tribler-ci · 2015-03-12T16:57:42Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester-isolated_devel/292/
Test PASSed.

NielsZeilemaker · 2015-03-13T13:03:15Z

Tribler/community/tunnel/tunnel_community.py

@@ -643,6 +657,10 @@ def check_create(self, messages):
            if self.crypto.key.pub().key_to_bin() != message.payload.node_public_key:
                yield DropMessage(message, "TunnelCommunity: public keys do not match")
                continue
+
+            if not self.settings.become_exitnode and message.payload.become_exit:
+                yield DropMessage(message, "Exit-node functionality disabled")


I would reply with a message saying denied or something like that, then choose a different peer if you receive a denied response. Otherwise creating a tunnel might be ridicously difficult.

Moreover, how are you going to link two tunnels? Are introduction/rendezvous points exit points? Because they do exit_data.

NielsZeilemaker · 2015-03-13T13:06:14Z

@whirm same thing is happening here, look it asserts http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester-isolated_devel/292/artifact/output//Screenshot-33.png but "passes" passes all the tests

whirm · 2015-03-13T13:54:18Z

@NielsZeilemaker it's weird.

I'm only hooking on the sys.excepthook to capture unhandled exceptions I don't see how that can happen (If the test raises an exception my wrapper does nothing else than raising the exception again. It's when the test passes that I check for unhandled exceptions: https://github.com/Tribler/tribler/blob/devel/Tribler/Test/test_as_server.py#L56

I can't think of a way this would fail.

Any idea?

NielsZeilemaker · 2015-03-13T13:58:03Z

Maybe because callconditional is setting do_assert to False? Hence, it doesn't assert, it only quits.

LipuFei · 2015-03-13T14:01:28Z

@whirm I agree with @NielsZeilemaker, can be the do_assert thing.

whirm · 2015-03-13T15:51:51Z

Well, then why is that test not setting it to True? I thought you where saying this has something to do with the unhandled exception catcher stuff I made.

NielsZeilemaker · 2015-03-13T16:47:02Z

Because we didn't want to assert on the wx thread, but maybe that's OK now. You've modified this code

whirm · 2015-03-13T16:50:01Z

@NielsZeilemaker and how was this supposed to work? how where the tests marked as failed if you weren't asserting? To me it looks like they have been always broken. the last time the assert thing was modified was in 2013...

NielsZeilemaker · 2015-03-13T16:51:04Z

You removed the assert list, we did the asserts after closing tribler

NielsZeilemaker · 2015-03-13T17:11:18Z

@whirm its this one ebe98d6

whirm · 2015-03-13T17:35:50Z

Oh man, I don't know how I went over that one. So we just need to remove the assertion skip everywhere. I'll make a PR for that.

NielsZeilemaker · 2015-03-13T17:38:50Z

I guess, simply remove the option. But have a look if everything still works. Maybe you need to rewrite it such that if do_assert is false it adds the assert to the list and still only asserts after tribler quit

LipuFei · 2015-03-13T17:41:02Z

@whirm I think we still need the option. For example, in remote_search test, the CallConditional is used to check if we have enough connections or not. In this case, we may still want to continue the test even if we don't have enough connections because the search function may still work.

tribler-ci · 2015-04-17T18:19:00Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3981/
Test PASSed.

NielsZeilemaker · 2015-04-17T20:21:42Z

retest this please

rjruigrok · 2015-04-17T20:34:45Z

The crypto bug still occasionly occurs, let's wait for the test result... Hopefully it passes again but it's a wild guess :)

When looking to the output in jenkins, receiving/decoding data on the circuits to the rp / ip sometimes fails. The error was the same as the error when the same peer is used more than once in a circuit, but with the current code I don't see a place where this is still possible. What's the best way of debugging this kind of crypto problems?

tribler-ci · 2015-04-17T20:52:09Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3982/
Test FAILed.

tribler-ci · 2015-04-18T12:30:21Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3983/
Test FAILed.

rjruigrok · 2015-04-18T12:45:38Z

retest this please

tribler-ci · 2015-04-18T13:14:55Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3984/
Test PASSed.

tribler-ci · 2015-04-18T13:22:55Z

Merged build finished. Test FAILed.

tribler-ci · 2015-04-18T13:22:55Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3985/
Test FAILed.

tribler-ci · 2015-04-18T14:14:16Z

Merged build finished. Test PASSed.

tribler-ci · 2015-04-18T14:14:17Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3986/
Test PASSed.

synctext · 2015-04-19T17:32:08Z

retest this please

tribler-ci · 2015-04-19T18:03:08Z

Refer to this link for build results (access rights to CI server needed):
http://jenkins.tribler.org//job/GH_Tribler_pull-request-tester_devel/3987/
Test FAILed.

rjruigrok · 2015-04-19T18:38:08Z

I tried to figure out the crypto bug but really have no clue. It seems it always occurs in the rp /ip circuit (not the linking one) with length 2, while peeling off data from the first layer. sometimes the first cell sent for the circuit succeeds and the next cells all fail. And sometimes, like the last test, all tunnel tests run without issues. Anyone having a suggestion or idea what the cause could be? Problem started after moving the exit node choice to the create_circuit method.

NielsZeilemaker · 2015-04-19T18:40:57Z

It's decrypting with the wrong key, that's why we get the error. I'll have
a look what causing it when I have some time. But for now, I don't think
it related to your changes
Op 19 apr. 2015 8:38 PM schreef "Rob Ruigrok" [email protected]:

I tried to figure out the crypto bug but really have no clue. It seems it
always occurs in the rp /ip circuit (not the linking one) with length 2,
while peeling off data from the first layer. sometimes the first cell sent
for the circuit succeeds and the next cells all fail. And sometimes, like
the last test, all tunnel tests run without issues. Anyone having a
suggestion or idea what the cause could be? Problem started after moving
the exit node choice to the create_circuit method.

—
Reply to this email directly or view it on GitHub
#1325 (comment).

synctext · 2015-04-19T20:12:15Z

Thanks for the insight @NielsZeilemaker. Does that mean this PR can be pushed through? Plus newer ones are opened for connectability, move the 4-8 creation magic into community.py etc..

NielsZeilemaker · 2015-04-19T20:14:06Z

Jup, that would be my suggestion
Op 19 apr. 2015 10:12 PM schreef "Johan Pouwelse" <[email protected]

:

Thanks for the insight @NielsZeilemaker
https://github.com/NielsZeilemaker. Does that mean this PR can be
pushed through? Plus newer ones are opened for connectability, move the 4-8
creation magic into community.py etc..

—
Reply to this email directly or view it on GitHub
#1325 (comment).

Exiting data over tunnels is only allowed after opt-in via settings

rjruigrok · 2015-04-20T22:29:26Z

@NielsZeilemaker Finally resolved the issue! You were wrong - It was related to my changes. :-)

The problem was caused by the selection of the first_hop in create_circuit. I discovered the issue after a lot of debugging with the tunnel messages send along. It was selecting the first hop without checking whether it would also become the required_exit, hence the circuit got the same hop twice in the circuit (as first hop and as exit node). It's clear to see in the following log. Note the using port 17410, and a few lines further the send created is sent to candidate 17410.

Exit-node enabled
ERROR   1429565522.64   LibtorrentMgr:109   could not restore dht state, starting from scratch
Using port 17410
ERROR   1429565563.45 tunnel_community:868   Got create (2270751903) from ('10.0.2.15', 45321)
ERROR   1429565564.30 tunnel_community:677   send created to 1 candidates: ['{10.0.2.15:45321 84.105.131.99:57987}']
ERROR   1429565565.28 tunnel_community:868   Got extend (2270751903) from ('10.0.2.15', 45321)
ERROR   1429565565.28 tunnel_community:1211  Decrypt layer for relaying own circuit 2270751903 with session key 적�Ŏ@:Q��?�P$ 
ERROR   1429565565.28 tunnel_community:677   send create to 1 candidates: ['{10.0.2.15:17410}']
ERROR   1429565565.28 tunnel_community:868   Got create (2584804787) from ('10.0.2.15', 17410)
ERROR   1429565565.29 tunnel_community:677   send created to 1 candidates: ['{10.0.2.15:17410}']
ERROR   1429565565.31 tunnel_community:868   Got created (2584804787) from ('10.0.2.15', 17410)
ERROR   1429565565.31 tunnel_community:1189  Encrypt layer for relaying own circuit 2270751903 with session key �����?7�����6
ERROR   1429565565.31 tunnel_community:677   send extended to 1 candidates: ['{10.0.2.15:45321}']
ERROR   1429565569.17 tunnel_community:1226  Decrypt layer for relaying data to exit node for circuit 2270751903 with session key 적�Ŏ@:Q��?�P$   
ERROR   1429565569.17 tunnel_community:677   send data to 1 candidates: ['{10.0.2.15:17410}']
ERROR   1429565569.17 tunnel_community:1221  Encrypt layer for relaying data to origin for circuit 2584804787 with session key 8�K������{������
ERROR   1429565569.17 tunnel_community:677   send data to 1 candidates: ['{10.0.2.15:45321}']

Although it should theoretically be possible to use the same hop multiple times in a circuit, the crypto in its current state will crash. You might solve this bug, but it eventually helped us to identify the problem of reusing hops in the same circuit... As long as everybody knows that failures in the crypto can be caused by reuse of hops, this should be fine. What do you think?

BTW I'm already preparing a pull request which resolves this issue (and which adds some debugging info)

NielsZeilemaker · 2015-04-21T04:52:32Z

For now, avoiding the same member twice seems to be a solution. I'll have a look if I can fix this bug later on.

rjruigrok force-pushed the exitnode-switch branch from 67bdeac to 75fe51e Compare March 12, 2015 15:22

Exiting data over tunnels is only allowed after switching the opt-in …

d33a2e9

…setting

rjruigrok force-pushed the exitnode-switch branch from 75fe51e to d33a2e9 Compare March 12, 2015 16:24

NielsZeilemaker reviewed Mar 13, 2015
View reviewed changes

rjruigrok force-pushed the exitnode-switch branch from cfd1599 to 23b8d85 Compare April 17, 2015 17:49

rjruigrok force-pushed the exitnode-switch branch from 23b8d85 to 9275a72 Compare April 18, 2015 11:59

rjruigrok force-pushed the exitnode-switch branch from 9275a72 to a40070b Compare April 18, 2015 13:12

rjruigrok force-pushed the exitnode-switch branch from a40070b to 9d911b5 Compare April 18, 2015 13:25

Bugfixes and refactoring on circuit creation

b3439a4

rjruigrok force-pushed the exitnode-switch branch from 9d911b5 to b3439a4 Compare April 18, 2015 13:43

NielsZeilemaker mentioned this pull request Apr 20, 2015

Add test to verify "same member used in circuit twice crypto bug" #1370

Closed

NielsZeilemaker added a commit that referenced this pull request Apr 20, 2015

Merge pull request #1325 from rjruigrok/exitnode-switch

17ac69b

Exiting data over tunnels is only allowed after opt-in via settings

NielsZeilemaker merged commit 17ac69b into Tribler:devel Apr 20, 2015

rjruigrok mentioned this pull request Apr 20, 2015

Bugfix for crypto hidden services + hidden services related updates in GUI #1376

Merged

rjruigrok deleted the exitnode-switch branch April 22, 2015 06:20

Exiting data over tunnels is only allowed after opt-in via settings #1325

Exiting data over tunnels is only allowed after opt-in via settings #1325

Conversation

rjruigrok commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

rjruigrok commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

rjruigrok commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

tribler-ci commented Mar 12, 2015

NielsZeilemaker Mar 13, 2015

Choose a reason for hiding this comment

NielsZeilemaker commented Mar 13, 2015

whirm commented Mar 13, 2015

NielsZeilemaker commented Mar 13, 2015

LipuFei commented Mar 13, 2015

whirm commented Mar 13, 2015

NielsZeilemaker commented Mar 13, 2015

whirm commented Mar 13, 2015

NielsZeilemaker commented Mar 13, 2015

NielsZeilemaker commented Mar 13, 2015

whirm commented Mar 13, 2015

NielsZeilemaker commented Mar 13, 2015

LipuFei commented Mar 13, 2015

tribler-ci commented Apr 17, 2015

NielsZeilemaker commented Apr 17, 2015

rjruigrok commented Apr 17, 2015

tribler-ci commented Apr 17, 2015

tribler-ci commented Apr 18, 2015

rjruigrok commented Apr 18, 2015

tribler-ci commented Apr 18, 2015

tribler-ci commented Apr 18, 2015

tribler-ci commented Apr 18, 2015

tribler-ci commented Apr 18, 2015

tribler-ci commented Apr 18, 2015

synctext commented Apr 19, 2015

tribler-ci commented Apr 19, 2015

rjruigrok commented Apr 19, 2015

NielsZeilemaker commented Apr 19, 2015

synctext commented Apr 19, 2015

NielsZeilemaker commented Apr 19, 2015

rjruigrok commented Apr 20, 2015

NielsZeilemaker commented Apr 21, 2015