Commit Graph

4697 Commits

Author SHA1 Message Date
Olaoluwa Osuntokun
702bf7be29
server: Broadcast now takes a set of peers to skip
In this commit, we modify the Broadcast to take a *set* of peers to
skip, rather than just a single peer. We make this modification as when
a new channel is discovered, it’s likely the case that we get the
announcement from several peers rather than a single peer. With this
change, we’ll ensure that the caller (who is aware of the set of
senders) is able to properly avoid wasting bandwidth by re-sending the
message to all peers that sent it to us originally.
2017-12-26 16:25:40 +01:00
Olaoluwa Osuntokun
f45502f890
discovery: update message processing to emit networkMsg rather than lnwire.Message
In this commit, we make an incremental step towards page of the new
feature of deDupedAnnoucnements to return the set of senders for each
message. All methods the process new channel announcements, will now
return an instance of networkMsg rather than lnwire.Message. This will
allow passing the returned announcement directly into
deDupedAnnoucnements.AddMsg().
2017-12-26 16:23:12 +01:00
Olaoluwa Osuntokun
f1b40e0b4d
discovery: modify deDupedAnnouncements to keep track of senders for each msg
In this commit, we modify the deDupedAnnouncements struct slightly. The
element of this struct will now keep track of the set of senders that
sent a particular message. Each time a message is added, we’ll replace
the new message with the old (as normal), but we’ll also add the new
sender to the set of known senders.

With this new feature, we’ll be able to avoid re-sending a message to
the peer that sent it to us in the first place.
2017-12-26 16:19:03 +01:00
Olaoluwa Osuntokun
7e54b4ae46
lnwire: if unable to parse encapsulated error, log length and raw bytes
In this commit, we add additional detail in the error return hen were’e
unable to properly decode an encapsulated error. This additional
logging was added with the goal of being able to track down a lingering
bug, where at times lnd cannot decode a TemporaryChannelFailure
message.
2017-12-26 16:17:22 +01:00
Matt Drollette
d6172ca8b4 discovery: correctly format ipv6 addresses from dns seed 2017-12-22 20:12:35 +01:00
Olaoluwa Osuntokun
b74b9fc16e
test: extend time outs in flaky tests due to Travis CI instance speed variance 2017-12-22 19:43:50 +01:00
Olaoluwa Osuntokun
9777176d7d
lnwallet: don't use persistent pointer to funding tx within channel state machine
This commit fixes a lingering bug that could at times cause
incompatibilities with other implementations when attempting a
cooperative channel close. Before this commit, we would use a pointer
to the funding txin everywhere. As a result, each time we made a new
state, or verified one, we would modify the sequence field of the main
txin of the commitment transaction. Due to this if we updated the
channel, then went to do a cooperative channel closure, the sequence of
the txin would still be set to the value we used as the state hint.

To remedy this, we now copy the txin each time when making the
commitment transaction, and also the cooperative closure transaction.
This avoids accidentally mutating the txin itself.

Fixes #502.
2017-12-22 19:26:35 +01:00
Olaoluwa Osuntokun
3f2a5241c1
Merge pull request #509 from halseth/fix-sendtopeer-deadlock
[trivial] server: fix deadlock in sendToPeer at peer shutdown
2017-12-22 17:48:26 +01:00
Olaoluwa Osuntokun
151a4325b1
htlcswitch: fix alignment of the packetQueue's fields for 32-bit systems (#507)
In this commit, we fix an existing issue that would cause lnd to panic
on 32-bit systems. Within the packetQueue we utilize atomics heavily.
However, it's the caller's job to ensure 64-bit alignment of 64-bit words
accessed atomically. This is documented within the sync/atomic package
as a set of known bugs.

The old alignment of this struct was:

⛰  structlayout github.com/lightningnetwork/lnd/htlcswitch packetQueue
packetQueue.queueLen int32: 0-4 (size 4, align 4)
padding: 4-8 (size 4, align 0)
packetQueue.totalHtlcAmt int64: 8-16 (size 8, align 8)
packetQueue.queueCond *sync.Cond: 16-24 (size 8, align 8)
packetQueue.queueMtx.state int32: 24-28 (size 4, align 4)
packetQueue.queueMtx.sema uint32: 28-32 (size 4, align 4)
packetQueue.queue []*github.com/lightningnetwork/lnd/htlcswitch.htlcPacket: 32-56 (size 24, align 8)
packetQueue.outgoingPkts chan *github.com/lightningnetwork/lnd/htlcswitch.htlcPacket: 56-64 (size 8, align 8)
packetQueue.freeSlots chan struct{}: 64-72 (size 8, align 8)
packetQueue.wg.noCopy sync.noCopy: 72-72 (size 0, align 1)
packetQueue.wg.state1 [12]byte: 72-84 (size 12, align 1)
packetQueue.wg.sema uint32: 84-88 (size 4, align 4)
packetQueue.quit chan struct{}: 88-96 (size 8, align 8)

After this commit, the new alignment of this sturct is:

⛰  structlayout  -json github.com/lightningnetwork/lnd/htlcswitch packetQueue | structlayout-optimize
packetQueue.queue []*github.com/lightningnetwork/lnd/htlcswitch.htlcPacket: 0-24 (size 24, align 8)
packetQueue.wg struct: 24-40 (size 16, align 8)
packetQueue.freeSlots chan struct{}: 40-48 (size 8, align 8)
packetQueue.queueCond *sync.Cond: 48-56 (size 8, align 8)
packetQueue.queueMtx struct: 56-64 (size 8, align 8)
packetQueue.outgoingPkts chan *github.com/lightningnetwork/lnd/htlcswitch.htlcPacket: 64-72 (size 8, align 8)
packetQueue.totalHtlcAmt int64: 72-80 (size 8, align 8)
packetQueue.quit chan struct{}: 80-88 (size 8, align 8)
packetQueue.queueLen int32: 88-92 (size 4, align 8)
padding: 92-96 (size 4, align 0)

Fixes #505, and #463.
2017-12-22 16:32:11 +01:00
Johan T. Halseth
db8b4cad4a
server: fix deadlock in sendToPeer at peer shutdown
This commit fixes a deadlock that could occur when
a peer disconnected during a call to sentToPeer. In
This particular case, a message would successfully
be queued, the peer would shutdown, and we would
block waiting for an error to be returned on the
message's error channel, which would deadlock.
This fixes that by also checking for peer shutdown.
2017-12-21 23:32:39 +01:00
Olaoluwa Osuntokun
3b986b4c14
htlcswitch: extend timeouts within bandwidth consistency tests 2017-12-21 11:47:45 +01:00
Olaoluwa Osuntokun
a93be76d16
test: extend timeouts within testRevokedCloseRetribution tests 2017-12-21 11:44:46 +01:00
Olaoluwa Osuntokun
b56dcc988b
lntest: re-write loop in ConnectNodes to use WaitPredicate 2017-12-21 11:44:20 +01:00
Olaoluwa Osuntokun
6992822865
lntest: CloseChannel will wait for both nodes to show chan as active if not force
In this commit, we modify the CloseChannel to wait for both nodes to
detect that channel as being active before we attempt to close it. This
should serve to reduce many of the flakes that we’ve been seeing on
travis which were caused by node A detecting the channel as active, but
node B not, leading to a test flake under certain timing conditions.
The new function uses the recently added WaitPredicate method.
2017-12-21 11:43:17 +01:00
Olaoluwa Osuntokun
c3f84b1a47
lntest: re-write AssertChannelExists to use WaitPredicate internally
Recent changes to the funding manger’s state machine have resulted in
some additional database calls during the funding process. This has
slowed down the tests by a few ms here and there. Recent integration
test runs have begun to fail due to AssertChannelExists returning an
error as the channel hasn’t fully propagated yet. In order to remedy
this, we’ll now use WaitPredicate to poll repeatedly to ensure. This
should serve to reduce flakes encountered within the integration tests.
2017-12-21 11:35:29 +01:00
Olaoluwa Osuntokun
72147fe1e5
lntest: add new WaitPredicate helper function 2017-12-21 11:33:34 +01:00
Olaoluwa Osuntokun
2154ec130f
lntest: add new LookUpNodeByPub method to lookup active nodes by pubkey
In this commit, we add a new helper function to the NetworkHarness
struct. This helper function serves to allow test authors to look up
pointer to an active node based on its current public key.

Each time a new node is started, its public key will be re-registered
within the global nodesByPub map.
2017-12-21 11:33:04 +01:00
Olaoluwa Osuntokun
ef2838fdd3
lnwire: log type of failure upon decode failure 2017-12-21 11:31:15 +01:00
Johan T. Halseth
2f926f1a47 server: add NotifyWhenOnline method to gossiper 2017-12-19 13:01:59 -06:00
Johan T. Halseth
c3aa23f061 discovery test: update gossiper test for new database structure 2017-12-19 13:01:59 -06:00
Johan T. Halseth
e3ae204fcb discovery: make gossiper able to resend failed AnnounceSignatures
This commit makes the gossiper track the state of a local
AnnounceSignature message, such that it can retry sending
it to the remote peer if needed. It will also persist this
state in the WaitingProofStore, such that it can resume
from this state at startup.
2017-12-19 13:01:59 -06:00
Johan T. Halseth
8a79bbf383 server: return error from sendPeerMessages
This commit adds a return error to sendPeerMessages,
making it possible to observe if a message sent to
a peer fails or succeeds.
2017-12-19 13:01:59 -06:00
Johan T. Halseth
25b77a0aee peer: add error chan to queueMsg 2017-12-19 13:01:59 -06:00
Daniel Coonce
c36b1c3992 lncli: Fix "invalid handle" error
Uses syscall.Stdin rather than 0 for terminal.ReadPassword() calls.
2017-12-19 14:56:42 +01:00
Olaoluwa Osuntokun
e0c292d960
Merge pull request #486 from cfromknecht/invoice-itest-cancel
lnd_test: invoice subscription cancel lingering goroutine
2017-12-17 18:42:36 -08:00
Olaoluwa Osuntokun
3bf4b5a830
Merge pull request #489 from cfromknecht/force-close-complete-payreq
lnd_test: use completePaymentRequests in channel force close itest
2017-12-17 18:41:01 -08:00
Matt Drollette
adf0d98194 multi: fix several typos in godoc comments 2017-12-17 18:40:05 -08:00
Johan T. Halseth
b74a281b57 discovery test: test for remote ChannelUpdate for unknown channel
This commit adds a test that ensures that if we receive a
ChannelUpdate for a channel we don't know about, it will
be reprocessed after we receive a ChannelAnnouncement for
that channel.
2017-12-17 18:35:34 -08:00
Johan T. Halseth
db829cd0c5 discovery: reprocess premature ChannelUpdates
This commit makes the gossiper store received ChannelUpdates
that is not for any known channel in a map, such that they
can be reprocessed when the ChannelAnnouncement arrives.

This is done to handle the case where we receive a ChannelUpdate
from our channel counterparty before we have been able to process
our own local ChannelAnnouncement.
2017-12-17 18:35:34 -08:00
Johan T. Halseth
8e120d1e62 lnwallet test: add FFAnnounceChan flag to reservation init 2017-12-17 18:35:34 -08:00
Johan T. Halseth
fa5b5521bf lnd_test: update integration tests for 6 block channel announcements 2017-12-17 18:35:34 -08:00
Johan T. Halseth
d7072c4f8f lntest/harness: set Private=false in OpenChannel 2017-12-17 18:35:34 -08:00
Johan T. Halseth
b3948d352e funding tests: update tests for the 6 block and private channel changes 2017-12-17 18:35:34 -08:00
Johan T. Halseth
9af7cc9b99 funding: cleanup of 6 conf and 'private' channel logic
This commits slightly rewrites the newly introduced
logic for private channels. Instead of keeping the
channel announce preference in a database within
fundingManager, it is stored as part of the
OpenChannel struct.

In addition, the ChanOpenStatus_Open update is now
sent after the channel is added to the router, instead
of waiting until the 6 blocks confirmation has passed.
2017-12-17 18:35:34 -08:00
Johan T. Halseth
fffe15f0fd lnwallet: add FundingFlag parameter to InitChannelReservation 2017-12-17 18:35:34 -08:00
Johan T. Halseth
10b838e9b4 gossiper: private channel cleanup + comments
This commit adds some comments and does some cleanup
of the logic that makes sure non-public channels
(channels with no AuthProof) are not broadcasted
to the network.
2017-12-17 18:35:34 -08:00
Johan T. Halseth
48e22219c0 channeldb: add ChannelFlags to OpenChannel struct
This commit adds the ChannelFlags field, of type
lnwire.FundingFlags, to the OpenChannel struct,
including serialization for database storage.
This is done to preserve the flags that were
sent during channel opening, currently used
to determine whether a channel should be made
public or not after opening.
2017-12-17 18:35:34 -08:00
Johan T. Halseth
17d1d5dc9c lncli: add 'private' flag to openchannel command
This make lncli openchannel take a --private
parameter, set to false by default.
2017-12-17 18:35:34 -08:00
Johan T. Halseth
464f57d3a5 lnrpc: update godoc for 'private' flag 2017-12-17 18:35:34 -08:00
nsa
f06f6bc869 funding test: test private channels.
Tests in fundingmanager_test.go that
assert the privacy guarantees of non-broadcasting.
2017-12-17 18:35:34 -08:00
nsa
c954ca1f21 funding: adding to graph, private chans, chanAnn after 6 confs
This commit introduces some new interdependent functionality. As
soon as the fundingLocked message is sent, the channel is
immediately added to the ChannelRouter's internal topology.

Finally, channels are now only broadcasted to the greater
network after six confirmations on the funding transaction
has been reached.
2017-12-17 18:35:34 -08:00
nsa
dab7f31e44 gossiper: Only send channel updates for edges with AuthProof. 2017-12-17 18:35:34 -08:00
nsa
3beef359a2 rpcserver: forward private parameter to OpenChannel 2017-12-17 18:35:34 -08:00
nsa
5c4add8e0e pilot: pass private=false to OpenChannel 2017-12-17 18:35:34 -08:00
nsa
9fcb845bd2 server: use private parameter when opening channel 2017-12-17 18:35:34 -08:00
nsa
17c98473d9 lnrpc: add --private arg to OpenChanRequest 2017-12-17 18:35:34 -08:00
Olaoluwa Osuntokun
8411a5b8ce
rpc: on chan close, fallback to default fee rate if fee estimate too low
On testate as times the fee estimation can swing widely. As we
currently don’t yet use vsize everywhere internally, we’re forced to
manually scale to weight for the moment. If the returned fee rate is
too low, then it can cause our estimate to go to zero. This also has
the effect of meaning that the chanCloser doesn’t currently advance if
the initial starting fee is zero.
2017-12-16 16:37:41 -08:00
Jim Posen
84d7c7ee2c lntest: Refactor set up calls to NetworkHarness. 2017-12-14 20:06:23 -08:00
Jim Posen
02177a3f1c lntest: Improve HarnessNode stop logic and remove restart(). 2017-12-14 20:06:23 -08:00
Jim Posen
19ed1fb8db lntest: Add ShutdownNode method to harness.
This is preferable to calling Shutdown on the node directly so that
the harness manages the entire lifecycle of an lnd process.
2017-12-14 20:06:23 -08:00