We weren't killing it. Eventually it would die, and peer_owner_finished()
would access subd->peer->owner, but that peer was freed already.
Closes: #261
Reported-by: Christian Decker <decker.christian@gmail.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
To reproduce the next bug, I had to ensure that one node keeps thinking it's
disconnected, then the other node reconnects, then the first node realizes
it's disconnected.
This code does that, adding a '0' dev-disconnect modifier. That means
we fork off a process which (due to pipebuf) will accept a little
data, but when the dev_disconnect file is truncated (a hacky, but
effective, signalling mechanism) will exit, as if the socket finally
realized it's not connected any more.
The python tests hang waiting for the daemon to terminate if you leave
the blackhole around; to give a clue as to what's happening in this
case I moved the log dump to before killing the daemon.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
jl777 reported a crash when we try to pay past reserve. Fix that (and
a whole class of related bugs) and add tests.
In test_lightning.py I had to make non-async path for sendpay() non-threaded
to get the exception passed through for testing.
Closes: #236
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
I was hoping to defer HTLC updates until we actually store HTLCs, but
we need to flush to DB whenever balances update as well.
Signed-off-by: Christian Decker <decker.christian@gmail.com>
This is the big one, and it's completely anticlimactic: it loads all
channels that have reached opening and are not marked as
closingd_complete into memory, that's it.
Signed-off-by: Christian Decker <decker.christian@gmail.com>
I was hoping to trigger on more things from the bitcoind process, but
stuff like mempool is hard to trigger on. Reducing to info so we can
work a bit easier with pdb and the log becomes less noisy.
We'll need this for testing nodes going down during payment.
However, there's no good way to silence the threads that I can tell,
so we get a nasty backtrace from it.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
I tracked down a bug, and couldn't figure out why valgrind wasn't
finding it.
From man valgrind:
--log-file=<filename>
Specifies that Valgrind should send all of its messages to the
specified file. If the file name is empty, it causes an abort.
There are three special format specifiers that can be used in the
file name.
%p is replaced with the current process ID. This is very useful for
program that invoke multiple processes. WARNING: If you use
--trace-children=yes and your program invokes multiple processes OR
your program forks without calling exec afterwards, and you don't
use this specifier (or the %q specifier below), the Valgrind output
from all those processes will go into one file, possibly jumbled
up, and possibly incomplete.
"possibly incomplete" indeed!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We weren't registering reconnecting peers for broadcasts. Just
starting a timer is enough. Also added an integration test to check
that the gossip sync is being resumed.
This is a transitional state, while we're waiting to see the
closing tx onchain (which is To Be Implemented).
The simplest way to do re-transmission is to re-use closingd, and just
disallow any updates.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We actually don't need to transition if we're reconnecting, and logic
to go to CHANNELD_NORMAL was wrong: we checked that we'd seen funding tx
locked, but not that we'd received a msg from the remote peer.
We need to fix the tests now we no longer double-transition, too.
Fixes: #188
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
When we drop an HTLC_ADD packet, sometimes the commit timer fires
before we try to read from the fd. In this case, the payment is
considered committed and we don't fail.
We need manual commit to work around this, and also we'd need to do
the pay command asynchronously from python because it will block.
That's a bit out of scope for now, so just handle either way.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We keep the scriptpubkey to send until after a commitment_signed (or,
in the corner case, if there's no pending commitment). When we
receive a shutdown from the peer, we pass it up to the master.
It's up to the master not to add any more HTLCs, which works because
we move from CHANNELD_NORMAL to CHANNELD_SHUTTING_DOWN.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Currently it's fairly ad-hoc, but we need to tell it to channeld when
it restarts, so we define it as the non-HTLC balance.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
When adding their HTLCs, it needs all the information. When failing,
it needs the id as key and the failure reason. When fulfilling, it
needs the id and payment preimage.
It also needs to know when we have received an revoke_and_ack or a
commitment_signed, to place in the database.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
I have a test which waits for multiple occurrences of the same string,
but doesn't want them to overlap. Make wait_for_log() do the right thing,
so that it only looks for log entries since the last wait_for_log.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
On my laptop under load, 5 seconds was no longer enough for legacy.
But this breaks async (they all see mempool increase, and fire
prematurely), so stop doing that.
I can't get this test to work at all, in fact, without this patch.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
I actually hit this very hard to reproduce race: if we haven't process
the channeld message when block #6 comes in, we won't send the gossip
message. We wait for logs, but don't generate new blocks, and timeout
on l1.daemon.wait_for_log('peer_out WIRE_ANNOUNCEMENT_SIGNATURES').
The solution, which also tests that we don't send announcement signatures
immediately, is to generate a single block, wait for CHANNELD_NORMAL,
then (in gossip tests), generate 5 more.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
1. We explicitly assert what state we're coming from, to make transitions
clearer.
2. Every transition has a state, even between owners while waiting for HSM.
3. Explictly step though getting the HSM signature on the funding tx
before starting channeld, rather than doing it in parallel: makes
states clearer.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We need to do this on every connection, whether reconnecting or not,
so it makes sense for the handshake daemon to handle it and return
the feature fields.
Longer term I'm considering having the handshake daemon handle the
listening and connecting, and simply hand the fds back once the peers
are ready.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We currently create a peer struct, then complete handshake to find out
who it is. This means we have a half-formed peer, and worse: if it's
a reconnect we get two peers the same.
Add an explicit 'struct connection' for the handshake phase, and
construct a 'struct peer' once that's done.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Rusty Russell
Christian Decker