Currently, anything which doesn't have a live channel is considered transient.
We free this first under stress, and also if they're still connecting.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
If we don't find one searching from our random spot in the peer table,
we're supposed to wrap, not crash!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We use a crude heuristic: if we were trying to contact them, it's a
"deliberate" connection, and should be preserved.
Changelog-Changed: connectd: prioritize peers with channels (and log!) if we run low on file descriptors.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
I thought I was going to want to have a convenient way of counting
these, but it turns out unnecessary. Still, this is slightly more
efficient and simple, so I am including it.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This can happen if we're totally out of fds, but previously we gave
no log message indicating this!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This has the benefit of being shorter, as well as more reliable (you
will get a link error if we can't print it, not a runtime one!).
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This code was trying to check that the address type is not one of the ADDR_TYPE_TOR*
types, but the is_toraddr() function checks a domain name! The cast should have been
a clue that this was wrong!
Anyway, wireaddr_to_addrinfo() aborts on these cases already, so the asserts here are
superfluous.
Found in unrelated CI run:
```
Valgrind error file: valgrind-errors.20610
==20610== Conditional jump or move depends on uninitialised value(s)
==20610== at 0x484ED28: strlen (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==20610== by 0x138FA3: is_toraddr (wireaddr.c:344)
==20610== by 0x11499B: conn_init (connectd.c:729)
==20610== by 0x28FD73: next_plan (io.c:59)
==20610== by 0x28FF94: io_new_conn_ (io.c:116)
==20610== by 0x11531B: try_connect_one_addr (connectd.c:927)
==20610== by 0x1182A8: try_connect_peer (connectd.c:1781)
==20610== by 0x11834E: connect_to_peer (connectd.c:1797)
==20610== by 0x119241: recv_req (connectd.c:2074)
==20610== by 0x12836F: handle_read (daemon_conn.c:35)
==20610== by 0x28FD73: next_plan (io.c:59)
==20610== by 0x2909A8: do_plan (io.c:407)
==20610==
```
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
On Mac most tests report BROKEN because sodium creating an untracked fd pointing to /dev/random. dev_report_fd’s finds it at tear down and reports a BROKEN message.
We allow a single “char special” fd without reporting it as broken improving QOL for Mac developers.
While we’re here we added the fd mode to the log to help with future rogue fd issues.
ChangeLog-None
This makes it easier to use outside simple subds, and now lightningd can
simply dump to log rather than returning JSON.
JSON formatting was a lot of work, and we only did it for lightningd, not for
subdaemons. Easier to use the logs in all cases.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We still refuse to run dev commands if lightningd sends it to us
despite us not being in developer mode, but that's mainly paranoia.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Also requires us to expose memleak when !DEVELOPER, however we only
ever used the memleak tracking when the LIGHTNINGD_DEV_MEMLEAK
environment variable was set, so keep that.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We will access the freed connection to gossipd. This is weird to track
down when the *actual* issue is that gossipd died!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
I never really liked this hack: websockets are useful, advertizing
them not so much.
Note that we never actually documented that we would advertize these!
Changelog-EXPERIMENTAL: Protocol: Removed support for advertizing websocket addresses in gossip.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This contained cut & paste code, and it wasn't clear to me that
the first loop included DNS entries with IPv6 entries.
Instead, allow the iterator to take multiple types, and use
a switch statement so compile will break as new types are added.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We accept that we will fail to listen if we bind both IPv6 and IPv4 to
the same socket on a dual-stack machine (e.g. normal Linux), but we weren't
closing the fd.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Most of this is piping the flag through so we know it's a websocket!
Reported-by: @ShahanaFarooqui
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
ccan/io stores the context pointer for io_new_conn, but we were using
`daemon->listeners` which we reallocate, so it can use a stale pointer.
```
0x3e1700 call_error
ccan/ccan/tal/tal.c:93
0x3e1700 check_bounds
ccan/ccan/tal/tal.c:165
0x3e1700 to_tal_hdr
ccan/ccan/tal/tal.c:174
0x3e1211 to_tal_hdr_or_null
ccan/ccan/tal/tal.c:186
0x3e1211 tal_alloc_
ccan/ccan/tal/tal.c:426
0x3db8f4 io_new_conn_
ccan/ccan/io/io.c:91
0x3dd2e1 accept_conn
ccan/ccan/io/poll.c:277
0x3dd2e1 io_loop
ccan/ccan/io/poll.c:444
0x3419fa main
connectd/connectd.c:2081
```
Fixes: #6060
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
The `tmpctx` is free'd before the error is read out/sent over the wire;
there's a call that will copy the array before sending it, let's use
that instead and take() the object?
------------------------------- Valgrind errors --------------------------------
Valgrind error file: valgrind-errors.2181501
==2181501== Syscall param write(buf) points to unaddressable byte(s)
==2181501== at 0x49E4077: write (write.c:26)
==2181501== by 0x1C79A3: do_write (io.c:189)
==2181501== by 0x1C80AB: do_plan (io.c:394)
==2181501== by 0x1C81BA: io_ready (io.c:423)
==2181501== by 0x1CA45B: io_loop (poll.c:453)
==2181501== by 0x118593: main (connectd.c:2053)
==2181501== Address 0x4afb158 is 40 bytes inside a block of size 140 free'd
==2181501== at 0x483F0C3: free (vg_replace_malloc.c:872)
==2181501== by 0x1D103C: del_tree (tal.c:421)
==2181501== by 0x1D130A: tal_free (tal.c:486)
==2181501== by 0x1364B8: clean_tmpctx (utils.c:172)
==2181501== by 0x1266DD: daemon_poll (daemon.c:87)
==2181501== by 0x1CA334: io_loop (poll.c:420)
==2181501== by 0x118593: main (connectd.c:2053)
==2181501== Block was alloc'd at
==2181501== at 0x483C855: malloc (vg_replace_malloc.c:381)
==2181501== by 0x1D0AC5: allocate (tal.c:250)
==2181501== by 0x1D1086: tal_alloc_ (tal.c:428)
==2181501== by 0x1D124F: tal_alloc_arr_ (tal.c:471)
==2181501== by 0x126204: cryptomsg_encrypt_msg (cryptomsg.c:161)
==2181501== by 0x11335F: peer_connected (connectd.c:318)
==2181501== by 0x118A8A: peer_init_received (peer_exchange_initmsg.c:135)
==2181501== by 0x1C751E: next_plan (io.c:59)
==2181501== by 0x1C8126: do_plan (io.c:407)
==2181501== by 0x1C8168: io_ready (io.c:417)
==2181501== by 0x1CA45B: io_loop (poll.c:453)
==2181501== by 0x118593: main (connectd.c:2053)
==2181501==
{
<insert_a_suppression_name_here>
Memcheck:Param
write(buf)
fun:write
fun:do_write
fun:do_plan
fun:io_ready
fun:io_loop
fun:main
}
--------------------------------------------------------------------------------
We were stressing the servers if node cannot be found. Only do lookup
on manual connect commands.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Protocol: lightningd: Only use DNS server address lookup on manual `connect` commands, not normal reconnection attempts.
Add memleak_ignore_children() so callers can do exclusions themselves.
Having two exclusions was always such a hack!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This allows GDB to print values, but also allows us to use them in
'case' statements. This wasn't allowed before because they're not
constant terms.
This also made it clear there's a clash between two error codes,
so move one.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Changed: JSON-RPC: Error code from bcli plugin changed from 400 to 500.
We used to tell connectd to remember our connect delay, and hand it
back (increased if necessary).
Instead, simply record when we last tried to connect. If it was less
than 10 minutes ago, double delay (up to 5 minutes max), otherwise
reset delay to 1 second.
This covers all scenarios: whether we reconnect then immediately
disconnect, or never successfully connect, it doesn't matter.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Fixes: #5453
Don't log_io final messages twice (multiplex_final_message already does
this, so it's confusing to see us send e.g. WIRE_ERROR twice!).
And report that the peer has failed to connect out *before* telling
lightningd, otherwise we get a very confusing ordering, e.g.:
```
2022-07-23T05:17:36.096Z DEBUG 027d0de66d08f956a8d606c0d1c34e59bda38c05a3b1cc738fdd6378716c644997-lightningd: Reconnecting in 4 seconds
2022-07-23T05:17:36.096Z DEBUG 027d0de66d08f956a8d606c0d1c34e59bda38c05a3b1cc738fdd6378716c644997-lightningd: Will try reconnect in 4 seconds
2022-07-23T05:17:36.096Z DEBUG 027d0de66d08f956a8d606c0d1c34e59bda38c05a3b1cc738fdd6378716c644997-connectd: Failed connected out:
```
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Connectd already does this when we *receive* an error or warning, but
now do it on send. This causes some slight behavior change: we don't
disconnect when we close a channel, for example (our behaviour here
has been inconsistent across versions, depending on the code).
When connectd is told to disconnect, it now does so immediately, and
doesn't wait for subds to drain etc. That simplifies the manual
disconnect case, which now cleans up as it would from any other
disconnection when connectd says it's disconnected.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We want to avoid lost messages in the common cases.
This generalizes our drain code, by giving the subds each 5 seconds to
close themselves, but continue to allow them to send us traffic (if
peer is still connected) and continue to send them traffic.
We continue to send traffic *out* to the peer (if it's still
connected), until all subds are gone. We still have a 5 second timer
to close the connection to peer.
On reconnects, we don't do this "drain period" on reconnects: we kill
immediately.
We fix up one test which was looking for the "disconnect" message
explicitly.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This allows us to detect when lightningd hasn't seen our latest
disconnect/reconnect; in particular, we would hit the following pattern:
1. lightningd says to connect a subd.
2. connectd disconnects and reconnects.
3. connectd reads message, connects subd.
4. lightningd reads disconnect and reconnect, sends msg to connect to subd again.
5. connectd asserts because subd is alreacy connected.
This way connectd can tell if lightningd is talking about the previous
connection, and ignoere it.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Before this patch:
1. connectd says it's connected (peer_connected)
2. we tell connectd we want to talk about each channel (peer_make_active)
3. connectd gives us an fd for each channel, and we connect it to a subd (peer_active)
4. OR, connectd says it sent something about a channel we didn't tell it about, with an fd (peer_active)
Now:
1. connectd says it's connected (peer_connected)
2. we start all appropriate subds and tell connectd to what channels/fds (peer_connect_subd).
3. if connectd says it sent something about a channel we didn't tell it about, we either tell
it to hang up (peer_final_msg), or connect a new opening daemon (peer_connect_subd).
This is the minimal-size patch, which is why we create socket pairs in
so many places to use the existing functions. Many cleanups are
possible, since the new flow is so simple.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We don't have to put aside a peer which is reconnecting and wait for
lightningd to remove the old peer, we can now simply free the old
and add the new.
Fixes: #5240
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Now we have separate peer draining logic, we can simply use it when
connectd tells us to release the peer, without waiting. (We could
simply free the peer, but that's a bit rude, as messages can get
lost).
This removes various complex flags and logic we had before.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Fixed: `connectd`: various crashes and issues fixed by simplification and rewrite.
This removes it from the hashtable, and forces it to do nothing but
send out any remaining packets, then close.
It is, in effect, reduced to a stub, with no further interactions
with the rest of the system (all subds are freed already).
Also removes the need for an explicit "final_msg" too.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Got complaints about us hanging up on some nodes because they don't respond
to pings in a timely manner (e.g. ACINQ?), but that turned out to be something
else.
Nonetheless, we've had reports in the past of LND badly prioritizing gossip
traffic, and thus important messages can get queued behind gossip dumps!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Changed: connectd: give busy peers more time to respond to pings.
