Add SSL support to P2P #8996

vtnerd · 2023-09-19T23:24:27Z

First, my apologies to @moneromooo-monero for NACKing his/her SSL proposal years ago, only for me to bring it back.

This does exactly as the title suggests, with options to disable P2P encryption entirely, and an option to re-use a SSL certificate between "runs". The default generates a new SSL certificate each monerod run, so that the node cannot be tracked across IP address changes.

Nodes do not trust encryption information from peers. Instead, every peer is assumed to be in autodetect mode, unless overridden on the CLI or via handshake/ping messages directly from the peer.

Possibly bad:
If a node chooses to re-use SSL certificates, a change in certificates will cause connection failures until the node is removed from the white+gray lists OR the node makes a direct connection and provides the new SSL certificate.

EDIT:
I will also attempt some unit test changes; there is enough to review here that I expect this diff to be up a while.

vtnerd · 2023-10-06T20:30:57Z

Added sodium to the dependency list, hopefully builds pass now.

src/cryptonote_protocol/cryptonote_protocol_defs.h

vtnerd · 2023-11-16T17:34:24Z

Force pushed a change based on my last comment (ssl_support_t doesn't specify numeric values).

vtnerd · 2023-11-16T17:36:27Z

Did another force-push with a rebase, to get rid of the external changes that were not intended.

vtnerd · 2023-11-19T00:25:44Z

I added a crude functional test for SSL. It just tests that autodetection and disabling SSL works.

j-berman · 2024-04-19T00:44:30Z

Started a review and have an initial comment. I didn't get around to fixing connect_async as we discussed over here, and that needs to be done for this PR now that connect_async can get called with an SSL setting.

Running this code on a node that can get incoming connections, you can see that the back ping to an incoming node always fails (log level 2 you'll see back ping invoke failed, and no PING SUCCESS) after an incoming node connects to you.

I'm happy to do this since I said I would over in that PR, my bad for not getting to it sooner.

vtnerd · 2024-04-28T21:05:42Z

@j-berman Fixed SSL connections with connect_async.

vtnerd · 2024-04-28T21:24:48Z

Forgot to include newest code in last push. Trying again!

vtnerd · 2024-04-29T02:04:54Z

Rebased, and trying to fix a linking issue that I cannot reproduce locally.

SyntheticBird45 · 2024-06-17T17:49:31Z

Please excuse my ignorance, I'm not familiar with the boost library and I'm therefore wondering what cipher are actually available ? afaiu, SSL namespace include TLS up to TLSv1.3: https://www.boost.org/doc/libs/1_74_0/boost/asio/ssl/context_base.hpp

I'm probably blind but i don't find in the commit changes where are cipher options. That would be good to enforce at least TLSv1.2

vtnerd · 2024-06-17T21:19:25Z

The code already forces TLS 1.2+. It also limits the cipher list a few lines below that.

vtnerd · 2024-06-18T01:36:46Z

Bad rebase, will fix.

vtnerd · 2024-06-18T14:08:27Z

The test that just failed should be spurious.

vtnerd · 2024-06-18T14:08:50Z

I rebased against latest master changes, which included a test merge and a seed node merge.

vtnerd · 2025-03-03T01:19:40Z

Force pushed a rebase. Nothing should have changed. Looking at the diff again.

vtnerd · 2025-03-03T01:23:12Z

Force pushed another small update removing some unintended changes.

vtnerd · 2025-03-03T14:38:01Z

The failed test is not related. May push a dummy update so that all unit tests get a chance to run.

vtnerd · 2025-08-20T01:49:51Z

Force pushed a rebase merge for p2p functional tests.

j-berman

This generally looks ok to me, though I'm not particularly well versed with SSL. Mostly nits and some questions

j-berman · 2024-04-18T19:39:34Z

docs/LEVIN_PROTOCOL.md

+> The server/responder must always be in autodetect SSL mode, so that it can
+handle any possible state that the client/initiator is in.
+
+> If a node changes certificates, "old" peers will be unable to connect until


It looks as though if monerod generates a new temporary cert every run, that the node won't share the fingerprint with its peers. Therefore this should only affect nodes that have a persistent cert and then change it, ya?

Correct, if a fingerprint is sent in the peerlist exchange, it is removed by the receiving node. The fingerprint is also exchanged in the handshake, iff the user has a fixed certificate. In this situation, the peer caches it and is equivalent to a unique ip/port pair in terms of treatment and recovery.

j-berman · 2024-04-18T19:48:35Z

src/p2p/net_node.cpp

-    const command_line::arg_descriptor<std::vector<std::string> > arg_p2p_add_exclusive_node   = {"add-exclusive-node", "Specify list of peers to connect to only."
-                                                                                                  " If this option is given the options add-priority-node and seed-node are ignored"};
-    const command_line::arg_descriptor<std::vector<std::string> > arg_p2p_seed_node   = {"seed-node", "Connect to a node to retrieve peer addresses, and disconnect"};
+    const command_line::arg_descriptor<std::vector<std::string> > arg_p2p_add_peer   = {"add-peer", "Manually add peer to local peerlist. Append \",sha-256 fingerprint\" to make authenticated connection." 


Nits:

Suggested change

const command_line::arg_descriptor<std::vector<std::string> > arg_p2p_add_peer = {"add-peer", "Manually add peer to local peerlist. Append \",sha-256 fingerprint\" to make authenticated connection."

const command_line::arg_descriptor<std::vector<std::string> > arg_p2p_add_peer = {"add-peer", "Manually add peer to local peerlist. Append \",<sha-256 fingerprint>\" to make authenticated connection. "

Same for other references to the fingerprint

j-berman · 2024-04-18T22:59:22Z

contrib/epee/include/net/net_utils_base.h

                               m_is_income(false),
                               m_started(time(NULL)),
-                               m_ssl(false),
+			       m_ssl(ssl_support_t::e_ssl_support_disabled),


nit: the whitespace has tabs here and doesn't line up in my editor

Suggested change

m_ssl(ssl_support_t::e_ssl_support_disabled),

m_ssl(ssl_support_t::e_ssl_support_disabled),

I think I fixed this. Always annoying in editors.

j-berman · 2024-04-19T00:23:09Z

contrib/epee/include/net/connection_basic.hpp


-		template<typename ConstBufferSequence, typename WriteHandler>
-		void async_write_some(const ConstBufferSequence &buffers, WriteHandler &&handler)
+		bool server_handshake(boost::asio::const_buffer buffer)


This is unused

docs/LEVIN_PROTOCOL.md

contrib/epee/include/net/abstract_tcp_server2.inl

contrib/epee/include/net/levin_protocol_handler_async.h

j-berman · 2025-10-24T02:14:55Z

tests/unit_tests/levin.cpp

            auto notification = receiver_.get_notification<cryptonote::NOTIFY_NEW_TRANSACTIONS>().second;
            EXPECT_EQ(txs, notification.txs);
-            EXPECT_FALSE(notification._.empty());
+            EXPECT_TRUE(notification._.empty());


Checking my understanding: because the padding doesn't get serialized into _ as part of the NOTIFY_NEW_TRANSACTIONS request anymore, and instead gets stuffed into the buffer, the _ field is now empty even though this request should have been padded. And the reason you don't pad into a _ field anymore is because you want padding on every message, rather than strictly NOTIFY_NEW_TRANSACTIONS

It would be nice if there was an explicit way to see/test the padding at this level, but not worth changing the internals for it

A comment explaining how these _with_padding tests are just testing smooth notifications even with pad true would be solid, since they're no longer explicitly testing the padding

Commented the best I could in the test file.

j-berman · 2025-10-24T02:34:34Z

contrib/epee/src/levin_base.cpp

    head.m_return_code = SWAP32LE(return_code);

    std::memcpy(buffer.tellp() - buffer.size(), std::addressof(head), sizeof(head));
+    buffer.put_n(0, pad_bytes);


Just to check my understanding: when de-serializing a message, the de-serializer is unaffected by these padded bytes because these are just dangling bytes that have no associated key, ya? The de-serializer can safely just ignore them

Yes, the epee parser always ignored extra data after the root object. The "noise" mode in Tor uses this "feature" already - each message is padded to a specific length with possible trailing. So we have a working example at least.

Any future parser already has to support this feature due to the noise mode.

src/p2p/net_node.inl

vtnerd · 2025-10-25T00:56:58Z

Force pushed a rebase and @j-berman changes.

jonathancross · 2025-12-02T16:27:19Z

@vtnerd Looks like there are conflicts to resolve.

vtnerd · 2025-12-02T18:09:48Z

Force pushed a rebase to solve merge conflicts.

vtnerd force-pushed the feature/p2p_ssl_squashed branch from 22ca764 to 5208f25 Compare October 6, 2023 20:30

vtnerd mentioned this pull request Oct 18, 2023

Update Wallet SSL Policies #7089

Open

vtnerd commented Nov 16, 2023

View reviewed changes

src/cryptonote_protocol/cryptonote_protocol_defs.h Outdated Show resolved Hide resolved

vtnerd force-pushed the feature/p2p_ssl_squashed branch from 5208f25 to d8e410c Compare November 16, 2023 17:33

vtnerd force-pushed the feature/p2p_ssl_squashed branch from d8e410c to 6f1e246 Compare November 16, 2023 17:35

vtnerd force-pushed the feature/p2p_ssl_squashed branch 3 times, most recently from af7d5c0 to 5d35a9a Compare November 18, 2023 22:22

vtnerd force-pushed the feature/p2p_ssl_squashed branch from 5d35a9a to da4fc6c Compare April 28, 2024 21:04

vtnerd force-pushed the feature/p2p_ssl_squashed branch from da4fc6c to 3d88cc0 Compare April 28, 2024 21:24

vtnerd force-pushed the feature/p2p_ssl_squashed branch from 3d88cc0 to a7c49c9 Compare April 29, 2024 02:03

vtnerd mentioned this pull request Jun 9, 2024

vtnerd/Lee Clagett 2024 Q1/Q2 month 2 completion notes MAGICGrants/Monero-Fund#29

Closed

vtnerd force-pushed the feature/p2p_ssl_squashed branch from a7c49c9 to b3c8322 Compare June 17, 2024 23:25

vtnerd force-pushed the feature/p2p_ssl_squashed branch from b3c8322 to 60925f1 Compare June 18, 2024 02:44

vtnerd mentioned this pull request Jul 10, 2024

vtnerd/Lee Clagett 2024 Q1/Q2 month 3 completion notes MAGICGrants/Monero-Fund#31

Closed

Rucknium mentioned this pull request Sep 11, 2024

Monero Research Lab Meeting - Wed 11 September 2024, 17:00 UTC monero-project/meta#1070

Closed

vtnerd force-pushed the feature/p2p_ssl_squashed branch from 60925f1 to 0b18c43 Compare March 3, 2025 01:18

vtnerd force-pushed the feature/p2p_ssl_squashed branch from 0b18c43 to 0bc4259 Compare March 3, 2025 01:22

plowsof mentioned this pull request Jun 30, 2025

Monero Community Workgroup Meeting: Jun 21st 16:00 UTC monero-project/meta#1224

Closed

vtnerd force-pushed the feature/p2p_ssl_squashed branch from 0bc4259 to eb47b3f Compare August 20, 2025 01:47

vtnerd force-pushed the feature/p2p_ssl_squashed branch 2 times, most recently from c814bf3 to ba7a086 Compare August 30, 2025 21:27

vtnerd mentioned this pull request Sep 22, 2025

Tx Relay v2: New Relay Logic with threshold-based peer dropping #9933

Open

j-berman reviewed Oct 24, 2025

View reviewed changes

vtnerd force-pushed the feature/p2p_ssl_squashed branch from ba7a086 to b2c323f Compare October 25, 2025 00:56

j-berman approved these changes Oct 27, 2025

View reviewed changes

Rucknium mentioned this pull request Nov 4, 2025

Monero Research Lab Meeting - Wed 29 October 2025, 17:00 UTC monero-project/meta#1289

Closed

Add SSL support to P2P

bb1eaab

vtnerd force-pushed the feature/p2p_ssl_squashed branch from b2c323f to bb1eaab Compare December 2, 2025 18:05

	const command_line::arg_descriptor<std::vector<std::string> > arg_p2p_add_peer = {"add-peer", "Manually add peer to local peerlist. Append \",sha-256 fingerprint\" to make authenticated connection."
	const command_line::arg_descriptor<std::vector<std::string> > arg_p2p_add_peer = {"add-peer", "Manually add peer to local peerlist. Append \",<sha-256 fingerprint>\" to make authenticated connection. "

	m_ssl(ssl_support_t::e_ssl_support_disabled),
	m_ssl(ssl_support_t::e_ssl_support_disabled),

Uh oh!

Add SSL support to P2P #8996

Are you sure you want to change the base?

Add SSL support to P2P #8996

Uh oh!

Conversation

vtnerd commented Sep 19, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

vtnerd commented Oct 6, 2023

Uh oh!

Uh oh!

vtnerd commented Nov 16, 2023

Uh oh!

vtnerd commented Nov 16, 2023

Uh oh!

vtnerd commented Nov 19, 2023

Uh oh!

j-berman commented Apr 19, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

vtnerd commented Apr 28, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

vtnerd commented Apr 28, 2024

Uh oh!

vtnerd commented Apr 29, 2024

Uh oh!

SyntheticBird45 commented Jun 17, 2024

Uh oh!

vtnerd commented Jun 17, 2024

Uh oh!

vtnerd commented Jun 18, 2024

Uh oh!

vtnerd commented Jun 18, 2024

Uh oh!

vtnerd commented Jun 18, 2024

Uh oh!

vtnerd commented Mar 3, 2025

Uh oh!

vtnerd commented Mar 3, 2025

Uh oh!

vtnerd commented Mar 3, 2025

Uh oh!

vtnerd commented Aug 20, 2025

Uh oh!

j-berman left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

vtnerd commented Oct 25, 2025

Uh oh!

jonathancross commented Dec 2, 2025

vtnerd commented Sep 19, 2023 •

edited

Loading

j-berman commented Apr 19, 2024 •

edited

Loading

vtnerd commented Apr 28, 2024 •

edited

Loading