nim-libp2p-experimental/libp2p/connmanager.nim

573 lines
16 KiB
Nim
Raw Normal View History

2022-07-01 18:19:57 +00:00
# Nim-LibP2P
2023-01-20 14:47:40 +00:00
# Copyright (c) 2023 Status Research & Development GmbH
2022-07-01 18:19:57 +00:00
# Licensed under either of
# * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
# * MIT license ([LICENSE-MIT](LICENSE-MIT))
# at your option.
# This file may not be copied, modified, or distributed except according to
# those terms.
when (NimMajor, NimMinor) < (1, 4):
{.push raises: [Defect].}
else:
{.push raises: [].}
import std/[options, tables, sequtils, sets]
2021-05-25 17:06:04 +00:00
import pkg/[chronos, chronicles, metrics]
import peerinfo,
peerstore,
stream/connection,
2020-09-23 14:07:16 +00:00
muxers/muxer,
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
utils/semaphore,
2020-09-23 14:07:16 +00:00
errors
logScope:
2020-12-01 17:34:27 +00:00
topics = "libp2p connmanager"
declareGauge(libp2p_peers, "total connected peers")
const
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
MaxConnections* = 50
MaxConnectionsPerPeer* = 1
type
TooManyConnectionsError* = object of LPError
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
2020-09-23 14:07:16 +00:00
ConnEventKind* {.pure.} = enum
Connected, # A connection was made and securely upgraded - there may be
# more than one concurrent connection thus more than one upgrade
# event per peer.
Disconnected # Peer disconnected - this event is fired once per upgrade
# when the associated connection is terminated.
ConnEvent* = object
case kind*: ConnEventKind
of ConnEventKind.Connected:
incoming*: bool
else:
discard
ConnEventHandler* =
proc(peerId: PeerId, event: ConnEvent): Future[void]
{.gcsafe, raises: [Defect].}
2020-09-23 14:07:16 +00:00
PeerEventKind* {.pure.} = enum
2020-09-23 14:07:16 +00:00
Left,
Identified,
2020-09-23 14:07:16 +00:00
Joined
PeerEvent* = object
case kind*: PeerEventKind
of PeerEventKind.Joined:
initiator*: bool
else:
discard
2020-09-23 14:07:16 +00:00
PeerEventHandler* =
proc(peerId: PeerId, event: PeerEvent): Future[void] {.gcsafe, raises: [Defect].}
2020-09-23 14:07:16 +00:00
MuxerHolder = object
muxer: Muxer
handle: Future[void]
ConnManager* = ref object of RootObj
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
maxConnsPerPeer: int
inSema*: AsyncSemaphore
outSema*: AsyncSemaphore
2021-12-16 10:05:20 +00:00
conns: Table[PeerId, HashSet[Connection]]
muxed: Table[Connection, MuxerHolder]
connEvents: array[ConnEventKind, OrderedSet[ConnEventHandler]]
peerEvents: array[PeerEventKind, OrderedSet[PeerEventHandler]]
expectedConnections: Table[PeerId, Future[Connection]]
peerStore*: PeerStore
ConnectionSlot* = object
connManager: ConnManager
direction: Direction
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
proc newTooManyConnectionsError(): ref TooManyConnectionsError {.inline.} =
result = newException(TooManyConnectionsError, "Too many connections")
proc new*(C: type ConnManager,
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
maxConnsPerPeer = MaxConnectionsPerPeer,
maxConnections = MaxConnections,
maxIn = -1,
maxOut = -1): ConnManager =
var inSema, outSema: AsyncSemaphore
if maxIn > 0 or maxOut > 0:
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
inSema = newAsyncSemaphore(maxIn)
outSema = newAsyncSemaphore(maxOut)
elif maxConnections > 0:
inSema = newAsyncSemaphore(maxConnections)
outSema = inSema
else:
raiseAssert "Invalid connection counts!"
C(maxConnsPerPeer: maxConnsPerPeer,
inSema: inSema,
outSema: outSema)
2021-12-16 10:05:20 +00:00
proc connCount*(c: ConnManager, peerId: PeerId): int =
2020-09-23 14:07:16 +00:00
c.conns.getOrDefault(peerId).len
proc connectedPeers*(c: ConnManager, dir: Direction): seq[PeerId] =
var peers = newSeq[PeerId]()
for peerId, conns in c.conns:
if conns.anyIt(it.dir == dir):
peers.add(peerId)
return peers
2020-09-23 14:07:16 +00:00
proc addConnEventHandler*(c: ConnManager,
handler: ConnEventHandler,
kind: ConnEventKind) =
2020-09-23 14:07:16 +00:00
## Add peer event handler - handlers must not raise exceptions!
##
try:
if isNil(handler): return
c.connEvents[kind].incl(handler)
except Exception as exc:
# TODO: there is an Exception being raised
# somewhere in the depths of the std.
# Might be related to https://github.com/nim-lang/Nim/issues/17382
raiseAssert exc.msg
2020-09-23 14:07:16 +00:00
proc removeConnEventHandler*(c: ConnManager,
handler: ConnEventHandler,
kind: ConnEventKind) =
try:
c.connEvents[kind].excl(handler)
except Exception as exc:
# TODO: there is an Exception being raised
# somewhere in the depths of the std.
# Might be related to https://github.com/nim-lang/Nim/issues/17382
raiseAssert exc.msg
2020-09-23 14:07:16 +00:00
proc triggerConnEvent*(c: ConnManager,
peerId: PeerId,
event: ConnEvent) {.async, gcsafe.} =
2020-09-23 14:07:16 +00:00
try:
trace "About to trigger connection events", peer = peerId
if c.connEvents[event.kind].len() > 0:
trace "triggering connection events", peer = peerId, event = $event.kind
2020-09-23 14:07:16 +00:00
var connEvents: seq[Future[void]]
for h in c.connEvents[event.kind]:
connEvents.add(h(peerId, event))
2020-09-23 14:07:16 +00:00
checkFutures(await allFinished(connEvents))
except CancelledError as exc:
raise exc
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
except CatchableError as exc:
2020-09-23 14:07:16 +00:00
warn "Exception in triggerConnEvents",
msg = exc.msg, peer = peerId, event = $event
2020-09-23 14:07:16 +00:00
proc addPeerEventHandler*(c: ConnManager,
handler: PeerEventHandler,
kind: PeerEventKind) =
2020-09-23 14:07:16 +00:00
## Add peer event handler - handlers must not raise exceptions!
##
if isNil(handler): return
try:
c.peerEvents[kind].incl(handler)
except Exception as exc:
# TODO: there is an Exception being raised
# somewhere in the depths of the std.
# Might be related to https://github.com/nim-lang/Nim/issues/17382
raiseAssert exc.msg
2020-09-23 14:07:16 +00:00
proc removePeerEventHandler*(c: ConnManager,
handler: PeerEventHandler,
kind: PeerEventKind) =
try:
c.peerEvents[kind].excl(handler)
except Exception as exc:
# TODO: there is an Exception being raised
# somewhere in the depths of the std.
# Might be related to https://github.com/nim-lang/Nim/issues/17382
raiseAssert exc.msg
2020-09-23 14:07:16 +00:00
proc triggerPeerEvents*(c: ConnManager,
peerId: PeerId,
2020-09-23 14:07:16 +00:00
event: PeerEvent) {.async, gcsafe.} =
trace "About to trigger peer events", peer = peerId
if c.peerEvents[event.kind].len == 0:
2020-09-23 14:07:16 +00:00
return
try:
let count = c.connCount(peerId)
if event.kind == PeerEventKind.Joined and count != 1:
trace "peer already joined", peer = peerId, event = $event
2020-09-23 14:07:16 +00:00
return
elif event.kind == PeerEventKind.Left and count != 0:
trace "peer still connected or already left", peer = peerId, event = $event
2020-09-23 14:07:16 +00:00
return
trace "triggering peer events", peer = peerId, event = $event
2020-09-23 14:07:16 +00:00
var peerEvents: seq[Future[void]]
for h in c.peerEvents[event.kind]:
peerEvents.add(h(peerId, event))
2020-09-23 14:07:16 +00:00
checkFutures(await allFinished(peerEvents))
except CancelledError as exc:
raise exc
except CatchableError as exc: # handlers should not raise!
warn "Exception in triggerPeerEvents", exc = exc.msg, peer = peerId
2020-09-23 14:07:16 +00:00
proc expectConnection*(c: ConnManager, p: PeerId): Future[Connection] {.async.} =
## Wait for a peer to connect to us. This will bypass the `MaxConnectionsPerPeer`
if p in c.expectedConnections:
raise LPError.newException("Already expecting a connection from that peer")
let future = newFuture[Connection]()
c.expectedConnections[p] = future
try:
return await future
finally:
c.expectedConnections.del(p)
proc contains*(c: ConnManager, conn: Connection): bool =
## checks if a connection is being tracked by the
## connection manager
##
if isNil(conn):
return
return conn in c.conns.getOrDefault(conn.peerId)
2021-12-16 10:05:20 +00:00
proc contains*(c: ConnManager, peerId: PeerId): bool =
peerId in c.conns
proc contains*(c: ConnManager, muxer: Muxer): bool =
## checks if a muxer is being tracked by the connection
## manager
##
if isNil(muxer):
return
let conn = muxer.connection
if conn notin c:
return
if conn notin c.muxed:
return
return muxer == c.muxed.getOrDefault(conn).muxer
proc closeMuxerHolder(muxerHolder: MuxerHolder) {.async.} =
trace "Cleaning up muxer", m = muxerHolder.muxer
await muxerHolder.muxer.close()
if not(isNil(muxerHolder.handle)):
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
try:
await muxerHolder.handle # TODO noraises?
except CatchableError as exc:
trace "Exception in close muxer handler", exc = exc.msg
trace "Cleaned up muxer", m = muxerHolder.muxer
proc delConn(c: ConnManager, conn: Connection) =
let peerId = conn.peerId
c.conns.withValue(peerId, peerConns):
peerConns[].excl(conn)
if peerConns[].len == 0:
c.conns.del(peerId) # invalidates `peerConns`
libp2p_peers.set(c.conns.len.int64)
trace "Removed connection", conn
proc cleanupConn(c: ConnManager, conn: Connection) {.async.} =
## clean connection's resources such as muxers and streams
if isNil(conn):
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
trace "Wont cleanup a nil connection"
return
# Remove connection from all tables without async breaks
var muxer = some(MuxerHolder())
if not c.muxed.pop(conn, muxer.get()):
muxer = none(MuxerHolder)
delConn(c, conn)
try:
if muxer.isSome:
await closeMuxerHolder(muxer.get())
finally:
await conn.close()
trace "Connection cleaned up", conn
proc onConnUpgraded(c: ConnManager, conn: Connection) {.async.} =
try:
trace "Triggering connect events", conn
conn.upgrade()
let peerId = conn.peerId
await c.triggerPeerEvents(
peerId, PeerEvent(kind: PeerEventKind.Joined, initiator: conn.dir == Direction.Out))
await c.triggerConnEvent(
peerId, ConnEvent(kind: ConnEventKind.Connected, incoming: conn.dir == Direction.In))
except CatchableError as exc:
# This is top-level procedure which will work as separate task, so it
# do not need to propagate CancelledError and should handle other errors
warn "Unexpected exception in switch peer connection cleanup",
conn, msg = exc.msg
proc peerCleanup(c: ConnManager, conn: Connection) {.async.} =
try:
trace "Triggering disconnect events", conn
let peerId = conn.peerId
await c.triggerConnEvent(
peerId, ConnEvent(kind: ConnEventKind.Disconnected))
await c.triggerPeerEvents(peerId, PeerEvent(kind: PeerEventKind.Left))
if not(c.peerStore.isNil):
c.peerStore.cleanup(peerId)
except CatchableError as exc:
# This is top-level procedure which will work as separate task, so it
# do not need to propagate CancelledError and should handle other errors
warn "Unexpected exception peer cleanup handler",
conn, msg = exc.msg
proc onClose(c: ConnManager, conn: Connection) {.async.} =
## connection close even handler
##
## triggers the connections resource cleanup
##
try:
await conn.join()
trace "Connection closed, cleaning up", conn
await c.cleanupConn(conn)
except CancelledError:
# This is top-level procedure which will work as separate task, so it
# do not need to propagate CancelledError.
debug "Unexpected cancellation in connection manager's cleanup", conn
except CatchableError as exc:
debug "Unexpected exception in connection manager's cleanup",
errMsg = exc.msg, conn
finally:
trace "Triggering peerCleanup", conn
asyncSpawn c.peerCleanup(conn)
proc selectConn*(c: ConnManager,
2021-12-16 10:05:20 +00:00
peerId: PeerId,
dir: Direction): Connection =
## Select a connection for the provided peer and direction
##
let conns = toSeq(
c.conns.getOrDefault(peerId))
.filterIt( it.dir == dir )
if conns.len > 0:
return conns[0]
2021-12-16 10:05:20 +00:00
proc selectConn*(c: ConnManager, peerId: PeerId): Connection =
## Select a connection for the provided giving priority
## to outgoing connections
##
var conn = c.selectConn(peerId, Direction.Out)
if isNil(conn):
conn = c.selectConn(peerId, Direction.In)
if isNil(conn):
trace "connection not found", peerId
return conn
proc selectMuxer*(c: ConnManager, conn: Connection): Muxer =
## select the muxer for the provided connection
##
if isNil(conn):
return
if conn in c.muxed:
return c.muxed.getOrDefault(conn).muxer
else:
debug "no muxer for connection", conn
proc storeConn*(c: ConnManager, conn: Connection)
{.raises: [Defect, LPError].} =
## store a connection
##
if isNil(conn):
raise newException(LPError, "Connection cannot be nil")
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
if conn.closed or conn.atEof:
raise newException(LPError, "Connection closed or EOF")
let peerId = conn.peerId
# we use getOrDefault in the if below instead of [] to avoid the KeyError
if peerId in c.expectedConnections and
not(c.expectedConnections.getOrDefault(peerId).finished):
c.expectedConnections.getOrDefault(peerId).complete(conn)
elif c.conns.getOrDefault(peerId).len > c.maxConnsPerPeer:
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
debug "Too many connections for peer",
conn, conns = c.conns.getOrDefault(peerId).len
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
raise newTooManyConnectionsError()
c.conns.mgetOrPut(peerId, HashSet[Connection]()).incl(conn)
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
libp2p_peers.set(c.conns.len.int64)
# Launch on close listener
# All the errors are handled inside `onClose()` procedure.
asyncSpawn c.onClose(conn)
trace "Stored connection",
conn, direction = $conn.dir, connections = c.conns.len
proc getIncomingSlot*(c: ConnManager): Future[ConnectionSlot] {.async.} =
await c.inSema.acquire()
return ConnectionSlot(connManager: c, direction: In)
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
2023-01-05 14:02:52 +00:00
proc getOutgoingSlot*(c: ConnManager, forceDial = false): ConnectionSlot {.raises: [Defect, TooManyConnectionsError].} =
2022-02-24 16:31:47 +00:00
if forceDial:
c.outSema.forceAcquire()
elif not c.outSema.tryAcquire():
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
trace "Too many outgoing connections!", count = c.outSema.count,
max = c.outSema.size
raise newTooManyConnectionsError()
return ConnectionSlot(connManager: c, direction: Out)
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
proc release*(cs: ConnectionSlot) =
if cs.direction == In:
cs.connManager.inSema.release()
else:
cs.connManager.outSema.release()
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
proc trackConnection*(cs: ConnectionSlot, conn: Connection) =
if isNil(conn):
cs.release()
return
proc semaphoreMonitor() {.async.} =
try:
await conn.join()
except CatchableError as exc:
trace "Exception in semaphore monitor, ignoring", exc = exc.msg
cs.release()
asyncSpawn semaphoreMonitor()
proc storeMuxer*(c: ConnManager,
muxer: Muxer,
handle: Future[void] = nil)
{.raises: [Defect, CatchableError].} =
## store the connection and muxer
##
if isNil(muxer):
raise newException(CatchableError, "muxer cannot be nil")
if isNil(muxer.connection):
raise newException(CatchableError, "muxer's connection cannot be nil")
if muxer.connection notin c:
raise newException(CatchableError, "cant add muxer for untracked connection")
c.muxed[muxer.connection] = MuxerHolder(
muxer: muxer,
handle: handle)
trace "Stored muxer",
muxer, handle = not handle.isNil, connections = c.conns.len
asyncSpawn c.onConnUpgraded(muxer.connection)
proc getStream*(c: ConnManager,
2021-12-16 10:05:20 +00:00
peerId: PeerId,
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
dir: Direction): Future[Connection] {.async, gcsafe.} =
## get a muxed stream for the provided peer
## with the given direction
##
let muxer = c.selectMuxer(c.selectConn(peerId, dir))
if not(isNil(muxer)):
return await muxer.newStream()
proc getStream*(c: ConnManager,
2021-12-16 10:05:20 +00:00
peerId: PeerId): Future[Connection] {.async, gcsafe.} =
## get a muxed stream for the passed peer from any connection
##
let muxer = c.selectMuxer(c.selectConn(peerId))
if not(isNil(muxer)):
return await muxer.newStream()
proc getStream*(c: ConnManager,
Connection limits (#384) * master merge * wip * avoid deadlocks * tcp limits * expose client field in chronosstream * limit incoming connections * update with new listen api * fix release * don't override peerinfo in connection * rework transport with accept * use semaphore to track resource ussage * rework with new transport accept api * move events to conn manager (#373) * use semaphore to track resource ussage * merge master * expose api to acquire conn slots * don't fail expensive metrics * allow tracking and updating connections * set global connection limits to 80 * add per peer connection limits * make sure conn is closed if tracking failed * more descriptive naming for handle * rework with new transport accept api * add `getStream` hide `selectConn` * add TransportClosedError * make nil explicit * don't make unnecessary copies of message * logging * error handling * cleanup semaphore * track connections properly * throw `TooManyConnections` when tracking outgoing * use proper exception and handle conventions * check onCloseHandle for nil * revert internalConnect changes * adding upgraded flag * await stream before closing * simplify tracking * wip * logging * split connection limits into incoming and outgoing * further streamline connection limits split counts * don't use closeWithEOF * move peer and conn event triggers from switch * wip * wip * wip * merge master * handle nil connections properly * add clarifying comment * don't raise exc on nil * no finally * add proper min/max connections logic * rebase master * merge master * master merge * remove request timeout should be addressed in separate PR * merge master * share semaphore when in/out limits arent enforced * merge master * use import * pass semaphore to trackConn * don't close last conn * use storeConn * merge master * use storeConn
2021-01-21 04:00:24 +00:00
conn: Connection): Future[Connection] {.async, gcsafe.} =
## get a muxed stream for the passed connection
##
let muxer = c.selectMuxer(conn)
if not(isNil(muxer)):
return await muxer.newStream()
2021-12-16 10:05:20 +00:00
proc dropPeer*(c: ConnManager, peerId: PeerId) {.async.} =
## drop connections and cleanup resources for peer
##
trace "Dropping peer", peerId
let conns = c.conns.getOrDefault(peerId)
for conn in conns:
trace "Removing connection", conn
delConn(c, conn)
var muxers: seq[MuxerHolder]
for conn in conns:
if conn in c.muxed:
muxers.add c.muxed[conn]
c.muxed.del(conn)
for muxer in muxers:
await closeMuxerHolder(muxer)
for conn in conns:
await conn.close()
trace "Dropped peer", peerId
trace "Peer dropped", peerId
proc close*(c: ConnManager) {.async.} =
## cleanup resources for the connection
## manager
##
trace "Closing ConnManager"
let conns = c.conns
c.conns.clear()
let muxed = c.muxed
c.muxed.clear()
let expected = c.expectedConnections
c.expectedConnections.clear()
for _, fut in expected:
await fut.cancelAndWait()
for _, muxer in muxed:
await closeMuxerHolder(muxer)
for _, conns2 in conns:
for conn in conns2:
await conn.close()
trace "Closed ConnManager"