Commit Graph

852 Commits

Author SHA1 Message Date
Giovanni Petrantoni 98d82fce5c
fix opportunistic graft in internal 11 testing (#390) 2020-10-05 11:35:03 +09:00
Giovanni Petrantoni 0f2435f551
better opportunistic grafting score (when score is disabled) (#389) 2020-10-03 09:26:45 +09:00
Dean Eigenmann 853238a215
feature/expose-matcher (#387)
* exposes matcher

* might work

* fix

* fix
2020-10-02 08:59:15 -06:00
Giovanni Petrantoni 4a98a8af5a
gossip pruning fixes related to #371 (#385)
* gossip pruning fixes related to #371

* better trace for grafted/pruned

* shorted azure testing again
2020-10-02 13:09:31 +09:00
Mamy Ratsimbazafy 03f5bbba6d
saner logging (#381) 2020-09-29 09:40:06 -06:00
Giovanni Petrantoni 98d0cc3a16
defaultMsgIdProvider alternative/test anonymize (#379)
* defaultMsgIdProvider alternative/test anonymize

* avoid freeze during flood tests

* avoid `empty message, skipping` situation

* test observers

* avoid double initPubSub

* fix gossip testing (specially when anonymize is on)

* make azure tests shorter
2020-09-28 09:11:18 +02:00
Jacek Sieka 8ecef46738
reencode gossipsub messages with anonymization (#378)
This helps protect against clients sending more data than they should
and thus getting penalized on topics that require anonymity
2020-09-25 18:39:34 +02:00
Jacek Sieka 17e00e642a
limit write queue length (#376)
To break a potential read/write deadlock, gossipsub uses an unbounded
queue for writes - when peers are too slow to process this queue, it may
end up growing without bounds causing high memory usage.

Here, we introduce a maximum write queue length after which the peer is
disconnected - the queue is generous enough that any "normal" usage
should be fine - writes that are `await`:ed are not affected, only
writes that are launched in an `asyncSpawn` task or similar.

* avoid unnecessary copy of message when there are no send observers
* release message memory earlier in gossipsub
* simplify pubsubpeer logging
2020-09-24 18:43:20 +02:00
Jacek Sieka 58f26d4164
temporarily skip broken test on windows (#375) 2020-09-24 14:50:04 +02:00
Jacek Sieka 25bd0a18f4
small fixes (#374)
* add helper to read EOF marker after closing stream (else stream stay
alive until timeout/reset)
* don't assert on empty channel message
* don't loop when writing to chronos (no need)
2020-09-24 07:30:19 +02:00
Giovanni Petrantoni ec322124ac
allow to omit peerId and seqno (#372)
* allow to omit peerId and seqno

* small refactor

* wip

* fix message encoding

* improve rpc signature logic

* remove peerid from verify

* trace fixes

* fix message test

* fix gossip 1.0 tests
2020-09-23 17:56:33 +02:00
Dmitriy Ryajov abd234601b
move events to conn manager (#373) 2020-09-23 08:07:16 -06:00
Jacek Sieka 471e5906f6
fix gossipsub memory leak on disconnected peer (#371)
When messages can't be sent to peer, we try to establish a send
connection - this causes messages to stack up as more and more unsent
messages are blocked on the dial lock.

* remove dial lock
* run reconnection loop in background task
2020-09-22 09:05:53 +02:00
Jacek Sieka 49a12e619d
channel close race and deadlock fixes (#368)
* channel close race and deadlock fixes

* remove send lock, write chunks in one go
* push some of half-closed implementation to BufferStream
* fix some hangs where LPChannel readers and writers would not always
wake up
* simplify lazy channels
* fix close happening more than once in some orderings
* reenable connection tracking tests
* close channels first on mplex close such that consumers can read bytes

A notable difference is that BufferedStream is no longer considered EOF
until someone has actually read the EOF marker.

* docs, simplification
2020-09-21 19:48:19 +02:00
Giovanni Petrantoni b99d2039a8
Gossip one one (#240)
* allow multiple codecs per protocol (without breaking things)

* add 1.1 protocol to gossip

* explicit peering part 1

* explicit peering part 2

* explicit peering part 3

* PeerInfo and ControlPrune protocols

* fix encodePrune

* validated always, even explicit peers

* prune by score (score is stub still)

* add a way to pass parameters to gossip

* standard setup fixes

* take into account explicit direct peers in publish

* add floodPublish logic

* small fixes, publish still half broken

* make sure to waitsub in sparse test

* use var semantics to optimize table access

* wip... lvalues don't work properly sadly...

* big publish refactor, replenish and balance

* fix internal tests

* use g.peers for fanout (todo: don't include flood peers)

* exclude non gossip from fanout

* internal test fixes

* fix flood tests

* fix test's trypublish

* test interop fixes

* make sure to not remove peers from gossip table

* restore old replenishFanout

* cleanups

* restore utility module import

* restore trace vs debug in gossip

* improve fanout replenish behavior further

* triage publish nil peers (issue is on master too but just hidden behind a if/in)

* getGossipPeers fixes

* remove topics from pubsubpeer (was unused)

* simplify rebalanceMesh (following spec) and make it finally reach D_high

* better diagnostics

* merge new pubsubpeer, copy 1.1 to new module

* fix up merge

* conditional enable gossip11 module

* add back topics in peers, re-enable flood publish

* add more heartbeat locking to prevent races

* actually lock the heartbeat

* minor fixes

* with sugar

* merge 1.0

* remove assertion in publish

* fix multistream 1.1 multi proto

* Fix merge oops

* wip

* fix gossip 11 upstream

* gossipsub11 -> gossipsub

* support interop testing

* tests fixing

* fix directchat build

* control prune updates (pb)

* wip parameters

* gossip internal tests fixes

* parameters wip

* finishup with params

* cleanups/wip

* small sugar

* grafted and pruned procs

* wip updateScores

* wip

* fix logging issue

* pubsubpeer, chronicles explicit override

* fix internal gossip tests

* wip

* tables troubleshooting

* score wip

* score wip

* fixes

* fix test utils generateNodes

* don't delete while iterating in score update

* fix grafted defect

* add a handleConnect in subscribeTopic

* pruning improvements

* wip

* score fixes

* post merge - builds gossip tests

* further merge fixes

* rebalance improvements and opportunistic grafting

* fix test for now

* restore explicit peering

* implement peer exchange graft message

* add an hard cap to PX

* backoff time management

* IWANT cap/budget

* Adaptive gossip dissemination

* outbound mesh quota, internal tests fixing

* oversub prune score based, finish outbound quota

* finishup with score and ihave budget

* use go daemon 0.3.0

* import fixes

* byScore cleanup score sorting

* remove pointless scaling in `/` Duration operator

* revert using libp2p org for daemon

* interop fixes

* fixes and cleanup

* remove heartbeat assertion, minor debug fixes

* logging improvements and cleaning up

* (to revert) add some traces

* add explicit topic to gossip rpcs

* pubsub merge fixes and type fix in switch

* Revert "(to revert) add some traces"

This reverts commit 4663eaab6cc336c81cee50bc54025cf0b7bcbd99.

* cleanup some now irrelevant todo

* shuffle peers anyway as score might be disabled

* add missing shuffle

* old merge fix

* more merge fixes

* debug improvements

* re-enable gossip internal tests

* add gossip10 fallback (dormant but tested)

* split gossipsub internal tests into 1.0 and 1.1

Co-authored-by: Dmitriy Ryajov <dryajov@gmail.com>
2020-09-21 11:16:29 +02:00
Jacek Sieka b7e5d1122c
cleanups (#366)
* reuse connection timeout for noise handshake (avoid extra timer)
* enforce nbytes > 0 for readOnce
* avoid some unnecessary memory zeroing
* simplify noise
* fix dumping when noise splits message
2020-09-16 11:55:25 +02:00
Dmitriy Ryajov b0d86b95dd
add peer lifecycle events (#357)
* add peer lifecycle events

* rework peer events to not use connection events

* don't use result in pubsub and switch init

* wip

* use ordered hashes and remove logscope

* logging

* add missing test

* small fixes
2020-09-15 14:19:22 -06:00
Giovanni Petrantoni a6007be428
avoid sending empty seqno and/or fromPeer (gossip rpc) (#364) 2020-09-15 12:33:18 +02:00
Eugene Kabanov e2d46c6b53
Add `libp2p_dump` and `libp2p_dump_dir` compiler time options. (#359)
* Add `libp2p_dump` and `libp2p_dump_dir` compiler time options.

* Add tools/pbcap_parser.

* Add mplex control messages decoding.
2020-09-15 11:16:43 +02:00
Oskar Thorén 5e66f6fbd8
Add logScope to connmanager and pubsubprotobuf (#363) 2020-09-15 08:03:53 +02:00
Jacek Sieka 0db45462cd
mplex fixes (#362)
* remove almost-empty types module
* lock when writing message (that's the only place the lock matters, and
only when the message is > max msg size)
* logging updates (log in consistent order, makes reading logs easier)
* raise EOF from readExactly only if no bytes have been read (to signal
that _no_ bytes were lost)
2020-09-14 10:19:54 +02:00
Jacek Sieka 96d4c44fec
refactor bufferstream to use a queue (#346)
This change modifies how the backpressure algorithm in bufferstream
works - in particular, instead of working byte-by-byte, it will now work
seq-by-seq.

When data arrives, it usually does so in packets - in the current
bufferstream, the packet is read then split into bytes which are fed one
by one to the bufferstream. On the reading side, the bytes are popped of
the bufferstream, again byte by byte, to satisfy `readOnce` requests -
this introduces a lot of synchronization traffic because the checks for
full buffer and for async event handling must be done for every byte.

In this PR, a queue of length 1 is used instead - this means there will
at most exist one "packet" in `pushTo`, one in the queue and one in the
slush buffer that is used to store incomplete reads.

* avoid byte-by-byte copy to buffer, with synchronization in-between
* reuse AsyncQueue synchronization logic instead of rolling own
* avoid writeHandler callback - implement `write` method instead
* simplify EOF signalling by only setting EOF flag in queue reader (and
reset)
* remove BufferStream pipes (unused)
* fixes drainBuffer deadlock when drain is called from within read loop
and thus blocks draining
* fix lpchannel init order
2020-09-10 08:19:13 +02:00
Jacek Sieka 5b347adf58
logging fixes and small cleanups (#361)
In particular, allow longer multistream select reads
2020-09-09 19:12:08 +02:00
Jacek Sieka 63b38734bd
fix poor performance in LRU cache (#360)
it turns out (in NBC) a heap is sufficiently slow becuase of all the
deletes that it makes more sense to go with a linked list
2020-09-09 18:28:46 +02:00
Jacek Sieka 82c179db9e
mplex fixes (#356)
* close the right connection when channel send fails
* don't crash on channel id that is not unique
2020-09-08 08:24:28 +02:00
Jacek Sieka 2b72d485a3
a few more log fixes (#355) 2020-09-07 14:15:11 +02:00
Jacek Sieka c1856fda53
simplify and unify logging (#353)
* use short format for logging peerid
* log peerid:oid for connections
2020-09-06 10:31:47 +02:00
Jacek Sieka 9b815efe8f
gossipsub: don't subscribe to floodsub also (#352) 2020-09-04 22:53:03 +02:00
Jacek Sieka 16a008db75
fix connection event order when connection dies early (#351)
if the connection is already closed (because the remote closes during
identfiy for example), an exception would be raised which would leave
the connection in limbo, beacuse it would not go through the rest of
internalConnect.

Also, if the connection is already closed, the disconnect event would be
scheduled before the connect event :/
2020-09-04 20:30:26 +02:00
Jacek Sieka 6d91d61844
small cleanups & docs (#347)
* simplify gossipsub heartbeat start / stop
* avoid alloc in peerid check
* stop iterating over seq after unsubscribing item (could crash)
* don't crash on missing private key with enabled sigs (shouldn't happen
but...)
2020-09-04 18:31:43 +02:00
Eugene Kabanov 0b85192119
Remove asyncCheck from codebase. (#345)
* Remove asyncCheck from codebase.

* Replace all `discard` statements with new `asyncSpawn`.

* Bump `nim-chronos` requirement.
2020-09-04 18:30:45 +02:00
Jacek Sieka 5819c6a9a7
gossipsub / floodsub fixes (#348)
* mcache fixes

* remove timed cache - the window shifting already removes old messages
* ref -> object
* avoid unnecessary allocations with `[]` operator

* simplify init

* fix several gossipsub/floodsub issues

* floodsub, gossipsub: don't rebroadcast messages that fail validation
(!)
* floodsub, gossipsub: don't crash when unsubscribing from unknown
topics (!)
* gossipsub: don't send message to peers that are not interested in the
topic, when messages don't share topic list
* floodsub: don't repeat all messages for each message when
rebroadcasting
* floodsub: allow sending empty data
* floodsub: fix inefficient unsubscribe
* sync floodsub/gossipsub logging
* gossipsub: include incoming messages in mcache (!)
* gossipsub: don't rebroadcast already-seen messages (!)
* pubsubpeer: remove incoming/outgoing seen caches - these are already
handled in gossipsub, floodsub and will cause trouble when peers try to
resubscribe / regraft topics (because control messages will have same
digest)
* timedcache: reimplement without timers (fixes timer leaks and extreme
inefficiency due to per-message closures, futures etc)
* timedcache: ref -> obj
2020-09-04 08:10:32 +02:00
Eugene Kabanov c0bc73ddac
Fix Azure CI x86 problems. (#350) 2020-09-03 20:13:37 +03:00
Jacek Sieka cd1c68dbc5
avoid send deadlock by not allowing send to block (#342)
* avoid send deadlock by not allowing send to block

* handle message issues more consistently
2020-09-01 09:33:03 +02:00
Dmitriy Ryajov d3182c4dba
No raise send (#339)
* dont raise in send

* check that the lock is acquire on release
2020-08-20 20:50:33 -06:00
Giovanni Petrantoni 840a76915e warn -> debug log levels in errors.nim 2020-08-20 16:53:28 +09:00
Jacek Sieka eb13845f65 work around send that may raise
`send` can raise exceptions that together with asyncCheck will
crash NBC
2020-08-19 14:25:30 +03:00
Zahary Karadjov af0955c58b
Add comments explaning a possible deadlock 2020-08-18 13:51:41 +03:00
Zahary Karadjov 60122a044c
Restore interop with Lighthouse by preventing concurrent meshsub dials 2020-08-17 22:40:58 +03:00
Jacek Sieka 833a5b8e57
add muxer nil check 2020-08-17 13:32:02 +02:00
Jacek Sieka cfcda3c3ef
work around race conditions between identify and other protocols
when identify is run on incoming connections, the connmanager tables are
updated too late for incoming connections to properly be handled

this is a quickfix that will eventually need cleaning up
2020-08-17 13:29:45 +02:00
Jacek Sieka 790b67c923
work around bufferstream deadlock (#332)
mplex backpressure handling deadlocks with something
2020-08-17 12:45:54 +02:00
Jacek Sieka 53877e97bd
trace logs 2020-08-17 12:39:25 +02:00
Jacek Sieka f46bf0faa4
remove send lock (#334)
* remove send lock

When mplex receives data it will block until a reader has processed the
data. Thus, when a large message is received, such as a gossipsub
subscription table, all of mplex will be blocked until all reading is
finished.

However, if at the same time a `dial` to establish a gossipsub send
connection is ongoing, that `dial` will be blocked because mplex is no
longer reading data - specifically, it might indeed be the connection
that's processing the previous data that is waiting for a send
connection.

There are other problems with the current code:
* If an exception is raised, it is not necessarily raised for the same
connection as `p.sendConn`, so resetting `p.sendConn` in the exception
handling is wrong
* `p.isConnected` is checked before taking the lock - thus, if it
returns false, a new dial will be started. If a new task enters `send`
before dial is finished, it will also determine `p.isConnected` is
false, then get stuck on the lock - when the previous task finishes and
releases the lock, the new task will _also_ dial and thus reset
`p.sendConn` causing a leak.

* prefer existing connection

simplifies flow
2020-08-17 12:38:27 +02:00
Jacek Sieka b12145dff7
avoid crash when subscribe is received (#333)
...by making subscribeTopic synchronous, avoiding a peer table lookup
completely.

rebalanceMesh will be called a second later - it's fine
2020-08-17 12:10:22 +02:00
Jacek Sieka ab864fc747
logging cleanups and small fixes (#331) 2020-08-15 21:50:31 +02:00
Jacek Sieka 397f9edfd4
simplify mplex (#327)
* less async
* less copying of data
* less redundant cleanup
2020-08-15 07:58:30 +02:00
Jacek Sieka 9c7e055310
set activity flag on noise / secio (#330) 2020-08-15 07:36:15 +02:00
Dmitriy Ryajov d1f1e1b31e
add missing mplex half closed test (#326) 2020-08-12 07:23:49 +02:00
Dmitriy Ryajov b76b3e0e9b
Rework pubsub (#322)
* move pubsub of off switch, pass switch into pubsub

* use join on lpstreams

* properly cleanup up failed peers

* fix tests

* fix peertable hasPeerId

* fix tests

* rework sending, remove helpers from pubsubpeer, unify in broadcast

* further split broadcast into send

* use send where appropriate

* use formatIt

* improve trace

Co-authored-by: Giovanni Petrantoni <giovanni@fragcolor.xyz>
2020-08-11 18:05:49 -06:00