bad response to getutxout (bad sats amount)

[kirelagin]

Issue and Steps to Reproduce

I have the following setup:

• A “public” lightning node (v0.8.1) connected to public nodes with (currently) one public channel open.
• A “private” lightning node connected (not yet) only to my public node.
• Each lightning has its own bitcoin node (Bitcoin Core version v0.19.0.1-g1bc9988993ee84bc814e5a7f33cc90f670a19f6a).
• All this on Bitcoin testnet3.

The setup was performed as follows:

• Fully sync the first BTC node, start the first lightning node.
• Connect the first lightning node to a public node, fund the channel.
• Fully sync the second BTC node, start the second lightning node.
• Connect the first lightning node to the second one.

Output from the second (“private”) node:

{
   "peers": [
      {
         "id": "021c38ebb30aa17901aec609e98da063b4e433bfbe45656bddb34ea688192bd734",
         "connected": true,
         "netaddr": [
            "<ip>:53160"
         ],
         "globalfeatures": "",
         "localfeatures": "02aaa2",
         "features": "02aaa2",
         "channels": []
      }
   ]
}

The problem is that after a couple of seconds the second node crashes with:

<path>/bin/../libexec/c-lightning/plugins/bcli error: bad response to getutxout (bad sats amount), response was {"jsonrpc":"2.0","id":1853,"result":{"amount":"0msat","script":"0020c07569ccab2b"} }
2020-03-30T11:27:08.482Z **BROKEN** lightningd: <path>/bin/../libexec/c-lightning/plugins/bcli error: bad response to getutxout (bad sats amount), response was {"jsonrpc":"2.0","id":1853,"result":{"amount":"0msat","script":"0020c07569ccab2b"} }
FATAL SIGNAL 6 (version v0.8.1)

The same happens if I try to connect in the opposite direction, that is from the second node to the first.

I also tried wiping the state of c-lightnind and bitcoind on the second server and resyncing from scratch, same result.

getinfo output

First lightning node:

{
   "id": "021c38ebb30aa17901aec609e98da063b4e433bfbe45656bddb34ea688192bd734",
   "alias": "STRANGESET",
   "color": "021c38",
   "num_peers": 1,
   "num_pending_channels": 0,
   "num_active_channels": 1,
   "num_inactive_channels": 0,
   "address": [],
   "binding": [
      {
         "type": "ipv4",
         "address": "0.0.0.0",
         "port": 9735
      }
   ],
   "version": "v0.8.1",
   "blockheight": 1692037,
   "network": "testnet",
   "msatoshi_fees_collected": 0,
   "fees_collected_msat": "0msat",
   "lightning-dir": "/var/lib/clightning/testnet"
}

Second lightning node:

{
   "id": "03d44f0feafb17d8ad785ea7e9a96b591bcf8a727c735d51ee16d564f779dbc7b5",
   "alias": "LOUDCALENDAR",
   "color": "03d44f",
   "num_peers": 0,
   "num_pending_channels": 0,
   "num_active_channels": 0,
   "num_inactive_channels": 0,
   "address": [],
   "binding": [
      {
         "type": "ipv4",
         "address": "0.0.0.0",
         "port": 9735
      }
   ],
   "version": "v0.8.1",
   "blockheight": 1692040,
   "network": "testnet",
   "msatoshi_fees_collected": 0,
   "fees_collected_msat": "0msat",
   "lightning-dir": "/var/lib/clightning/testnet"
}

---

I am still trying to figure out what exactly is going on, any pointers would be very appreciated. For now I can’t even make sense of the error message as 0msat sounds like a valid amount, although I am not really sure what format it expects.

[kirelagin]

This is the request that causes it to crash:

$ bitcoin-cli gettxout "9112fdaee5566f27343b466b868041e33c539dbac8ee9e4e50b6af7bc60d255e" 0
{
  "bestblock": "0000000000000e961f7fe3f8100282a7117285124a67c40977d08e3619ee8e24",
  "confirmations": 399396,
  "value": 0.00000000,
  "scriptPubKey": {
    "asm": "0 5e3fdfa2ac033fb24c31ae0fd92a6bdfe313bb7c687f4d25f58a7581bb328128",
    "hex": "00205e3fdfa2ac033fb24c31ae0fd92a6bdfe313bb7c687f4d25f58a7581bb328128",
    "reqSigs": 1,
    "type": "witness_v0_scripthash",
    "addresses": [
      "tb1qtclalg4vqvlmynp34c8aj2ntml338wmudpl56f043f6crwejsy5qnd79kg"
    ]
  },
  "coinbase": false
}

[kirelagin]

The weirdest thing of all is that there is no packet in the capture that would contain amount 0msat as the error says.

[kirelagin]

Putting the misleading error aside, according to the comment at parse_amount_sat, 0.00000000 is indeed not a valid input:

/* Valid strings:
 *  [0-9]+ => satoshi.
 *  [0-9]+sat => satoshi.
 *  [0-9]+000msat => satoshi.
 *  [0-9]+.[0-9]{1,8}btc => satoshi.
 */

[kirelagin]

cc @rustyrussell since you wrote this code

[kirelagin]

Ah, ok, so bcli.c converts this bitcoind amount to some other form. So the problem seems to be somewhere around json_to_bitcoin_amount and json_add_amount_sat_only.

[kirelagin]

Probably cc @darosior as it seems like the bug might have been introduced in #3488, although I am not 100% sure.

[darosior]

@kirelagin you are right, Bitcoin #3605 fixes this.

[darosior]

And I'm currently testing Rusty's latest patch on testnet ;)

[kirelagin]

Alright, so tldr:

bcli.c receives amount in sat and then calls json_add_amount_sat_only, which converts it to msat and adds the msat suffix. Now, when parsing it, the parse_amount_sat function sees the msat suffix and verifies that the number ends with three zeroes, aka, it is a whole number of satoshis.

Which works almost great, except that when you multiply 0 by whatever, you still get 0. So it sounds like adding a special case for 0msat is the way to go.

[kirelagin]

Oh ok cool.

[darosior]

Will commit his patch and push it so you can pull this branch.