We still close the channel if we *send* an error, but we seem to have hit
another case where LND sends an error which seems transient, so this will
make a best-effort attempt to preserve our channel in that case.
Some test have to be modified, since they don't terminate as they did
previously :(
Changelog-Changed: quirks: We'll now reconnect and retry if we get an error on an established channel. This works around lnd sending error messages that may be non-fatal.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
By combining set_state() with selected_peer() we can give a single log
line describing what we're asking for, from whom.
We also add more verbosity to a few key areas, such as gossip rotation
and when gossipd tells a peer to send an error. And move a comment which
was above the wrong function (due to rebase?).
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This means we intercept the peer's gossip_timestamp_filter request
in the per-peer subdaemon itself. The rest of the semantics are fairly
simple however.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Keeping the uintmap ordering all the broadcastable messages is expensive:
130MB for the million-channels project. But now we delete obsolete entries
from the store, we can have the per-peer daemons simply read that sequentially
and stream the gossip itself.
This is the most primitive version, where all gossip is streamed;
successive patches will bring back proper handling of timestamp filtering
and initial_routing_sync.
We add a gossip_state field to track what's happening with our gossip
streaming: it's initialized in gossipd, and currently always set, but
once we handle timestamps the per-peer daemon may do it when the first
filter is sent.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
They already send *us* gossip messages, so they have to be distinct anyway.
Why make us both do extra work?
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Encapsulating the peer state was a win for lightningd; not surprisingly,
it's even more of a win for the other daemons, especially as we want
to add a little gossip information.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We ask gossipd for the channel_update for the outgoing channel; any other
messages it sends us get queued for later processing.
But this is overzealous: we can shunt those msgs to the peer while
we're waiting. This fixes a nasty case where we have to handle
WIRE_GOSSIPD_NEW_STORE_FD messages by queuing the fd for later.
This then means that WIRE_GOSSIPD_NEW_STORE_FD can be handled
internally inside handle_gossip_msg(), since it's always dealt with
the same, simplifying all callers.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Instead of reading the store ourselves, we can just send them an
offset. This saves gossipd a lot of work, putting it where it belongs
(in the daemon responsible for the specific peer).
MCP bench results:
store_load_msec:28509-31001(29206.6+/-9.4e+02)
vsz_kb:580004-580016(580006+/-4.8)
store_rewrite_sec:11.640000-12.730000(11.908+/-0.41)
listnodes_sec:1.790000-1.880000(1.83+/-0.032)
listchannels_sec:21.180000-21.950000(21.476+/-0.27)
routing_sec:2.210000-11.160000(7.126+/-3.1)
peer_write_all_sec:36.270000-41.200000(38.168+/-1.9)
Signficant savings in streaming gossip:
-peer_write_all_sec:48.160000-51.480000(49.608+/-1.1)
+peer_write_all_sec:35.780000-37.980000(36.43+/-0.81)
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Similar to the previous "handle peer input before gossip input", this
fixes similar potential deadlock for closingd and openingd which use
peer_or_gossip_sync_read.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This avoids some very ugly switch() statements which mixed the two,
but we also take the chance to rename 'towire_gossip_' to
'towire_gossipd_' for those inter-daemon messages; they're messages to
gossipd, not gossip messages.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Also means we simplify the handle_gossip_msg() since everyone wants it to
use sync_crypto_write().
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
The One Big API is confusing, and has enough corner cases that we should
ditch it rather than add more.
See: https://www.sandimetz.com/blog/2016/1/20/the-wrong-abstraction
In particular, when openingd is changed to chat to peers even when
it's not actively opening a channel, it wants to handle (most) errors
by continuing, not calling peer_failed().
This exposes the constituent parts.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
That was the cause of the bad gossip order failures: gossipd thought our
channel was live, but the other end didn't receive message last time.
Now gossipd doesn't use fd to kill us (connectd tells master to do so), we
can implement read_peer_msg_nogossip().
Fixes: #1706
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
structeq() is too dangerous: if a structure has padding, it can fail
silently.
The new ccan/structeq instead provides a macro to define foo_eq(),
which does the right thing in case of padding (which none of our
structures currently have anyway).
Upgrade ccan, and use it everywhere. Except run-peer-wire.c, which
is only testing code and can use raw memcmp(): valgrind will tell us
if padding exists.
Interestingly, we still declared short_channel_id_eq, even though
we didn't define it any more!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This means that openingd and closingd now forward our gossip. But the real
reason we want to do this is that it gives an easy way for gossipd to kill
any active daemon, by closing its fd: previously closingd and openingd didn't
read the fd, so tended not to notice.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
(This was sitting in my gossip-enchancement patch queue, but it simplifies
this set too, so I moved it here).
In 94711969f we added an explicit gossip_index so when gossipd gets
peers back from other daemons, it knows what gossip it has sent (since
gossipd can send gossip after the other daemon is already complete).
This solution is insufficient for the more general case where gossipd
wants to send other messages reliably, so replace it with the other
solution: have gossipd drain the "gossip fd" which the daemon returns.
This turns out to be quite simple, and is probably how I should have
done it originally :(
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
If we're going to simply take() a pointer, don't allocate it off a random
object. Using NULL makes our intent clear, particularly with allocating
packets we're going to take() onto a queue.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Because peer_failed would previously drop the connection, we had a
special 'negotiation_failed' message which made the master hand it
back to gossipd. We don't need that any more.
This also meant we no longer need a special hook in read_peer_msg
for openingd to send this message.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Several daemons (onchaind, hsm) want to use the status messages, but
don't communicate with peers. The coming changes made them drag in
more code they didn't need, so instead we have a different
non-overlapping type.
We combine the status_received_errmsg and status_sent_errmsg
into a single status_peer_error, with the presence or not of the
'error_for_them' field indicating direction.
We also rename status_fatal_connection_lost() to
peer_failed_connection_lost() to fit in.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
It's a bit ugly because each caller has slightly different needs, but
we have three hooks and standard helpers.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Rusty Russell
darosior
practicalswift