Reduce fingerprinting through timestamps in 'addr' messages. #5860

sipa commented at 1:21 PM on March 6, 2015: member

sipa force-pushed on Mar 7, 2015

fanquake commented at 7:37 AM on March 8, 2015: member

Travis failure

Cleaning up
terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::lock_error> >'
  what():  boost: mutex lock failed in pthread_mutex_lock: Invalid argument
terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::lock_error> >'
  what():  boost: mutex lock failed in pthread_mutex_lock: Invalid argument
*** glibc detected *** /home/travis/build/bitcoin/bitcoin/bitcoin-x86_64-unknown-linux-gnu/src/bitcoind: malloc(): memory corruption: 0x00007fd8131c3ab1 ***

laanwj added this to the milestone 0.10.0 on Mar 9, 2015

laanwj added the label P2P on Mar 9, 2015

Diapolo commented at 10:46 AM on March 9, 2015: none

@laanwj Isn't 0.10.0 released ;)? Why did you tag this 0.10.0 then?

laanwj commented at 10:48 AM on March 9, 2015: member

@diapolo Don't be so pedantic, I'm using it as marker for the next version on the 0.10 branch.

Diapolo commented at 10:52 AM on March 9, 2015: none

@laanwj Just asking, no offense :).

laanwj commented at 11:22 AM on March 9, 2015: member

@diapolo okay :) @sipa @fanquake Tried a Travis respin, still same problem on same configuration.

laanwj commented at 8:19 AM on March 11, 2015: member

I'm also getting these errors when running the RPC tests locally with this patch:

bitcoind: /usr/include/boost/thread/pthread/recursive_mutex.hpp:110: void boost::recursive_mutex::lock(): Assertion `!pthread_mutex_lock(&m)' failed.
bitcoind: /usr/include/boost/thread/pthread/recursive_mutex.hpp:110: void boost::recursive_mutex::lock(): Assertion `!pthread_mutex_lock(&m)' failed.

This appears to be a problem with destruction order. By moving addrman.Connected(addr); to the CNode destructor, addrman is called when each node is deleted. This may cause a use-after-free of a mutex?

jgarzik commented at 1:54 PM on March 11, 2015: contributor

concept ACK

sipa force-pushed on Mar 11, 2015

sipa commented at 2:03 PM on March 11, 2015: member

Changed approach.

jonasnick commented at 11:42 PM on March 11, 2015: contributor

To give a bit of context, without this PR it is possible to learn all outbound peers of a node by repeatedly sending 'getaddr' messages. See the histogram for the timestamps of a full node and a client client .

sipa force-pushed on Mar 12, 2015

laanwj commented at 7:47 AM on March 13, 2015: member

utACK

in src/main.cpp:None in d88011e435 outdated

3637 | @@ -3628,6 +3638,12 @@ bool static ProcessMessage(CNode* pfrom, string strCommand, CDataStream& vRecv,
3638 |      else if (strCommand == "verack")
3639 |      {
3640 |          pfrom->SetRecvVersion(min(pfrom->nVersion, PROTOCOL_VERSION));
3641 | +
3642 | +        // Update the last seen time for this node's address

laanwj commented at 1:00 PM on March 16, 2015:

This comment is no longer correct, should probably be Update the connected status for this node

sipa commented at 10:09 AM on March 17, 2015:

Done.

jonasnick commented at 9:40 PM on March 16, 2015: contributor

untested ack

Reduce fingerprinting through timestamps in 'addr' messages.

Suggested by Jonas Nick.

9c2737901b

sipa force-pushed on Mar 17, 2015

laanwj merged this on Mar 17, 2015

laanwj closed this on Mar 17, 2015

laanwj referenced this in commit 93a8c46807 on Mar 17, 2015

sipa referenced this in commit ca301bf98c on Mar 17, 2015

laanwj commented at 3:41 PM on March 17, 2015: member

Cherry-picked to 0.10 as ca301bf

laanwj referenced this in commit 149c1d890d on Apr 6, 2015

reddink referenced this in commit bc820a93ff on Jul 11, 2020

reddink referenced this in commit 5ddae186a0 on Jul 14, 2020

DrahtBot locked this on Sep 8, 2021