Merge branch 'master' into penalty-non-priority-queue

Co-authored-by: diegomrsantos <diego@status.im>

.github/workflows/bumper.yml

@@ -2,8 +2,7 @@ name: Bumper
 on:
   push:
     branches:
-      - unstable
-      - bumper
+      - master
   workflow_dispatch:
 
 jobs:

.github/workflows/daily.yml

@@ -0,0 +1,12 @@
+name: Daily
+on:
+  schedule:
+    - cron: "30 6 * * *"
+  workflow_dispatch:
+
+jobs:
+  call-multi-nim-common:
+    uses: ./.github/workflows/daily_common.yml
+    with:
+      nim-branch: "['version-1-6','version-2-0']"
+      cpu: "['amd64']"

.github/workflows/daily_common.yml

@@ -0,0 +1,84 @@
+name: daily-common
+
+on:
+  workflow_call:
+    inputs:
+      nim-branch:
+        description: 'Nim branch'
+        required: true
+        type: string
+      cpu:
+        description: 'CPU'
+        required: true
+        type: string
+      exclude:
+        description: 'Exclude matrix configurations'
+        required: false
+        type: string
+        default: "[]"
+
+jobs:
+  delete-cache:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: snnaplab/delete-branch-cache-action@v1
+
+  build:
+    needs: delete-cache
+    timeout-minutes: 120
+    strategy:
+      fail-fast: false
+      matrix:
+        platform:
+          - os: linux
+            builder: ubuntu-20
+            shell: bash
+          - os: macos
+            builder: macos-12
+            shell: bash
+          - os: windows
+            builder: windows-2019
+            shell: msys2 {0}
+        branch: ${{ fromJSON(inputs.nim-branch) }}
+        cpu: ${{ fromJSON(inputs.cpu) }}
+        exclude: ${{ fromJSON(inputs.exclude) }}
+
+    defaults:
+      run:
+        shell: ${{ matrix.platform.shell }}
+
+    name: '${{ matrix.platform.os }}-${{ matrix.cpu }} (Nim ${{ matrix.branch }})'
+    runs-on: ${{ matrix.platform.builder }}
+    continue-on-error: ${{ matrix.branch == 'devel' || matrix.branch == 'version-2-0' }}
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Nim
+        uses: "./.github/actions/install_nim"
+        with:
+          os: ${{ matrix.platform.os }}
+          shell: ${{ matrix.platform.shell }}
+          nim_branch: ${{ matrix.branch }}
+          cpu: ${{ matrix.cpu }}
+
+      - name: Setup Go
+        uses: actions/setup-go@v4
+        with:
+          go-version: '~1.15.5'
+          cache: false
+
+      - name: Install p2pd
+        run: |
+          V=1 bash scripts/build_p2pd.sh p2pdCache 124530a3
+
+      - name: Run tests
+        run: |
+          nim --version
+          nimble --version
+          nimble install -y --depsOnly
+          NIMFLAGS="${NIMFLAGS} --mm:refc" nimble test
+          if [[ "${{ matrix.branch }}" == "devel" ]]; then
+            echo -e "\nTesting with '--mm:orc':\n"
+            NIMFLAGS="${NIMFLAGS} --mm:orc" nimble test
+          fi

.github/workflows/daily_i386.yml

@@ -0,0 +1,13 @@
+name: Daily i386
+on:
+  schedule:
+    - cron: "30 6 * * *"
+  workflow_dispatch:
+
+jobs:
+  call-multi-nim-common:
+    uses: ./.github/workflows/daily_common.yml
+    with:
+      nim-branch: "['version-1-6','version-2-0', 'devel']"
+      cpu: "['i386']"
+      exclude: "[{'platform': {'os':'macos'}}, {'platform': {'os':'windows'}}]"

.github/workflows/daily_nim_devel.yml

@@ -0,0 +1,12 @@
+name: Daily Nim Devel
+on:
+  schedule:
+    - cron: "30 6 * * *"
+  workflow_dispatch:
+
+jobs:
+  call-multi-nim-common:
+    uses: ./.github/workflows/daily_common.yml
+    with:
+      nim-branch: "['devel']"
+      cpu: "['amd64']"

.github/workflows/doc.yml

@@ -19,13 +19,13 @@ jobs:
 
       - uses: jiro4989/setup-nim-action@v1
         with:
-          nim-version: 'stable'
+          nim-version: '1.6.x'
 
       - name: Generate doc
         run: |
           nim --version
           nimble --version
-          nimble install_pinned -y
+          nimble install_pinned
           # nim doc can "fail", but the doc is still generated
           nim doc --git.url:https://github.com/status-im/nim-libp2p --git.commit:${GITHUB_REF##*/} --outdir:${GITHUB_REF##*/} --project libp2p || true
 

.github/workflows/interop.yml

@@ -11,44 +11,30 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  run-multidim-interop:
-    name: Run multidimensional interoperability tests
+  run-transport-interop:
+    name: Run transport interoperability tests
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
-        with:
-          repository: libp2p/test-plans
-          submodules: true
-          fetch-depth: 0
-
+      - uses: actions/checkout@v4
+      - uses: docker/setup-buildx-action@v3
       - name: Build image
-        run: >
-          cd transport-interop/impl/nim/v1.0 &&
-          make commitSha=$GITHUB_SHA image_name=nim-libp2p-head
-
-      - name: Create ping-version.json
-        run: >
-          (cat << EOF
-            {
-                "id": "nim-libp2p-head",
-                "containerImageID": "nim-libp2p-head",
-                "transports": [
-                    "tcp",
-                    "ws"
-                ],
-                "secureChannels": [
-                    "noise"
-                ],
-                "muxers": [
-                    "mplex",
-                    "yamux"
-                ]
-            }
-          EOF
-
-          ) > ${{ github.workspace }}/test_head.json
-
-      - uses: libp2p/test-plans/.github/actions/run-transport-interop-test@master
+        run: docker buildx build --load -t nim-libp2p-head -f tests/transport-interop/Dockerfile .
+      - name: Run tests
+        uses: libp2p/test-plans/.github/actions/run-transport-interop-test@master
         with:
           test-filter: nim-libp2p-head
-          extra-versions: ${{ github.workspace }}/test_head.json
+          extra-versions: ${{ github.workspace }}/tests/transport-interop/version.json
+
+  run-hole-punching-interop:
+    name: Run hole-punching interoperability tests
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v4
+      - uses: docker/setup-buildx-action@v3
+      - name: Build image
+        run: docker buildx build --load -t nim-libp2p-head -f tests/hole-punching-interop/Dockerfile .
+      - name: Run tests
+        uses: libp2p/test-plans/.github/actions/run-interop-hole-punch-test@master
+        with:
+          test-filter: nim-libp2p-head
+          extra-versions: ${{ github.workspace }}/tests/hole-punching-interop/version.json

.github/workflows/multi_nim.yml

@@ -1,82 +0,0 @@
-name: Daily
-on:
-  schedule:
-    - cron: "30 6 * * *"
-  workflow_dispatch:
-
-jobs:
-  delete-cache:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: snnaplab/delete-branch-cache-action@v1
-
-  build:
-    needs: delete-cache
-    timeout-minutes: 120
-    strategy:
-      fail-fast: false
-      matrix:
-        target:
-          - os: linux
-            cpu: amd64
-          - os: linux
-            cpu: i386
-          - os: macos
-            cpu: amd64
-          - os: windows
-            cpu: amd64
-          #- os: windows
-            #cpu: i386
-        branch: [version-1-6, version-2-0, devel]
-        include:
-          - target:
-              os: linux
-            builder: ubuntu-20.04
-            shell: bash
-          - target:
-              os: macos
-            builder: macos-12
-            shell: bash
-          - target:
-              os: windows
-            builder: windows-2019
-            shell: msys2 {0}
-
-    defaults:
-      run:
-        shell: ${{ matrix.shell }}
-
-    name: '${{ matrix.target.os }}-${{ matrix.target.cpu }} (Nim ${{ matrix.branch }})'
-    runs-on: ${{ matrix.builder }}
-    continue-on-error: ${{ matrix.branch == 'devel' || matrix.branch == 'version-2-0' }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v2
-
-      - name: Setup Nim
-        uses: "./.github/actions/install_nim"
-        with:
-          os: ${{ matrix.target.os }}
-          shell: ${{ matrix.shell }}
-          nim_branch: ${{ matrix.branch }}
-          cpu: ${{ matrix.target.cpu }}
-
-      - name: Setup Go
-        uses: actions/setup-go@v2
-        with:
-          go-version: '~1.15.5'
-
-      - name: Install p2pd
-        run: |
-          V=1 bash scripts/build_p2pd.sh p2pdCache 124530a3
-
-      - name: Run tests
-        run: |
-          nim --version
-          nimble --version
-          nimble install -y --depsOnly
-          NIMFLAGS="${NIMFLAGS} --gc:refc" nimble test
-          if [[ "${{ matrix.branch }}" == "devel" ]]; then
-            echo -e "\nTesting with '--gc:orc':\n"
-            NIMFLAGS="${NIMFLAGS} --gc:orc" nimble test
-          fi

.gitignore

@@ -16,3 +16,4 @@ tests/pubsub/testgossipsub
 examples/*.md
 nimble.develop
 nimble.paths
+go-libp2p-daemon/

.pinned

@@ -1,6 +1,6 @@
 bearssl;https://github.com/status-im/nim-bearssl@#e4157639db180e52727712a47deaefcbbac6ec86
 chronicles;https://github.com/status-im/nim-chronicles@#32ac8679680ea699f7dbc046e8e0131cac97d41a
-chronos;https://github.com/status-im/nim-chronos@#ba143e029f35fd9b4cd3d89d007cc834d0d5ba3c
+chronos;https://github.com/status-im/nim-chronos@#672db137b7cad9b384b8f4fb551fb6bbeaabfe1b
 dnsclient;https://github.com/ba0f3/dnsclient.nim@#23214235d4784d24aceed99bbfe153379ea557c8
 faststreams;https://github.com/status-im/nim-faststreams@#720fc5e5c8e428d9d0af618e1e27c44b42350309
 httputils;https://github.com/status-im/nim-http-utils@#3b491a40c60aad9e8d3407443f46f62511e63b18

README.md

@@ -20,6 +20,7 @@
 - [Background](#background)
 - [Install](#install)
 - [Getting Started](#getting-started)
+  - [Go-libp2p-daemon](#go-libp2p-daemon)
 - [Modules](#modules)
 - [Users](#users)
 - [Stability](#stability)
@@ -40,6 +41,8 @@ Learn more about libp2p at [**libp2p.io**](https://libp2p.io) and follow libp2p'
 ## Install
 **Prerequisite**
 - [Nim](https://nim-lang.org/install.html)
+> The currently supported Nim version is 1.6.18.
+
 ```
 nimble install libp2p
 ```
@@ -47,11 +50,11 @@ nimble install libp2p
 ## Getting Started
 You'll find the nim-libp2p documentation [here](https://status-im.github.io/nim-libp2p/docs/).
 
-**Go Daemon:**
-Please find the installation and usage intructions in [daemonapi.md](examples/go-daemon/daemonapi.md).
+### Testing
+Remember you'll need to build the `go-libp2p-daemon` binary to run the `nim-libp2p` tests.
+To do so, please follow the installation instructions in [daemonapi.md](examples/go-daemon/daemonapi.md).
 
 ## Modules
-
 List of packages modules implemented in nim-libp2p:
 
 | Name                                                       | Description                                                                                                      |
@@ -68,7 +71,6 @@ List of packages modules implemented in nim-libp2p:
 | [libp2p-ws](libp2p/transports/wstransport.nim)             | WebSocket & WebSocket Secure transport                                                                           |
 | [libp2p-tor](libp2p/transports/tortransport.nim)           | Tor Transport                                                                                                    |
 | **Secure Channels**                                        |                                                                                                                  |
-| [libp2p-secio](libp2p/protocols/secure/secio.nim)          | Secio secure channel                                                                                             |
 | [libp2p-noise](libp2p/protocols/secure/noise.nim)          | [Noise](https://docs.libp2p.io/concepts/secure-comm/noise/) secure channel                                       |
 | [libp2p-plaintext](libp2p/protocols/secure/plaintext.nim)  | Plain Text for development purposes                                                                              |
 | **Stream Multiplexers**                                    |                                                                                                                  |
@@ -132,7 +134,8 @@ The libp2p implementation in Nim is a work in progress. We welcome contributors
 - Go through the modules and **check out existing issues**. This would be especially useful for modules in active development. Some knowledge of IPFS/libp2p may be required, as well as the infrastructure behind it.
 - **Perform code reviews**. Feel free to let us know if you found anything that can a) speed up the project development b) ensure better quality and c) reduce possible future bugs.
 - **Add tests**. Help nim-libp2p to be more robust by adding more tests to the [tests folder](tests/).
-
+- **Small PRs**. Try to keep PRs atomic and digestible. This makes the review process and pinpointing bugs easier.
+- **Code format**. Please format code using [nph](https://github.com/arnetheduck/nph).
 The code follows the [Status Nim Style Guide](https://status-im.github.io/nim-style-guide/).
 
 ### Contributors

config.nims

@@ -7,7 +7,6 @@ if dirExists("nimbledeps/pkgs2"):
 switch("warning", "CaseTransition:off")
 switch("warning", "ObservableStores:off")
 switch("warning", "LockLevel:off")
---define:chronosStrictException
 --styleCheck:usages
 switch("warningAsError", "UseBase:on")
 --styleCheck:error

examples/go-daemon/daemonapi.md

@@ -1,6 +1,8 @@
 # Table of Contents
 - [Introduction](#introduction)
+- [Prerequisites](#prerequisites)
 - [Installation](#installation)
+  - [Script](#script)
 - [Usage](#usage)
   - [Example](#example)
   - [Getting Started](#getting-started)
@@ -8,26 +10,29 @@
 # Introduction
 This is a libp2p-backed daemon wrapping the functionalities of go-libp2p for use in Nim. <br>
 For more information about the go daemon, check out [this repository](https://github.com/libp2p/go-libp2p-daemon).
+> **Required only** for running the tests.
+
+# Prerequisites
+Go with version `1.15.15`.
+> You will *likely* be able to build `go-libp2p-daemon` with different Go versions, but **they haven't been tested**.
 
 # Installation
+Follow one of the methods below:
+
+## Script
+Run the build script while having the `go` command pointing to the correct Go version.
+We recommend using `1.15.15`, as previously stated.
 ```sh
-# clone and install dependencies
-git clone https://github.com/status-im/nim-libp2p
-cd nim-libp2p
-nimble install
-
-# perform unit tests
-nimble test
-
-# update the git submodule to install the go daemon 
-git submodule update --init --recursive
-go version
-git clone https://github.com/libp2p/go-libp2p-daemon
-cd go-libp2p-daemon
-git checkout v0.0.1
-go install ./...
-cd ..
+./scripts/build_p2pd.sh
 ```
+If everything goes correctly, the binary (`p2pd`) should be built and placed in the correct directory.
+If you find any issues, please head into our discord and ask for our asistance.
+
+After successfully building the binary, remember to add it to your path so it can be found. You can do that by running:
+```sh
+export PATH="$PATH:$HOME/go/bin"
+```
+> **Tip:** To make this change permanent, add the command above to your `.bashrc` file.
 
 # Usage 
 

examples/helloworld.nim

@@ -13,14 +13,14 @@ type
 proc new(T: typedesc[TestProto]): T =
 
   # every incoming connections will be in handled in this closure
-  proc handle(conn: Connection, proto: string) {.async, gcsafe.} =
+  proc handle(conn: Connection, proto: string) {.async.} =
     echo "Got from remote - ", string.fromBytes(await conn.readLp(1024))
     await conn.writeLp("Roger p2p!")
 
     # We must close the connections ourselves when we're done with it
     await conn.close()
 
-  return T(codecs: @[TestCodec], handler: handle)
+  return T.new(codecs = @[TestCodec], handler = handle)
 
 ##
 # Helper to create a switch/node
@@ -40,7 +40,7 @@ proc createSwitch(ma: MultiAddress, rng: ref HmacDrbgContext): Switch =
 ##
 # The actual application
 ##
-proc main() {.async, gcsafe.} =
+proc main() {.async.} =
   let
     rng = newRng() # Single random number source for the whole application
     # port 0 will take a random available port

examples/tutorial_1_connect.nim

@@ -53,7 +53,7 @@ proc createSwitch(ma: MultiAddress, rng: ref HmacDrbgContext): Switch =
 ##
 ##
 ## Let's now start to create our main procedure:
-proc main() {.async, gcsafe.} =
+proc main() {.async.} =
   let
     rng = newRng()
     localAddress = MultiAddress.init("/ip4/0.0.0.0/tcp/0").tryGet()

examples/tutorial_2_customproto.nim

@@ -25,7 +25,7 @@ type TestProto = ref object of LPProtocol
 
 proc new(T: typedesc[TestProto]): T =
   # every incoming connections will in be handled in this closure
-  proc handle(conn: Connection, proto: string) {.async, gcsafe.} =
+  proc handle(conn: Connection, proto: string) {.async.} =
     # Read up to 1024 bytes from this connection, and transform them into
     # a string
     echo "Got from remote - ", string.fromBytes(await conn.readLp(1024))
@@ -44,7 +44,7 @@ proc hello(p: TestProto, conn: Connection) {.async.} =
 ## Again, pretty straight-forward, we just send a message on the connection.
 ##
 ## We can now create our main procedure:
-proc main() {.async, gcsafe.} =
+proc main() {.async.} =
   let
     rng = newRng()
     testProto = TestProto.new()

examples/tutorial_3_protobuf.nim

@@ -108,7 +108,7 @@ type
 
 proc new(_: typedesc[MetricProto], cb: MetricCallback): MetricProto =
   var res: MetricProto
-  proc handle(conn: Connection, proto: string) {.async, gcsafe.} =
+  proc handle(conn: Connection, proto: string) {.async.} =
     let
       metrics = await res.metricGetter()
       asProtobuf = metrics.encode()
@@ -126,7 +126,7 @@ proc fetch(p: MetricProto, conn: Connection): Future[MetricList] {.async.} =
   return MetricList.decode(protobuf).tryGet()
 
 ## We can now create our main procedure:
-proc main() {.async, gcsafe.} =
+proc main() {.async.} =
   let rng = newRng()
   proc randomMetricGenerator: Future[MetricList] {.async.} =
     let metricCount = rng[].generate(uint32) mod 16

examples/tutorial_5_discovery.nim

@@ -33,7 +33,7 @@ proc createSwitch(rdv: RendezVous = RendezVous.new()): Switch =
 const DumbCodec = "/dumb/proto/1.0.0"
 type DumbProto = ref object of LPProtocol
 proc new(T: typedesc[DumbProto], nodeNumber: int): T =
-  proc handle(conn: Connection, proto: string) {.async, gcsafe.} =
+  proc handle(conn: Connection, proto: string) {.async.} =
     echo "Node", nodeNumber, " received: ", string.fromBytes(await conn.readLp(1024))
     await conn.close()
   return T.new(codecs = @[DumbCodec], handler = handle)
@@ -49,7 +49,7 @@ proc new(T: typedesc[DumbProto], nodeNumber: int): T =
 ## (rendezvous in this case) as a bootnode. For this example, we'll
 ## create a bootnode, and then every peer will advertise itself on the
 ## bootnode, and use it to find other peers
-proc main() {.async, gcsafe.} =
+proc main() {.async.} =
   let bootNode = createSwitch()
   await bootNode.start()
 

examples/tutorial_6_game.nim

@@ -143,7 +143,7 @@ proc draw(g: Game) =
 ## peer know that we are available, check that he is also available,
 ## and launch the game.
 proc new(T: typedesc[GameProto], g: Game): T =
-  proc handle(conn: Connection, proto: string) {.async, gcsafe.} =
+  proc handle(conn: Connection, proto: string) {.async.} =
     defer: await conn.closeWithEof()
     if g.peerFound.finished or g.hasCandidate:
       await conn.close()

libp2p.nimble

@@ -12,12 +12,12 @@ requires "nim >= 1.6.0",
          "dnsclient >= 0.3.0 & < 0.4.0",
          "bearssl >= 0.1.4",
          "chronicles >= 0.10.2",
-         "chronos >= 3.0.6",
+         "chronos >= 4.0.0",
          "metrics",
          "secp256k1",
          "stew#head",
          "websock",
-         "unittest2 >= 0.0.5 & <= 0.1.0"
+         "unittest2"
 
 let nimc = getEnv("NIMC", "nim") # Which nim compiler to use
 let lang = getEnv("NIMLANG", "c") # Which backend (c/cpp/js)

libp2p/builders.nim

@@ -25,7 +25,7 @@ import
   muxers/[muxer, mplex/mplex, yamux/yamux],
   protocols/[identify, secure/secure, secure/noise, rendezvous],
   protocols/connectivity/[autonat/server, relay/relay, relay/client, relay/rtransport],
-  connmanager, upgrademngrs/muxedupgrade,
+  connmanager, upgrademngrs/muxedupgrade, observedaddrmanager,
   nameresolving/nameresolver,
   errors, utility
 
@@ -36,8 +36,7 @@ type
   TransportProvider* {.public.} = proc(upgr: Upgrade): Transport {.gcsafe, raises: [].}
 
   SecureProtocol* {.pure.} = enum
-    Noise,
-    Secio {.deprecated.}
+    Noise
 
   SwitchBuilder* = ref object
     privKey: Option[PrivateKey]
@@ -59,6 +58,7 @@ type
     circuitRelay: Relay
     rdv: RendezVous
     services: seq[Service]
+    observedAddrManager: ObservedAddrManager
 
 proc new*(T: type[SwitchBuilder]): T {.public.} =
   ## Creates a SwitchBuilder
@@ -121,8 +121,12 @@ proc withMplex*(
   b.muxers.add(MuxerProvider.new(newMuxer, MplexCodec))
   b
 
-proc withYamux*(b: SwitchBuilder): SwitchBuilder =
-  proc newMuxer(conn: Connection): Muxer = Yamux.new(conn)
+proc withYamux*(b: SwitchBuilder,
+    windowSize: int = YamuxDefaultWindowSize,
+    inTimeout: Duration = 5.minutes,
+    outTimeout: Duration = 5.minutes): SwitchBuilder =
+  proc newMuxer(conn: Connection): Muxer =
+    Yamux.new(conn, windowSize, inTimeout = inTimeout, outTimeout = outTimeout)
 
   assert b.muxers.countIt(it.codec == YamuxCodec) == 0, "Yamux build multiple times"
   b.muxers.add(MuxerProvider.new(newMuxer, YamuxCodec))
@@ -201,6 +205,10 @@ proc withServices*(b: SwitchBuilder, services: seq[Service]): SwitchBuilder =
   b.services = services
   b
 
+proc withObservedAddrManager*(b: SwitchBuilder, observedAddrManager: ObservedAddrManager): SwitchBuilder =
+  b.observedAddrManager = observedAddrManager
+  b
+
 proc build*(b: SwitchBuilder): Switch
   {.raises: [LPError], public.} =
 
@@ -223,8 +231,13 @@ proc build*(b: SwitchBuilder): Switch
       protoVersion = b.protoVersion,
       agentVersion = b.agentVersion)
 
+  let identify =
+    if b.observedAddrManager != nil:
+      Identify.new(peerInfo, b.sendSignedPeerRecord, b.observedAddrManager)
+    else:
+      Identify.new(peerInfo, b.sendSignedPeerRecord)
+
   let
-    identify = Identify.new(peerInfo, b.sendSignedPeerRecord)
     connManager = ConnManager.new(b.maxConnsPerPeer, b.maxConnections, b.maxIn, b.maxOut)
     ms = MultistreamSelect.new()
     muxedUpgrade = MuxedUpgrade.new(b.muxers, secureManagerInstances, ms)
@@ -277,29 +290,25 @@ proc build*(b: SwitchBuilder): Switch
   return switch
 
 proc newStandardSwitch*(
-  privKey = none(PrivateKey),
-  addrs: MultiAddress | seq[MultiAddress] = MultiAddress.init("/ip4/127.0.0.1/tcp/0").tryGet(),
-  secureManagers: openArray[SecureProtocol] = [
+    privKey = none(PrivateKey),
+    addrs: MultiAddress | seq[MultiAddress] =
+      MultiAddress.init("/ip4/127.0.0.1/tcp/0").expect("valid address"),
+    secureManagers: openArray[SecureProtocol] = [
       SecureProtocol.Noise,
     ],
-  transportFlags: set[ServerFlags] = {},
-  rng = newRng(),
-  inTimeout: Duration = 5.minutes,
-  outTimeout: Duration = 5.minutes,
-  maxConnections = MaxConnections,
-  maxIn = -1,
-  maxOut = -1,
-  maxConnsPerPeer = MaxConnectionsPerPeer,
-  nameResolver: NameResolver = nil,
-  sendSignedPeerRecord = false,
-  peerStoreCapacity = 1000): Switch
-  {.raises: [LPError], public.} =
+    transportFlags: set[ServerFlags] = {},
+    rng = newRng(),
+    inTimeout: Duration = 5.minutes,
+    outTimeout: Duration = 5.minutes,
+    maxConnections = MaxConnections,
+    maxIn = -1,
+    maxOut = -1,
+    maxConnsPerPeer = MaxConnectionsPerPeer,
+    nameResolver: NameResolver = nil,
+    sendSignedPeerRecord = false,
+    peerStoreCapacity = 1000
+): Switch {.raises: [LPError], public.} =
   ## Helper for common switch configurations.
-  {.push warning[Deprecated]:off.}
-  if SecureProtocol.Secio in secureManagers:
-      quit("Secio is deprecated!") # use of secio is unsafe
-  {.pop.}
-
   let addrs = when addrs is MultiAddress: @[addrs] else: addrs
   var b = SwitchBuilder
     .new()

libp2p/cid.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -261,12 +261,6 @@ proc write*(vb: var VBuffer, cid: Cid) {.inline.} =
   ## Write CID value ``cid`` to buffer ``vb``.
   vb.writeArray(cid.data.buffer)
 
-proc encode*(mbtype: typedesc[MultiBase], encoding: string,
-             cid: Cid): string {.inline.} =
-  ## Get MultiBase encoded representation of ``cid`` using encoding
-  ## ``encoding``.
-  result = MultiBase.encode(encoding, cid.data.buffer).tryGet()
-
 proc hash*(cid: Cid): Hash {.inline.} =
   hash(cid.data.buffer)
 

libp2p/connmanager.nim

@@ -128,7 +128,7 @@ proc removeConnEventHandler*(c: ConnManager,
 
 proc triggerConnEvent*(c: ConnManager,
                        peerId: PeerId,
-                       event: ConnEvent) {.async, gcsafe.} =
+                       event: ConnEvent) {.async.} =
   try:
     trace "About to trigger connection events", peer = peerId
     if c.connEvents[event.kind].len() > 0:
@@ -160,7 +160,7 @@ proc removePeerEventHandler*(c: ConnManager,
 
 proc triggerPeerEvents*(c: ConnManager,
                         peerId: PeerId,
-                        event: PeerEvent) {.async, gcsafe.} =
+                        event: PeerEvent) {.async.} =
 
   trace "About to trigger peer events", peer = peerId
   if c.peerEvents[event.kind].len == 0:
@@ -311,12 +311,14 @@ proc storeMuxer*(c: ConnManager,
 
       raise newTooManyConnectionsError()
 
-  assert muxer notin c.muxed.getOrDefault(peerId)
-
-  let
-    newPeer = peerId notin c.muxed
-  assert newPeer or c.muxed[peerId].len > 0
-  c.muxed.mgetOrPut(peerId, newSeq[Muxer]()).add(muxer)
+  var newPeer = false
+  c.muxed.withValue(peerId, muxers):
+    doAssert muxers[].len > 0
+    doAssert muxer notin muxers[]
+    muxers[].add(muxer)
+  do:
+    c.muxed[peerId] = @[muxer]
+    newPeer = true
   libp2p_peers.set(c.muxed.len.int64)
 
   asyncSpawn c.triggerConnEvent(
@@ -379,7 +381,7 @@ proc trackMuxer*(cs: ConnectionSlot, mux: Muxer) =
   cs.trackConnection(mux.connection)
 
 proc getStream*(c: ConnManager,
-                muxer: Muxer): Future[Connection] {.async, gcsafe.} =
+                muxer: Muxer): Future[Connection] {.async.} =
   ## get a muxed stream for the passed muxer
   ##
 
@@ -387,7 +389,7 @@ proc getStream*(c: ConnManager,
     return await muxer.newStream()
 
 proc getStream*(c: ConnManager,
-                peerId: PeerId): Future[Connection] {.async, gcsafe.} =
+                peerId: PeerId): Future[Connection] {.async.} =
   ## get a muxed stream for the passed peer from any connection
   ##
 
@@ -395,7 +397,7 @@ proc getStream*(c: ConnManager,
 
 proc getStream*(c: ConnManager,
                 peerId: PeerId,
-                dir: Direction): Future[Connection] {.async, gcsafe.} =
+                dir: Direction): Future[Connection] {.async.} =
   ## get a muxed stream for the passed peer from a connection with `dir`
   ##
 

libp2p/crypto/crypto.nim

@@ -924,59 +924,6 @@ proc selectBest*(order: int, p1, p2: string): string =
       if felement == selement:
         return felement
 
-proc createProposal*(nonce, pubkey: openArray[byte],
-                     exchanges, ciphers, hashes: string): seq[byte] =
-  ## Create SecIO proposal message using random ``nonce``, local public key
-  ## ``pubkey``, comma-delimieted list of supported exchange schemes
-  ## ``exchanges``, comma-delimeted list of supported ciphers ``ciphers`` and
-  ## comma-delimeted list of supported hashes ``hashes``.
-  var msg = initProtoBuffer({WithUint32BeLength})
-  msg.write(1, nonce)
-  msg.write(2, pubkey)
-  msg.write(3, exchanges)
-  msg.write(4, ciphers)
-  msg.write(5, hashes)
-  msg.finish()
-  msg.buffer
-
-proc decodeProposal*(message: seq[byte], nonce, pubkey: var seq[byte],
-                     exchanges, ciphers, hashes: var string): bool =
-  ## Parse incoming proposal message and decode remote random nonce ``nonce``,
-  ## remote public key ``pubkey``, comma-delimieted list of supported exchange
-  ## schemes ``exchanges``, comma-delimeted list of supported ciphers
-  ## ``ciphers`` and comma-delimeted list of supported hashes ``hashes``.
-  ##
-  ## Procedure returns ``true`` on success and ``false`` on error.
-  var pb = initProtoBuffer(message)
-  let r1 = pb.getField(1, nonce)
-  let r2 = pb.getField(2, pubkey)
-  let r3 = pb.getField(3, exchanges)
-  let r4 = pb.getField(4, ciphers)
-  let r5 = pb.getField(5, hashes)
-
-  r1.get(false) and r2.get(false) and r3.get(false) and
-  r4.get(false) and r5.get(false)
-
-proc createExchange*(epubkey, signature: openArray[byte]): seq[byte] =
-  ## Create SecIO exchange message using ephemeral public key ``epubkey`` and
-  ## signature of proposal blocks ``signature``.
-  var msg = initProtoBuffer({WithUint32BeLength})
-  msg.write(1, epubkey)
-  msg.write(2, signature)
-  msg.finish()
-  msg.buffer
-
-proc decodeExchange*(message: seq[byte],
-                     pubkey, signature: var seq[byte]): bool =
-  ## Parse incoming exchange message and decode remote ephemeral public key
-  ## ``pubkey`` and signature ``signature``.
-  ##
-  ## Procedure returns ``true`` on success and ``false`` on error.
-  var pb = initProtoBuffer(message)
-  let r1 = pb.getField(1, pubkey)
-  let r2 = pb.getField(2, signature)
-  r1.get(false) and r2.get(false)
-
 ## Serialization/Deserialization helpers
 
 proc write*(vb: var VBuffer, pubkey: PublicKey) {.

libp2p/daemon/daemonapi.nim

@@ -553,7 +553,7 @@ proc getSocket(pattern: string,
     closeSocket(sock)
 
 # This is forward declaration needed for newDaemonApi()
-proc listPeers*(api: DaemonAPI): Future[seq[PeerInfo]] {.async, gcsafe.}
+proc listPeers*(api: DaemonAPI): Future[seq[PeerInfo]] {.async.}
 
 proc copyEnv(): StringTableRef =
   ## This procedure copy all environment variables into StringTable.
@@ -755,7 +755,7 @@ proc newDaemonApi*(flags: set[P2PDaemonFlags] = {},
 
   # Starting daemon process
   # echo "Starting ", cmd, " ", args.join(" ")
-  api.process = 
+  api.process =
     exceptionToAssert:
       startProcess(cmd, "", args, env, {poParentStreams})
   # Waiting until daemon will not be bound to control socket.
@@ -1032,7 +1032,7 @@ proc enterDhtMessage(pb: ProtoBuffer, rt: DHTResponseType): ProtoBuffer
     var value: seq[byte]
     if pbDhtResponse.getRequiredField(3, value).isErr():
       raise newException(DaemonLocalError, "Missing required DHT field `value`!")
-    
+
     return initProtoBuffer(value)
   else:
     raise newException(DaemonLocalError, "Wrong message type!")

libp2p/debugutils.nim

@@ -247,7 +247,7 @@ proc toString*(msg: ProtoMessage, dump = true): string =
         else: "[REMOTE]"
       local & direction & remote
   let seqid = block:
-    msg.seqID.wihValue(seqid): "seqID = " & $seqid & " "
+    msg.seqID.withValue(seqid): "seqID = " & $seqid & " "
     else: ""
   let mtype = block:
     msg.mtype.withValue(typ): "type = " & $typ & " "

libp2p/dial.nim

@@ -26,7 +26,7 @@ method connect*(
   addrs: seq[MultiAddress],
   forceDial = false,
   reuseConnection = true,
-  upgradeDir = Direction.Out) {.async, base.} =
+  dir = Direction.Out) {.async, base.} =
   ## connect remote peer without negotiating
   ## a protocol
   ##

libp2p/dialer.nim

@@ -53,7 +53,7 @@ proc dialAndUpgrade(
   peerId: Opt[PeerId],
   hostname: string,
   address: MultiAddress,
-  upgradeDir = Direction.Out):
+  dir = Direction.Out):
   Future[Muxer] {.async.} =
 
   for transport in self.transports: # for each transport
@@ -75,23 +75,29 @@ proc dialAndUpgrade(
 
       let mux =
         try:
-          dialed.transportDir = upgradeDir
-          await transport.upgrade(dialed, upgradeDir, peerId)
+          # This is for the very specific case of a simultaneous dial during DCUtR. In this case, both sides will have
+          # an Outbound direction at the transport level. Therefore we update the DCUtR initiator transport direction to Inbound.
+          # The if below is more general and might handle other use cases in the future.
+          if dialed.dir != dir:
+            dialed.dir = dir
+          await transport.upgrade(dialed, peerId)
+        except CancelledError as exc:
+          await dialed.close()
+          raise exc
         except CatchableError as exc:
           # If we failed to establish the connection through one transport,
           # we won't succeeded through another - no use in trying again
           await dialed.close()
-          debug "Upgrade failed", err = exc.msg, peerId = peerId.get(default(PeerId))
-          if exc isnot CancelledError:
-            if upgradeDir == Direction.Out:
-              libp2p_failed_upgrades_outgoing.inc()
-            else:
-              libp2p_failed_upgrades_incoming.inc()
+          debug "Connection upgrade failed", err = exc.msg, peerId = peerId.get(default(PeerId))
+          if dialed.dir == Direction.Out:
+            libp2p_failed_upgrades_outgoing.inc()
+          else:
+            libp2p_failed_upgrades_incoming.inc()
 
           # Try other address
           return nil
 
-      doAssert not isNil(mux), "connection died after upgrade " & $upgradeDir
+      doAssert not isNil(mux), "connection died after upgrade " & $dialed.dir
       debug "Dial successful", peerId = mux.connection.peerId
       return mux
   return nil
@@ -128,7 +134,7 @@ proc dialAndUpgrade(
   self: Dialer,
   peerId: Opt[PeerId],
   addrs: seq[MultiAddress],
-  upgradeDir = Direction.Out):
+  dir = Direction.Out):
   Future[Muxer] {.async.} =
 
   debug "Dialing peer", peerId = peerId.get(default(PeerId))
@@ -146,7 +152,7 @@ proc dialAndUpgrade(
           else: await self.nameResolver.resolveMAddress(expandedAddress)
 
       for resolvedAddress in resolvedAddresses:
-        result = await self.dialAndUpgrade(addrPeerId, hostname, resolvedAddress, upgradeDir)
+        result = await self.dialAndUpgrade(addrPeerId, hostname, resolvedAddress, dir)
         if not isNil(result):
           return result
 
@@ -164,7 +170,7 @@ proc internalConnect(
   addrs: seq[MultiAddress],
   forceDial: bool,
   reuseConnection = true,
-  upgradeDir = Direction.Out):
+  dir = Direction.Out):
   Future[Muxer] {.async.} =
   if Opt.some(self.localPeerId) == peerId:
     raise newException(CatchableError, "can't dial self!")
@@ -182,7 +188,7 @@ proc internalConnect(
     let slot = self.connManager.getOutgoingSlot(forceDial)
     let muxed =
       try:
-        await self.dialAndUpgrade(peerId, addrs, upgradeDir)
+        await self.dialAndUpgrade(peerId, addrs, dir)
       except CatchableError as exc:
         slot.release()
         raise exc
@@ -209,7 +215,7 @@ method connect*(
   addrs: seq[MultiAddress],
   forceDial = false,
   reuseConnection = true,
-  upgradeDir = Direction.Out) {.async.} =
+  dir = Direction.Out) {.async.} =
   ## connect remote peer without negotiating
   ## a protocol
   ##
@@ -217,7 +223,7 @@ method connect*(
   if self.connManager.connCount(peerId) > 0 and reuseConnection:
     return
 
-  discard await self.internalConnect(Opt.some(peerId), addrs, forceDial, reuseConnection, upgradeDir)
+  discard await self.internalConnect(Opt.some(peerId), addrs, forceDial, reuseConnection, dir)
 
 method connect*(
   self: Dialer,

libp2p/errors.nim

@@ -19,7 +19,8 @@ func toException*(e: string): ref LPError =
 # sadly nim needs more love for hygienic templates
 # so here goes the macro, its based on the proc/template version
 # and uses quote do so it's quite readable
-macro checkFutures*[T](futs: seq[Future[T]], exclude: untyped = []): untyped =
+# TODO https://github.com/nim-lang/Nim/issues/22936
+macro checkFutures*[F](futs: seq[F], exclude: untyped = []): untyped =
   let nexclude = exclude.len
   case nexclude
   of 0:
@@ -43,34 +44,3 @@ macro checkFutures*[T](futs: seq[Future[T]], exclude: untyped = []): untyped =
             # We still don't abort but warn
             debug "A future has failed, enable trace logging for details", error=exc.name
             trace "Exception details", msg=exc.msg
-
-proc allFuturesThrowing*[T](args: varargs[Future[T]]): Future[void] =
-  var futs: seq[Future[T]]
-  for fut in args:
-    futs &= fut
-  proc call() {.async.} =
-    var first: ref CatchableError = nil
-    futs = await allFinished(futs)
-    for fut in futs:
-      if fut.failed:
-        let err = fut.readError()
-        if err of Defect:
-          raise err
-        else:
-          if err of CancelledError:
-            raise err
-          if isNil(first):
-            first = err
-    if not isNil(first):
-      raise first
-
-  return call()
-
-template tryAndWarn*(message: static[string]; body: untyped): untyped =
-  try:
-    body
-  except CancelledError as exc:
-    raise exc
-  except CatchableError as exc:
-    debug "An exception has ocurred, enable trace logging for details", name = exc.name, msg = message
-    trace "Exception details", exc = exc.msg

libp2p/multiaddress.nim

@@ -13,8 +13,8 @@
 {.push public.}
 
 import pkg/chronos, chronicles
-import std/[nativesockets, hashes]
-import tables, strutils, sets, stew/shims/net
+import std/[nativesockets, net, hashes]
+import tables, strutils, sets
 import multicodec, multihash, multibase, transcoder, vbuffer, peerid,
        protobuf/minprotobuf, errors, utility
 import stew/[base58, base32, endians2, results]
@@ -119,23 +119,46 @@ proc ip6VB(vb: var VBuffer): bool =
   if vb.readArray(a.address_v6) == 16:
     result = true
 
-proc ip6zoneStB(s: string, vb: var VBuffer): bool =
-  ## IPv6 stringToBuffer() implementation.
+template pathStringToBuffer(s: string, vb: var VBuffer): bool =
   if len(s) > 0:
     vb.writeSeq(s)
-    result = true
+    true
+  else:
+    false
+
+template pathBufferToString(vb: var VBuffer, s: var string): bool =
+  s = ""
+  if (vb.readSeq(s) > 0) and (len(s) > 0):
+    true
+  else:
+    false
+
+template pathBufferToStringNoSlash(vb: var VBuffer, s: var string): bool =
+  s = ""
+  if (vb.readSeq(s) > 0) and (len(s) > 0) and (s.find('/') == -1):
+    true
+  else:
+    false
+
+template pathValidateBuffer(vb: var VBuffer): bool =
+  var s = ""
+  pathBufferToString(vb, s)
+
+template pathValidateBufferNoSlash(vb: var VBuffer): bool =
+  var s = ""
+  pathBufferToStringNoSlash(vb, s)
+
+proc ip6zoneStB(s: string, vb: var VBuffer): bool =
+  ## IPv6 stringToBuffer() implementation.
+  pathStringToBuffer(s, vb)
 
 proc ip6zoneBtS(vb: var VBuffer, s: var string): bool =
   ## IPv6 bufferToString() implementation.
-  if vb.readSeq(s) > 0:
-    result = true
+  pathBufferToStringNoSlash(vb, s)
 
 proc ip6zoneVB(vb: var VBuffer): bool =
   ## IPv6 validateBuffer() implementation.
-  var s = ""
-  if vb.readSeq(s) > 0:
-    if s.find('/') == -1:
-      result = true
+  pathValidateBufferNoSlash(vb)
 
 proc portStB(s: string, vb: var VBuffer): bool =
   ## Port number stringToBuffer() implementation.
@@ -154,7 +177,8 @@ proc portBtS(vb: var VBuffer, s: var string): bool =
   ## Port number bufferToString() implementation.
   var port: array[2, byte]
   if vb.readArray(port) == 2:
-    var nport = (safeConvert[uint16](port[0]) shl 8) or safeConvert[uint16](port[1])
+    let nport =
+      (safeConvert[uint16](port[0]) shl 8) or safeConvert[uint16](port[1])
     s = $nport
     result = true
 
@@ -214,7 +238,8 @@ proc onionBtS(vb: var VBuffer, s: var string): bool =
   ## ONION address bufferToString() implementation.
   var buf: array[12, byte]
   if vb.readArray(buf) == 12:
-    var nport = (safeConvert[uint16](buf[10]) shl 8) or safeConvert[uint16](buf[11])
+    let nport =
+      (safeConvert[uint16](buf[10]) shl 8) or safeConvert[uint16](buf[11])
     s = Base32Lower.encode(buf.toOpenArray(0, 9))
     s.add(":")
     s.add($nport)
@@ -248,7 +273,8 @@ proc onion3BtS(vb: var VBuffer, s: var string): bool =
   ## ONION address bufferToString() implementation.
   var buf: array[37, byte]
   if vb.readArray(buf) == 37:
-    var nport = (safeConvert[uint16](buf[35]) shl 8) or safeConvert[uint16](buf[36])
+    var nport =
+      (safeConvert[uint16](buf[35]) shl 8) or safeConvert[uint16](buf[36])
     s = Base32Lower.encode(buf.toOpenArray(0, 34))
     s.add(":")
     s.add($nport)
@@ -262,40 +288,27 @@ proc onion3VB(vb: var VBuffer): bool =
 
 proc unixStB(s: string, vb: var VBuffer): bool =
   ## Unix socket name stringToBuffer() implementation.
-  if len(s) > 0:
-    vb.writeSeq(s)
-    result = true
+  pathStringToBuffer(s, vb)
 
 proc unixBtS(vb: var VBuffer, s: var string): bool =
   ## Unix socket name bufferToString() implementation.
-  s = ""
-  if vb.readSeq(s) > 0:
-    result = true
+  pathBufferToString(vb, s)
 
 proc unixVB(vb: var VBuffer): bool =
   ## Unix socket name validateBuffer() implementation.
-  var s = ""
-  if vb.readSeq(s) > 0:
-    result = true
+  pathValidateBuffer(vb)
 
 proc dnsStB(s: string, vb: var VBuffer): bool =
   ## DNS name stringToBuffer() implementation.
-  if len(s) > 0:
-    vb.writeSeq(s)
-    result = true
+  pathStringToBuffer(s, vb)
 
 proc dnsBtS(vb: var VBuffer, s: var string): bool =
   ## DNS name bufferToString() implementation.
-  s = ""
-  if vb.readSeq(s) > 0:
-    result = true
+  pathBufferToStringNoSlash(vb, s)
 
 proc dnsVB(vb: var VBuffer): bool =
   ## DNS name validateBuffer() implementation.
-  var s = ""
-  if vb.readSeq(s) > 0:
-    if s.find('/') == -1:
-      result = true
+  pathValidateBufferNoSlash(vb)
 
 proc mapEq*(codec: string): MaPattern =
   ## ``Equal`` operator for pattern
@@ -398,6 +411,9 @@ const
     MAProtocol(
       mcodec: multiCodec("quic"), kind: Marker, size: 0
     ),
+    MAProtocol(
+      mcodec: multiCodec("quic-v1"), kind: Marker, size: 0
+    ),
     MAProtocol(
       mcodec: multiCodec("ip6zone"), kind: Length, size: 0,
       coder: TranscoderIP6Zone
@@ -657,7 +673,8 @@ proc getPart(ma: MultiAddress, index: int): MaResult[MultiAddress] =
     inc(offset)
   ok(res)
 
-proc getParts[U, V](ma: MultiAddress, slice: HSlice[U, V]): MaResult[MultiAddress] =
+proc getParts[U, V](ma: MultiAddress,
+                    slice: HSlice[U, V]): MaResult[MultiAddress] =
   when slice.a is BackwardsIndex or slice.b is BackwardsIndex:
     let maLength = ? len(ma)
   template normalizeIndex(index): int =
@@ -671,7 +688,8 @@ proc getParts[U, V](ma: MultiAddress, slice: HSlice[U, V]): MaResult[MultiAddres
     ? res.append(? ma[i])
   ok(res)
 
-proc `[]`*(ma: MultiAddress, i: int | BackwardsIndex): MaResult[MultiAddress] {.inline.} =
+proc `[]`*(ma: MultiAddress,
+           i: int | BackwardsIndex): MaResult[MultiAddress] {.inline.} =
   ## Returns part with index ``i`` of MultiAddress ``ma``.
   when i is BackwardsIndex:
     let maLength = ? len(ma)
@@ -766,7 +784,7 @@ proc toString*(value: MultiAddress): MaResult[string] =
       if not proto.coder.bufferToString(vb.data, part):
         return err("multiaddress: Decoding protocol error")
       parts.add($(proto.mcodec))
-      if proto.kind == Path and part[0] == '/':
+      if len(part) > 0 and (proto.kind == Path) and (part[0] == '/'):
         parts.add(part[1..^1])
       else:
         parts.add(part)
@@ -955,7 +973,7 @@ proc init*(mtype: typedesc[MultiAddress]): MultiAddress =
   ## Initialize empty MultiAddress.
   result.data = initVBuffer()
 
-proc init*(mtype: typedesc[MultiAddress], address: ValidIpAddress,
+proc init*(mtype: typedesc[MultiAddress], address: IpAddress,
            protocol: IpTransportProtocol, port: Port): MultiAddress =
   var res: MultiAddress
   res.data = initVBuffer()
@@ -1117,16 +1135,20 @@ proc getField*(pb: ProtoBuffer, field: int,
   if not(res):
     ok(false)
   else:
-    value = MultiAddress.init(buffer).valueOr: return err(ProtoError.IncorrectBlob)
+    value = MultiAddress.init(buffer).valueOr:
+      return err(ProtoError.IncorrectBlob)
     ok(true)
 
 proc getRepeatedField*(pb: ProtoBuffer, field: int,
                        value: var seq[MultiAddress]): ProtoResult[bool] {.
      inline.} =
-  ## Read repeated field from protobuf message. ``field`` is field number. If the message is malformed, an error is returned.
-  ## If field is not present in message, then ``ok(false)`` is returned and value is empty. If field is present,
-  ## but no items could be parsed, then ``err(ProtoError.IncorrectBlob)`` is returned and value is empty.
-  ## If field is present and some item could be parsed, then ``true`` is returned and value contains the parsed values.
+  ## Read repeated field from protobuf message. ``field`` is field number.
+  ## If the message is malformed, an error is returned. If field is not present
+  ## in message, then ``ok(false)`` is returned and value is empty. If field is
+  ## present, but no items could be parsed, then
+  ## ``err(ProtoError.IncorrectBlob)`` is returned and value is empty.
+  ## If field is present and some item could be parsed, then ``true`` is
+  ## returned and value contains the parsed values.
   var items: seq[seq[byte]]
   value.setLen(0)
   let res = ? pb.getRepeatedField(field, items)

libp2p/multicodec.nim

@@ -193,6 +193,7 @@ const MultiCodecList = [
   ("https", 0x01BB),
   ("tls", 0x01C0),
   ("quic", 0x01CC),
+  ("quic-v1", 0x01CD),
   ("ws", 0x01DD),
   ("wss", 0x01DE),
   ("p2p-websocket-star", 0x01DF), # not in multicodec list

libp2p/multistream.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -45,15 +45,18 @@ proc new*(T: typedesc[MultistreamSelect]): T =
   )
 
 template validateSuffix(str: string): untyped =
-    if str.endsWith("\n"):
-      str.removeSuffix("\n")
-    else:
-      raise newException(MultiStreamError, "MultistreamSelect failed, malformed message")
+  if str.endsWith("\n"):
+    str.removeSuffix("\n")
+  else:
+    raise (ref MultiStreamError)(msg:
+      "MultistreamSelect failed, malformed message")
 
-proc select*(_: MultistreamSelect | type MultistreamSelect,
-             conn: Connection,
-             proto: seq[string]):
-             Future[string] {.async.} =
+proc select*(
+    _: MultistreamSelect | type MultistreamSelect,
+    conn: Connection,
+    proto: seq[string]
+): Future[string] {.async: (raises: [
+    CancelledError, LPStreamError, MultiStreamError]).} =
   trace "initiating handshake", conn, codec = Codec
   ## select a remote protocol
   await conn.writeLp(Codec & "\n") # write handshake
@@ -66,7 +69,7 @@ proc select*(_: MultistreamSelect | type MultistreamSelect,
 
   if s != Codec:
     notice "handshake failed", conn, codec = s
-    raise newException(MultiStreamError, "MultistreamSelect handshake failed")
+    raise (ref MultiStreamError)(msg: "MultistreamSelect handshake failed")
   else:
     trace "multistream handshake success", conn
 
@@ -98,19 +101,29 @@ proc select*(_: MultistreamSelect | type MultistreamSelect,
       # No alternatives, fail
       return ""
 
-proc select*(_: MultistreamSelect | type MultistreamSelect,
-             conn: Connection,
-             proto: string): Future[bool] {.async.} =
+proc select*(
+    _: MultistreamSelect | type MultistreamSelect,
+    conn: Connection,
+    proto: string
+): Future[bool] {.async: (raises: [
+    CancelledError, LPStreamError, MultiStreamError]).} =
   if proto.len > 0:
-    return (await MultistreamSelect.select(conn, @[proto])) == proto
+    (await MultistreamSelect.select(conn, @[proto])) == proto
   else:
-    return (await MultistreamSelect.select(conn, @[])) == Codec
+    (await MultistreamSelect.select(conn, @[])) == Codec
 
-proc select*(m: MultistreamSelect, conn: Connection): Future[bool] =
+proc select*(
+    m: MultistreamSelect,
+    conn: Connection
+): Future[bool] {.async: (raises: [
+    CancelledError, LPStreamError, MultiStreamError], raw: true).} =
   m.select(conn, "")
 
-proc list*(m: MultistreamSelect,
-           conn: Connection): Future[seq[string]] {.async.} =
+proc list*(
+    m: MultistreamSelect,
+    conn: Connection
+): Future[seq[string]] {.async: (raises: [
+    CancelledError, LPStreamError, MultiStreamError]).} =
   ## list remote protos requests on connection
   if not await m.select(conn):
     return
@@ -126,12 +139,13 @@ proc list*(m: MultistreamSelect,
   result = list
 
 proc handle*(
-  _: type MultistreamSelect,
-  conn: Connection,
-  protos: seq[string],
-  matchers = newSeq[Matcher](),
-  active: bool = false,
-  ): Future[string] {.async, gcsafe.} =
+    _: type MultistreamSelect,
+    conn: Connection,
+    protos: seq[string],
+    matchers = newSeq[Matcher](),
+    active: bool = false
+): Future[string] {.async: (raises: [
+    CancelledError, LPStreamError, MultiStreamError]).} =
   trace "Starting multistream negotiation", conn, handshaked = active
   var handshaked = active
   while not conn.atEof:
@@ -140,8 +154,8 @@ proc handle*(
 
     if not handshaked and ms != Codec:
       debug "expected handshake message", conn, instead=ms
-      raise newException(CatchableError,
-                         "MultistreamSelect handling failed, invalid first message")
+      raise (ref MultiStreamError)(msg:
+        "MultistreamSelect handling failed, invalid first message")
 
     trace "handle: got request", conn, ms
     if ms.len() <= 0:
@@ -172,14 +186,16 @@ proc handle*(
       trace "no handlers", conn, protocol = ms
       await conn.writeLp(Na)
 
-proc handle*(m: MultistreamSelect, conn: Connection, active: bool = false) {.async, gcsafe.} =
+proc handle*(
+    m: MultistreamSelect,
+    conn: Connection,
+    active: bool = false) {.async: (raises: [CancelledError]).} =
   trace "Starting multistream handler", conn, handshaked = active
   var
-    handshaked = active
     protos: seq[string]
     matchers: seq[Matcher]
   for h in m.handlers:
-    if not isNil(h.match):
+    if h.match != nil:
       matchers.add(h.match)
     for proto in h.protos:
       protos.add(proto)
@@ -187,12 +203,13 @@ proc handle*(m: MultistreamSelect, conn: Connection, active: bool = false) {.asy
   try:
     let ms = await MultistreamSelect.handle(conn, protos, matchers, active)
     for h in m.handlers:
-      if (not isNil(h.match) and h.match(ms)) or h.protos.contains(ms):
+      if (h.match != nil and h.match(ms)) or h.protos.contains(ms):
         trace "found handler", conn, protocol = ms
 
         var protocolHolder = h
         let maxIncomingStreams = protocolHolder.protocol.maxIncomingStreams
-        if protocolHolder.openedStreams.getOrDefault(conn.peerId) >= maxIncomingStreams:
+        if protocolHolder.openedStreams.getOrDefault(conn.peerId) >=
+            maxIncomingStreams:
           debug "Max streams for protocol reached, blocking new stream",
             conn, protocol = ms, maxIncomingStreams
           return
@@ -229,10 +246,14 @@ proc addHandler*(m: MultistreamSelect,
                  matcher: Matcher = nil) =
   addHandler(m, @[codec], protocol, matcher)
 
-proc addHandler*(m: MultistreamSelect,
-                 codec: string,
-                 handler: LPProtoHandler,
-                 matcher: Matcher = nil) =
+proc addHandler*[E](
+    m: MultistreamSelect,
+    codec: string,
+    handler: LPProtoHandler |
+      proc (
+        conn: Connection,
+        proto: string): InternalRaisesFuture[void, E],
+    matcher: Matcher = nil) =
   ## helper to allow registering pure handlers
   trace "registering proto handler", proto = codec
   let protocol = new LPProtocol
@@ -243,8 +264,34 @@ proc addHandler*(m: MultistreamSelect,
                                protocol: protocol,
                                match: matcher))
 
-proc start*(m: MultistreamSelect) {.async.} =
-  await allFutures(m.handlers.mapIt(it.protocol.start()))
+proc start*(m: MultistreamSelect) {.async: (raises: [CancelledError]).} =
+  # Nim 1.6.18: Using `mapIt` results in a seq of `.Raising([])`
+  # TODO https://github.com/nim-lang/Nim/issues/23445
+  var futs = newSeqOfCap[Future[void].Raising([CancelledError])](m.handlers.len)
+  for it in m.handlers:
+    futs.add it.protocol.start()
+  try:
+    await allFutures(futs)
+    for fut in futs:
+      await fut
+  except CancelledError as exc:
+    var pending: seq[Future[void].Raising([])]
+    doAssert m.handlers.len == futs.len, "Handlers modified while starting"
+    for i, fut in futs:
+      if not fut.finished:
+        pending.add fut.cancelAndWait()
+      elif fut.completed:
+        pending.add m.handlers[i].protocol.stop()
+      else:
+        static: doAssert typeof(fut).E is (CancelledError,)
+    await noCancel allFutures(pending)
+    raise exc
 
-proc stop*(m: MultistreamSelect) {.async.} =
-  await allFutures(m.handlers.mapIt(it.protocol.stop()))
+proc stop*(m: MultistreamSelect) {.async: (raises: []).} =
+  # Nim 1.6.18: Using `mapIt` results in a seq of `.Raising([CancelledError])`
+  var futs = newSeqOfCap[Future[void].Raising([])](m.handlers.len)
+  for it in m.handlers:
+    futs.add it.protocol.stop()
+  await noCancel allFutures(futs)
+  for fut in futs:
+    await fut

libp2p/muxers/mplex/coder.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -42,7 +42,10 @@ const MaxMsgSize* = 1 shl 20 # 1mb
 proc newInvalidMplexMsgType*(): ref InvalidMplexMsgType =
   newException(InvalidMplexMsgType, "invalid message type")
 
-proc readMsg*(conn: Connection): Future[Msg] {.async, gcsafe.} =
+proc readMsg*(
+    conn: Connection
+): Future[Msg] {.async: (raises: [
+    CancelledError, LPStreamError, MuxerError]).} =
   let header = await conn.readVarint()
   trace "read header varint", varint = header, conn
 
@@ -55,10 +58,13 @@ proc readMsg*(conn: Connection): Future[Msg] {.async, gcsafe.} =
 
   return (header shr 3, MessageType(msgType), data)
 
-proc writeMsg*(conn: Connection,
-               id: uint64,
-               msgType: MessageType,
-               data: seq[byte] = @[]): Future[void] =
+proc writeMsg*(
+    conn: Connection,
+    id: uint64,
+    msgType: MessageType,
+    data: seq[byte] = @[]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
   var
     left = data.len
     offset = 0
@@ -84,8 +90,11 @@ proc writeMsg*(conn: Connection,
   # message gets written before some of the chunks
   conn.write(buf.buffer)
 
-proc writeMsg*(conn: Connection,
-               id: uint64,
-               msgType: MessageType,
-               data: string): Future[void] =
+proc writeMsg*(
+    conn: Connection,
+    id: uint64,
+    msgType: MessageType,
+    data: string
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
   conn.writeMsg(id, msgType, data.toBytes())

libp2p/muxers/mplex/lpchannel.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -28,7 +28,8 @@ when defined(libp2p_mplex_metrics):
   declareHistogram libp2p_mplex_qtime, "message queuing time"
 
 when defined(libp2p_network_protocols_metrics):
-  declareCounter libp2p_protocols_bytes, "total sent or received bytes", ["protocol", "direction"]
+  declareCounter libp2p_protocols_bytes,
+    "total sent or received bytes", ["protocol", "direction"]
 
 ## Channel half-closed states
 ##
@@ -64,16 +65,16 @@ type
 
 func shortLog*(s: LPChannel): auto =
   try:
-    if s.isNil: "LPChannel(nil)"
+    if s == nil: "LPChannel(nil)"
     elif s.name != $s.oid and s.name.len > 0:
       &"{shortLog(s.conn.peerId)}:{s.oid}:{s.name}"
     else: &"{shortLog(s.conn.peerId)}:{s.oid}"
   except ValueError as exc:
-    raise newException(Defect, exc.msg)
+    raiseAssert(exc.msg)
 
 chronicles.formatIt(LPChannel): shortLog(it)
 
-proc open*(s: LPChannel) {.async, gcsafe.} =
+proc open*(s: LPChannel) {.async: (raises: [CancelledError, LPStreamError]).} =
   trace "Opening channel", s, conn = s.conn
   if s.conn.isClosed:
     return
@@ -82,20 +83,20 @@ proc open*(s: LPChannel) {.async, gcsafe.} =
     s.isOpen = true
   except CancelledError as exc:
     raise exc
-  except CatchableError as exc:
+  except LPStreamError as exc:
     await s.conn.close()
     raise exc
 
 method closed*(s: LPChannel): bool =
   s.closedLocal
 
-proc closeUnderlying(s: LPChannel): Future[void] {.async.} =
+proc closeUnderlying(s: LPChannel): Future[void] {.async: (raises: []).} =
   ## Channels may be closed for reading and writing in any order - we'll close
   ## the underlying bufferstream when both directions are closed
   if s.closedLocal and s.atEof():
     await procCall BufferStream(s).close()
 
-proc reset*(s: LPChannel) {.async, gcsafe.} =
+proc reset*(s: LPChannel) {.async: (raises: []).} =
   if s.isClosed:
     trace "Already closed", s
     return
@@ -108,22 +109,21 @@ proc reset*(s: LPChannel) {.async, gcsafe.} =
 
   if s.isOpen and not s.conn.isClosed:
     # If the connection is still active, notify the other end
-    proc resetMessage() {.async.} =
+    proc resetMessage() {.async: (raises: []).} =
       try:
         trace "sending reset message", s, conn = s.conn
-        await s.conn.writeMsg(s.id, s.resetCode) # write reset
-      except CatchableError as exc:
-        # No cancellations
-        await s.conn.close()
+        await noCancel s.conn.writeMsg(s.id, s.resetCode) # write reset
+      except LPStreamError as exc:
         trace "Can't send reset message", s, conn = s.conn, msg = exc.msg
+        await s.conn.close()
 
     asyncSpawn resetMessage()
 
-  await s.closeImpl() # noraises, nocancels
+  await s.closeImpl()
 
   trace "Channel reset", s
 
-method close*(s: LPChannel) {.async, gcsafe.} =
+method close*(s: LPChannel) {.async: (raises: []).} =
   ## Close channel for writing - a message will be sent to the other peer
   ## informing them that the channel is closed and that we're waiting for
   ## their acknowledgement.
@@ -137,10 +137,9 @@ method close*(s: LPChannel) {.async, gcsafe.} =
   if s.isOpen and not s.conn.isClosed:
     try:
       await s.conn.writeMsg(s.id, s.closeCode) # write close
-    except CancelledError as exc:
+    except CancelledError:
       await s.conn.close()
-      raise exc
-    except CatchableError as exc:
+    except LPStreamError as exc:
       # It's harmless that close message cannot be sent - the connection is
       # likely down already
       await s.conn.close()
@@ -154,16 +153,17 @@ method initStream*(s: LPChannel) =
   if s.objName.len == 0:
     s.objName = LPChannelTrackerName
 
-  s.timeoutHandler = proc(): Future[void] {.gcsafe.} =
+  s.timeoutHandler = proc(): Future[void] {.async: (raises: [], raw: true).} =
     trace "Idle timeout expired, resetting LPChannel", s
     s.reset()
 
   procCall BufferStream(s).initStream()
 
-method readOnce*(s: LPChannel,
-                 pbytes: pointer,
-                 nbytes: int):
-                 Future[int] {.async.} =
+method readOnce*(
+    s: LPChannel,
+    pbytes: pointer,
+    nbytes: int
+): Future[int] {.async: (raises: [CancelledError, LPStreamError]).} =
   ## Mplex relies on reading being done regularly from every channel, or all
   ## channels are blocked - in particular, this means that reading from one
   ## channel must not be done from within a callback / read handler of another
@@ -186,15 +186,19 @@ method readOnce*(s: LPChannel,
     if bytes == 0:
       await s.closeUnderlying()
     return bytes
-  except CatchableError as exc:
-    # readOnce in BufferStream generally raises on EOF or cancellation - for
-    # the former, resetting is harmless, for the latter it's necessary because
-    # data has been lost in s.readBuf and there's no way to gracefully recover /
-    # use the channel any more
+  except CancelledError as exc:
+    await s.reset()
+    raise exc
+  except LPStreamError as exc:
+    # Resetting is necessary because data has been lost in s.readBuf and
+    # there's no way to gracefully recover / use the channel any more
     await s.reset()
     raise newLPStreamConnDownError(exc)
 
-proc prepareWrite(s: LPChannel, msg: seq[byte]): Future[void] {.async.} =
+proc prepareWrite(
+    s: LPChannel,
+    msg: seq[byte]
+): Future[void] {.async: (raises: [CancelledError, LPStreamError]).} =
   # prepareWrite is the slow path of writing a message - see conditions in
   # write
   if s.remoteReset:
@@ -222,7 +226,10 @@ proc prepareWrite(s: LPChannel, msg: seq[byte]): Future[void] {.async.} =
   await s.conn.writeMsg(s.id, s.msgCode, msg)
 
 proc completeWrite(
-    s: LPChannel, fut: Future[void], msgLen: int): Future[void] {.async.} =
+    s: LPChannel,
+    fut: Future[void].Raising([CancelledError, LPStreamError]),
+    msgLen: int
+): Future[void] {.async: (raises: [CancelledError, LPStreamError]).} =
   try:
     s.writes += 1
 
@@ -235,7 +242,10 @@ proc completeWrite(
 
     when defined(libp2p_network_protocols_metrics):
       if s.protocol.len > 0:
-        libp2p_protocols_bytes.inc(msgLen.int64, labelValues=[s.protocol, "out"])
+        # This crashes on Nim 2.0.2 with `--mm:orc` during `nimble test`
+        # https://github.com/status-im/nim-metrics/issues/79
+        libp2p_protocols_bytes.inc(
+          msgLen.int64, labelValues = [s.protocol, "out"])
 
     s.activity = true
   except CancelledError as exc:
@@ -247,7 +257,7 @@ proc completeWrite(
     raise exc
   except LPStreamEOFError as exc:
     raise exc
-  except CatchableError as exc:
+  except LPStreamError as exc:
     trace "exception in lpchannel write handler", s, msg = exc.msg
     await s.reset()
     await s.conn.close()
@@ -255,7 +265,11 @@ proc completeWrite(
   finally:
     s.writes -= 1
 
-method write*(s: LPChannel, msg: seq[byte]): Future[void] =
+method write*(
+    s: LPChannel,
+    msg: seq[byte]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
   ## Write to mplex channel - there may be up to MaxWrite concurrent writes
   ## pending after which the peer is disconnected
 
@@ -276,13 +290,12 @@ method write*(s: LPChannel, msg: seq[byte]): Future[void] =
 method getWrapped*(s: LPChannel): Connection = s.conn
 
 proc init*(
-  L: type LPChannel,
-  id: uint64,
-  conn: Connection,
-  initiator: bool,
-  name: string = "",
-  timeout: Duration = DefaultChanTimeout): LPChannel =
-
+    L: type LPChannel,
+    id: uint64,
+    conn: Connection,
+    initiator: bool,
+    name: string = "",
+    timeout: Duration = DefaultChanTimeout): LPChannel =
   let chann = L(
     id: id,
     name: name,

libp2p/muxers/mplex/mplex.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -56,7 +56,7 @@ proc newTooManyChannels(): ref TooManyChannels =
 proc newInvalidChannelIdError(): ref InvalidChannelIdError =
   newException(InvalidChannelIdError, "max allowed channel count exceeded")
 
-proc cleanupChann(m: Mplex, chann: LPChannel) {.async, inline.} =
+proc cleanupChann(m: Mplex, chann: LPChannel) {.async: (raises: []), inline.} =
   ## remove the local channel from the internal tables
   ##
   try:
@@ -68,19 +68,19 @@ proc cleanupChann(m: Mplex, chann: LPChannel) {.async, inline.} =
       libp2p_mplex_channels.set(
         m.channels[chann.initiator].len.int64,
         labelValues = [$chann.initiator, $m.connection.peerId])
-  except CatchableError as exc:
+  except CancelledError as exc:
     warn "Error cleaning up mplex channel", m, chann, msg = exc.msg
 
-proc newStreamInternal*(m: Mplex,
-                        initiator: bool = true,
-                        chanId: uint64 = 0,
-                        name: string = "",
-                        timeout: Duration): LPChannel
-                        {.gcsafe, raises: [InvalidChannelIdError].} =
+proc newStreamInternal*(
+    m: Mplex,
+    initiator: bool = true,
+    chanId: uint64 = 0,
+    name: string = "",
+    timeout: Duration): LPChannel {.gcsafe, raises: [InvalidChannelIdError].} =
   ## create new channel/stream
   ##
-  let id = if initiator:
-    m.currentId.inc(); m.currentId
+  let id =
+    if initiator: m.currentId.inc(); m.currentId
     else: chanId
 
   if id in m.channels[initiator]:
@@ -111,18 +111,14 @@ proc newStreamInternal*(m: Mplex,
       m.channels[initiator].len.int64,
       labelValues = [$initiator, $m.connection.peerId])
 
-proc handleStream(m: Mplex, chann: LPChannel) {.async.} =
+proc handleStream(m: Mplex, chann: LPChannel) {.async: (raises: []).} =
   ## call the muxer stream handler for this channel
   ##
-  try:
-    await m.streamHandler(chann)
-    trace "finished handling stream", m, chann
-    doAssert(chann.closed, "connection not closed by handler!")
-  except CatchableError as exc:
-    trace "Exception in mplex stream handler", m, chann, msg = exc.msg
-    await chann.reset()
+  await m.streamHandler(chann)
+  trace "finished handling stream", m, chann
+  doAssert(chann.closed, "connection not closed by handler!")
 
-method handle*(m: Mplex) {.async, gcsafe.} =
+method handle*(m: Mplex) {.async: (raises: []).} =
   trace "Starting mplex handler", m
   try:
     while not m.connection.atEof:
@@ -150,7 +146,7 @@ method handle*(m: Mplex) {.async, gcsafe.} =
         else:
           if m.channels[false].len > m.maxChannCount - 1:
             warn "too many channels created by remote peer",
-                  allowedMax = MaxChannelCount, m
+                 allowedMax = MaxChannelCount, m
             raise newTooManyChannels()
 
           let name = string.fromBytes(data)
@@ -159,59 +155,65 @@ method handle*(m: Mplex) {.async, gcsafe.} =
       trace "Processing channel message", m, channel, data = data.shortLog
 
       case msgType:
-        of MessageType.New:
-          trace "created channel", m, channel
+      of MessageType.New:
+        trace "created channel", m, channel
 
-          if not isNil(m.streamHandler):
-            # Launch handler task
-            # All the errors are handled inside `handleStream()` procedure.
-            asyncSpawn m.handleStream(channel)
+        if m.streamHandler != nil:
+          # Launch handler task
+          # All the errors are handled inside `handleStream()` procedure.
+          asyncSpawn m.handleStream(channel)
 
-        of MessageType.MsgIn, MessageType.MsgOut:
-          if data.len > MaxMsgSize:
-            warn "attempting to send a packet larger than allowed",
-                 allowed = MaxMsgSize, channel
-            raise newLPStreamLimitError()
+      of MessageType.MsgIn, MessageType.MsgOut:
+        if data.len > MaxMsgSize:
+          warn "attempting to send a packet larger than allowed",
+                allowed = MaxMsgSize, channel
+          raise newLPStreamLimitError()
 
-          trace "pushing data to channel", m, channel, len = data.len
-          try:
-            await channel.pushData(data)
-            trace "pushed data to channel", m, channel, len = data.len
-          except LPStreamClosedError as exc:
-            # Channel is being closed, but `cleanupChann` was not yet triggered.
-            trace "pushing data to channel failed", m, channel, len = data.len,
-              msg = exc.msg
-            discard  # Ignore message, same as if `cleanupChann` had completed.
+        trace "pushing data to channel", m, channel, len = data.len
+        try:
+          await channel.pushData(data)
+          trace "pushed data to channel", m, channel, len = data.len
+        except LPStreamClosedError as exc:
+          # Channel is being closed, but `cleanupChann` was not yet triggered.
+          trace "pushing data to channel failed", m, channel, len = data.len,
+            msg = exc.msg
+          discard  # Ignore message, same as if `cleanupChann` had completed.
 
-        of MessageType.CloseIn, MessageType.CloseOut:
-          await channel.pushEof()
-        of MessageType.ResetIn, MessageType.ResetOut:
-          channel.remoteReset = true
-          await channel.reset()
+      of MessageType.CloseIn, MessageType.CloseOut:
+        await channel.pushEof()
+      of MessageType.ResetIn, MessageType.ResetOut:
+        channel.remoteReset = true
+        await channel.reset()
   except CancelledError:
     debug "Unexpected cancellation in mplex handler", m
   except LPStreamEOFError as exc:
     trace "Stream EOF", m, msg = exc.msg
-  except CatchableError as exc:
-    debug "Unexpected exception in mplex read loop", m, msg = exc.msg
+  except LPStreamError as exc:
+    debug "Unexpected stream exception in mplex read loop", m, msg = exc.msg
+  except MuxerError as exc:
+    debug "Unexpected muxer exception in mplex read loop", m, msg = exc.msg
   finally:
     await m.close()
   trace "Stopped mplex handler", m
 
-proc new*(M: type Mplex,
-           conn: Connection,
-           inTimeout: Duration = DefaultChanTimeout,
-           outTimeout: Duration = DefaultChanTimeout,
-           maxChannCount: int = MaxChannelCount): Mplex =
+proc new*(
+    M: type Mplex,
+    conn: Connection,
+    inTimeout: Duration = DefaultChanTimeout,
+    outTimeout: Duration = DefaultChanTimeout,
+    maxChannCount: int = MaxChannelCount): Mplex =
   M(connection: conn,
     inChannTimeout: inTimeout,
     outChannTimeout: outTimeout,
     oid: genOid(),
     maxChannCount: maxChannCount)
 
-method newStream*(m: Mplex,
-                  name: string = "",
-                  lazy: bool = false): Future[Connection] {.async, gcsafe.} =
+method newStream*(
+    m: Mplex,
+    name: string = "",
+    lazy: bool = false
+): Future[Connection] {.async: (raises: [
+    CancelledError, LPStreamError, MuxerError]).} =
   let channel = m.newStreamInternal(timeout = m.inChannTimeout)
 
   if not lazy:
@@ -219,7 +221,7 @@ method newStream*(m: Mplex,
 
   return Connection(channel)
 
-method close*(m: Mplex) {.async, gcsafe.} =
+method close*(m: Mplex) {.async: (raises: []).} =
   if m.isClosed:
     trace "Already closed", m
     return

libp2p/muxers/muxer.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -23,16 +23,17 @@ type
   MuxerError* = object of LPError
   TooManyChannels* = object of MuxerError
 
-  StreamHandler* = proc(conn: Connection): Future[void] {.gcsafe, raises: [].}
-  MuxerHandler* = proc(muxer: Muxer): Future[void] {.gcsafe, raises: [].}
+  StreamHandler* = proc(conn: Connection): Future[void] {.async: (raises: []).}
+  MuxerHandler* = proc(muxer: Muxer): Future[void] {.async: (raises: []).}
 
   Muxer* = ref object of RootObj
     streamHandler*: StreamHandler
-    handler*: Future[void]
+    handler*: Future[void].Raising([])
     connection*: Connection
 
   # user provider proc that returns a constructed Muxer
-  MuxerConstructor* = proc(conn: Connection): Muxer {.gcsafe, closure, raises: [].}
+  MuxerConstructor* =
+    proc(conn: Connection): Muxer {.gcsafe, closure, raises: [].}
 
   # this wraps a creator proc that knows how to make muxers
   MuxerProvider* = object
@@ -40,24 +41,32 @@ type
     codec*: string
 
 func shortLog*(m: Muxer): auto =
-  if isNil(m): "nil"
+  if m == nil: "nil"
   else: shortLog(m.connection)
+
 chronicles.formatIt(Muxer): shortLog(it)
 
 # muxer interface
-method newStream*(m: Muxer, name: string = "", lazy: bool = false):
-  Future[Connection] {.base, async, gcsafe.} = discard
-method close*(m: Muxer) {.base, async, gcsafe.} =
-  if not isNil(m.connection):
+method newStream*(
+    m: Muxer,
+    name: string = "",
+    lazy: bool = false
+): Future[Connection] {.base, async: (raises: [
+    CancelledError, LPStreamError, MuxerError], raw: true).} =
+  raiseAssert("Not implemented!")
+
+method close*(m: Muxer) {.base, async: (raises: []).} =
+  if m.connection != nil:
     await m.connection.close()
-method handle*(m: Muxer): Future[void] {.base, async, gcsafe.} = discard
+
+method handle*(m: Muxer): Future[void] {.base, async: (raises: []).} = discard
 
 proc new*(
-  T: typedesc[MuxerProvider],
-  creator: MuxerConstructor,
-  codec: string): T {.gcsafe.} =
-
+    T: typedesc[MuxerProvider],
+    creator: MuxerConstructor,
+    codec: string): T {.gcsafe.} =
   let muxerProvider = T(newMuxer: creator, codec: codec)
   muxerProvider
 
-method getStreams*(m: Muxer): seq[Connection] {.base.} = doAssert false, "not implemented"
+method getStreams*(m: Muxer): seq[Connection] {.base.} =
+  raiseAssert("Not implemented!")

libp2p/muxers/yamux/yamux.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -22,18 +22,22 @@ logScope:
 const
   YamuxCodec* = "/yamux/1.0.0"
   YamuxVersion = 0.uint8
-  DefaultWindowSize = 256000
+  YamuxDefaultWindowSize* = 256000
+  MaxSendQueueSize = 256000
   MaxChannelCount = 200
 
 when defined(libp2p_yamux_metrics):
-  declareGauge(libp2p_yamux_channels, "yamux channels", labels = ["initiator", "peer"])
-  declareHistogram libp2p_yamux_send_queue, "message send queue length (in byte)",
-    buckets = [0.0, 100.0, 250.0, 1000.0, 2000.0, 1600.0, 6400.0, 25600.0, 256000.0]
-  declareHistogram libp2p_yamux_recv_queue, "message recv queue length (in byte)",
-    buckets = [0.0, 100.0, 250.0, 1000.0, 2000.0, 1600.0, 6400.0, 25600.0, 256000.0]
+  declareGauge libp2p_yamux_channels,
+    "yamux channels", labels = ["initiator", "peer"]
+  declareHistogram libp2p_yamux_send_queue,
+    "message send queue length (in byte)", buckets = [
+      0.0, 100.0, 250.0, 1000.0, 2000.0, 3200.0, 6400.0, 25600.0, 256000.0]
+  declareHistogram libp2p_yamux_recv_queue,
+    "message recv queue length (in byte)", buckets = [
+      0.0, 100.0, 250.0, 1000.0, 2000.0, 3200.0, 6400.0, 25600.0, 256000.0]
 
 type
-  YamuxError* = object of CatchableError
+  YamuxError* = object of MuxerError
 
   MsgType = enum
     Data = 0x0
@@ -59,7 +63,10 @@ type
     streamId: uint32
     length: uint32
 
-proc readHeader(conn: LPStream): Future[YamuxHeader] {.async, gcsafe.} =
+proc readHeader(
+    conn: LPStream
+): Future[YamuxHeader] {.async: (raises: [
+    CancelledError, LPStreamError, MuxerError]).} =
   var buffer: array[12, byte]
   await conn.readExactly(addr buffer[0], 12)
 
@@ -73,10 +80,10 @@ proc readHeader(conn: LPStream): Future[YamuxHeader] {.async, gcsafe.} =
   return result
 
 proc `$`(header: YamuxHeader): string =
-  result = "{" & $header.msgType & ", "
-  result &= "{" & header.flags.foldl(if a != "": a & ", " & $b else: $b, "") & "}, "
-  result &= "streamId: " & $header.streamId & ", "
-  result &= "length: " & $header.length & "}"
+  "{" & $header.msgType & ", " &
+  "{" & header.flags.foldl(if a != "": a & ", " & $b else: $b, "") & "}, " &
+  "streamId: " & $header.streamId & ", " &
+  "length: " & $header.length & "}"
 
 proc encode(header: YamuxHeader): array[12, byte] =
   result[0] = header.version
@@ -85,10 +92,14 @@ proc encode(header: YamuxHeader): array[12, byte] =
   result[4..7] = toBytesBE(header.streamId)
   result[8..11] = toBytesBE(header.length)
 
-proc write(conn: LPStream, header: YamuxHeader): Future[void] {.gcsafe.} =
+proc write(
+    conn: LPStream,
+    header: YamuxHeader
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
   trace "write directly on stream", h = $header
   var buffer = header.encode()
-  return conn.write(@buffer)
+  conn.write(@buffer)
 
 proc ping(T: type[YamuxHeader], flag: MsgFlags, pingData: uint32): T =
   T(
@@ -106,11 +117,10 @@ proc goAway(T: type[YamuxHeader], status: GoAwayStatus): T =
   )
 
 proc data(
-  T: type[YamuxHeader],
-  streamId: uint32,
-  length: uint32 = 0,
-  flags: set[MsgFlags] = {},
-  ): T =
+    T: type[YamuxHeader],
+    streamId: uint32,
+    length: uint32 = 0,
+    flags: set[MsgFlags] = {}): T =
   T(
     version: YamuxVersion,
     msgType: MsgType.Data,
@@ -120,11 +130,10 @@ proc data(
   )
 
 proc windowUpdate(
-  T: type[YamuxHeader],
-  streamId: uint32,
-  delta: uint32,
-  flags: set[MsgFlags] = {},
-  ): T =
+    T: type[YamuxHeader],
+    streamId: uint32,
+    delta: uint32,
+    flags: set[MsgFlags] = {}): T =
   T(
     version: YamuxVersion,
     msgType: MsgType.WindowUpdate,
@@ -137,12 +146,13 @@ type
   ToSend = tuple
     data: seq[byte]
     sent: int
-    fut: Future[void]
+    fut: Future[void].Raising([CancelledError, LPStreamError])
   YamuxChannel* = ref object of Connection
     id: uint32
     recvWindow: int
     sendWindow: int
     maxRecvWindow: int
+    maxSendQueueSize: int
     conn: Connection
     isSrc: bool
     opened: bool
@@ -151,16 +161,15 @@ type
     recvQueue: seq[byte]
     isReset: bool
     remoteReset: bool
-    closedRemotely: Future[void]
+    closedRemotely: Future[void].Raising([])
     closedLocally: bool
     receivedData: AsyncEvent
-    returnedEof: bool
 
 proc `$`(channel: YamuxChannel): string =
   result = if channel.conn.dir == Out: "=> " else: "<= "
   result &= $channel.id
   var s: seq[string] = @[]
-  if channel.closedRemotely.done():
+  if channel.closedRemotely.completed():
     s.add("ClosedRemotely")
   if channel.closedLocally:
     s.add("ClosedLocally")
@@ -169,29 +178,44 @@ proc `$`(channel: YamuxChannel): string =
   if s.len > 0:
     result &= " {" & s.foldl(if a != "": a & ", " & b else: b, "") & "}"
 
-proc sendQueueBytes(channel: YamuxChannel, limit: bool = false): int =
-  for (elem, sent, _) in channel.sendQueue:
-    result.inc(min(elem.len - sent, if limit: channel.maxRecvWindow div 3 else: elem.len - sent))
+proc lengthSendQueue(channel: YamuxChannel): int =
+  ## Returns the length of what remains to be sent
+  ##
+  channel.sendQueue.foldl(a + b.data.len - b.sent, 0)
 
-proc actuallyClose(channel: YamuxChannel) {.async.} =
+proc lengthSendQueueWithLimit(channel: YamuxChannel): int =
+  ## Returns the length of what remains to be sent, but limit the size of big messages.
+  ##
+  # For leniency, limit big messages size to the third of maxSendQueueSize
+  # This value is arbitrary, it's not in the specs, it permits to store up to
+  # 3 big messages if the peer is stalling.
+  channel.sendQueue.foldl(a + min(b.data.len - b.sent, channel.maxSendQueueSize div 3), 0)
+
+proc actuallyClose(channel: YamuxChannel) {.async: (raises: []).} =
   if channel.closedLocally and channel.sendQueue.len == 0 and
-     channel.closedRemotely.done():
+     channel.closedRemotely.completed():
     await procCall Connection(channel).closeImpl()
 
-proc remoteClosed(channel: YamuxChannel) {.async.} =
-  if not channel.closedRemotely.done():
+proc remoteClosed(channel: YamuxChannel) {.async: (raises: []).} =
+  if not channel.closedRemotely.completed():
     channel.closedRemotely.complete()
     await channel.actuallyClose()
 
-method closeImpl*(channel: YamuxChannel) {.async, gcsafe.} =
+method closeImpl*(channel: YamuxChannel) {.async: (raises: []).} =
   if not channel.closedLocally:
+    trace "Closing yamux channel locally", streamId = channel.id, conn = channel.conn
     channel.closedLocally = true
 
-    if channel.isReset == false and channel.sendQueue.len == 0:
-      await channel.conn.write(YamuxHeader.data(channel.id, 0, {Fin}))
+    if not channel.isReset and channel.sendQueue.len == 0:
+      try: await channel.conn.write(YamuxHeader.data(channel.id, 0, {Fin}))
+      except CancelledError, LPStreamError: discard
     await channel.actuallyClose()
 
-proc reset(channel: YamuxChannel, isLocal: bool = false) {.async.} =
+proc reset(
+    channel: YamuxChannel, isLocal: bool = false) {.async: (raises: []).} =
+  # If we reset locally, we want to flush up to a maximum of recvWindow
+  # bytes. It's because the peer we're connected to can send us data before
+  # it receives the reset.
   if channel.isReset:
     return
   trace "Reset channel"
@@ -203,20 +227,24 @@ proc reset(channel: YamuxChannel, isLocal: bool = false) {.async.} =
   channel.recvQueue = @[]
   channel.sendWindow = 0
   if not channel.closedLocally:
-    if isLocal:
+    if isLocal and not channel.isSending:
       try: await channel.conn.write(YamuxHeader.data(channel.id, 0, {Rst}))
-      except LPStreamEOFError as exc: discard
-      except LPStreamClosedError as exc: discard
+      except CancelledError, LPStreamError: discard
     await channel.close()
-  if not channel.closedRemotely.done():
+  if not channel.closedRemotely.completed():
     await channel.remoteClosed()
   channel.receivedData.fire()
   if not isLocal:
-    # If we reset locally, we want to flush up to a maximum of recvWindow
-    # bytes. We use the recvWindow in the proc cleanupChann.
+    # If the reset is remote, there's no reason to flush anything.
     channel.recvWindow = 0
 
-proc updateRecvWindow(channel: YamuxChannel) {.async.} =
+proc updateRecvWindow(
+    channel: YamuxChannel
+) {.async: (raises: [CancelledError, LPStreamError]).} =
+  ## Send to the peer a window update when the recvWindow is empty enough
+  ##
+  # In order to avoid spamming a window update everytime a byte is read,
+  # we send it everytime half of the maxRecvWindow is read.
   let inWindow = channel.recvWindow + channel.recvQueue.len
   if inWindow > channel.maxRecvWindow div 2:
     return
@@ -230,31 +258,36 @@ proc updateRecvWindow(channel: YamuxChannel) {.async.} =
   trace "increasing the recvWindow", delta
 
 method readOnce*(
-  channel: YamuxChannel,
-  pbytes: pointer,
-  nbytes: int):
-  Future[int] {.async.} =
+    channel: YamuxChannel,
+    pbytes: pointer,
+    nbytes: int
+): Future[int] {.async: (raises: [CancelledError, LPStreamError]).} =
+  ## Read from a yamux channel
 
   if channel.isReset:
-    raise if channel.remoteReset:
+    raise
+      if channel.remoteReset:
         newLPStreamResetError()
       elif channel.closedLocally:
         newLPStreamClosedError()
       else:
         newLPStreamConnDownError()
-  if channel.returnedEof:
+  if channel.isEof:
     raise newLPStreamRemoteClosedError()
   if channel.recvQueue.len == 0:
     channel.receivedData.clear()
-    await channel.closedRemotely or channel.receivedData.wait()
-    if channel.closedRemotely.done() and channel.recvQueue.len == 0:
-      channel.returnedEof = true
-      return 0
+    try:  # https://github.com/status-im/nim-chronos/issues/516
+      discard await race(channel.closedRemotely, channel.receivedData.wait())
+    except ValueError: raiseAssert("Futures list is not empty")
+    if channel.closedRemotely.completed() and channel.recvQueue.len == 0:
+      channel.isEof = true
+      return 0 # we return 0 to indicate that the channel is closed for reading from now on
 
   let toRead = min(channel.recvQueue.len, nbytes)
 
   var p = cast[ptr UncheckedArray[byte]](pbytes)
-  toOpenArray(p, 0, nbytes - 1)[0..<toRead] = channel.recvQueue.toOpenArray(0, toRead - 1)
+  toOpenArray(p, 0, nbytes - 1)[0..<toRead] =
+    channel.recvQueue.toOpenArray(0, toRead - 1)
   channel.recvQueue = channel.recvQueue[toRead..^1]
 
   # We made some room in the recv buffer let the peer know
@@ -262,7 +295,9 @@ method readOnce*(
   channel.activity = true
   return toRead
 
-proc gotDataFromRemote(channel: YamuxChannel, b: seq[byte]) {.async.} =
+proc gotDataFromRemote(
+    channel: YamuxChannel,
+    b: seq[byte]) {.async: (raises: [CancelledError, LPStreamError]).} =
   channel.recvWindow -= b.len
   channel.recvQueue = channel.recvQueue.concat(b)
   channel.receivedData.fire()
@@ -273,26 +308,27 @@ proc gotDataFromRemote(channel: YamuxChannel, b: seq[byte]) {.async.} =
 proc setMaxRecvWindow*(channel: YamuxChannel, maxRecvWindow: int) =
   channel.maxRecvWindow = maxRecvWindow
 
-proc trySend(channel: YamuxChannel) {.async.} =
+proc trySend(
+    channel: YamuxChannel
+) {.async: (raises: [CancelledError, LPStreamError]).} =
   if channel.isSending:
     return
   channel.isSending = true
   defer: channel.isSending = false
+
   while channel.sendQueue.len != 0:
     channel.sendQueue.keepItIf(not (it.fut.cancelled() and it.sent == 0))
     if channel.sendWindow == 0:
-      trace "send window empty"
-      if channel.sendQueueBytes(true) > channel.maxRecvWindow:
-        debug "channel send queue too big, resetting", maxSendWindow=channel.maxRecvWindow,
-          currentQueueSize = channel.sendQueueBytes(true)
-        try:
-          await channel.reset(true)
-        except CatchableError as exc:
-          debug "failed to reset", msg=exc.msg
+      trace "trying to send while the sendWindow is empty"
+      if channel.lengthSendQueueWithLimit() > channel.maxSendQueueSize:
+        trace "channel send queue too big, resetting",
+          maxSendQueueSize = channel.maxSendQueueSize,
+          currentQueueSize = channel.lengthSendQueueWithLimit()
+        await channel.reset(isLocal = true)
       break
 
     let
-      bytesAvailable = channel.sendQueueBytes()
+      bytesAvailable = channel.lengthSendQueue()
       toSend = min(channel.sendWindow, bytesAvailable)
     var
       sendBuffer = newSeqUninitialized[byte](toSend + 12)
@@ -305,22 +341,33 @@ proc trySend(channel: YamuxChannel) {.async.} =
 
     sendBuffer[0..<12] = header.encode()
 
-    var futures: seq[Future[void]]
+    var futures: seq[Future[void].Raising([CancelledError, LPStreamError])]
     while inBuffer < toSend:
+      # concatenate the different message we try to send into one buffer
       let (data, sent, fut) = channel.sendQueue[0]
       let bufferToSend = min(data.len - sent, toSend - inBuffer)
       sendBuffer.toOpenArray(12, 12 + toSend - 1)[inBuffer..<(inBuffer+bufferToSend)] =
         channel.sendQueue[0].data.toOpenArray(sent, sent + bufferToSend - 1)
       channel.sendQueue[0].sent.inc(bufferToSend)
       if channel.sendQueue[0].sent >= data.len:
+        # if every byte of the message is in the buffer, add the write future to the
+        # sequence of futures to be completed (or failed) when the buffer is sent
         futures.add(fut)
         channel.sendQueue.delete(0)
       inBuffer.inc(bufferToSend)
 
-    trace "build send buffer", h = $header, msg=string.fromBytes(sendBuffer[12..^1])
+    trace "try to send the buffer", h = $header
     channel.sendWindow.dec(toSend)
-    try: await channel.conn.write(sendBuffer)
-    except CatchableError as exc:
+    try:
+      await channel.conn.write(sendBuffer)
+    except CancelledError:
+      trace "cancelled sending the buffer"
+      for fut in futures.items():
+        fut.cancelSoon()
+      await channel.reset()
+      break
+    except LPStreamError as exc:
+      trace "failed to send the buffer"
       let connDown = newLPStreamConnDownError(exc)
       for fut in futures.items():
         fut.fail(connDown)
@@ -330,7 +377,13 @@ proc trySend(channel: YamuxChannel) {.async.} =
       fut.complete()
     channel.activity = true
 
-method write*(channel: YamuxChannel, msg: seq[byte]): Future[void] =
+method write*(
+    channel: YamuxChannel,
+    msg: seq[byte]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
+  ## Write to yamux channel
+  ##
   result = newFuture[void]("Yamux Send")
   if channel.remoteReset:
     result.fail(newLPStreamResetError())
@@ -343,15 +396,22 @@ method write*(channel: YamuxChannel, msg: seq[byte]): Future[void] =
     return result
   channel.sendQueue.add((msg, 0, result))
   when defined(libp2p_yamux_metrics):
-    libp2p_yamux_recv_queue.observe(channel.sendQueueBytes().int64)
+    libp2p_yamux_send_queue.observe(channel.lengthSendQueue().int64)
   asyncSpawn channel.trySend()
 
-proc open*(channel: YamuxChannel) {.async, gcsafe.} =
+proc open(
+    channel: YamuxChannel
+) {.async: (raises: [CancelledError, LPStreamError]).} =
+  ## Open a yamux channel by sending a window update with Syn or Ack flag
+  ##
   if channel.opened:
     trace "Try to open channel twice"
     return
   channel.opened = true
-  await channel.conn.write(YamuxHeader.data(channel.id, 0, {if channel.isSrc: Syn else: Ack}))
+  await channel.conn.write(YamuxHeader.windowUpdate(
+    channel.id,
+    uint32(max(channel.maxRecvWindow - YamuxDefaultWindowSize, 0)),
+    {if channel.isSrc: Syn else: Ack}))
 
 method getWrapped*(channel: YamuxChannel): Connection = channel.conn
 
@@ -362,48 +422,73 @@ type
     currentId: uint32
     isClosed: bool
     maxChannCount: int
+    windowSize: int
+    maxSendQueueSize: int
+    inTimeout: Duration
+    outTimeout: Duration
 
 proc lenBySrc(m: Yamux, isSrc: bool): int =
   for v in m.channels.values():
     if v.isSrc == isSrc: result += 1
 
-proc cleanupChann(m: Yamux, channel: YamuxChannel) {.async.} =
-  await channel.join()
+proc cleanupChannel(m: Yamux, channel: YamuxChannel) {.async: (raises: []).} =
+  try:
+    await channel.join()
+  except CancelledError:
+    discard
   m.channels.del(channel.id)
   when defined(libp2p_yamux_metrics):
-    libp2p_yamux_channels.set(m.lenBySrc(channel.isSrc).int64, [$channel.isSrc, $channel.peerId])
+    libp2p_yamux_channels.set(
+      m.lenBySrc(channel.isSrc).int64, [$channel.isSrc, $channel.peerId])
   if channel.isReset and channel.recvWindow > 0:
     m.flushed[channel.id] = channel.recvWindow
 
-proc createStream(m: Yamux, id: uint32, isSrc: bool): YamuxChannel =
-  result = YamuxChannel(
+proc createStream(
+    m: Yamux, id: uint32, isSrc: bool,
+    recvWindow: int, maxSendQueueSize: int): YamuxChannel =
+  # During initialization, recvWindow can be larger than maxRecvWindow.
+  # This is because the peer we're connected to will always assume
+  # that the initial recvWindow is 256k.
+  # To solve this contradiction, no updateWindow will be sent until
+  # recvWindow is less than maxRecvWindow
+  proc newClosedRemotelyFut(): Future[void] {.async: (raises: [], raw: true).} =
+    newFuture[void]()
+  var stream = YamuxChannel(
     id: id,
-    maxRecvWindow: DefaultWindowSize,
-    recvWindow: DefaultWindowSize,
-    sendWindow: DefaultWindowSize,
+    maxRecvWindow: recvWindow,
+    recvWindow: if recvWindow > YamuxDefaultWindowSize: recvWindow else: YamuxDefaultWindowSize,
+    sendWindow: YamuxDefaultWindowSize,
+    maxSendQueueSize: maxSendQueueSize,
     isSrc: isSrc,
     conn: m.connection,
     receivedData: newAsyncEvent(),
-    closedRemotely: newFuture[void]()
+    closedRemotely: newClosedRemotelyFut()
   )
-  result.objName = "YamuxStream"
-  result.dir = if isSrc: Direction.Out else: Direction.In
-  result.timeoutHandler = proc(): Future[void] {.gcsafe.} =
-    trace "Idle timeout expired, resetting YamuxChannel"
-    result.reset()
-  result.initStream()
-  result.peerId = m.connection.peerId
-  result.observedAddr = m.connection.observedAddr
-  result.transportDir = m.connection.transportDir
+  stream.objName = "YamuxStream"
+  if isSrc:
+    stream.dir = Direction.Out
+    stream.timeout = m.outTimeout
+  else:
+    stream.dir = Direction.In
+    stream.timeout = m.inTimeout
+  stream.timeoutHandler =
+    proc(): Future[void] {.async: (raises: [], raw: true).} =
+      trace "Idle timeout expired, resetting YamuxChannel"
+      stream.reset(isLocal = true)
+  stream.initStream()
+  stream.peerId = m.connection.peerId
+  stream.observedAddr = m.connection.observedAddr
+  stream.transportDir = m.connection.transportDir
   when defined(libp2p_agents_metrics):
-    result.shortAgent = m.connection.shortAgent
-  m.channels[id] = result
-  asyncSpawn m.cleanupChann(result)
+    stream.shortAgent = m.connection.shortAgent
+  m.channels[id] = stream
+  asyncSpawn m.cleanupChannel(stream)
   trace "created channel", id, pid=m.connection.peerId
   when defined(libp2p_yamux_metrics):
-    libp2p_yamux_channels.set(m.lenBySrc(isSrc).int64, [$isSrc, $result.peerId])
+    libp2p_yamux_channels.set(m.lenBySrc(isSrc).int64, [$isSrc, $stream.peerId])
+  return stream
 
-method close*(m: Yamux) {.async.} =
+method close*(m: Yamux) {.async: (raises: []).} =
   if m.isClosed == true:
     trace "Already closed"
     return
@@ -412,24 +497,21 @@ method close*(m: Yamux) {.async.} =
   trace "Closing yamux"
   let channels = toSeq(m.channels.values())
   for channel in channels:
-    await channel.reset(true)
+    await channel.reset(isLocal = true)
   try: await m.connection.write(YamuxHeader.goAway(NormalTermination))
-  except CatchableError as exc: trace "failed to send goAway", msg=exc.msg
+  except CancelledError as exc: trace "cancelled sending goAway", msg = exc.msg
+  except LPStreamError as exc: trace "failed to send goAway", msg = exc.msg
   await m.connection.close()
   trace "Closed yamux"
 
-proc handleStream(m: Yamux, channel: YamuxChannel) {.async.} =
-  ## call the muxer stream handler for this channel
+proc handleStream(m: Yamux, channel: YamuxChannel) {.async: (raises: []).} =
+  ## Call the muxer stream handler for this channel
   ##
-  try:
-    await m.streamHandler(channel)
-    trace "finished handling stream"
-    doAssert(channel.isClosed, "connection not closed by handler!")
-  except CatchableError as exc:
-    trace "Exception in yamux stream handler", msg = exc.msg
-    await channel.reset()
+  await m.streamHandler(channel)
+  trace "finished handling stream"
+  doAssert(channel.isClosed, "connection not closed by handler!")
 
-method handle*(m: Yamux) {.async, gcsafe.} =
+method handle*(m: Yamux) {.async: (raises: []).} =
   trace "Starting yamux handler", pid=m.connection.peerId
   try:
     while not m.connection.atEof:
@@ -453,28 +535,39 @@ method handle*(m: Yamux) {.async, gcsafe.} =
           else:
             if header.streamId in m.flushed:
               m.flushed.del(header.streamId)
+
             if header.streamId mod 2 == m.currentId mod 2:
+              debug "Peer used our reserved stream id, skipping", id=header.streamId, currentId=m.currentId, peerId=m.connection.peerId
               raise newException(YamuxError, "Peer used our reserved stream id")
-            let newStream = m.createStream(header.streamId, false)
+            let newStream = m.createStream(header.streamId, false, m.windowSize, m.maxSendQueueSize)
             if m.channels.len >= m.maxChannCount:
               await newStream.reset()
               continue
             await newStream.open()
             asyncSpawn m.handleStream(newStream)
         elif header.streamId notin m.channels:
-          if header.streamId notin m.flushed:
-            raise newException(YamuxError, "Unknown stream ID: " & $header.streamId)
-          elif header.msgType == Data:
-            # Flush the data
-            m.flushed[header.streamId].dec(int(header.length))
-            if m.flushed[header.streamId] < 0:
-              raise newException(YamuxError, "Peer exhausted the recvWindow after reset")
-            if header.length > 0:
-              var buffer = newSeqUninitialized[byte](header.length)
-              await m.connection.readExactly(addr buffer[0], int(header.length))
+          # Flush the data
+          m.flushed.withValue(header.streamId, flushed):
+            if header.msgType == Data:
+              flushed[].dec(int(header.length))
+              if flushed[] < 0:
+                raise newException(YamuxError,
+                  "Peer exhausted the recvWindow after reset")
+              if header.length > 0:
+                var buffer = newSeqUninitialized[byte](header.length)
+                await m.connection.readExactly(
+                  addr buffer[0], int(header.length))
+          do:
+            raise newException(YamuxError,
+              "Unknown stream ID: " & $header.streamId)
           continue
 
-        let channel = m.channels[header.streamId]
+        let channel =
+          try:
+            m.channels[header.streamId]
+          except KeyError:
+            raise newException(YamuxError,
+              "Stream was cleaned up before handling data: " & $header.streamId)
 
         if header.msgType == WindowUpdate:
           channel.sendWindow += int(header.length)
@@ -487,7 +580,7 @@ method handle*(m: Yamux) {.async, gcsafe.} =
           if header.length > 0:
             var buffer = newSeqUninitialized[byte](header.length)
             await m.connection.readExactly(addr buffer[0], int(header.length))
-            trace "Msg Rcv", msg=string.fromBytes(buffer)
+            trace "Msg Rcv", msg=shortLog(buffer)
             await channel.gotDataFromRemote(buffer)
 
         if MsgFlags.Fin in header.flags:
@@ -496,11 +589,24 @@ method handle*(m: Yamux) {.async, gcsafe.} =
         if MsgFlags.Rst in header.flags:
           trace "remote reset channel"
           await channel.reset()
+  except CancelledError as exc:
+    debug "Unexpected cancellation in yamux handler", msg = exc.msg
   except LPStreamEOFError as exc:
     trace "Stream EOF", msg = exc.msg
+  except LPStreamError as exc:
+    debug "Unexpected stream exception in yamux read loop", msg = exc.msg
   except YamuxError as exc:
     trace "Closing yamux connection", error=exc.msg
-    await m.connection.write(YamuxHeader.goAway(ProtocolError))
+    try:
+      await m.connection.write(YamuxHeader.goAway(ProtocolError))
+    except CancelledError, LPStreamError:
+      discard
+  except MuxerError as exc:
+    debug "Unexpected muxer exception in yamux read loop", msg = exc.msg
+    try:
+      await m.connection.write(YamuxHeader.goAway(ProtocolError))
+    except CancelledError, LPStreamError:
+      discard
   finally:
     await m.close()
   trace "Stopped yamux handler"
@@ -509,21 +615,32 @@ method getStreams*(m: Yamux): seq[Connection] =
   for c in m.channels.values: result.add(c)
 
 method newStream*(
-  m: Yamux,
-  name: string = "",
-  lazy: bool = false): Future[Connection] {.async, gcsafe.} =
-
+    m: Yamux,
+    name: string = "",
+    lazy: bool = false
+): Future[Connection] {.async: (raises: [
+    CancelledError, LPStreamError, MuxerError]).} =
   if m.channels.len > m.maxChannCount - 1:
     raise newException(TooManyChannels, "max allowed channel count exceeded")
-  let stream = m.createStream(m.currentId, true)
+  let stream = m.createStream(m.currentId, true, m.windowSize, m.maxSendQueueSize)
   m.currentId += 2
   if not lazy:
     await stream.open()
   return stream
 
-proc new*(T: type[Yamux], conn: Connection, maxChannCount: int = MaxChannelCount): T =
+proc new*(
+    T: type[Yamux], conn: Connection,
+    maxChannCount: int = MaxChannelCount,
+    windowSize: int = YamuxDefaultWindowSize,
+    maxSendQueueSize: int = MaxSendQueueSize,
+    inTimeout: Duration = 5.minutes,
+    outTimeout: Duration = 5.minutes): T =
   T(
     connection: conn,
     currentId: if conn.dir == Out: 1 else: 2,
-    maxChannCount: maxChannCount
+    maxChannCount: maxChannCount,
+    windowSize: windowSize,
+    maxSendQueueSize: maxSendQueueSize,
+    inTimeout: inTimeout,
+    outTimeout: outTimeout
   )

libp2p/nameresolving/nameresolver.nim

@@ -52,7 +52,7 @@ proc resolveOneAddress(
   ma: MultiAddress,
   domain: Domain = Domain.AF_UNSPEC,
   prefix = ""): Future[seq[MultiAddress]]
-  {.async, raises: [MaError, TransportAddressError].} =
+  {.async.} =
   #Resolve a single address
   var pbuf: array[2, byte]
 

libp2p/protocols/connectivity/autonat/server.nim

@@ -140,7 +140,7 @@ proc handleDial(autonat: Autonat, conn: Connection, msg: AutonatMsg): Future[voi
 
 proc new*(T: typedesc[Autonat], switch: Switch, semSize: int = 1, dialTimeout = 15.seconds): T =
   let autonat = T(switch: switch, sem: newAsyncSemaphore(semSize), dialTimeout: dialTimeout)
-  proc handleStream(conn: Connection, proto: string) {.async, gcsafe.} =
+  proc handleStream(conn: Connection, proto: string) {.async.} =
     try:
       let msg = AutonatMsg.decode(await conn.readLp(1024)).valueOr:
         raise newException(AutonatError, "Received malformed message")

libp2p/protocols/connectivity/autonat/service.nim

@@ -162,7 +162,7 @@ proc schedule(service: AutonatService, switch: Switch, interval: Duration) {.asy
 proc addressMapper(
   self: AutonatService,
   peerStore: PeerStore,
-  listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.gcsafe, async.} =
+  listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.async.} =
 
   if self.networkReachability != NetworkReachability.Reachable:
     return listenAddrs
@@ -179,7 +179,7 @@ proc addressMapper(
   return addrs
 
 method setup*(self: AutonatService, switch: Switch): Future[bool] {.async.} =
-  self.addressMapper = proc (listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.gcsafe, async.} =
+  self.addressMapper = proc (listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.async.} =
     return await addressMapper(self, switch.peerStore, listenAddrs)
 
   info "Setting up AutonatService"

libp2p/protocols/connectivity/dcutr/client.nim

@@ -66,7 +66,7 @@ proc startSync*(self: DcutrClient, switch: Switch, remotePeerId: PeerId, addrs:
 
     if peerDialableAddrs.len > self.maxDialableAddrs:
         peerDialableAddrs = peerDialableAddrs[0..<self.maxDialableAddrs]
-    var futs = peerDialableAddrs.mapIt(switch.connect(stream.peerId, @[it], forceDial = true, reuseConnection = false))
+    var futs = peerDialableAddrs.mapIt(switch.connect(stream.peerId, @[it], forceDial = true, reuseConnection = false, dir = Direction.In))
     try:
       discard await anyCompleted(futs).wait(self.connectTimeout)
       debug "Dcutr initiator has directly connected to the remote peer."

libp2p/protocols/connectivity/dcutr/core.nim

@@ -56,5 +56,10 @@ proc send*(conn: Connection, msgType: MsgType, addrs: seq[MultiAddress]) {.async
   let pb = DcutrMsg(msgType: msgType, addrs: addrs).encode()
   await conn.writeLp(pb.buffer)
 
-proc getHolePunchableAddrs*(addrs: seq[MultiAddress]): seq[MultiAddress] =
-  addrs.filterIt(TCP.match(it))
+proc getHolePunchableAddrs*(addrs: seq[MultiAddress]): seq[MultiAddress] {.raises: [LPError]} =
+  var result = newSeq[MultiAddress]()
+  for a in addrs:
+      # This is necessary to also accept addrs like /ip4/198.51.100/tcp/1234/p2p/QmYyQSo1c1Ym7orWxLYvCrM2EmxFTANf8wXmmE7DWjhx5N
+      if [TCP, mapAnd(TCP_DNS, P2PPattern), mapAnd(TCP_IP, P2PPattern)].anyIt(it.match(a)):
+        result.add(a[0..1].tryGet())
+  return result

libp2p/protocols/connectivity/dcutr/server.nim

@@ -19,7 +19,6 @@ import ../../protocol,
        ../../../switch,
        ../../../utils/future
 
-export DcutrError
 export chronicles
 
 type Dcutr* = ref object of LPProtocol
@@ -29,7 +28,7 @@ logScope:
 
 proc new*(T: typedesc[Dcutr], switch: Switch, connectTimeout = 15.seconds, maxDialableAddrs = 8): T =
 
-  proc handleStream(stream: Connection, proto: string) {.async, gcsafe.} =
+  proc handleStream(stream: Connection, proto: string) {.async.} =
     var peerDialableAddrs: seq[MultiAddress]
     try:
       let connectMsg = DcutrMsg.decode(await stream.readLp(1024))
@@ -56,7 +55,7 @@ proc new*(T: typedesc[Dcutr], switch: Switch, connectTimeout = 15.seconds, maxDi
 
       if peerDialableAddrs.len > maxDialableAddrs:
         peerDialableAddrs = peerDialableAddrs[0..<maxDialableAddrs]
-      var futs = peerDialableAddrs.mapIt(switch.connect(stream.peerId, @[it], forceDial = true, reuseConnection = false, upgradeDir = Direction.In))
+      var futs = peerDialableAddrs.mapIt(switch.connect(stream.peerId, @[it], forceDial = true, reuseConnection = false, dir = Direction.Out))
       try:
         discard await anyCompleted(futs).wait(connectTimeout)
         debug "Dcutr receiver has directly connected to the remote peer."
@@ -65,14 +64,14 @@ proc new*(T: typedesc[Dcutr], switch: Switch, connectTimeout = 15.seconds, maxDi
     except CancelledError as err:
       raise err
     except AllFuturesFailedError as err:
-      debug "Dcutr receiver could not connect to the remote peer, all connect attempts failed", peerDialableAddrs, msg = err.msg
-      raise newException(DcutrError, "Dcutr receiver could not connect to the remote peer, all connect attempts failed", err)
+      debug "Dcutr receiver could not connect to the remote peer, " &
+        "all connect attempts failed", peerDialableAddrs, msg = err.msg
     except AsyncTimeoutError as err:
-      debug "Dcutr receiver could not connect to the remote peer, all connect attempts timed out", peerDialableAddrs, msg = err.msg
-      raise newException(DcutrError, "Dcutr receiver could not connect to the remote peer, all connect attempts timed out", err)
+      debug "Dcutr receiver could not connect to the remote peer, " &
+        "all connect attempts timed out", peerDialableAddrs, msg = err.msg
     except CatchableError as err:
-      warn "Unexpected error when Dcutr receiver tried to connect to the remote peer", msg = err.msg
-      raise newException(DcutrError, "Unexpected error when Dcutr receiver tried to connect to the remote peer", err)
+      warn "Unexpected error when Dcutr receiver tried to connect " &
+        "to the remote peer", msg = err.msg
 
   let self = T()
   self.handler = handleStream

libp2p/protocols/connectivity/relay/client.nim

@@ -189,7 +189,7 @@ proc dialPeerV2*(
   conn.limitData = msgRcvFromRelay.limit.data
   return conn
 
-proc handleStopStreamV2(cl: RelayClient, conn: Connection) {.async, gcsafe.} =
+proc handleStopStreamV2(cl: RelayClient, conn: Connection) {.async.} =
   let msg = StopMessage.decode(await conn.readLp(RelayClientMsgSize)).valueOr:
     await sendHopStatus(conn, MalformedMessage)
     return
@@ -201,7 +201,7 @@ proc handleStopStreamV2(cl: RelayClient, conn: Connection) {.async, gcsafe.} =
     trace "Unexpected client / relayv2 handshake", msgType=msg.msgType
     await sendStopError(conn, MalformedMessage)
 
-proc handleStop(cl: RelayClient, conn: Connection, msg: RelayMessage) {.async, gcsafe.} =
+proc handleStop(cl: RelayClient, conn: Connection, msg: RelayMessage) {.async.} =
   let src = msg.srcPeer.valueOr:
     await sendStatus(conn, StatusV1.StopSrcMultiaddrInvalid)
     return
@@ -226,7 +226,7 @@ proc handleStop(cl: RelayClient, conn: Connection, msg: RelayMessage) {.async, g
   if cl.onNewConnection != nil: await cl.onNewConnection(conn, 0, 0)
   else: await conn.close()
 
-proc handleStreamV1(cl: RelayClient, conn: Connection) {.async, gcsafe.} =
+proc handleStreamV1(cl: RelayClient, conn: Connection) {.async.} =
   let msg = RelayMessage.decode(await conn.readLp(RelayClientMsgSize)).valueOr:
     await sendStatus(conn, StatusV1.MalformedMessage)
     return
@@ -266,7 +266,7 @@ proc new*(T: typedesc[RelayClient], canHop: bool = false,
              maxCircuitPerPeer: maxCircuitPerPeer,
              msgSize: msgSize,
              isCircuitRelayV1: circuitRelayV1)
-  proc handleStream(conn: Connection, proto: string) {.async, gcsafe.} =
+  proc handleStream(conn: Connection, proto: string) {.async.} =
     try:
       case proto:
       of RelayV1Codec: await cl.handleStreamV1(conn)

libp2p/protocols/connectivity/relay/rconn.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -23,11 +23,15 @@ type
 method readOnce*(
     self: RelayConnection,
     pbytes: pointer,
-    nbytes: int): Future[int] {.async.} =
+    nbytes: int
+): Future[int] {.async: (raises: [CancelledError, LPStreamError], raw: true).} =
   self.activity = true
-  return await self.conn.readOnce(pbytes, nbytes)
+  self.conn.readOnce(pbytes, nbytes)
 
-method write*(self: RelayConnection, msg: seq[byte]): Future[void] {.async.} =
+method write*(
+    self: RelayConnection,
+    msg: seq[byte]
+): Future[void] {.async: (raises: [CancelledError, LPStreamError]).} =
   self.dataSent.inc(msg.len)
   if self.limitData != 0 and self.dataSent > self.limitData:
     await self.close()
@@ -35,24 +39,25 @@ method write*(self: RelayConnection, msg: seq[byte]): Future[void] {.async.} =
   self.activity = true
   await self.conn.write(msg)
 
-method closeImpl*(self: RelayConnection): Future[void] {.async.} =
+method closeImpl*(self: RelayConnection): Future[void] {.async: (raises: []).} =
   await self.conn.close()
   await procCall Connection(self).closeImpl()
 
 method getWrapped*(self: RelayConnection): Connection = self.conn
 
 proc new*(
-  T: typedesc[RelayConnection],
-  conn: Connection,
-  limitDuration: uint32,
-  limitData: uint64): T =
+    T: typedesc[RelayConnection],
+    conn: Connection,
+    limitDuration: uint32,
+    limitData: uint64): T =
   let rc = T(conn: conn, limitDuration: limitDuration, limitData: limitData)
+  rc.dir = conn.dir
   rc.initStream()
   if limitDuration > 0:
-    proc checkDurationConnection() {.async.} =
-      let sleep = sleepAsync(limitDuration.seconds())
-      await sleep or conn.join()
-      if sleep.finished: await conn.close()
-      else: sleep.cancel()
+    proc checkDurationConnection() {.async: (raises: []).} =
+      try:
+        await noCancel conn.join().wait(limitDuration.seconds())
+      except AsyncTimeoutError:
+        await conn.close()
     asyncSpawn checkDurationConnection()
   return rc

libp2p/protocols/connectivity/relay/relay.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -105,7 +105,7 @@ proc isRelayed*(conn: Connection): bool =
     wrappedConn = wrappedConn.getWrapped()
   return false
 
-proc handleReserve(r: Relay, conn: Connection) {.async, gcsafe.} =
+proc handleReserve(r: Relay, conn: Connection) {.async.} =
   if conn.isRelayed():
     trace "reservation attempt over relay connection", pid = conn.peerId
     await sendHopStatus(conn, PermissionDenied)
@@ -128,7 +128,7 @@ proc handleReserve(r: Relay, conn: Connection) {.async, gcsafe.} =
 
 proc handleConnect(r: Relay,
                    connSrc: Connection,
-                   msg: HopMessage) {.async, gcsafe.} =
+                   msg: HopMessage) {.async.} =
   if connSrc.isRelayed():
     trace "connection attempt over relay connection"
     await sendHopStatus(connSrc, PermissionDenied)
@@ -200,7 +200,7 @@ proc handleConnect(r: Relay,
     await rconnDst.close()
   await bridge(rconnSrc, rconnDst)
 
-proc handleHopStreamV2*(r: Relay, conn: Connection) {.async, gcsafe.} =
+proc handleHopStreamV2*(r: Relay, conn: Connection) {.async.} =
   let msg = HopMessage.decode(await conn.readLp(r.msgSize)).valueOr:
     await sendHopStatus(conn, MalformedMessage)
     return
@@ -214,7 +214,7 @@ proc handleHopStreamV2*(r: Relay, conn: Connection) {.async, gcsafe.} =
 
 # Relay V1
 
-proc handleHop*(r: Relay, connSrc: Connection, msg: RelayMessage) {.async, gcsafe.} =
+proc handleHop*(r: Relay, connSrc: Connection, msg: RelayMessage) {.async.} =
   r.streamCount.inc()
   defer: r.streamCount.dec()
   if r.streamCount + r.rsvp.len() >= r.maxCircuit:
@@ -293,7 +293,7 @@ proc handleHop*(r: Relay, connSrc: Connection, msg: RelayMessage) {.async, gcsaf
   trace "relaying connection", src, dst
   await bridge(connSrc, connDst)
 
-proc handleStreamV1(r: Relay, conn: Connection) {.async, gcsafe.} =
+proc handleStreamV1(r: Relay, conn: Connection) {.async.} =
   let msg = RelayMessage.decode(await conn.readLp(r.msgSize)).valueOr:
     await sendStatus(conn, StatusV1.MalformedMessage)
     return
@@ -336,7 +336,7 @@ proc new*(T: typedesc[Relay],
             msgSize: msgSize,
             isCircuitRelayV1: circuitRelayV1)
 
-  proc handleStream(conn: Connection, proto: string) {.async, gcsafe.} =
+  proc handleStream(conn: Connection, proto: string) {.async.} =
     try:
       case proto:
       of RelayV2HopCodec: await r.handleHopStreamV2(conn)
@@ -361,17 +361,25 @@ proc deletesReservation(r: Relay) {.async.} =
       if n > r.rsvp[k]:
         r.rsvp.del(k)
 
-method start*(r: Relay) {.async.} =
+method start*(
+    r: Relay
+): Future[void] {.async: (raises: [CancelledError], raw: true).} =
+  let fut = newFuture[void]()
+  fut.complete()
   if not r.reservationLoop.isNil:
     warn "Starting relay twice"
-    return
+    return fut
   r.reservationLoop = r.deletesReservation()
   r.started = true
+  fut
 
-method stop*(r: Relay) {.async.} =
+method stop*(r: Relay): Future[void] {.async: (raises: [], raw: true).} =
+  let fut = newFuture[void]()
+  fut.complete()
   if r.reservationLoop.isNil:
     warn "Stopping relay without starting it"
-    return
+    return fut
   r.started = false
   r.reservationLoop.cancel()
   r.reservationLoop = nil
+  fut

libp2p/protocols/connectivity/relay/rtransport.nim

@@ -37,24 +37,24 @@ method start*(self: RelayTransport, ma: seq[MultiAddress]) {.async.} =
   self.client.onNewConnection = proc(
     conn: Connection,
     duration: uint32 = 0,
-    data: uint64 = 0) {.async, gcsafe, raises: [].} =
+    data: uint64 = 0) {.async.} =
       await self.queue.addLast(RelayConnection.new(conn, duration, data))
       await conn.join()
   self.selfRunning = true
   await procCall Transport(self).start(ma)
   trace "Starting Relay transport"
 
-method stop*(self: RelayTransport) {.async, gcsafe.} =
+method stop*(self: RelayTransport) {.async.} =
   self.running = false
   self.selfRunning = false
   self.client.onNewConnection = nil
   while not self.queue.empty():
     await self.queue.popFirstNoWait().close()
 
-method accept*(self: RelayTransport): Future[Connection] {.async, gcsafe.} =
+method accept*(self: RelayTransport): Future[Connection] {.async.} =
   result = await self.queue.popFirst()
 
-proc dial*(self: RelayTransport, ma: MultiAddress): Future[Connection] {.async, gcsafe.} =
+proc dial*(self: RelayTransport, ma: MultiAddress): Future[Connection] {.async.} =
   let
     sma = toSeq(ma.items())
     relayAddrs = sma[0..sma.len-4].mapIt(it.tryGet()).foldl(a & b)
@@ -90,7 +90,7 @@ method dial*(
   self: RelayTransport,
   hostname: string,
   ma: MultiAddress,
-  peerId: Opt[PeerId] = Opt.none(PeerId)): Future[Connection] {.async, gcsafe.} =
+  peerId: Opt[PeerId] = Opt.none(PeerId)): Future[Connection] {.async.} =
   peerId.withValue(pid):
     let address = MultiAddress.init($ma & "/p2p/" & $pid).tryGet()
     result = await self.dial(address)

libp2p/protocols/connectivity/relay/utils.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -21,63 +21,77 @@ const
   RelayV2HopCodec* = "/libp2p/circuit/relay/0.2.0/hop"
   RelayV2StopCodec* = "/libp2p/circuit/relay/0.2.0/stop"
 
-proc sendStatus*(conn: Connection, code: StatusV1) {.async, gcsafe.} =
+proc sendStatus*(
+    conn: Connection,
+    code: StatusV1
+) {.async: (raises: [CancelledError, LPStreamError], raw: true).} =
   trace "send relay/v1 status", status = $code & "(" & $ord(code) & ")"
   let
-    msg = RelayMessage(msgType: Opt.some(RelayType.Status), status: Opt.some(code))
+    msg = RelayMessage(
+      msgType: Opt.some(RelayType.Status), status: Opt.some(code))
     pb = encode(msg)
-  await conn.writeLp(pb.buffer)
+  conn.writeLp(pb.buffer)
 
-proc sendHopStatus*(conn: Connection, code: StatusV2) {.async, gcsafe.} =
+proc sendHopStatus*(
+    conn: Connection,
+    code: StatusV2
+) {.async: (raises: [CancelledError, LPStreamError], raw: true).} =
   trace "send hop relay/v2 status", status = $code & "(" & $ord(code) & ")"
   let
     msg = HopMessage(msgType: HopMessageType.Status, status: Opt.some(code))
     pb = encode(msg)
-  await conn.writeLp(pb.buffer)
+  conn.writeLp(pb.buffer)
 
-proc sendStopStatus*(conn: Connection, code: StatusV2) {.async.} =
+proc sendStopStatus*(
+    conn: Connection,
+    code: StatusV2
+) {.async: (raises: [CancelledError, LPStreamError], raw: true).} =
   trace "send stop relay/v2 status", status = $code & " (" & $ord(code) & ")"
   let
     msg = StopMessage(msgType: StopMessageType.Status, status: Opt.some(code))
     pb = encode(msg)
-  await conn.writeLp(pb.buffer)
+  conn.writeLp(pb.buffer)
 
-proc bridge*(connSrc: Connection, connDst: Connection) {.async.} =
+proc bridge*(
+    connSrc: Connection,
+    connDst: Connection) {.async: (raises: [CancelledError]).} =
   const bufferSize = 4096
   var
     bufSrcToDst: array[bufferSize, byte]
     bufDstToSrc: array[bufferSize, byte]
-    futSrc = connSrc.readOnce(addr bufSrcToDst[0], bufSrcToDst.high + 1)
-    futDst = connDst.readOnce(addr bufDstToSrc[0], bufDstToSrc.high + 1)
-    bytesSendFromSrcToDst = 0
-    bytesSendFromDstToSrc = 0
+    futSrc = connSrc.readOnce(addr bufSrcToDst[0], bufSrcToDst.len)
+    futDst = connDst.readOnce(addr bufDstToSrc[0], bufDstToSrc.len)
+    bytesSentFromSrcToDst = 0
+    bytesSentFromDstToSrc = 0
     bufRead: int
 
   try:
     while not connSrc.closed() and not connDst.closed():
-      await futSrc or futDst
+      try:  # https://github.com/status-im/nim-chronos/issues/516
+        discard await race(futSrc, futDst)
+      except ValueError: raiseAssert("Futures list is not empty")
       if futSrc.finished():
         bufRead = await futSrc
         if bufRead > 0:
-          bytesSendFromSrcToDst.inc(bufRead)
-          await connDst.write(@bufSrcToDst[0..<bufRead])
-          zeroMem(addr(bufSrcToDst), bufSrcToDst.high + 1)
-        futSrc = connSrc.readOnce(addr bufSrcToDst[0], bufSrcToDst.high + 1)
+          bytesSentFromSrcToDst.inc(bufRead)
+          await connDst.write(@bufSrcToDst[0 ..< bufRead])
+          zeroMem(addr bufSrcToDst[0], bufSrcToDst.len)
+        futSrc = connSrc.readOnce(addr bufSrcToDst[0], bufSrcToDst.len)
       if futDst.finished():
         bufRead = await futDst
         if bufRead > 0:
-          bytesSendFromDstToSrc += bufRead
-          await connSrc.write(bufDstToSrc[0..<bufRead])
-          zeroMem(addr(bufDstToSrc), bufDstToSrc.high + 1)
-        futDst = connDst.readOnce(addr bufDstToSrc[0], bufDstToSrc.high + 1)
+          bytesSentFromDstToSrc += bufRead
+          await connSrc.write(bufDstToSrc[0 ..< bufRead])
+          zeroMem(addr bufDstToSrc[0], bufDstToSrc.len)
+        futDst = connDst.readOnce(addr bufDstToSrc[0], bufDstToSrc.len)
   except CancelledError as exc:
     raise exc
-  except CatchableError as exc:
+  except LPStreamError as exc:
     if connSrc.closed() or connSrc.atEof():
       trace "relay src closed connection", src = connSrc.peerId
     if connDst.closed() or connDst.atEof():
       trace "relay dst closed connection", dst = connDst.peerId
     trace "relay error", exc=exc.msg
-  trace "end relaying", bytesSendFromSrcToDst, bytesSendFromDstToSrc
+  trace "end relaying", bytesSentFromSrcToDst, bytesSentFromDstToSrc
   await futSrc.cancelAndWait()
   await futDst.cancelAndWait()

libp2p/protocols/identify.nim

@@ -21,6 +21,7 @@ import ../protobuf/minprotobuf,
        ../peerid,
        ../crypto/crypto,
        ../multiaddress,
+       ../multicodec,
        ../protocols/protocol,
        ../utility,
        ../errors,
@@ -77,7 +78,7 @@ chronicles.expandIt(IdentifyInfo):
   signedPeerRecord =
     # The SPR contains the same data as the identify message
     # would be cumbersome to log
-    if iinfo.signedPeerRecord.isSome(): "Some"
+    if it.signedPeerRecord.isSome(): "Some"
     else: "None"
 
 proc encodeMsg(peerInfo: PeerInfo, observedAddr: Opt[MultiAddress], sendSpr: bool): ProtoBuffer
@@ -133,24 +134,24 @@ proc decodeMsg*(buf: seq[byte]): Opt[IdentifyInfo] =
   if ? pb.getField(6, agentVersion).toOpt():
     iinfo.agentVersion = some(agentVersion)
 
-  debug "decodeMsg: decoded identify", iinfo
   Opt.some(iinfo)
 
 proc new*(
   T: typedesc[Identify],
   peerInfo: PeerInfo,
-  sendSignedPeerRecord = false
+  sendSignedPeerRecord = false,
+  observedAddrManager = ObservedAddrManager.new(),
   ): T =
   let identify = T(
     peerInfo: peerInfo,
     sendSignedPeerRecord: sendSignedPeerRecord,
-    observedAddrManager: ObservedAddrManager.new(),
+    observedAddrManager: observedAddrManager,
   )
   identify.init()
   identify
 
 method init*(p: Identify) =
-  proc handle(conn: Connection, proto: string) {.async, gcsafe, closure.} =
+  proc handle(conn: Connection, proto: string) {.async.} =
     try:
       trace "handling identify request", conn
       var pb = encodeMsg(p.peerInfo, conn.observedAddr, p.sendSignedPeerRecord)
@@ -168,7 +169,7 @@ method init*(p: Identify) =
 
 proc identify*(self: Identify,
                conn: Connection,
-               remotePeerId: PeerId): Future[IdentifyInfo] {.async, gcsafe.} =
+               remotePeerId: PeerId): Future[IdentifyInfo] {.async.} =
   trace "initiating identify", conn
   var message = await conn.readLp(64*1024)
   if len(message) == 0:
@@ -176,6 +177,7 @@ proc identify*(self: Identify,
     raise newException(IdentityInvalidMsgError, "Empty message received!")
 
   var info = decodeMsg(message).valueOr: raise newException(IdentityInvalidMsgError, "Incorrect message received!")
+  debug "identify: decoded message", conn, info
   let
     pubkey = info.pubkey.valueOr: raise newException(IdentityInvalidMsgError, "No pubkey in identify")
     peer = PeerId.init(pubkey).valueOr: raise newException(IdentityInvalidMsgError, $error)
@@ -186,8 +188,12 @@ proc identify*(self: Identify,
   info.peerId = peer
 
   info.observedAddr.withValue(observed):
-    if not self.observedAddrManager.addObservation(observed):
-      debug "Observed address is not valid", observedAddr = observed
+    # Currently, we use the ObservedAddrManager only to find our dialable external NAT address. Therefore, addresses
+    # like "...\p2p-circuit\p2p\..." and "\p2p\..." are not useful to us.
+    if observed.contains(multiCodec("p2p-circuit")).get(false) or P2PPattern.matchPartial(observed):
+      trace "Not adding address to ObservedAddrManager.", observed
+    elif not self.observedAddrManager.addObservation(observed):
+      trace "Observed address is not valid.", observedAddr = observed
   return info
 
 proc new*(T: typedesc[IdentifyPush], handler: IdentifyPushHandler = nil): T {.public.} =
@@ -198,13 +204,14 @@ proc new*(T: typedesc[IdentifyPush], handler: IdentifyPushHandler = nil): T {.pu
   identifypush
 
 proc init*(p: IdentifyPush) =
-  proc handle(conn: Connection, proto: string) {.async, gcsafe, closure.} =
+  proc handle(conn: Connection, proto: string) {.async.} =
     trace "handling identify push", conn
     try:
       var message = await conn.readLp(64*1024)
 
       var identInfo = decodeMsg(message).valueOr:
         raise newException(IdentityInvalidMsgError, "Incorrect message received!")
+      debug "identify push: decoded message", conn, identInfo
 
       identInfo.pubkey.withValue(pubkey):
         let receivedPeerId = PeerId.init(pubkey).tryGet()

libp2p/protocols/perf/server.nim

@@ -27,7 +27,7 @@ type Perf* = ref object of LPProtocol
 
 proc new*(T: typedesc[Perf]): T {.public.} =
   var p = T()
-  proc handle(conn: Connection, proto: string) {.async, gcsafe, closure.} =
+  proc handle(conn: Connection, proto: string) {.async.} =
     var bytesRead = 0
     try:
       trace "Received benchmark performance check", conn

libp2p/protocols/ping.nim

@@ -51,12 +51,12 @@ proc new*(T: typedesc[Ping], handler: PingHandler = nil, rng: ref HmacDrbgContex
   ping
 
 method init*(p: Ping) =
-  proc handle(conn: Connection, proto: string) {.async, gcsafe, closure.} =
+  proc handle(conn: Connection, proto: string) {.async.} =
     try:
       trace "handling ping", conn
       var buf: array[PingSize, byte]
       await conn.readExactly(addr buf[0], PingSize)
-      trace "echoing ping", conn
+      trace "echoing ping", conn, pingData = @buf
       await conn.write(@buf)
       if not isNil(p.pingHandler):
         await p.pingHandler(conn.peerId)
@@ -71,7 +71,7 @@ method init*(p: Ping) =
 proc ping*(
   p: Ping,
   conn: Connection,
-  ): Future[Duration] {.async, gcsafe, public.} =
+  ): Future[Duration] {.async, public.} =
   ## Sends ping to `conn`, returns the delay
 
   trace "initiating ping", conn

libp2p/protocols/protocol.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -19,20 +19,29 @@ const
 
 type
   LPProtoHandler* = proc (
-    conn: Connection,
-    proto: string):
-    Future[void]
-    {.gcsafe, raises: [].}
+      conn: Connection,
+      proto: string): Future[void] {.async.}
 
   LPProtocol* = ref object of RootObj
     codecs*: seq[string]
-    handler*: LPProtoHandler ## this handler gets invoked by the protocol negotiator
+    handlerImpl: LPProtoHandler  ## invoked by the protocol negotiator
     started*: bool
     maxIncomingStreams: Opt[int]
 
 method init*(p: LPProtocol) {.base, gcsafe.} = discard
-method start*(p: LPProtocol) {.async, base.} = p.started = true
-method stop*(p: LPProtocol) {.async, base.} = p.started = false
+
+method start*(
+    p: LPProtocol) {.async: (raises: [CancelledError], raw: true), base.} =
+  let fut = newFuture[void]()
+  fut.complete()
+  p.started = true
+  fut
+
+method stop*(p: LPProtocol) {.async: (raises: [], raw: true), base.} =
+  let fut = newFuture[void]()
+  fut.complete()
+  p.started = false
+  fut
 
 proc maxIncomingStreams*(p: LPProtocol): int =
   p.maxIncomingStreams.get(DefaultMaxIncomingStreams)
@@ -41,7 +50,7 @@ proc `maxIncomingStreams=`*(p: LPProtocol, val: int) =
   p.maxIncomingStreams = Opt.some(val)
 
 func codec*(p: LPProtocol): string =
-  assert(p.codecs.len > 0, "Codecs sequence was empty!")
+  doAssert(p.codecs.len > 0, "Codecs sequence was empty!")
   p.codecs[0]
 
 func `codec=`*(p: LPProtocol, codec: string) =
@@ -49,15 +58,51 @@ func `codec=`*(p: LPProtocol, codec: string) =
   # if we use this abstraction
   p.codecs.insert(codec, 0)
 
+template `handler`*(p: LPProtocol): LPProtoHandler =
+  p.handlerImpl
+
+template `handler`*(
+    p: LPProtocol, conn: Connection, proto: string): Future[void] =
+  p.handlerImpl(conn, proto)
+
+func `handler=`*(p: LPProtocol, handler: LPProtoHandler) =
+  p.handlerImpl = handler
+
+# Callbacks that are annotated with `{.async: (raises).}` explicitly
+# document the types of errors that they may raise, but are not compatible
+# with `LPProtoHandler` and need to use a custom `proc` type.
+# They are internally wrapped into a `LPProtoHandler`, but still allow the
+# compiler to check that their `{.async: (raises).}` annotation is correct.
+# https://github.com/nim-lang/Nim/issues/23432
+func `handler=`*[E](
+    p: LPProtocol,
+    handler: proc (
+      conn: Connection,
+      proto: string): InternalRaisesFuture[void, E]) =
+  proc wrap(conn: Connection, proto: string): Future[void] {.async.} =
+    await handler(conn, proto)
+  p.handlerImpl = wrap
+
 proc new*(
-  T: type LPProtocol,
-  codecs: seq[string],
-  handler: LPProtoHandler,
-  maxIncomingStreams: Opt[int] | int = Opt.none(int)): T =
+    T: type LPProtocol,
+    codecs: seq[string],
+    handler: LPProtoHandler,
+    maxIncomingStreams: Opt[int] | int = Opt.none(int)): T =
   T(
     codecs: codecs,
-    handler: handler,
+    handlerImpl: handler,
     maxIncomingStreams:
       when maxIncomingStreams is int: Opt.some(maxIncomingStreams)
       else: maxIncomingStreams
   )
+
+proc new*[E](
+    T: type LPProtocol,
+    codecs: seq[string],
+    handler: proc (
+      conn: Connection,
+      proto: string): InternalRaisesFuture[void, E],
+    maxIncomingStreams: Opt[int] | int = Opt.none(int)): T =
+  proc wrap(conn: Connection, proto: string): Future[void] {.async.} =
+    await handler(conn, proto)
+  T.new(codec, wrap, maxIncomingStreams)

libp2p/protocols/pubsub/floodsub.nim

@@ -16,6 +16,7 @@ import ./pubsub,
        ./timedcache,
        ./peertable,
        ./rpc/[message, messages, protobuf],
+       nimcrypto/[hash, sha2],
        ../../crypto/crypto,
        ../../stream/connection,
        ../../peerid,
@@ -32,25 +33,34 @@ const FloodSubCodec* = "/floodsub/1.0.0"
 type
   FloodSub* {.public.} = ref object of PubSub
     floodsub*: PeerTable      # topic to remote peer map
-    seen*: TimedCache[MessageId] # message id:s already seen on the network
-    seenSalt*: seq[byte]
+    seen*: TimedCache[SaltedId]
+      # Early filter for messages recently observed on the network
+      # We use a salted id because the messages in this cache have not yet
+      # been validated meaning that an attacker has greater control over the
+      # hash key and therefore could poison the table
+    seenSalt*: sha256
+      # The salt in this case is a partially updated SHA256 context pre-seeded
+      # with some random data
 
-proc hasSeen*(f: FloodSub, msgId: MessageId): bool =
-  f.seenSalt & msgId in f.seen
+proc salt*(f: FloodSub, msgId: MessageId): SaltedId =
+  var tmp = f.seenSalt
+  tmp.update(msgId)
+  SaltedId(data: tmp.finish())
 
-proc addSeen*(f: FloodSub, msgId: MessageId): bool =
-  # Salting the seen hash helps avoid attacks against the hash function used
-  # in the nim hash table
+proc hasSeen*(f: FloodSub, saltedId: SaltedId): bool =
+  saltedId in f.seen
+
+proc addSeen*(f: FloodSub, saltedId: SaltedId): bool =
   # Return true if the message has already been seen
-  f.seen.put(f.seenSalt & msgId)
+  f.seen.put(saltedId)
 
-proc firstSeen*(f: FloodSub, msgId: MessageId): Moment =
-  f.seen.addedAt(f.seenSalt & msgId)
+proc firstSeen*(f: FloodSub, saltedId: SaltedId): Moment =
+  f.seen.addedAt(saltedId)
 
-proc handleSubscribe*(f: FloodSub,
-                      peer: PubSubPeer,
-                      topic: string,
-                      subscribe: bool) =
+proc handleSubscribe(f: FloodSub,
+                     peer: PubSubPeer,
+                     topic: string,
+                     subscribe: bool) =
   logScope:
     peer
     topic
@@ -96,10 +106,9 @@ method unsubscribePeer*(f: FloodSub, peer: PeerId) =
 method rpcHandler*(f: FloodSub,
                    peer: PubSubPeer,
                    data: seq[byte]) {.async.} =
-
   var rpcMsg = decodeRpcMsg(data).valueOr:
     debug "failed to decode msg from peer", peer, err = error
-    raise newException(CatchableError, "")
+    raise newException(CatchableError, "Peer msg couldn't be decoded")
 
   trace "decoded msg from peer", peer, msg = rpcMsg.shortLog
   # trigger hooks
@@ -117,9 +126,11 @@ method rpcHandler*(f: FloodSub,
       # TODO: descore peers due to error during message validation (malicious?)
       continue
 
-    let msgId = msgIdResult.get
+    let
+      msgId = msgIdResult.get
+      saltedId = f.salt(msgId)
 
-    if f.addSeen(msgId):
+    if f.addSeen(saltedId):
       trace "Dropping already-seen message", msgId, peer
       continue
 
@@ -148,16 +159,19 @@ method rpcHandler*(f: FloodSub,
       discard
 
     var toSendPeers = initHashSet[PubSubPeer]()
-    for t in msg.topicIds:                     # for every topic in the message
-      if t notin f.topics:
-        continue
-      f.floodsub.withValue(t, peers): toSendPeers.incl(peers[])
+    let topic = msg.topic
+    if topic notin f.topics:
+      debug "Dropping message due to topic not in floodsub topics", topic, msgId, peer
+      continue
 
-      await handleData(f, t, msg.data)
+    f.floodsub.withValue(topic, peers):
+      toSendPeers.incl(peers[])
+
+    await handleData(f, topic, msg.data)
 
     # In theory, if topics are the same in all messages, we could batch - we'd
     # also have to be careful to only include validated messages
-    f.broadcast(toSendPeers, RPCMsg(messages: @[msg]))
+    f.broadcast(toSendPeers, RPCMsg(messages: @[msg]), isHighPriority = false)
     trace "Forwared message to peers", peers = toSendPeers.len
 
   f.updateMetrics(rpcMsg)
@@ -213,13 +227,13 @@ method publish*(f: FloodSub,
   trace "Created new message",
     msg = shortLog(msg), peers = peers.len, topic, msgId
 
-  if f.addSeen(msgId):
+  if f.addSeen(f.salt(msgId)):
     # custom msgid providers might cause this
     trace "Dropping already-seen message", msgId, topic
     return 0
 
   # Try to send to all peers that are known to be interested
-  f.broadcast(peers, RPCMsg(messages: @[msg]))
+  f.broadcast(peers, RPCMsg(messages: @[msg]), isHighPriority = true)
 
   when defined(libp2p_expensive_metrics):
     libp2p_pubsub_messages_published.inc(labelValues = [topic])
@@ -231,8 +245,11 @@ method publish*(f: FloodSub,
 method initPubSub*(f: FloodSub)
   {.raises: [InitializationError].} =
   procCall PubSub(f).initPubSub()
-  f.seen = TimedCache[MessageId].init(2.minutes)
-  f.seenSalt = newSeqUninitialized[byte](sizeof(Hash))
-  hmacDrbgGenerate(f.rng[], f.seenSalt)
+  f.seen = TimedCache[SaltedId].init(2.minutes)
+  f.seenSalt.init()
+
+  var tmp: array[32, byte]
+  hmacDrbgGenerate(f.rng[], tmp)
+  f.seenSalt.update(tmp)
 
   f.init()

libp2p/protocols/pubsub/gossipsub.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -46,42 +46,82 @@ declareCounter(libp2p_gossipsub_saved_bytes, "bytes saved by gossipsub optimizat
 declareCounter(libp2p_gossipsub_duplicate, "number of duplicates received")
 declareCounter(libp2p_gossipsub_received, "number of messages received (deduplicated)")
 
-proc init*(_: type[GossipSubParams]): GossipSubParams =
+when defined(libp2p_expensive_metrics):
+  declareCounter(libp2p_pubsub_received_messages, "number of messages received", labels = ["id", "topic"])
+
+proc init*(
+  _: type[GossipSubParams],
+  pruneBackoff = 1.minutes,
+  unsubscribeBackoff = 5.seconds,
+  floodPublish = true,
+  gossipFactor: float64 = 0.25,
+  d = GossipSubD,
+  dLow = GossipSubDlo,
+  dHigh = GossipSubDhi,
+  dScore = GossipSubDlo,
+  dOut = GossipSubDlo - 1, # DLow - 1
+  dLazy = GossipSubD, # Like D,
+  heartbeatInterval = GossipSubHeartbeatInterval,
+  historyLength = GossipSubHistoryLength,
+  historyGossip = GossipSubHistoryGossip,
+  fanoutTTL = GossipSubFanoutTTL,
+  seenTTL = 2.minutes,
+  gossipThreshold = -100.0,
+  publishThreshold = -1000.0,
+  graylistThreshold = -10000.0,
+  opportunisticGraftThreshold = 0.0,
+  decayInterval = 1.seconds,
+  decayToZero = 0.01,
+  retainScore = 2.minutes,
+  appSpecificWeight = 0.0,
+  ipColocationFactorWeight = 0.0,
+  ipColocationFactorThreshold = 1.0,
+  behaviourPenaltyWeight = -1.0,
+  behaviourPenaltyDecay = 0.999,
+  directPeers = initTable[PeerId, seq[MultiAddress]](),
+  disconnectBadPeers = false,
+  enablePX = false,
+  bandwidthEstimatebps = 100_000_000, # 100 Mbps or 12.5 MBps
+  overheadRateLimit = Opt.none(tuple[bytes: int, interval: Duration]),
+  disconnectPeerAboveRateLimit = false,
+  maxNumElementsInNonPriorityQueue = DefaultMaxNumElementsInNonPriorityQueue): GossipSubParams =
+
   GossipSubParams(
       explicit: true,
-      pruneBackoff: 1.minutes,
-      unsubscribeBackoff: 5.seconds,
-      floodPublish: true,
-      gossipFactor: 0.25,
-      d: GossipSubD,
-      dLow: GossipSubDlo,
-      dHigh: GossipSubDhi,
-      dScore: GossipSubDlo,
-      dOut: GossipSubDlo - 1, # DLow - 1
-      dLazy: GossipSubD, # Like D
-      heartbeatInterval: GossipSubHeartbeatInterval,
-      historyLength: GossipSubHistoryLength,
-      historyGossip: GossipSubHistoryGossip,
-      fanoutTTL: GossipSubFanoutTTL,
-      seenTTL: 2.minutes,
-      gossipThreshold: -100,
-      publishThreshold: -1000,
-      graylistThreshold: -10000,
-      opportunisticGraftThreshold: 0,
-      decayInterval: 1.seconds,
-      decayToZero: 0.01,
-      retainScore: 2.minutes,
-      appSpecificWeight: 0.0,
-      ipColocationFactorWeight: 0.0,
-      ipColocationFactorThreshold: 1.0,
-      behaviourPenaltyWeight: -1.0,
-      behaviourPenaltyDecay: 0.999,
-      disconnectBadPeers: false,
-      enablePX: false,
-      bandwidthEstimatebps: 100_000_000, # 100 Mbps or 12.5 MBps
-      iwantTimeout: 3 * GossipSubHeartbeatInterval,
-      overheadRateLimit: Opt.none(tuple[bytes: int, interval: Duration]),
-      disconnectPeerAboveRateLimit: false
+      pruneBackoff: pruneBackoff,
+      unsubscribeBackoff: unsubscribeBackoff,
+      floodPublish: floodPublish,
+      gossipFactor: gossipFactor,
+      d: d,
+      dLow: dLow,
+      dHigh: dHigh,
+      dScore: dScore,
+      dOut: dOut,
+      dLazy: dLazy,
+      heartbeatInterval: heartbeatInterval,
+      historyLength: historyLength,
+      historyGossip: historyGossip,
+      fanoutTTL: fanoutTTL,
+      seenTTL: seenTTL,
+      gossipThreshold: gossipThreshold,
+      publishThreshold: publishThreshold,
+      graylistThreshold: graylistThreshold,
+      opportunisticGraftThreshold: opportunisticGraftThreshold,
+      decayInterval: decayInterval,
+      decayToZero: decayToZero,
+      retainScore: retainScore,
+      appSpecificWeight: appSpecificWeight,
+      ipColocationFactorWeight: ipColocationFactorWeight,
+      ipColocationFactorThreshold: ipColocationFactorThreshold,
+      behaviourPenaltyWeight: behaviourPenaltyWeight,
+      behaviourPenaltyDecay: behaviourPenaltyDecay,
+      directPeers: directPeers,
+      disconnectBadPeers: disconnectBadPeers,
+      enablePX: enablePX,
+      bandwidthEstimatebps: bandwidthEstimatebps,
+      overheadRateLimit: overheadRateLimit,
+      disconnectPeerAboveRateLimit: disconnectPeerAboveRateLimit,
+      maxNumElementsInNonPriorityQueue: maxNumElementsInNonPriorityQueue
     )
 
 proc validateParameters*(parameters: GossipSubParams): Result[void, cstring] =
@@ -112,6 +152,8 @@ proc validateParameters*(parameters: GossipSubParams): Result[void, cstring] =
     err("gossipsub: behaviourPenaltyWeight parameter error, Must be negative")
   elif parameters.behaviourPenaltyDecay < 0 or parameters.behaviourPenaltyDecay >= 1:
     err("gossipsub: behaviourPenaltyDecay parameter error, Must be between 0 and 1")
+  elif parameters.maxNumElementsInNonPriorityQueue <= 0:
+    err("gossipsub: maxNumElementsInNonPriorityQueue parameter error, Must be > 0")
   else:
     ok()
 
@@ -170,10 +212,10 @@ method onNewPeer*(g: GossipSub, peer: PubSubPeer) =
 
 method onPubSubPeerEvent*(p: GossipSub, peer: PubSubPeer, event: PubSubPeerEvent) {.gcsafe.} =
   case event.kind
-  of PubSubPeerEventKind.Connected:
+  of PubSubPeerEventKind.StreamOpened:
     discard
-  of PubSubPeerEventKind.Disconnected:
-    # If a send connection is lost, it's better to remove peer from the mesh -
+  of PubSubPeerEventKind.StreamClosed:
+    # If a send stream is lost, it's better to remove peer from the mesh -
     # if it gets reestablished, the peer will be readded to the mesh, and if it
     # doesn't, well.. then we hope the peer is going away!
     for topic, peers in p.mesh.mpairs():
@@ -181,6 +223,8 @@ method onPubSubPeerEvent*(p: GossipSub, peer: PubSubPeer, event: PubSubPeerEvent
       peers.excl(peer)
     for _, peers in p.fanout.mpairs():
       peers.excl(peer)
+  of PubSubPeerEventKind.DisconnectionRequested:
+    asyncSpawn p.disconnectPeer(peer) # this should unsubscribePeer the peer too
 
   procCall FloodSub(p).onPubSubPeerEvent(peer, event)
 
@@ -218,12 +262,14 @@ method unsubscribePeer*(g: GossipSub, peer: PeerId) =
     for topic, info in stats[].topicInfos.mpairs:
       info.firstMessageDeliveries = 0
 
+  pubSubPeer.stopSendNonPriorityTask()
+
   procCall FloodSub(g).unsubscribePeer(peer)
 
-proc handleSubscribe*(g: GossipSub,
-                      peer: PubSubPeer,
-                      topic: string,
-                      subscribe: bool) =
+proc handleSubscribe(g: GossipSub,
+                     peer: PubSubPeer,
+                     topic: string,
+                     subscribe: bool) =
   logScope:
     peer
     topic
@@ -272,46 +318,55 @@ proc handleControl(g: GossipSub, peer: PubSubPeer, control: ControlMessage) =
   var respControl: ControlMessage
   g.handleIDontWant(peer, control.idontwant)
   let iwant = g.handleIHave(peer, control.ihave)
-  if iwant.messageIds.len > 0:
+  if iwant.messageIDs.len > 0:
     respControl.iwant.add(iwant)
   respControl.prune.add(g.handleGraft(peer, control.graft))
   let messages = g.handleIWant(peer, control.iwant)
 
-  if
-    respControl.prune.len > 0 or
-    respControl.iwant.len > 0 or
-    messages.len > 0:
-    # iwant and prunes from here, also messages
+  let
+    isPruneNotEmpty = respControl.prune.len > 0
+    isIWantNotEmpty = respControl.iwant.len > 0
 
-    for smsg in messages:
-      for topic in smsg.topicIds:
-        if g.knownTopics.contains(topic):
-          libp2p_pubsub_broadcast_messages.inc(labelValues = [topic])
+  if isPruneNotEmpty or isIWantNotEmpty:
+
+    if isIWantNotEmpty:
+      libp2p_pubsub_broadcast_iwant.inc(respControl.iwant.len.int64)
+
+    if isPruneNotEmpty:
+      for prune in respControl.prune:
+        if g.knownTopics.contains(prune.topicID):
+          libp2p_pubsub_broadcast_prune.inc(labelValues = [prune.topicID])
         else:
-          libp2p_pubsub_broadcast_messages.inc(labelValues = ["generic"])
-
-    libp2p_pubsub_broadcast_iwant.inc(respControl.iwant.len.int64)
-
-    for prune in respControl.prune:
-      if g.knownTopics.contains(prune.topicId):
-        libp2p_pubsub_broadcast_prune.inc(labelValues = [prune.topicId])
-      else:
-        libp2p_pubsub_broadcast_prune.inc(labelValues = ["generic"])
+          libp2p_pubsub_broadcast_prune.inc(labelValues = ["generic"])
 
     trace "sending control message", msg = shortLog(respControl), peer
     g.send(
       peer,
-      RPCMsg(control: some(respControl), messages: messages))
+      RPCMsg(control: some(respControl)), isHighPriority = true)
+
+  if messages.len > 0:
+    for smsg in messages:
+      let topic = smsg.topic
+      if g.knownTopics.contains(topic):
+        libp2p_pubsub_broadcast_messages.inc(labelValues = [topic])
+      else:
+        libp2p_pubsub_broadcast_messages.inc(labelValues = ["generic"])
+
+    # iwant replies have lower priority
+    trace "sending iwant reply messages", peer
+    g.send(
+      peer,
+      RPCMsg(messages: messages), isHighPriority = false)
 
 proc validateAndRelay(g: GossipSub,
                       msg: Message,
-                      msgId, msgIdSalted: MessageId,
+                      msgId: MessageId, saltedId: SaltedId,
                       peer: PubSubPeer) {.async.} =
   try:
     let validation = await g.validate(msg)
 
     var seenPeers: HashSet[PubSubPeer]
-    discard g.validationSeen.pop(msgIdSalted, seenPeers)
+    discard g.validationSeen.pop(saltedId, seenPeers)
     libp2p_gossipsub_duplicate_during_validation.inc(seenPeers.len.int64)
     libp2p_gossipsub_saved_bytes.inc((msg.data.len * seenPeers.len).int64, labelValues = ["validation_duplicate"])
 
@@ -319,7 +374,7 @@ proc validateAndRelay(g: GossipSub,
     of ValidationResult.Reject:
       debug "Dropping message after validation, reason: reject",
         msgId = shortLog(msgId), peer
-      g.punishInvalidMessage(peer, msg)
+      await g.punishInvalidMessage(peer, msg)
       return
     of ValidationResult.Ignore:
       debug "Dropping message after validation, reason: ignore",
@@ -331,18 +386,16 @@ proc validateAndRelay(g: GossipSub,
     # store in cache only after validation
     g.mcache.put(msgId, msg)
 
-    g.rewardDelivered(peer, msg.topicIds, true)
+    let topic = msg.topic
+    g.rewardDelivered(peer, topic, true)
 
     var toSendPeers = HashSet[PubSubPeer]()
-    for t in msg.topicIds:                      # for every topic in the message
-      if t notin g.topics:
-        continue
+    if topic notin g.topics:
+      return
 
-      g.floodsub.withValue(t, peers): toSendPeers.incl(peers[])
-      g.mesh.withValue(t, peers): toSendPeers.incl(peers[])
-
-      # add direct peers
-      toSendPeers.incl(g.subscribedDirectPeers.getOrDefault(t))
+    g.floodsub.withValue(topic, peers): toSendPeers.incl(peers[])
+    g.mesh.withValue(topic, peers): toSendPeers.incl(peers[])
+    g.subscribedDirectPeers.withValue(topic, peers): toSendPeers.incl(peers[])
 
     # Don't send it to source peer, or peers that
     # sent it during validation
@@ -353,66 +406,55 @@ proc validateAndRelay(g: GossipSub,
     # bigger than the messageId
     if msg.data.len > msgId.len * 10:
       g.broadcast(toSendPeers, RPCMsg(control: some(ControlMessage(
-          idontwant: @[ControlIWant(messageIds: @[msgId])]
-        ))))
+          idontwant: @[ControlIWant(messageIDs: @[msgId])]
+        ))), isHighPriority = true)
 
     for peer in toSendPeers:
       for heDontWant in peer.heDontWants:
-        if msgId in heDontWant:
+        if saltedId in heDontWant:
           seenPeers.incl(peer)
           libp2p_gossipsub_idontwant_saved_messages.inc
           libp2p_gossipsub_saved_bytes.inc(msg.data.len.int64, labelValues = ["idontwant"])
           break
     toSendPeers.excl(seenPeers)
 
-
     # In theory, if topics are the same in all messages, we could batch - we'd
     # also have to be careful to only include validated messages
-    g.broadcast(toSendPeers, RPCMsg(messages: @[msg]))
+    g.broadcast(toSendPeers, RPCMsg(messages: @[msg]), isHighPriority = false)
     trace "forwarded message to peers", peers = toSendPeers.len, msgId, peer
-    for topic in msg.topicIds:
-      if topic notin g.topics: continue
 
-      if g.knownTopics.contains(topic):
-        libp2p_pubsub_messages_rebroadcasted.inc(toSendPeers.len.int64, labelValues = [topic])
-      else:
-        libp2p_pubsub_messages_rebroadcasted.inc(toSendPeers.len.int64, labelValues = ["generic"])
+    if g.knownTopics.contains(topic):
+      libp2p_pubsub_messages_rebroadcasted.inc(toSendPeers.len.int64, labelValues = [topic])
+    else:
+      libp2p_pubsub_messages_rebroadcasted.inc(toSendPeers.len.int64, labelValues = ["generic"])
 
-      await handleData(g, topic, msg.data)
+    await handleData(g, topic, msg.data)
   except CatchableError as exc:
     info "validateAndRelay failed", msg=exc.msg
 
 proc dataAndTopicsIdSize(msgs: seq[Message]): int =
-  msgs.mapIt(it.data.len + it.topicIds.mapIt(it.len).foldl(a + b, 0)).foldl(a + b, 0)
+  msgs.mapIt(it.data.len + it.topic.len).foldl(a + b, 0)
 
-proc rateLimit*(g: GossipSub, peer: PubSubPeer, rpcMsgOpt: Opt[RPCMsg], msgSize: int) {.raises:[PeerRateLimitError, CatchableError], async.} =
+proc messageOverhead(g: GossipSub, msg: RPCMsg, msgSize: int): int =
   # In this way we count even ignored fields by protobuf
+  let
+    payloadSize =
+      if g.verifySignature:
+        byteSize(msg.messages)
+      else:
+        dataAndTopicsIdSize(msg.messages)
+    controlSize = msg.control.withValue(control):
+      byteSize(control.ihave) + byteSize(control.iwant)
+    do: # no control message
+      0
 
-  var rmsg = rpcMsgOpt.valueOr:
-    peer.overheadRateLimitOpt.withValue(overheadRateLimit):
-      if not overheadRateLimit.tryConsume(msgSize):
-        libp2p_gossipsub_peers_rate_limit_hits.inc(labelValues = [peer.getAgent()]) # let's just measure at the beginning for test purposes.
-        debug "Peer sent a msg that couldn't be decoded and it's above rate limit.", peer, uselessAppBytesNum = msgSize
-        if g.parameters.disconnectPeerAboveRateLimit:
-          await g.disconnectPeer(peer)
-          raise newException(PeerRateLimitError, "Peer disconnected because it's above rate limit.")
-
-    raise newException(CatchableError, "Peer msg couldn't be decoded")
-
-  let usefulMsgBytesNum =
-    if g.verifySignature:
-      byteSize(rmsg.messages)
-    else:
-      dataAndTopicsIdSize(rmsg.messages)
-
-  var uselessAppBytesNum = msgSize - usefulMsgBytesNum
-  rmsg.control.withValue(control):
-    uselessAppBytesNum -= (byteSize(control.ihave) + byteSize(control.iwant))
+  msgSize - payloadSize - controlSize
 
+proc rateLimit*(g: GossipSub, peer: PubSubPeer, overhead: int) {.async.} =
   peer.overheadRateLimitOpt.withValue(overheadRateLimit):
-    if not overheadRateLimit.tryConsume(uselessAppBytesNum):
+    if not overheadRateLimit.tryConsume(overhead):
       libp2p_gossipsub_peers_rate_limit_hits.inc(labelValues = [peer.getAgent()]) # let's just measure at the beginning for test purposes.
-      debug "Peer sent too much useless application data and it's above rate limit.", peer, msgSize, uselessAppBytesNum, rmsg
+      debug "Peer sent too much useless application data and it's above rate limit.", peer, overhead
       if g.parameters.disconnectPeerAboveRateLimit:
         await g.disconnectPeer(peer)
         raise newException(PeerRateLimitError, "Peer disconnected because it's above rate limit.")
@@ -420,22 +462,31 @@ proc rateLimit*(g: GossipSub, peer: PubSubPeer, rpcMsgOpt: Opt[RPCMsg], msgSize:
 method rpcHandler*(g: GossipSub,
                   peer: PubSubPeer,
                   data: seq[byte]) {.async.} =
-
   let msgSize = data.len
   var rpcMsg = decodeRpcMsg(data).valueOr:
     debug "failed to decode msg from peer", peer, err = error
-    await rateLimit(g, peer, Opt.none(RPCMsg), msgSize)
-    return
+    await rateLimit(g, peer, msgSize)
+    # Raising in the handler closes the gossipsub connection (but doesn't
+    # disconnect the peer!)
+    # TODO evaluate behaviour penalty values
+    peer.behaviourPenalty += 0.1
+
+    raise newException(CatchableError, "Peer msg couldn't be decoded")
+
+  when defined(libp2p_expensive_metrics):
+    for m in rpcMsg.messages:
+      libp2p_pubsub_received_messages.inc(labelValues = [$peer.peerId, m.topic])
 
   trace "decoded msg from peer", peer, msg = rpcMsg.shortLog
-  await rateLimit(g, peer, Opt.some(rpcMsg), msgSize)
+  await rateLimit(g, peer, g.messageOverhead(rpcMsg, msgSize))
 
-  # trigger hooks
+  # trigger hooks - these may modify the message
   peer.recvObservers(rpcMsg)
 
   if rpcMsg.ping.len in 1..<64 and peer.pingBudget > 0:
-    g.send(peer, RPCMsg(pong: rpcMsg.ping))
+    g.send(peer, RPCMsg(pong: rpcMsg.ping), isHighPriority = true)
     peer.pingBudget.dec
+
   for i in 0..<min(g.topicsHigh, rpcMsg.subscriptions.len):
     template sub: untyped = rpcMsg.subscriptions[i]
     g.handleSubscribe(peer, sub.topic, sub.subscribe)
@@ -455,19 +506,15 @@ method rpcHandler*(g: GossipSub,
     if msgIdResult.isErr:
       debug "Dropping message due to failed message id generation",
         error = msgIdResult.error
-      # TODO: descore peers due to error during message validation (malicious?)
+      await g.punishInvalidMessage(peer, msg)
       continue
 
     let
       msgId = msgIdResult.get
-      msgIdSalted = msgId & g.seenSalt
-    g.outstandingIWANTs.withValue(msgId, iwantRequest):
-      if iwantRequest.peer.peerId == peer.peerId:
-        g.outstandingIWANTs.del(msgId)
+      msgIdSalted = g.salt(msgId)
+      topic = msg.topic
 
-    # addSeen adds salt to msgId to avoid
-    # remote attacking the hash function
-    if g.addSeen(msgId):
+    if g.addSeen(msgIdSalted):
       trace "Dropping already-seen message", msgId = shortLog(msgId), peer
 
       var alreadyReceived = false
@@ -477,8 +524,8 @@ method rpcHandler*(g: GossipSub,
           alreadyReceived = true
 
       if not alreadyReceived:
-        let delay = Moment.now() - g.firstSeen(msgId)
-        g.rewardDelivered(peer, msg.topicIds, false, delay)
+        let delay = Moment.now() - g.firstSeen(msgIdSalted)
+        g.rewardDelivered(peer, topic, false, delay)
 
       libp2p_gossipsub_duplicate.inc()
 
@@ -488,7 +535,7 @@ method rpcHandler*(g: GossipSub,
     libp2p_gossipsub_received.inc()
 
     # avoid processing messages we are not interested in
-    if msg.topicIds.allIt(it notin g.topics):
+    if topic notin g.topics:
       debug "Dropping message of topic without subscription", msgId = shortLog(msgId), peer
       continue
 
@@ -496,14 +543,14 @@ method rpcHandler*(g: GossipSub,
       # always validate if signature is present or required
       debug "Dropping message due to failed signature verification",
         msgId = shortLog(msgId), peer
-      g.punishInvalidMessage(peer, msg)
+      await g.punishInvalidMessage(peer, msg)
       continue
 
     if msg.seqno.len > 0 and msg.seqno.len != 8:
       # if we have seqno should be 8 bytes long
       debug "Dropping message due to invalid seqno length",
         msgId = shortLog(msgId), peer
-      g.punishInvalidMessage(peer, msg)
+      await g.punishInvalidMessage(peer, msg)
       continue
 
     # g.anonymize needs no evaluation when receiving messages
@@ -547,32 +594,31 @@ method onTopicSubscription*(g: GossipSub, topic: string, subscribed: bool) =
             topicID: topic,
             peers: g.peerExchangeList(topic),
             backoff: g.parameters.unsubscribeBackoff.seconds.uint64)])))
-    g.broadcast(mpeers, msg)
+    g.broadcast(mpeers, msg, isHighPriority = true)
 
     for peer in mpeers:
       g.pruned(peer, topic, backoff = some(g.parameters.unsubscribeBackoff))
 
     g.mesh.del(topic)
 
-
     # Send unsubscribe (in reverse order to sub/graft)
     procCall PubSub(g).onTopicSubscription(topic, subscribed)
 
 method publish*(g: GossipSub,
                 topic: string,
                 data: seq[byte]): Future[int] {.async.} =
-  # base returns always 0
-  discard await procCall PubSub(g).publish(topic, data)
-
   logScope:
     topic
 
-  trace "Publishing message on topic", data = data.shortLog
-
   if topic.len <= 0: # data could be 0/empty
     debug "Empty topic, skipping publish"
     return 0
 
+  # base returns always 0
+  discard await procCall PubSub(g).publish(topic, data)
+
+  trace "Publishing message on topic", data = data.shortLog
+
   var peers: HashSet[PubSubPeer]
 
   # add always direct peers
@@ -585,38 +631,39 @@ method publish*(g: GossipSub,
     # With flood publishing enabled, the mesh is used when propagating messages from other peers,
     # but a peer's own messages will always be published to all known peers in the topic, limited
     # to the amount of peers we can send it to in one heartbeat
-    var maxPeersToFlodOpt: Opt[int64]
-    if g.parameters.bandwidthEstimatebps > 0:
-      let
-        bandwidth = (g.parameters.bandwidthEstimatebps) div 8 div 1000 # Divisions are to convert it to Bytes per ms TODO replace with bandwidth estimate
-        msToTransmit = max(data.len div bandwidth, 1)
-      maxPeersToFlodOpt = Opt.some(max(g.parameters.heartbeatInterval.milliseconds div msToTransmit, g.parameters.dLow))
+
+    let maxPeersToFlood =
+      if g.parameters.bandwidthEstimatebps > 0:
+        let
+          bandwidth = (g.parameters.bandwidthEstimatebps) div 8 div 1000 # Divisions are to convert it to Bytes per ms TODO replace with bandwidth estimate
+          msToTransmit = max(data.len div bandwidth, 1)
+        max(g.parameters.heartbeatInterval.milliseconds div msToTransmit, g.parameters.dLow)
+      else:
+        int.high() # unlimited
 
     for peer in g.gossipsub.getOrDefault(topic):
-      maxPeersToFlodOpt.withValue(maxPeersToFlod):
-        if peers.len >= maxPeersToFlod: break
+      if peers.len >= maxPeersToFlood: break
+
       if peer.score >= g.parameters.publishThreshold:
         trace "publish: including flood/high score peer", peer
         peers.incl(peer)
 
-  if peers.len < g.parameters.dLow:
-    # not subscribed, or bad mesh, send to fanout peers
-    var fanoutPeers = g.fanout.getOrDefault(topic).toSeq()
-    if fanoutPeers.len < g.parameters.dLow:
-      g.replenishFanout(topic)
-      fanoutPeers = g.fanout.getOrDefault(topic).toSeq()
+  elif peers.len < g.parameters.dLow:
+    # not subscribed or bad mesh, send to fanout peers
+    # when flood-publishing, fanout won't help since all potential peers have
+    # already been added
 
+    g.replenishFanout(topic) # Make sure fanout is populated
+
+    var fanoutPeers = g.fanout.getOrDefault(topic).toSeq()
     g.rng.shuffle(fanoutPeers)
 
     for fanPeer in fanoutPeers:
       peers.incl(fanPeer)
       if peers.len > g.parameters.d: break
 
-    # even if we couldn't publish,
-    # we still attempted to publish
-    # on the topic, so it makes sense
-    # to update the last topic publish
-    # time
+    # Attempting to publish counts as fanout send (even if the message
+    # ultimately is not sent)
     g.lastFanoutPubSub[topic] = Moment.fromNow(g.parameters.fanoutTTL)
 
   if peers.len == 0:
@@ -644,14 +691,16 @@ method publish*(g: GossipSub,
 
   trace "Created new message", msg = shortLog(msg), peers = peers.len
 
-  if g.addSeen(msgId):
-    # custom msgid providers might cause this
+  if g.addSeen(g.salt(msgId)):
+    # If the message was received or published recently, don't re-publish it -
+    # this might happen when not using sequence id:s and / or with a custom
+    # message id provider
     trace "Dropping already-seen message"
     return 0
 
   g.mcache.put(msgId, msg)
 
-  g.broadcast(peers, RPCMsg(messages: @[msg]))
+  g.broadcast(peers, RPCMsg(messages: @[msg]), isHighPriority = true)
 
   if g.knownTopics.contains(topic):
     libp2p_pubsub_messages_published.inc(peers.len.int64, labelValues = [topic])
@@ -686,30 +735,40 @@ proc maintainDirectPeers(g: GossipSub) {.async.} =
     for id, addrs in g.parameters.directPeers:
       await g.addDirectPeer(id, addrs)
 
-method start*(g: GossipSub) {.async.} =
+method start*(
+    g: GossipSub
+): Future[void] {.async: (raises: [CancelledError], raw: true).} =
+  let fut = newFuture[void]()
+  fut.complete()
+
   trace "gossipsub start"
 
   if not g.heartbeatFut.isNil:
     warn "Starting gossipsub twice"
-    return
+    return fut
 
   g.heartbeatFut = g.heartbeat()
   g.scoringHeartbeatFut = g.scoringHeartbeat()
   g.directPeersLoop = g.maintainDirectPeers()
   g.started = true
+  fut
+
+method stop*(g: GossipSub): Future[void] {.async: (raises: [], raw: true).} =
+  let fut = newFuture[void]()
+  fut.complete()
 
-method stop*(g: GossipSub) {.async.} =
   trace "gossipsub stop"
   g.started = false
   if g.heartbeatFut.isNil:
     warn "Stopping gossipsub without starting it"
-    return
+    return fut
 
   # stop heartbeat interval
   g.directPeersLoop.cancel()
   g.scoringHeartbeatFut.cancel()
   g.heartbeatFut.cancel()
   g.heartbeatFut = nil
+  fut
 
 method initPubSub*(g: GossipSub)
   {.raises: [InitializationError].} =
@@ -723,7 +782,7 @@ method initPubSub*(g: GossipSub)
     raise newException(InitializationError, $validationRes.error)
 
   # init the floodsub stuff here, we customize timedcache in gossip!
-  g.seen = TimedCache[MessageId].init(g.parameters.seenTTL)
+  g.seen = TimedCache[SaltedId].init(g.parameters.seenTTL)
 
   # init gossip stuff
   g.mcache = MCache.init(g.parameters.historyGossip, g.parameters.historyLength)
@@ -736,4 +795,5 @@ method getOrCreatePeer*(
   let peer = procCall PubSub(g).getOrCreatePeer(peerId, protos)
   g.parameters.overheadRateLimit.withValue(overheadRateLimit):
     peer.overheadRateLimitOpt = Opt.some(TokenBucket.new(overheadRateLimit.bytes, overheadRateLimit.interval))
+  peer.maxNumElementsInNonPriorityQueue = g.parameters.maxNumElementsInNonPriorityQueue
   return peer

libp2p/protocols/pubsub/gossipsub/behavior.nim

@@ -30,7 +30,7 @@ declareGauge(libp2p_gossipsub_healthy_peers_topics, "number of topics in mesh wi
 declareCounter(libp2p_gossipsub_above_dhigh_condition, "number of above dhigh pruning branches ran", labels = ["topic"])
 declareGauge(libp2p_gossipsub_received_iwants, "received iwants", labels = ["kind"])
 
-proc grafted*(g: GossipSub, p: PubSubPeer, topic: string) {.raises: [].} =
+proc grafted*(g: GossipSub, p: PubSubPeer, topic: string) =
   g.withPeerStats(p.peerId) do (stats: var PeerStats):
     var info = stats.topicInfos.getOrDefault(topic)
     info.graftTime = Moment.now()
@@ -46,7 +46,7 @@ proc pruned*(g: GossipSub,
              p: PubSubPeer,
              topic: string,
              setBackoff: bool = true,
-             backoff = none(Duration)) {.raises: [].} =
+             backoff = none(Duration)) =
   if setBackoff:
     let
       backoffDuration = backoff.get(g.parameters.pruneBackoff)
@@ -70,7 +70,7 @@ proc pruned*(g: GossipSub,
 
       trace "pruned", peer=p, topic
 
-proc handleBackingOff*(t: var BackoffTable, topic: string) {.raises: [].} =
+proc handleBackingOff*(t: var BackoffTable, topic: string) =
   let now = Moment.now()
   var expired = toSeq(t.getOrDefault(topic).pairs())
   expired.keepIf do (pair: tuple[peer: PeerId, expire: Moment]) -> bool:
@@ -79,7 +79,7 @@ proc handleBackingOff*(t: var BackoffTable, topic: string) {.raises: [].} =
     t.withValue(topic, v):
       v[].del(peer)
 
-proc peerExchangeList*(g: GossipSub, topic: string): seq[PeerInfoMsg] {.raises: [].} =
+proc peerExchangeList*(g: GossipSub, topic: string): seq[PeerInfoMsg] =
   if not g.parameters.enablePX:
     return @[]
   var peers = g.gossipsub.getOrDefault(topic, initHashSet[PubSubPeer]()).toSeq()
@@ -100,11 +100,11 @@ proc peerExchangeList*(g: GossipSub, topic: string): seq[PeerInfoMsg] {.raises:
 
 proc handleGraft*(g: GossipSub,
                  peer: PubSubPeer,
-                 grafts: seq[ControlGraft]): seq[ControlPrune] = # {.raises: [Defect].} TODO chronicles exception on windows
+                 grafts: seq[ControlGraft]): seq[ControlPrune] =
   var prunes: seq[ControlPrune]
   for graft in grafts:
-    let topic = graft.topicId
-    trace "peer grafted topic", peer, topic
+    let topic = graft.topicID
+    trace "peer grafted topicID", peer, topic
 
     # It is an error to GRAFT on a direct peer
     if peer.peerId in g.parameters.directPeers:
@@ -204,12 +204,11 @@ proc getPeers(prune: ControlPrune, peer: PubSubPeer): seq[(PeerId, Option[PeerRe
 
   routingRecords
 
-
-proc handlePrune*(g: GossipSub, peer: PubSubPeer, prunes: seq[ControlPrune]) {.raises: [].} =
+proc handlePrune*(g: GossipSub, peer: PubSubPeer, prunes: seq[ControlPrune]) =
   for prune in prunes:
-    let topic = prune.topicId
+    let topic = prune.topicID
 
-    trace "peer pruned topic", peer, topic
+    trace "peer pruned topicID", peer, topic
 
     # add peer backoff
     if prune.backoff > 0:
@@ -239,7 +238,7 @@ proc handlePrune*(g: GossipSub, peer: PubSubPeer, prunes: seq[ControlPrune]) {.r
 
 proc handleIHave*(g: GossipSub,
                  peer: PubSubPeer,
-                 ihaves: seq[ControlIHave]): ControlIWant {.raises: [].} =
+                 ihaves: seq[ControlIHave]): ControlIWant =
   var res: ControlIWant
   if peer.score < g.parameters.gossipThreshold:
     trace "ihave: ignoring low score peer", peer, score = peer.score
@@ -248,34 +247,32 @@ proc handleIHave*(g: GossipSub,
   else:
     for ihave in ihaves:
       trace "peer sent ihave",
-        peer, topic = ihave.topicId, msgs = ihave.messageIds
-      if ihave.topicId in g.topics:
-        for msgId in ihave.messageIds:
-          if not g.hasSeen(msgId):
+        peer, topicID = ihave.topicID, msgs = ihave.messageIDs
+      if ihave.topicID in g.topics:
+        for msgId in ihave.messageIDs:
+          if not g.hasSeen(g.salt(msgId)):
             if peer.iHaveBudget <= 0:
               break
-            elif msgId notin res.messageIds and msgId notin g.outstandingIWANTs:
-              g.outstandingIWANTs[msgId] = IWANTRequest(messageId: msgId, peer: peer, timestamp: Moment.now())
-              res.messageIds.add(msgId)
+            elif msgId notin res.messageIDs:
+              res.messageIDs.add(msgId)
               dec peer.iHaveBudget
               trace "requested message via ihave", messageID=msgId
     # shuffling res.messageIDs before sending it out to increase the likelihood
     # of getting an answer if the peer truncates the list due to internal size restrictions.
-    g.rng.shuffle(res.messageIds)
+    g.rng.shuffle(res.messageIDs)
     return res
 
 proc handleIDontWant*(g: GossipSub,
                       peer: PubSubPeer,
                       iDontWants: seq[ControlIWant]) =
   for dontWant in iDontWants:
-    for messageId in dontWant.messageIds:
+    for messageId in dontWant.messageIDs:
       if peer.heDontWants[^1].len > 1000: break
-      if messageId.len > 100: continue
-      peer.heDontWants[^1].incl(messageId)
+      peer.heDontWants[^1].incl(g.salt(messageId))
 
 proc handleIWant*(g: GossipSub,
                  peer: PubSubPeer,
-                 iwants: seq[ControlIWant]): seq[Message] {.raises: [].} =
+                 iwants: seq[ControlIWant]): seq[Message] =
   var
     messages: seq[Message]
     invalidRequests = 0
@@ -283,7 +280,7 @@ proc handleIWant*(g: GossipSub,
     trace "iwant: ignoring low score peer", peer, score = peer.score
   else:
     for iwant in iwants:
-      for mid in iwant.messageIds:
+      for mid in iwant.messageIDs:
         trace "peer sent iwant", peer, messageID = mid
         # canAskIWant will only return true once for a specific message
         if not peer.canAskIWant(mid):
@@ -301,18 +298,7 @@ proc handleIWant*(g: GossipSub,
         messages.add(msg)
   return messages
 
-proc checkIWANTTimeouts(g: GossipSub, timeoutDuration: Duration) {.raises: [].} =
-  let currentTime = Moment.now()
-  var idsToRemove = newSeq[MessageId]()
-  for msgId, request in g.outstandingIWANTs.pairs():
-    if currentTime - request.timestamp > timeoutDuration:
-      trace "IWANT request timed out", messageID=msgId, peer=request.peer
-      request.peer.behaviourPenalty += 0.1
-      idsToRemove.add(msgId)
-  for msgId in idsToRemove:
-      g.outstandingIWANTs.del(msgId)
-
-proc commitMetrics(metrics: var MeshMetrics) {.raises: [].} =
+proc commitMetrics(metrics: var MeshMetrics) =
   libp2p_gossipsub_low_peers_topics.set(metrics.lowPeersTopics)
   libp2p_gossipsub_no_peers_topics.set(metrics.noPeersTopics)
   libp2p_gossipsub_under_dout_topics.set(metrics.underDoutTopics)
@@ -321,7 +307,7 @@ proc commitMetrics(metrics: var MeshMetrics) {.raises: [].} =
   libp2p_gossipsub_peers_per_topic_fanout.set(metrics.otherPeersPerTopicFanout, labelValues = ["other"])
   libp2p_gossipsub_peers_per_topic_mesh.set(metrics.otherPeersPerTopicMesh, labelValues = ["other"])
 
-proc rebalanceMesh*(g: GossipSub, topic: string, metrics: ptr MeshMetrics = nil) {.raises: [].} =
+proc rebalanceMesh*(g: GossipSub, topic: string, metrics: ptr MeshMetrics = nil) =
   logScope:
     topic
     mesh = g.mesh.peers(topic)
@@ -542,16 +528,16 @@ proc rebalanceMesh*(g: GossipSub, topic: string, metrics: ptr MeshMetrics = nil)
   # Send changes to peers after table updates to avoid stale state
   if grafts.len > 0:
     let graft = RPCMsg(control: some(ControlMessage(graft: @[ControlGraft(topicID: topic)])))
-    g.broadcast(grafts, graft)
+    g.broadcast(grafts, graft, isHighPriority = true)
   if prunes.len > 0:
     let prune = RPCMsg(control: some(ControlMessage(
       prune: @[ControlPrune(
         topicID: topic,
         peers: g.peerExchangeList(topic),
         backoff: g.parameters.pruneBackoff.seconds.uint64)])))
-    g.broadcast(prunes, prune)
+    g.broadcast(prunes, prune, isHighPriority = true)
 
-proc dropFanoutPeers*(g: GossipSub) {.raises: [].} =
+proc dropFanoutPeers*(g: GossipSub) =
   # drop peers that we haven't published to in
   # GossipSubFanoutTTL seconds
   let now = Moment.now()
@@ -564,7 +550,7 @@ proc dropFanoutPeers*(g: GossipSub) {.raises: [].} =
   for topic in drops:
     g.lastFanoutPubSub.del topic
 
-proc replenishFanout*(g: GossipSub, topic: string) {.raises: [].} =
+proc replenishFanout*(g: GossipSub, topic: string) =
   ## get fanout peers for a topic
   logScope: topic
   trace "about to replenish fanout"
@@ -580,7 +566,7 @@ proc replenishFanout*(g: GossipSub, topic: string) {.raises: [].} =
 
   trace "fanout replenished with peers", peers = g.fanout.peers(topic)
 
-proc getGossipPeers*(g: GossipSub): Table[PubSubPeer, ControlMessage] {.raises: [].} =
+proc getGossipPeers*(g: GossipSub): Table[PubSubPeer, ControlMessage] =
   ## gossip iHave messages to peers
   ##
 
@@ -591,7 +577,7 @@ proc getGossipPeers*(g: GossipSub): Table[PubSubPeer, ControlMessage] {.raises:
   trace "getting gossip peers (iHave)", ntopics=topics.len
   for topic in topics:
     if topic notin g.gossipsub:
-      trace "topic not in gossip array, skipping", topicID = topic
+      trace "topic not in gossip array, skipping", topic = topic
       continue
 
     let mids = g.mcache.window(topic)
@@ -624,26 +610,25 @@ proc getGossipPeers*(g: GossipSub): Table[PubSubPeer, ControlMessage] {.raises:
       x notin gossipPeers and
       x.score >= g.parameters.gossipThreshold
 
-    var target = g.parameters.dLazy
-    let factor = (g.parameters.gossipFactor.float * allPeers.len.float).int
-    if factor > target:
-      target = min(factor, allPeers.len)
+    # https://github.com/libp2p/specs/blob/98c5aa9421703fc31b0833ad8860a55db15be063/pubsub/gossipsub/gossipsub-v1.1.md#adaptive-gossip-dissemination
+    let
+      factor = (g.parameters.gossipFactor.float * allPeers.len.float).int
+      target = max(g.parameters.dLazy, factor)
 
     if target < allPeers.len:
       g.rng.shuffle(allPeers)
       allPeers.setLen(target)
 
-    let msgIdsAsSet = ihave.messageIds.toHashSet()
-
     for peer in allPeers:
       control.mgetOrPut(peer, ControlMessage()).ihave.add(ihave)
-      peer.sentIHaves[^1].incl(msgIdsAsSet)
+      for msgId in ihave.messageIDs:
+        peer.sentIHaves[^1].incl(msgId)
 
   libp2p_gossipsub_cache_window_size.set(cacheWindowSize.int64)
 
   return control
 
-proc onHeartbeat(g: GossipSub) {.raises: [].} =
+proc onHeartbeat(g: GossipSub) =
     # reset IWANT budget
     # reset IHAVE cap
     block:
@@ -651,7 +636,7 @@ proc onHeartbeat(g: GossipSub) {.raises: [].} =
         peer.sentIHaves.addFirst(default(HashSet[MessageId]))
         if peer.sentIHaves.len > g.parameters.historyLength:
           discard peer.sentIHaves.popLast()
-        peer.heDontWants.addFirst(default(HashSet[MessageId]))
+        peer.heDontWants.addFirst(default(HashSet[SaltedId]))
         if peer.heDontWants.len > g.parameters.historyLength:
           discard peer.heDontWants.popLast()
         peer.iHaveBudget = IHavePeerBudget
@@ -675,13 +660,14 @@ proc onHeartbeat(g: GossipSub) {.raises: [].} =
           g.pruned(peer, t)
           g.mesh.removePeer(t, peer)
           prunes &= peer
+          peer.clearNonPriorityQueue()
       if prunes.len > 0:
         let prune = RPCMsg(control: some(ControlMessage(
           prune: @[ControlPrune(
             topicID: t,
             peers: g.peerExchangeList(t),
             backoff: g.parameters.pruneBackoff.seconds.uint64)])))
-        g.broadcast(prunes, prune)
+        g.broadcast(prunes, prune, isHighPriority = true)
 
       # pass by ptr in order to both signal we want to update metrics
       # and as well update the struct for each topic during this iteration
@@ -699,16 +685,14 @@ proc onHeartbeat(g: GossipSub) {.raises: [].} =
     for peer, control in peers:
       # only ihave from here
       for ihave in control.ihave:
-        if g.knownTopics.contains(ihave.topicId):
-          libp2p_pubsub_broadcast_ihave.inc(labelValues = [ihave.topicId])
+        if g.knownTopics.contains(ihave.topicID):
+          libp2p_pubsub_broadcast_ihave.inc(labelValues = [ihave.topicID])
         else:
           libp2p_pubsub_broadcast_ihave.inc(labelValues = ["generic"])
-      g.send(peer, RPCMsg(control: some(control)))
+      g.send(peer, RPCMsg(control: some(control)), isHighPriority = true)
 
     g.mcache.shift() # shift the cache
 
-# {.pop.} # raises []
-
 proc heartbeat*(g: GossipSub) {.async.} =
   heartbeat "GossipSub", g.parameters.heartbeatInterval:
     trace "running heartbeat", instance = cast[int](g)
@@ -717,5 +701,3 @@ proc heartbeat*(g: GossipSub) {.async.} =
     for trigger in g.heartbeatEvents:
       trace "firing heartbeat event", instance = cast[int](g)
       trigger.fire()
-
-    checkIWANTTimeouts(g, g.parameters.iwantTimeout)

libp2p/protocols/pubsub/gossipsub/scoring.nim

@@ -87,8 +87,6 @@ proc colocationFactor(g: GossipSub, peer: PubSubPeer): float64 =
   else:
     0.0
 
-{.pop.}
-
 proc disconnectPeer*(g: GossipSub, peer: PubSubPeer) {.async.} =
   try:
     await g.switch.disconnect(peer.peerId)
@@ -240,53 +238,52 @@ proc scoringHeartbeat*(g: GossipSub) {.async.} =
     trace "running scoring heartbeat", instance = cast[int](g)
     g.updateScores()
 
-proc punishInvalidMessage*(g: GossipSub, peer: PubSubPeer, msg: Message) =
+proc punishInvalidMessage*(g: GossipSub, peer: PubSubPeer, msg: Message) {.async.}  =
   let uselessAppBytesNum = msg.data.len
   peer.overheadRateLimitOpt.withValue(overheadRateLimit):
     if not overheadRateLimit.tryConsume(uselessAppBytesNum):
       debug "Peer sent invalid message and it's above rate limit", peer, uselessAppBytesNum
       libp2p_gossipsub_peers_rate_limit_hits.inc(labelValues = [peer.getAgent()]) # let's just measure at the beginning for test purposes.
-      # discard g.disconnectPeer(peer)
-      # debug "Peer disconnected", peer, uselessAppBytesNum
-      # raise newException(PeerRateLimitError, "Peer sent invalid message and it's above rate limit")
+      if g.parameters.disconnectPeerAboveRateLimit:
+        await g.disconnectPeer(peer)
+        raise newException(PeerRateLimitError, "Peer disconnected because it's above rate limit.")
 
+  let topic = msg.topic
+  if topic notin g.topics:
+    return
 
-  for tt in msg.topicIds:
-    let t = tt
-    if t notin g.topics:
-      continue
-
-    let tt = t
-    # update stats
-    g.withPeerStats(peer.peerId) do (stats: var PeerStats):
-      stats.topicInfos.mgetOrPut(tt, TopicInfo()).invalidMessageDeliveries += 1
+  # update stats
+  g.withPeerStats(peer.peerId) do(stats: var PeerStats):
+    stats.topicInfos.mgetOrPut(topic, TopicInfo()).invalidMessageDeliveries += 1
 
 proc addCapped*[T](stat: var T, diff, cap: T) =
   stat += min(diff, cap - stat)
 
 proc rewardDelivered*(
-    g: GossipSub, peer: PubSubPeer, topics: openArray[string], first: bool, delay = ZeroDuration) =
-  for tt in topics:
-    let t = tt
-    if t notin g.topics:
-      continue
+    g: GossipSub,
+    peer: PubSubPeer,
+    topic: string,
+    first: bool,
+    delay = ZeroDuration,
+) =
+  if topic notin g.topics:
+    return
 
-    let tt = t
-    let topicParams = g.topicParams.mgetOrPut(t, TopicParams.init())
-                                            # if in mesh add more delivery score
+  let topicParams = g.topicParams.mgetOrPut(topic, TopicParams.init())
+    # if in mesh add more delivery score
 
-    if delay > topicParams.meshMessageDeliveriesWindow:
-      # Too old
-      continue
+  if delay > topicParams.meshMessageDeliveriesWindow:
+    # Too old
+    return
 
-    g.withPeerStats(peer.peerId) do (stats: var PeerStats):
-      stats.topicInfos.withValue(tt, tstats):
-        if first:
-          tstats[].firstMessageDeliveries.addCapped(
-            1, topicParams.firstMessageDeliveriesCap)
+  g.withPeerStats(peer.peerId) do (stats: var PeerStats):
+    stats.topicInfos.withValue(topic, tstats):
+      if first:
+        tstats[].firstMessageDeliveries.addCapped(
+          1, topicParams.firstMessageDeliveriesCap)
 
-        if tstats[].inMesh:
-          tstats[].meshMessageDeliveries.addCapped(
-            1, topicParams.meshMessageDeliveriesCap)
-      do: # make sure we don't loose this information
-        stats.topicInfos[tt] = TopicInfo(meshMessageDeliveries: 1)
+      if tstats[].inMesh:
+        tstats[].meshMessageDeliveries.addCapped(
+          1, topicParams.meshMessageDeliveriesCap)
+    do: # make sure we don't lose this information
+      stats.topicInfos[topic] = TopicInfo(meshMessageDeliveries: 1)

libp2p/protocols/pubsub/gossipsub/types.nim

@@ -102,6 +102,11 @@ type
     behaviourPenalty*: float64 # the eventual penalty score
 
   GossipSubParams* {.public.} = object
+    # explicit is used to check if the GossipSubParams instance was created by the user either passing params to GossipSubParams(...)
+    # or GossipSubParams.init(...). In the first case explicit should be set to true when calling the Nim constructor.
+    # In the second case, the param isn't necessary and should be always be set to true by init.
+    # If none of those options were used, it means the instance was created using Nim default values.
+    # In this case, GossipSubParams.init() should be called when initing GossipSub to set the values to their default value defined by nim-libp2p.
     explicit*: bool
     pruneBackoff*: Duration
     unsubscribeBackoff*: Duration
@@ -143,13 +148,15 @@ type
     enablePX*: bool
 
     bandwidthEstimatebps*: int # This is currently used only for limting flood publishing. 0 disables flood-limiting completely
-    iwantTimeout*: Duration
 
     overheadRateLimit*: Opt[tuple[bytes: int, interval: Duration]]
     disconnectPeerAboveRateLimit*: bool
 
+    # Max number of elements allowed in the non-priority queue. When this limit has been reached, the peer will be disconnected.
+    maxNumElementsInNonPriorityQueue*: int
+
   BackoffTable* = Table[string, Table[PeerId, Moment]]
-  ValidationSeenTable* = Table[MessageId, HashSet[PubSubPeer]]
+  ValidationSeenTable* = Table[SaltedId, HashSet[PubSubPeer]]
 
   RoutingRecordsPair* = tuple[id: PeerId, record: Option[PeerRecord]]
   RoutingRecordsHandler* =
@@ -165,8 +172,6 @@ type
     subscribedDirectPeers*: PeerTable          # directpeers that we keep alive
     backingOff*: BackoffTable                  # peers to backoff from when replenishing the mesh
     lastFanoutPubSub*: Table[string, Moment]   # last publish time for fanout topics
-    gossip*: Table[string, seq[ControlIHave]]  # pending gossip
-    control*: Table[string, ControlMessage]    # pending control messages
     mcache*: MCache                            # messages cache
     validationSeen*: ValidationSeenTable       # peers who sent us message in validation
     heartbeatFut*: Future[void]                # cancellation future for heartbeat interval
@@ -181,7 +186,6 @@ type
     routingRecordsHandler*: seq[RoutingRecordsHandler] # Callback for peer exchange
 
     heartbeatEvents*: seq[AsyncEvent]
-    outstandingIWANTs*: Table[MessageId, IWANTRequest]
 
   MeshMetrics* = object
     # scratch buffers for metrics
@@ -192,8 +196,3 @@ type
     lowPeersTopics*: int64 # npeers < dlow
     healthyPeersTopics*: int64 # npeers >= dlow
     underDoutTopics*: int64
-
-  IWANTRequest* = object
-    messageId*: MessageId
-    peer*: PubSubPeer
-    timestamp*: Moment

libp2p/protocols/pubsub/mcache.nim

@@ -9,52 +9,57 @@
 
 {.push raises: [].}
 
-import std/[sets, tables, options]
+import std/[sets, tables]
 import rpc/[messages]
+import results
 
-export sets, tables, messages, options
+export sets, tables, messages, results
 
 type
   CacheEntry* = object
-    mid*: MessageId
-    topicIds*: seq[string]
+    msgId*: MessageId
+    topic*: string
 
   MCache* = object of RootObj
     msgs*: Table[MessageId, Message]
     history*: seq[seq[CacheEntry]]
+    pos*: int
     windowSize*: Natural
 
-func get*(c: MCache, mid: MessageId): Option[Message] =
-  if mid in c.msgs:
-    try: some(c.msgs[mid])
+func get*(c: MCache, msgId: MessageId): Opt[Message] =
+  if msgId in c.msgs:
+    try: Opt.some(c.msgs[msgId])
     except KeyError: raiseAssert "checked"
   else:
-    none(Message)
+    Opt.none(Message)
 
-func contains*(c: MCache, mid: MessageId): bool =
-  mid in c.msgs
+func contains*(c: MCache, msgId: MessageId): bool =
+  msgId in c.msgs
 
 func put*(c: var MCache, msgId: MessageId, msg: Message) =
   if not c.msgs.hasKeyOrPut(msgId, msg):
     # Only add cache entry if the message was not already in the cache
-    c.history[0].add(CacheEntry(mid: msgId, topicIds: msg.topicIds))
+    c.history[c.pos].add(CacheEntry(msgId: msgId, topic: msg.topic))
 
 func window*(c: MCache, topic: string): HashSet[MessageId] =
   let
     len = min(c.windowSize, c.history.len)
 
   for i in 0..<len:
-    for entry in c.history[i]:
-      for t in entry.topicIds:
-        if t == topic:
-          result.incl(entry.mid)
-          break
+    # Work backwards from `pos` in the circular buffer
+    for entry in c.history[(c.pos + c.history.len - i) mod c.history.len]:
+      if entry.topic == topic:
+        result.incl(entry.msgId)
 
 func shift*(c: var MCache) =
-  for entry in c.history.pop():
-    c.msgs.del(entry.mid)
+  # Shift circular buffer to write to a new position, clearing it from past
+  # iterations
+  c.pos = (c.pos + 1) mod c.history.len
 
-  c.history.insert(@[])
+  for entry in c.history[c.pos]:
+    c.msgs.del(entry.msgId)
+
+  reset(c.history[c.pos])
 
 func init*(T: type MCache, window, history: Natural): T =
   T(

libp2p/protocols/pubsub/pubsub.nim

@@ -30,7 +30,6 @@ import ./errors as pubsub_errors,
        ../../errors,
        ../../utility
 
-import metrics
 import stew/results
 export results
 
@@ -138,18 +137,34 @@ method unsubscribePeer*(p: PubSub, peerId: PeerId) {.base, gcsafe.} =
 
   libp2p_pubsub_peers.set(p.peers.len.int64)
 
-proc send*(p: PubSub, peer: PubSubPeer, msg: RPCMsg) {.raises: [].} =
-  ## Attempt to send `msg` to remote peer
+proc send*(p: PubSub, peer: PubSubPeer, msg: RPCMsg, isHighPriority: bool) {.raises: [].} =
+  ## This procedure attempts to send a `msg` (of type `RPCMsg`) to the specified remote peer in the PubSub network.
   ##
+  ## Parameters:
+  ## - `p`: The `PubSub` instance.
+  ## - `peer`: An instance of `PubSubPeer` representing the peer to whom the message should be sent.
+  ## - `msg`: The `RPCMsg` instance that contains the message to be sent.
+  ## - `isHighPriority`: A boolean indicating whether the message should be treated as high priority.
+  ## High priority messages are sent immediately, while low priority messages are queued and sent only after all high
+  ## priority messages have been sent.
 
   trace "sending pubsub message to peer", peer, msg = shortLog(msg)
-  peer.send(msg, p.anonymize)
+  peer.send(msg, p.anonymize, isHighPriority)
 
 proc broadcast*(
   p: PubSub,
   sendPeers: auto, # Iteratble[PubSubPeer]
-  msg: RPCMsg) {.raises: [].} =
-  ## Attempt to send `msg` to the given peers
+  msg: RPCMsg,
+  isHighPriority: bool) {.raises: [].} =
+  ## This procedure attempts to send a `msg` (of type `RPCMsg`) to a specified group of peers in the PubSub network.
+  ##
+  ## Parameters:
+  ## - `p`: The `PubSub` instance.
+  ## - `sendPeers`: An iterable of `PubSubPeer` instances representing the peers to whom the message should be sent.
+  ## - `msg`: The `RPCMsg` instance that contains the message to be broadcast.
+  ## - `isHighPriority`: A boolean indicating whether the message should be treated as high priority.
+  ## High priority messages are sent immediately, while low priority messages are queued and sent only after all high
+  ## priority messages have been sent.
 
   let npeers = sendPeers.len.int64
   for sub in msg.subscriptions:
@@ -165,28 +180,28 @@ proc broadcast*(
         libp2p_pubsub_broadcast_unsubscriptions.inc(npeers, labelValues = ["generic"])
 
   for smsg in msg.messages:
-    for topic in smsg.topicIds:
-      if p.knownTopics.contains(topic):
-        libp2p_pubsub_broadcast_messages.inc(npeers, labelValues = [topic])
-      else:
-        libp2p_pubsub_broadcast_messages.inc(npeers, labelValues = ["generic"])
+    let topic = smsg.topic
+    if p.knownTopics.contains(topic):
+      libp2p_pubsub_broadcast_messages.inc(npeers, labelValues = [topic])
+    else:
+      libp2p_pubsub_broadcast_messages.inc(npeers, labelValues = ["generic"])
 
   msg.control.withValue(control):
     libp2p_pubsub_broadcast_iwant.inc(npeers * control.iwant.len.int64)
 
     for ihave in control.ihave:
-      if p.knownTopics.contains(ihave.topicId):
-        libp2p_pubsub_broadcast_ihave.inc(npeers, labelValues = [ihave.topicId])
+      if p.knownTopics.contains(ihave.topicID):
+        libp2p_pubsub_broadcast_ihave.inc(npeers, labelValues = [ihave.topicID])
       else:
         libp2p_pubsub_broadcast_ihave.inc(npeers, labelValues = ["generic"])
     for graft in control.graft:
-      if p.knownTopics.contains(graft.topicId):
-        libp2p_pubsub_broadcast_graft.inc(npeers, labelValues = [graft.topicId])
+      if p.knownTopics.contains(graft.topicID):
+        libp2p_pubsub_broadcast_graft.inc(npeers, labelValues = [graft.topicID])
       else:
         libp2p_pubsub_broadcast_graft.inc(npeers, labelValues = ["generic"])
     for prune in control.prune:
-      if p.knownTopics.contains(prune.topicId):
-        libp2p_pubsub_broadcast_prune.inc(npeers, labelValues = [prune.topicId])
+      if p.knownTopics.contains(prune.topicID):
+        libp2p_pubsub_broadcast_prune.inc(npeers, labelValues = [prune.topicID])
       else:
         libp2p_pubsub_broadcast_prune.inc(npeers, labelValues = ["generic"])
 
@@ -195,19 +210,19 @@ proc broadcast*(
 
   if anyIt(sendPeers, it.hasObservers):
     for peer in sendPeers:
-      p.send(peer, msg)
+      p.send(peer, msg, isHighPriority)
   else:
     # Fast path that only encodes message once
     let encoded = encodeRpcMsg(msg, p.anonymize)
     for peer in sendPeers:
-      asyncSpawn peer.sendEncoded(encoded)
+      asyncSpawn peer.sendEncoded(encoded, isHighPriority)
 
 proc sendSubs*(p: PubSub,
                peer: PubSubPeer,
                topics: openArray[string],
                subscribe: bool) =
   ## send subscriptions to remote peer
-  p.send(peer, RPCMsg.withSubs(topics, subscribe))
+  p.send(peer, RPCMsg.withSubs(topics, subscribe), isHighPriority = true)
 
   for topic in topics:
     if subscribe:
@@ -236,29 +251,27 @@ proc updateMetrics*(p: PubSub, rpcMsg: RPCMsg) =
         libp2p_pubsub_received_unsubscriptions.inc(labelValues = ["generic"])
 
   for i in 0..<rpcMsg.messages.len():
-    template smsg: untyped = rpcMsg.messages[i]
-    for j in 0..<smsg.topicIds.len():
-      template topic: untyped = smsg.topicIds[j]
-      if p.knownTopics.contains(topic):
-        libp2p_pubsub_received_messages.inc(labelValues = [topic])
-      else:
-        libp2p_pubsub_received_messages.inc(labelValues = ["generic"])
+    let topic = rpcMsg.messages[i].topic
+    if p.knownTopics.contains(topic):
+      libp2p_pubsub_received_messages.inc(labelValues = [topic])
+    else:
+      libp2p_pubsub_received_messages.inc(labelValues = ["generic"])
 
   rpcMsg.control.withValue(control):
     libp2p_pubsub_received_iwant.inc(control.iwant.len.int64)
     for ihave in control.ihave:
-      if p.knownTopics.contains(ihave.topicId):
-        libp2p_pubsub_received_ihave.inc(labelValues = [ihave.topicId])
+      if p.knownTopics.contains(ihave.topicID):
+        libp2p_pubsub_received_ihave.inc(labelValues = [ihave.topicID])
       else:
         libp2p_pubsub_received_ihave.inc(labelValues = ["generic"])
     for graft in control.graft:
-      if p.knownTopics.contains(graft.topicId):
-        libp2p_pubsub_received_graft.inc(labelValues = [graft.topicId])
+      if p.knownTopics.contains(graft.topicID):
+        libp2p_pubsub_received_graft.inc(labelValues = [graft.topicID])
       else:
         libp2p_pubsub_received_graft.inc(labelValues = ["generic"])
     for prune in control.prune:
-      if p.knownTopics.contains(prune.topicId):
-        libp2p_pubsub_received_prune.inc(labelValues = [prune.topicId])
+      if p.knownTopics.contains(prune.topicID):
+        libp2p_pubsub_received_prune.inc(labelValues = [prune.topicID])
       else:
         libp2p_pubsub_received_prune.inc(labelValues = ["generic"])
 
@@ -273,11 +286,14 @@ method onNewPeer(p: PubSub, peer: PubSubPeer) {.base, gcsafe.} = discard
 method onPubSubPeerEvent*(p: PubSub, peer: PubSubPeer, event: PubSubPeerEvent) {.base, gcsafe.} =
   # Peer event is raised for the send connection in particular
   case event.kind
-  of PubSubPeerEventKind.Connected:
+  of PubSubPeerEventKind.StreamOpened:
     if p.topics.len > 0:
       p.sendSubs(peer, toSeq(p.topics.keys), true)
-  of PubSubPeerEventKind.Disconnected:
+  of PubSubPeerEventKind.StreamClosed:
     discard
+  of PubSubPeerEventKind.DisconnectionRequested:
+    discard
+
 
 method getOrCreatePeer*(
     p: PubSub,
@@ -501,7 +517,7 @@ method addValidator*(p: PubSub,
   ## will be sent to `hook`. `hook` can return either `Accept`,
   ## `Ignore` or `Reject` (which can descore the peer)
   for t in topic:
-    trace "adding validator for topic", topicId = t
+    trace "adding validator for topic", topic = t
     p.validators.mgetOrPut(t, HashSet[ValidatorHandler]()).incl(hook)
 
 method removeValidator*(p: PubSub,
@@ -516,13 +532,13 @@ method removeValidator*(p: PubSub,
 method validate*(p: PubSub, message: Message): Future[ValidationResult] {.async, base.} =
   var pending: seq[Future[ValidationResult]]
   trace "about to validate message"
-  for topic in message.topicIds:
-    trace "looking for validators on topic", topicId = topic,
-                                             registered = toSeq(p.validators.keys)
-    if topic in p.validators:
-      trace "running validators for topic", topicId = topic
-      for validator in p.validators[topic]:
-        pending.add(validator(topic, message))
+  let topic = message.topic
+  trace "looking for validators on topic",
+    topic = topic, registered = toSeq(p.validators.keys)
+  if topic in p.validators:
+    trace "running validators for topic", topic = topic
+    for validator in p.validators[topic]:
+      pending.add(validator(topic, message))
 
   result = ValidationResult.Accept
   let futs = await allFinished(pending)

libp2p/protocols/pubsub/pubsubpeer.nim

@@ -28,10 +28,18 @@ logScope:
 
 when defined(libp2p_expensive_metrics):
   declareCounter(libp2p_pubsub_sent_messages, "number of messages sent", labels = ["id", "topic"])
-  declareCounter(libp2p_pubsub_received_messages, "number of messages received", labels = ["id", "topic"])
   declareCounter(libp2p_pubsub_skipped_received_messages, "number of received skipped messages", labels = ["id"])
   declareCounter(libp2p_pubsub_skipped_sent_messages, "number of sent skipped messages", labels = ["id"])
 
+when defined(pubsubpeer_queue_metrics):
+  declareGauge(libp2p_gossipsub_priority_queue_size, "the number of messages in the priority queue", labels = ["id"])
+  declareGauge(libp2p_gossipsub_non_priority_queue_size, "the number of messages in the non-priority queue", labels = ["id"])
+
+const
+  DefaultMaxNumElementsInNonPriorityQueue* = 1024
+  BehaviourPenaltyFoNonPriorityQueueOverLimit = 0.0001  # this value is quite arbitrary and was found empirically
+  # to result in a behaviourPenalty around [0.1, 0.2] when the score is updated.
+
 type
   PeerRateLimitError* = object of CatchableError
 
@@ -40,8 +48,9 @@ type
     onSend*: proc(peer: PubSubPeer; msgs: var RPCMsg) {.gcsafe, raises: [].}
 
   PubSubPeerEventKind* {.pure.} = enum
-    Connected
-    Disconnected
+    StreamOpened
+    StreamClosed
+    DisconnectionRequested # tells gossipsub that the transport connection to the peer should be closed
 
   PubSubPeerEvent* = object
     kind*: PubSubPeerEventKind
@@ -50,6 +59,14 @@ type
   DropConn* = proc(peer: PubSubPeer) {.gcsafe, raises: [].} # have to pass peer as it's unknown during init
   OnEvent* = proc(peer: PubSubPeer, event: PubSubPeerEvent) {.gcsafe, raises: [].}
 
+  RpcMessageQueue* = ref object
+    # Tracks async tasks for sending high-priority peer-published messages.
+    sendPriorityQueue: Deque[Future[void]]
+    # Queue for lower-priority messages, like "IWANT" replies and relay messages.
+    nonPriorityQueue: AsyncQueue[seq[byte]]
+    # Task for processing non-priority message queue.
+    sendNonPriorityTask: Future[void]
+
   PubSubPeer* = ref object of RootObj
     getConn*: GetConn                   # callback to establish a new send connection
     onEvent*: OnEvent                   # Connectivity updates for peer
@@ -63,7 +80,10 @@ type
 
     score*: float64
     sentIHaves*: Deque[HashSet[MessageId]]
-    heDontWants*: Deque[HashSet[MessageId]]
+    heDontWants*: Deque[HashSet[SaltedId]]
+      ## IDONTWANT contains unvalidated message id:s which may be long and/or
+      ## expensive to look up, so we apply the same salting to them as during
+      ## unvalidated message processing
     iHaveBudget*: int
     pingBudget*: int
     maxMessageSize: int
@@ -71,6 +91,9 @@ type
     behaviourPenalty*: float64 # the eventual penalty score
     overheadRateLimitOpt*: Opt[TokenBucket]
 
+    rpcmessagequeue: RpcMessageQueue
+    maxNumElementsInNonPriorityQueue*: int # The max number of elements allowed in the non-priority queue.
+
   RPCHandler* = proc(peer: PubSubPeer, data: seq[byte]): Future[void]
     {.gcsafe, raises: [].}
 
@@ -83,6 +106,16 @@ when defined(libp2p_agents_metrics):
       #so we have to read the parents short agent..
       p.sendConn.getWrapped().shortAgent
 
+proc getAgent*(peer: PubSubPeer): string =
+  return
+    when defined(libp2p_agents_metrics):
+      if peer.shortAgent.len > 0:
+        peer.shortAgent
+      else:
+        "unknown"
+    else:
+      "unknown"
+
 func hash*(p: PubSubPeer): Hash =
   p.peerId.hash
 
@@ -138,12 +171,6 @@ proc handle*(p: PubSubPeer, conn: Connection) {.async.} =
           conn, peer = p, closed = conn.closed,
           data = data.shortLog
 
-        when defined(libp2p_expensive_metrics):
-          for m in rmsg.messages:
-            for t in m.topicIDs:
-              # metrics
-              libp2p_pubsub_received_messages.inc(labelValues = [$p.peerId, t])
-
         await p.handler(p, data)
         data = newSeq[byte]() # Release memory
     except PeerRateLimitError as exc:
@@ -164,6 +191,24 @@ proc handle*(p: PubSubPeer, conn: Connection) {.async.} =
     debug "exiting pubsub read loop",
       conn, peer = p, closed = conn.closed
 
+proc closeSendConn(p: PubSubPeer, event: PubSubPeerEventKind) {.async.} =
+  if p.sendConn != nil:
+    debug "Removing send connection", p, conn = p.sendConn
+    await p.sendConn.close()
+    p.sendConn = nil
+
+  if not p.connectedFut.finished:
+    p.connectedFut.complete()
+
+  try:
+    if p.onEvent != nil:
+      p.onEvent(p, PubSubPeerEvent(kind: event))
+  except CancelledError as exc:
+    raise exc
+  except CatchableError as exc:
+    debug "Errors during diconnection events", error = exc.msg
+  # don't cleanup p.address else we leak some gossip stat table
+
 proc connectOnce(p: PubSubPeer): Future[void] {.async.} =
   try:
     if p.connectedFut.finished:
@@ -176,7 +221,7 @@ proc connectOnce(p: PubSubPeer): Future[void] {.async.} =
     # remote peer - if we had multiple channels up and one goes down, all
     # stop working so we make an effort to only keep a single channel alive
 
-    trace "Get new send connection", p, newConn
+    debug "Get new send connection", p, newConn
 
     # Careful to race conditions here.
     # Topic subscription relies on either connectedFut
@@ -186,37 +231,21 @@ proc connectOnce(p: PubSubPeer): Future[void] {.async.} =
     p.address = if p.sendConn.observedAddr.isSome: some(p.sendConn.observedAddr.get) else: none(MultiAddress)
 
     if p.onEvent != nil:
-      p.onEvent(p, PubSubPeerEvent(kind: PubSubPeerEventKind.Connected))
+      p.onEvent(p, PubSubPeerEvent(kind: PubSubPeerEventKind.StreamOpened))
 
     await handle(p, newConn)
   finally:
-    if p.sendConn != nil:
-      trace "Removing send connection", p, conn = p.sendConn
-      await p.sendConn.close()
-      p.sendConn = nil
+    await p.closeSendConn(PubSubPeerEventKind.StreamClosed)
 
-    if not p.connectedFut.finished:
-      p.connectedFut.complete()
-
-    try:
-      if p.onEvent != nil:
-        p.onEvent(p, PubSubPeerEvent(kind: PubSubPeerEventKind.Disconnected))
-    except CancelledError as exc:
-      raise exc
-    except CatchableError as exc:
-      debug "Errors during diconnection events", error = exc.msg
-
-    # don't cleanup p.address else we leak some gossip stat table
-
-proc connectImpl(p: PubSubPeer) {.async.} =
+proc connectImpl(peer: PubSubPeer) {.async.} =
   try:
     # Keep trying to establish a connection while it's possible to do so - the
     # send connection might get disconnected due to a timeout or an unrelated
     # issue so we try to get a new on
     while true:
-      await connectOnce(p)
+      await connectOnce(peer)
   except CatchableError as exc: # never cancelled
-    debug "Could not establish send connection", msg = exc.msg
+    debug "Could not establish send connection", peer, msg = exc.msg
 
 proc connect*(p: PubSubPeer) =
   if p.connected:
@@ -230,35 +259,37 @@ proc hasSendConn*(p: PubSubPeer): bool =
 template sendMetrics(msg: RPCMsg): untyped =
   when defined(libp2p_expensive_metrics):
     for x in msg.messages:
-      for t in x.topicIDs:
-        # metrics
-        libp2p_pubsub_sent_messages.inc(labelValues = [$p.peerId, t])
+      # metrics
+      libp2p_pubsub_sent_messages.inc(labelValues = [$p.peerId, x.topic])
 
-proc sendEncoded*(p: PubSubPeer, msg: seq[byte]) {.raises: [], async.} =
-  doAssert(not isNil(p), "pubsubpeer nil!")
+proc clearSendPriorityQueue(p: PubSubPeer) =
+  if p.rpcmessagequeue.sendPriorityQueue.len == 0:
+    return # fast path
 
-  if msg.len <= 0:
-    debug "empty message, skipping", p, msg = shortLog(msg)
-    return
+  while p.rpcmessagequeue.sendPriorityQueue.len > 0 and
+      p.rpcmessagequeue.sendPriorityQueue[0].finished:
+    discard p.rpcmessagequeue.sendPriorityQueue.popFirst()
 
-  if msg.len > p.maxMessageSize:
-    info "trying to send a msg too big for pubsub", maxSize=p.maxMessageSize, msgSize=msg.len
-    return
+  while p.rpcmessagequeue.sendPriorityQueue.len > 0 and
+      p.rpcmessagequeue.sendPriorityQueue[^1].finished:
+    discard p.rpcmessagequeue.sendPriorityQueue.popLast()
 
-  if p.sendConn == nil:
-    # Wait for a send conn to be setup. `connectOnce` will
-    # complete this even if the sendConn setup failed
-    await p.connectedFut
+  when defined(pubsubpeer_queue_metrics):
+    libp2p_gossipsub_priority_queue_size.set(
+      value = p.rpcmessagequeue.sendPriorityQueue.len.int64,
+      labelValues = [$p.peerId])
 
-  var conn = p.sendConn
-  if conn == nil or conn.closed():
-    debug "No send connection", p, msg = shortLog(msg)
-    return
+proc clearNonPriorityQueue*(p: PubSubPeer) =
+  if len(p.rpcmessagequeue.nonPriorityQueue) > 0:
+    p.rpcmessagequeue.nonPriorityQueue.clear()
 
-  trace "sending encoded msgs to peer", conn, encoded = shortLog(msg)
+    when defined(pubsubpeer_queue_metrics):
+      libp2p_gossipsub_non_priority_queue_size.set(labelValues = [$p.peerId], value = 0)
 
+proc sendMsgContinue(conn: Connection, msgFut: Future[void]) {.async.} =
+  # Continuation for a pending `sendMsg` future from below
   try:
-    await conn.writeLp(msg)
+    await msgFut
     trace "sent pubsub message to remote", conn
   except CatchableError as exc: # never cancelled
     # Because we detach the send call from the currently executing task using
@@ -269,6 +300,80 @@ proc sendEncoded*(p: PubSubPeer, msg: seq[byte]) {.raises: [], async.} =
 
     await conn.close() # This will clean up the send connection
 
+proc sendMsgSlow(p: PubSubPeer, msg: seq[byte]) {.async.} =
+  # Slow path of `sendMsg` where msg is held in memory while send connection is
+  # being set up
+  if p.sendConn == nil:
+    # Wait for a send conn to be setup. `connectOnce` will
+    # complete this even if the sendConn setup failed
+    discard await race(p.connectedFut)
+
+  var conn = p.sendConn
+  if conn == nil or conn.closed():
+    debug "No send connection", p, msg = shortLog(msg)
+    return
+
+  trace "sending encoded msg to peer", conn, encoded = shortLog(msg)
+  await sendMsgContinue(conn, conn.writeLp(msg))
+
+proc sendMsg(p: PubSubPeer, msg: seq[byte]): Future[void] =
+  if p.sendConn != nil and not p.sendConn.closed():
+    # Fast path that avoids copying msg (which happens for {.async.})
+    let conn = p.sendConn
+
+    trace "sending encoded msg to peer", conn, encoded = shortLog(msg)
+    let f = conn.writeLp(msg)
+    if not f.completed():
+      sendMsgContinue(conn, f)
+    else:
+      f
+  else:
+    sendMsgSlow(p, msg)
+
+proc sendEncoded*(p: PubSubPeer, msg: seq[byte], isHighPriority: bool): Future[void] =
+  ## Asynchronously sends an encoded message to a specified `PubSubPeer`.
+  ##
+  ## Parameters:
+  ## - `p`: The `PubSubPeer` instance to which the message is to be sent.
+  ## - `msg`: The message to be sent, encoded as a sequence of bytes (`seq[byte]`).
+  ## - `isHighPriority`: A boolean indicating whether the message should be treated as high priority.
+  ## High priority messages are sent immediately, while low priority messages are queued and sent only after all high
+  ## priority messages have been sent.
+  doAssert(not isNil(p), "pubsubpeer nil!")
+
+  p.clearSendPriorityQueue()
+
+  # When queues are empty, skipping the non-priority queue for low priority
+  # messages reduces latency
+  let emptyQueues =
+   (p.rpcmessagequeue.sendPriorityQueue.len() +
+        p.rpcmessagequeue.nonPriorityQueue.len()) == 0
+
+  if msg.len <= 0:
+    debug "empty message, skipping", peer = p, msg = shortLog(msg)
+    Future[void].completed()
+  elif msg.len > p.maxMessageSize:
+    info "trying to send a msg too big for pubsub", maxSize=p.maxMessageSize, msgSize=msg.len
+    Future[void].completed()
+  elif isHighPriority or emptyQueues:
+    let f = p.sendMsg(msg)
+    if not f.finished:
+      p.rpcmessagequeue.sendPriorityQueue.addLast(f)
+      when defined(pubsubpeer_queue_metrics):
+        libp2p_gossipsub_priority_queue_size.inc(labelValues = [$p.peerId])
+    f
+  else:
+    if len(p.rpcmessagequeue.nonPriorityQueue) >= p.maxNumElementsInNonPriorityQueue:
+      p.behaviourPenalty += BehaviourPenaltyFoNonPriorityQueueOverLimit
+      trace "Peer has reached maxNumElementsInNonPriorityQueue. Discarding message and applying behaviour penalty.", peer = p, score = p.score,
+        behaviourPenalty = p.behaviourPenalty, agent = p.getAgent()
+      Future[void].completed()
+    else:
+      let f = p.rpcmessagequeue.nonPriorityQueue.addLast(msg)
+      when defined(pubsubpeer_queue_metrics):
+        libp2p_gossipsub_non_priority_queue_size.inc(labelValues = [$p.peerId])
+      f
+
 iterator splitRPCMsg(peer: PubSubPeer, rpcMsg: RPCMsg, maxSize: int, anonymize: bool): seq[byte] =
   ## This iterator takes an `RPCMsg` and sequentially repackages its Messages into new `RPCMsg` instances.
   ## Each new `RPCMsg` accumulates Messages until reaching the specified `maxSize`. If a single Message
@@ -304,7 +409,16 @@ iterator splitRPCMsg(peer: PubSubPeer, rpcMsg: RPCMsg, maxSize: int, anonymize:
   else:
     trace "message too big to sent", peer, rpcMsg = shortLog(currentRPCMsg)
 
-proc send*(p: PubSubPeer, msg: RPCMsg, anonymize: bool) {.raises: [].} =
+proc send*(p: PubSubPeer, msg: RPCMsg, anonymize: bool, isHighPriority: bool) {.raises: [].} =
+  ## Asynchronously sends an `RPCMsg` to a specified `PubSubPeer` with an option for anonymization.
+  ##
+  ## Parameters:
+  ## - `p`: The `PubSubPeer` instance to which the message is to be sent.
+  ## - `msg`: The `RPCMsg` instance representing the message to be sent.
+  ## - `anonymize`: A boolean flag indicating whether the message should be sent with anonymization.
+  ## - `isHighPriority`: A boolean flag indicating whether the message should be treated as high priority.
+  ## High priority messages are sent immediately, while low priority messages are queued and sent only after all high
+  ## priority messages have been sent.
   # When sending messages, we take care to re-encode them with the right
   # anonymization flag to ensure that we're not penalized for sending invalid
   # or malicious data on the wire - in particular, re-encoding protects against
@@ -324,11 +438,11 @@ proc send*(p: PubSubPeer, msg: RPCMsg, anonymize: bool) {.raises: [].} =
 
   if encoded.len > p.maxMessageSize and msg.messages.len > 1:
     for encodedSplitMsg in splitRPCMsg(p, msg, p.maxMessageSize, anonymize):
-      asyncSpawn p.sendEncoded(encodedSplitMsg)
+      asyncSpawn p.sendEncoded(encodedSplitMsg, isHighPriority)
   else:
     # If the message size is within limits, send it as is
     trace "sending msg to peer", peer = p, rpcMsg = shortLog(msg)
-    asyncSpawn p.sendEncoded(encoded)
+    asyncSpawn p.sendEncoded(encoded, isHighPriority)
 
 proc canAskIWant*(p: PubSubPeer, msgId: MessageId): bool =
   for sentIHave in p.sentIHaves.mitems():
@@ -337,6 +451,45 @@ proc canAskIWant*(p: PubSubPeer, msgId: MessageId): bool =
       return true
   return false
 
+proc sendNonPriorityTask(p: PubSubPeer) {.async.} =
+  while true:
+     # we send non-priority messages only if there are no pending priority messages
+     let msg = await p.rpcmessagequeue.nonPriorityQueue.popFirst()
+     while p.rpcmessagequeue.sendPriorityQueue.len > 0:
+       p.clearSendPriorityQueue()
+       # waiting for the last future minimizes the number of times we have to
+       # wait for something (each wait = performance cost) -
+       # clearSendPriorityQueue ensures we're not waiting for an already-finished
+       # future
+       if p.rpcmessagequeue.sendPriorityQueue.len > 0:
+        # `race` prevents `p.rpcmessagequeue.sendPriorityQueue[^1]` from being
+        # cancelled when this task is cancelled
+        discard await race(p.rpcmessagequeue.sendPriorityQueue[^1])
+     when defined(pubsubpeer_queue_metrics):
+       libp2p_gossipsub_non_priority_queue_size.dec(labelValues = [$p.peerId])
+     await p.sendMsg(msg)
+
+proc startSendNonPriorityTask(p: PubSubPeer) =
+  debug "starting sendNonPriorityTask", p
+  if p.rpcmessagequeue.sendNonPriorityTask.isNil:
+    p.rpcmessagequeue.sendNonPriorityTask = p.sendNonPriorityTask()
+
+proc stopSendNonPriorityTask*(p: PubSubPeer) =
+  if not p.rpcmessagequeue.sendNonPriorityTask.isNil:
+    debug "stopping sendNonPriorityTask", p
+    p.rpcmessagequeue.sendNonPriorityTask.cancelSoon()
+    p.rpcmessagequeue.sendNonPriorityTask = nil
+    p.rpcmessagequeue.sendPriorityQueue.clear()
+    when defined(pubsubpeer_queue_metrics):
+      libp2p_gossipsub_priority_queue_size.set(labelValues = [$p.peerId], value = 0)
+    p.clearNonPriorityQueue()
+
+proc new(T: typedesc[RpcMessageQueue]): T =
+  return T(
+    sendPriorityQueue: initDeque[Future[void]](),
+    nonPriorityQueue: newAsyncQueue[seq[byte]]()
+  )
+
 proc new*(
   T: typedesc[PubSubPeer],
   peerId: PeerId,
@@ -344,6 +497,7 @@ proc new*(
   onEvent: OnEvent,
   codec: string,
   maxMessageSize: int,
+  maxNumElementsInNonPriorityQueue: int = DefaultMaxNumElementsInNonPriorityQueue,
   overheadRateLimitOpt: Opt[TokenBucket] = Opt.none(TokenBucket)): T =
 
   result = T(
@@ -353,17 +507,10 @@ proc new*(
     peerId: peerId,
     connectedFut: newFuture[void](),
     maxMessageSize: maxMessageSize,
-    overheadRateLimitOpt: overheadRateLimitOpt
+    overheadRateLimitOpt: overheadRateLimitOpt,
+    rpcmessagequeue: RpcMessageQueue.new(),
+    maxNumElementsInNonPriorityQueue: maxNumElementsInNonPriorityQueue
   )
   result.sentIHaves.addFirst(default(HashSet[MessageId]))
-  result.heDontWants.addFirst(default(HashSet[MessageId]))
-
-proc getAgent*(peer: PubSubPeer): string =
-  return
-    when defined(libp2p_agents_metrics):
-      if peer.shortAgent.len > 0:
-        peer.shortAgent
-      else:
-        "unknown"
-    else:
-      "unknown"
+  result.heDontWants.addFirst(default(HashSet[SaltedId]))
+  result.startSendNonPriorityTask()

libp2p/protocols/pubsub/rpc/message.nim

@@ -63,7 +63,7 @@ proc init*(
     seqno: Option[uint64],
     sign: bool = true): Message
     {.gcsafe, raises: [LPError].} =
-  var msg = Message(data: data, topicIDs: @[topic])
+  var msg = Message(data: data, topic: topic)
 
   # order matters, we want to include seqno in the signature
   seqno.withValue(seqn):
@@ -87,7 +87,7 @@ proc init*(
     topic: string,
     seqno: Option[uint64]): Message
     {.gcsafe, raises: [LPError].} =
-  var msg = Message(data: data, topicIDs: @[topic])
+  var msg = Message(data: data, topic: topic)
   msg.fromPeer = peerId
 
   seqno.withValue(seqn):

libp2p/protocols/pubsub/rpc/messages.nim

@@ -9,8 +9,8 @@
 
 {.push raises: [].}
 
-import options, sequtils, sugar
-import "../../.."/[
+import options, sequtils
+import ../../../[
         peerid,
         routing_record,
         utility
@@ -27,52 +27,58 @@ proc expectedFields[T](t: typedesc[T], existingFieldNames: seq[string]) {.raises
     raise newException(CatchableError, $T & " fields changed, please search for and revise all relevant procs. New fields: " & $fieldNames)
 
 type
-    PeerInfoMsg* = object
-      peerId*: PeerId
-      signedPeerRecord*: seq[byte]
+  PeerInfoMsg* = object
+    peerId*: PeerId
+    signedPeerRecord*: seq[byte]
 
-    SubOpts* = object
-      subscribe*: bool
-      topic*: string
+  SubOpts* = object
+    subscribe*: bool
+    topic*: string
 
-    MessageId* = seq[byte]
+  MessageId* = seq[byte]
 
-    Message* = object
-      fromPeer*: PeerId
-      data*: seq[byte]
-      seqno*: seq[byte]
-      topicIds*: seq[string]
-      signature*: seq[byte]
-      key*: seq[byte]
+  SaltedId* = object
+    # Salted hash of message ID - used instead of the ordinary message ID to
+    # avoid hash poisoning attacks and to make memory usage more predictable
+    # with respect to the variable-length message id
+    data*: MDigest[256]
 
-    ControlMessage* = object
-      ihave*: seq[ControlIHave]
-      iwant*: seq[ControlIWant]
-      graft*: seq[ControlGraft]
-      prune*: seq[ControlPrune]
-      idontwant*: seq[ControlIWant]
+  Message* = object
+    fromPeer*: PeerId
+    data*: seq[byte]
+    seqno*: seq[byte]
+    topic*: string
+    signature*: seq[byte]
+    key*: seq[byte]
 
-    ControlIHave* = object
-      topicId*: string
-      messageIds*: seq[MessageId]
+  ControlMessage* = object
+    ihave*: seq[ControlIHave]
+    iwant*: seq[ControlIWant]
+    graft*: seq[ControlGraft]
+    prune*: seq[ControlPrune]
+    idontwant*: seq[ControlIWant]
 
-    ControlIWant* = object
-      messageIds*: seq[MessageId]
+  ControlIHave* = object
+    topicID*: string
+    messageIDs*: seq[MessageId]
 
-    ControlGraft* = object
-      topicId*: string
+  ControlIWant* = object
+    messageIDs*: seq[MessageId]
 
-    ControlPrune* = object
-      topicId*: string
-      peers*: seq[PeerInfoMsg]
-      backoff*: uint64
+  ControlGraft* = object
+    topicID*: string
 
-    RPCMsg* = object
-      subscriptions*: seq[SubOpts]
-      messages*: seq[Message]
-      control*: Option[ControlMessage]
-      ping*: seq[byte]
-      pong*: seq[byte]
+  ControlPrune* = object
+    topicID*: string
+    peers*: seq[PeerInfoMsg]
+    backoff*: uint64
+
+  RPCMsg* = object
+    subscriptions*: seq[SubOpts]
+    messages*: seq[Message]
+    control*: Option[ControlMessage]
+    ping*: seq[byte]
+    pong*: seq[byte]
 
 func withSubs*(
     T: type RPCMsg, topics: openArray[string], subscribe: bool): T =
@@ -81,23 +87,23 @@ func withSubs*(
 
 func shortLog*(s: ControlIHave): auto =
   (
-    topicId: s.topicId.shortLog,
-    messageIds: mapIt(s.messageIds, it.shortLog)
+    topic: s.topicID.shortLog,
+    messageIDs: mapIt(s.messageIDs, it.shortLog)
   )
 
 func shortLog*(s: ControlIWant): auto =
   (
-    messageIds: mapIt(s.messageIds, it.shortLog)
+    messageIDs: mapIt(s.messageIDs, it.shortLog)
   )
 
 func shortLog*(s: ControlGraft): auto =
   (
-    topicId: s.topicId.shortLog
+    topic: s.topicID.shortLog
   )
 
 func shortLog*(s: ControlPrune): auto =
   (
-    topicId: s.topicId.shortLog
+    topic: s.topicID.shortLog
   )
 
 func shortLog*(c: ControlMessage): auto =
@@ -113,7 +119,7 @@ func shortLog*(msg: Message): auto =
     fromPeer: msg.fromPeer.shortLog,
     data: msg.data.shortLog,
     seqno: msg.seqno.shortLog,
-    topicIds: $msg.topicIds,
+    topic: msg.topic,
     signature: msg.signature.shortLog,
     key: msg.key.shortLog
   )
@@ -133,35 +139,35 @@ static: expectedFields(SubOpts, @["subscribe", "topic"])
 proc byteSize(subOpts: SubOpts): int =
   1 + subOpts.topic.len # 1 byte for the bool
 
-static: expectedFields(Message, @["fromPeer", "data", "seqno", "topicIds", "signature", "key"])
+static: expectedFields(Message, @["fromPeer", "data", "seqno", "topic", "signature", "key"])
 proc byteSize*(msg: Message): int =
-  msg.fromPeer.len + msg.data.len + msg.seqno.len +
-         msg.signature.len + msg.key.len + msg.topicIds.foldl(a + b.len, 0)
+  msg.fromPeer.len + msg.data.len + msg.seqno.len + msg.signature.len + msg.key.len +
+    msg.topic.len
 
 proc byteSize*(msgs: seq[Message]): int =
   msgs.foldl(a + b.byteSize, 0)
 
-static: expectedFields(ControlIHave, @["topicId", "messageIds"])
+static: expectedFields(ControlIHave, @["topicID", "messageIDs"])
 proc byteSize(controlIHave: ControlIHave): int =
-  controlIHave.topicId.len + controlIHave.messageIds.foldl(a + b.len, 0)
+  controlIHave.topicID.len + controlIHave.messageIDs.foldl(a + b.len, 0)
 
 proc byteSize*(ihaves: seq[ControlIHave]): int =
   ihaves.foldl(a + b.byteSize, 0)
 
-static: expectedFields(ControlIWant, @["messageIds"])
+static: expectedFields(ControlIWant, @["messageIDs"])
 proc byteSize(controlIWant: ControlIWant): int =
-  controlIWant.messageIds.foldl(a + b.len, 0)
+  controlIWant.messageIDs.foldl(a + b.len, 0)
 
 proc byteSize*(iwants: seq[ControlIWant]): int =
   iwants.foldl(a + b.byteSize, 0)
 
-static: expectedFields(ControlGraft, @["topicId"])
+static: expectedFields(ControlGraft, @["topicID"])
 proc byteSize(controlGraft: ControlGraft): int =
-  controlGraft.topicId.len
+  controlGraft.topicID.len
 
-static: expectedFields(ControlPrune, @["topicId", "peers", "backoff"])
+static: expectedFields(ControlPrune, @["topicID", "peers", "backoff"])
 proc byteSize(controlPrune: ControlPrune): int =
-  controlPrune.topicId.len + controlPrune.peers.foldl(a + b.byteSize, 0) + 8 # 8 bytes for uint64
+  controlPrune.topicID.len + controlPrune.peers.foldl(a + b.byteSize, 0) + 8 # 8 bytes for uint64
 
 static: expectedFields(ControlMessage, @["ihave", "iwant", "graft", "prune", "idontwant"])
 proc byteSize(control: ControlMessage): int =

libp2p/protocols/pubsub/rpc/protobuf.nim

@@ -29,7 +29,7 @@ when defined(libp2p_protobuf_metrics):
 
 proc write*(pb: var ProtoBuffer, field: int, graft: ControlGraft) =
   var ipb = initProtoBuffer()
-  ipb.write(1, graft.topicId)
+  ipb.write(1, graft.topicID)
   ipb.finish()
   pb.write(field, ipb)
 
@@ -45,7 +45,7 @@ proc write*(pb: var ProtoBuffer, field: int, infoMsg: PeerInfoMsg) =
 
 proc write*(pb: var ProtoBuffer, field: int, prune: ControlPrune) =
   var ipb = initProtoBuffer()
-  ipb.write(1, prune.topicId)
+  ipb.write(1, prune.topicID)
   for peer in prune.peers:
     ipb.write(2, peer)
   ipb.write(3, prune.backoff)
@@ -57,8 +57,8 @@ proc write*(pb: var ProtoBuffer, field: int, prune: ControlPrune) =
 
 proc write*(pb: var ProtoBuffer, field: int, ihave: ControlIHave) =
   var ipb = initProtoBuffer()
-  ipb.write(1, ihave.topicId)
-  for mid in ihave.messageIds:
+  ipb.write(1, ihave.topicID)
+  for mid in ihave.messageIDs:
     ipb.write(2, mid)
   ipb.finish()
   pb.write(field, ipb)
@@ -68,7 +68,7 @@ proc write*(pb: var ProtoBuffer, field: int, ihave: ControlIHave) =
 
 proc write*(pb: var ProtoBuffer, field: int, iwant: ControlIWant) =
   var ipb = initProtoBuffer()
-  for mid in iwant.messageIds:
+  for mid in iwant.messageIDs:
     ipb.write(1, mid)
   if len(ipb.buffer) > 0:
     ipb.finish()
@@ -110,8 +110,7 @@ proc encodeMessage*(msg: Message, anonymize: bool): seq[byte] =
   pb.write(2, msg.data)
   if len(msg.seqno) > 0 and not anonymize:
     pb.write(3, msg.seqno)
-  for topic in msg.topicIds:
-    pb.write(4, topic)
+  pb.write(4, msg.topic)
   if len(msg.signature) > 0 and not anonymize:
     pb.write(5, msg.signature)
   if len(msg.key) > 0 and not anonymize:
@@ -133,10 +132,10 @@ proc decodeGraft*(pb: ProtoBuffer): ProtoResult[ControlGraft] {.
 
   trace "decodeGraft: decoding message"
   var control = ControlGraft()
-  if ? pb.getField(1, control.topicId):
-    trace "decodeGraft: read topicId", topic_id = control.topicId
+  if ? pb.getField(1, control.topicID):
+    trace "decodeGraft: read topicID", topicID = control.topicID
   else:
-    trace "decodeGraft: topicId is missing"
+    trace "decodeGraft: topicID is missing"
   ok(control)
 
 proc decodePeerInfoMsg*(pb: ProtoBuffer): ProtoResult[PeerInfoMsg] {.
@@ -160,10 +159,10 @@ proc decodePrune*(pb: ProtoBuffer): ProtoResult[ControlPrune] {.
 
   trace "decodePrune: decoding message"
   var control = ControlPrune()
-  if ? pb.getField(1, control.topicId):
-    trace "decodePrune: read topicId", topic_id = control.topicId
+  if ? pb.getField(1, control.topicID):
+    trace "decodePrune: read topicID", topic = control.topicID
   else:
-    trace "decodePrune: topicId is missing"
+    trace "decodePrune: topicID is missing"
   var bpeers: seq[seq[byte]]
   if ? pb.getRepeatedField(2, bpeers):
     for bpeer in bpeers:
@@ -179,12 +178,12 @@ proc decodeIHave*(pb: ProtoBuffer): ProtoResult[ControlIHave] {.
 
   trace "decodeIHave: decoding message"
   var control = ControlIHave()
-  if ? pb.getField(1, control.topicId):
-    trace "decodeIHave: read topicId", topic_id = control.topicId
+  if ? pb.getField(1, control.topicID):
+    trace "decodeIHave: read topicID", topic = control.topicID
   else:
-    trace "decodeIHave: topicId is missing"
-  if ? pb.getRepeatedField(2, control.messageIds):
-    trace "decodeIHave: read messageIDs", message_ids = control.messageIds
+    trace "decodeIHave: topicID is missing"
+  if ? pb.getRepeatedField(2, control.messageIDs):
+    trace "decodeIHave: read messageIDs", message_ids = control.messageIDs
   else:
     trace "decodeIHave: no messageIDs"
   ok(control)
@@ -195,8 +194,8 @@ proc decodeIWant*(pb: ProtoBuffer): ProtoResult[ControlIWant] {.inline.} =
 
   trace "decodeIWant: decoding message"
   var control = ControlIWant()
-  if ? pb.getRepeatedField(1, control.messageIds):
-    trace "decodeIWant: read messageIDs", message_ids = control.messageIds
+  if ? pb.getRepeatedField(1, control.messageIDs):
+    trace "decodeIWant: read messageIDs", message_ids = control.messageIDs
   else:
     trace "decodeIWant: no messageIDs"
   ok(control)
@@ -286,10 +285,11 @@ proc decodeMessage*(pb: ProtoBuffer): ProtoResult[Message] {.inline.} =
     trace "decodeMessage: read seqno", seqno = msg.seqno
   else:
     trace "decodeMessage: seqno is missing"
-  if ? pb.getRepeatedField(4, msg.topicIds):
-    trace "decodeMessage: read topics", topic_ids = msg.topicIds
+  if ?pb.getField(4, msg.topic):
+    trace "decodeMessage: read topic", topic = msg.topic
   else:
-    trace "decodeMessage: topics are missing"
+    trace "decodeMessage: topic is required"
+    return err(ProtoError.RequiredFieldMissing)
   if ? pb.getField(5, msg.signature):
     trace "decodeMessage: read signature", signature = msg.signature.shortLog()
   else:

libp2p/protocols/pubsub/timedcache.nim

@@ -9,12 +9,13 @@
 
 {.push raises: [].}
 
-import std/[tables]
-
+import std/[hashes, sets]
 import chronos/timer, stew/results
 
 import ../../utility
 
+export results
+
 const Timeout* = 10.seconds # default timeout in ms
 
 type
@@ -26,20 +27,38 @@ type
 
   TimedCache*[K] = object of RootObj
     head, tail: TimedEntry[K] # nim linked list doesn't allow inserting at pos
-    entries: Table[K, TimedEntry[K]]
+    entries: HashSet[TimedEntry[K]]
     timeout: Duration
 
+func `==`*[E](a, b: TimedEntry[E]): bool =
+  if isNil(a) == isNil(b):
+    isNil(a) or a.key == b.key
+  else:
+    false
+
+func hash*(a: TimedEntry): Hash =
+  if isNil(a):
+    default(Hash)
+  else:
+    hash(a[].key)
+
 func expire*(t: var TimedCache, now: Moment = Moment.now()) =
   while t.head != nil and t.head.expiresAt < now:
-    t.entries.del(t.head.key)
+    t.entries.excl(t.head)
     t.head.prev = nil
     t.head = t.head.next
     if t.head == nil: t.tail = nil
 
 func del*[K](t: var TimedCache[K], key: K): Opt[TimedEntry[K]] =
   # Removes existing key from cache, returning the previous value if present
-  var item: TimedEntry[K]
-  if t.entries.pop(key, item):
+  let tmp = TimedEntry[K](key: key)
+  if tmp in t.entries:
+    let item = try:
+      t.entries[tmp] # use the shared instance in the set
+    except KeyError:
+      raiseAssert "just checked"
+    t.entries.excl(item)
+
     if t.head == item: t.head = item.next
     if t.tail == item: t.tail = item.prev
 
@@ -55,14 +74,14 @@ func put*[K](t: var TimedCache[K], k: K, now = Moment.now()): bool =
   # refreshed.
   t.expire(now)
 
-  var previous = t.del(k) # Refresh existing item
-
-  var addedAt = now
-  previous.withValue(previous):
-    addedAt = previous.addedAt
+  let
+    previous = t.del(k) # Refresh existing item
+    addedAt = if previous.isSome():
+      previous[].addedAt
+    else:
+      now
 
   let node = TimedEntry[K](key: k, addedAt: addedAt, expiresAt: now + t.timeout)
-
   if t.head == nil:
     t.tail = node
     t.head = t.tail
@@ -83,16 +102,24 @@ func put*[K](t: var TimedCache[K], k: K, now = Moment.now()): bool =
       if cur == t.tail:
         t.tail = node
 
-  t.entries[k] = node
+  t.entries.incl(node)
 
   previous.isSome()
 
 func contains*[K](t: TimedCache[K], k: K): bool =
-  k in t.entries
+  let tmp = TimedEntry[K](key: k)
+  tmp in t.entries
 
-func addedAt*[K](t: TimedCache[K], k: K): Moment =
-  t.entries.getOrDefault(k).addedAt
+func addedAt*[K](t: var TimedCache[K], k: K): Moment =
+  let tmp = TimedEntry[K](key: k)
+  try:
+    if tmp in t.entries: # raising is slow
+      # Use shared instance from entries
+      return t.entries[tmp][].addedAt
+  except KeyError:
+    raiseAssert "just checked"
 
+  default(Moment)
 
 func init*[K](T: type TimedCache[K], timeout: Duration = Timeout): T =
   T(

libp2p/protocols/rendezvous.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -636,7 +636,7 @@ proc new*(T: typedesc[RendezVous],
     sema: newAsyncSemaphore(SemaphoreDefaultSize)
   )
   logScope: topics = "libp2p discovery rendezvous"
-  proc handleStream(conn: Connection, proto: string) {.async, gcsafe.} =
+  proc handleStream(conn: Connection, proto: string) {.async.} =
     try:
       let
         buf = await conn.readLp(4096)
@@ -678,17 +678,25 @@ proc deletesRegister(rdv: RendezVous) {.async.} =
     libp2p_rendezvous_registered.set(int64(total))
     libp2p_rendezvous_namespaces.set(int64(rdv.namespaces.len))
 
-method start*(rdv: RendezVous) {.async.} =
+method start*(
+    rdv: RendezVous
+): Future[void] {.async: (raises: [CancelledError], raw: true).} =
+  let fut = newFuture[void]()
+  fut.complete()
   if not rdv.registerDeletionLoop.isNil:
     warn "Starting rendezvous twice"
-    return
+    return fut
   rdv.registerDeletionLoop = rdv.deletesRegister()
   rdv.started = true
+  fut
 
-method stop*(rdv: RendezVous) {.async.} =
+method stop*(rdv: RendezVous): Future[void] {.async: (raises: [], raw: true).} =
+  let fut = newFuture[void]()
+  fut.complete()
   if rdv.registerDeletionLoop.isNil:
     warn "Stopping rendezvous without starting it"
-    return
+    return fut
   rdv.started = false
   rdv.registerDeletionLoop.cancel()
   rdv.registerDeletionLoop = nil
+  fut

libp2p/protocols/secure/noise.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -89,7 +89,7 @@ type
     readCs: CipherState
     writeCs: CipherState
 
-  NoiseError* = object of LPError
+  NoiseError* = object of LPStreamError
   NoiseHandshakeError* = object of NoiseError
   NoiseDecryptTagError* = object of NoiseError
   NoiseOversizedPayloadError* = object of NoiseError
@@ -99,10 +99,10 @@ type
 
 func shortLog*(conn: NoiseConnection): auto =
   try:
-    if conn.isNil: "NoiseConnection(nil)"
+    if conn == nil: "NoiseConnection(nil)"
     else: &"{shortLog(conn.peerId)}:{conn.oid}"
   except ValueError as exc:
-    raise newException(Defect, exc.msg)
+    raiseAssert(exc.msg)
 
 chronicles.formatIt(NoiseConnection): shortLog(it)
 
@@ -112,7 +112,7 @@ proc genKeyPair(rng: var HmacDrbgContext): KeyPair =
 
 proc hashProtocol(name: string): MDigest[256] =
   # If protocol_name is less than or equal to HASHLEN bytes in length,
-  # sets h equal to protocol_name with zero bytes appended to make HASHLEN bytes.
+  # sets h to protocol_name with zero bytes appended to make HASHLEN bytes.
   # Otherwise sets h = HASH(protocol_name).
 
   if name.len <= 32:
@@ -142,7 +142,7 @@ proc encrypt(
 
   inc state.n
   if state.n > NonceMax:
-    raise newException(NoiseNonceMaxError, "Noise max nonce value reached")
+    raise (ref NoiseNonceMaxError)(msg: "Noise max nonce value reached")
 
 proc encryptWithAd(state: var CipherState, ad, data: openArray[byte]): seq[byte]
   {.raises: [NoiseNonceMaxError].} =
@@ -168,10 +168,11 @@ proc decryptWithAd(state: var CipherState, ad, data: openArray[byte]): seq[byte]
   trace "decryptWithAd", tagIn = tagIn.shortLog, tagOut = tagOut.shortLog, nonce = state.n
   if tagIn != tagOut:
     debug "decryptWithAd failed", data = shortLog(data)
-    raise newException(NoiseDecryptTagError, "decryptWithAd failed tag authentication.")
+    raise (ref NoiseDecryptTagError)(msg:
+      "decryptWithAd failed tag authentication.")
   inc state.n
   if state.n > NonceMax:
-    raise newException(NoiseNonceMaxError, "Noise max nonce value reached")
+    raise (ref NoiseNonceMaxError)(msg: "Noise max nonce value reached")
 
 # Symmetricstate
 
@@ -181,8 +182,7 @@ proc init(_: type[SymmetricState]): SymmetricState =
   result.cs = CipherState(k: EmptyKey)
 
 proc mixKey(ss: var SymmetricState, ikm: ChaChaPolyKey) =
-  var
-    temp_keys: array[2, ChaChaPolyKey]
+  var temp_keys: array[2, ChaChaPolyKey]
   sha256.hkdf(ss.ck, ikm, [], temp_keys)
   ss.ck = temp_keys[0]
   ss.cs = CipherState(k: temp_keys[1])
@@ -198,8 +198,7 @@ proc mixHash(ss: var SymmetricState, data: openArray[byte]) =
 
 # We might use this for other handshake patterns/tokens
 proc mixKeyAndHash(ss: var SymmetricState, ikm: openArray[byte]) {.used.} =
-  var
-    temp_keys: array[3, ChaChaPolyKey]
+  var temp_keys: array[3, ChaChaPolyKey]
   sha256.hkdf(ss.ck, ikm, [], temp_keys)
   ss.ck = temp_keys[0]
   ss.mixHash(temp_keys[1])
@@ -234,7 +233,8 @@ proc init(_: type[HandshakeState]): HandshakeState =
 
 template write_e: untyped =
   trace "noise write e"
-  # Sets e (which must be empty) to GENERATE_KEYPAIR(). Appends e.public_key to the buffer. Calls MixHash(e.public_key).
+  # Sets e (which must be empty) to GENERATE_KEYPAIR().
+  # Appends e.public_key to the buffer. Calls MixHash(e.public_key).
   hs.e = genKeyPair(p.rng[])
   msg.add hs.e.publicKey
   hs.ss.mixHash(hs.e.publicKey)
@@ -275,33 +275,37 @@ template read_e: untyped =
   trace "noise read e", size = msg.len
 
   if msg.len < Curve25519Key.len:
-    raise newException(NoiseHandshakeError, "Noise E, expected more data")
+    raise (ref NoiseHandshakeError)(msg: "Noise E, expected more data")
 
-  # Sets re (which must be empty) to the next DHLEN bytes from the message. Calls MixHash(re.public_key).
+  # Sets re (which must be empty) to the next DHLEN bytes from the message.
+  # Calls MixHash(re.public_key).
   hs.re[0..Curve25519Key.high] = msg.toOpenArray(0, Curve25519Key.high)
   msg.consume(Curve25519Key.len)
   hs.ss.mixHash(hs.re)
 
 template read_s: untyped =
   trace "noise read s", size = msg.len
-  # Sets temp to the next DHLEN + 16 bytes of the message if HasKey() == True, or to the next DHLEN bytes otherwise.
+  # Sets temp to the next DHLEN + 16 bytes of the message if HasKey() == True,
+  # or to the next DHLEN bytes otherwise.
   # Sets rs (which must be empty) to DecryptAndHash(temp).
   let
     rsLen =
       if hs.ss.cs.hasKey:
         if msg.len < Curve25519Key.len + ChaChaPolyTag.len:
-          raise newException(NoiseHandshakeError, "Noise S, expected more data")
+          raise (ref NoiseHandshakeError)(msg: "Noise S, expected more data")
         Curve25519Key.len + ChaChaPolyTag.len
       else:
         if msg.len < Curve25519Key.len:
-          raise newException(NoiseHandshakeError, "Noise S, expected more data")
+          raise (ref NoiseHandshakeError)(msg: "Noise S, expected more data")
         Curve25519Key.len
   hs.rs[0..Curve25519Key.high] =
     hs.ss.decryptAndHash(msg.toOpenArray(0, rsLen - 1))
 
   msg.consume(rsLen)
 
-proc readFrame(sconn: Connection): Future[seq[byte]] {.async.} =
+proc readFrame(
+    sconn: Connection
+): Future[seq[byte]] {.async: (raises: [CancelledError, LPStreamError]).} =
   var besize {.noinit.}: array[2, byte]
   await sconn.readExactly(addr besize[0], besize.len)
   let size = uint16.fromBytesBE(besize).int
@@ -313,7 +317,11 @@ proc readFrame(sconn: Connection): Future[seq[byte]] {.async.} =
   await sconn.readExactly(addr buffer[0], buffer.len)
   return buffer
 
-proc writeFrame(sconn: Connection, buf: openArray[byte]): Future[void] =
+proc writeFrame(
+    sconn: Connection,
+    buf: openArray[byte]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
   doAssert buf.len <= uint16.high.int
   var
     lesize = buf.len.uint16
@@ -324,13 +332,24 @@ proc writeFrame(sconn: Connection, buf: openArray[byte]): Future[void] =
   outbuf &= buf
   sconn.write(outbuf)
 
-proc receiveHSMessage(sconn: Connection): Future[seq[byte]] = readFrame(sconn)
-proc sendHSMessage(sconn: Connection, buf: openArray[byte]): Future[void] =
+proc receiveHSMessage(
+    sconn: Connection
+): Future[seq[byte]] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
+  readFrame(sconn)
+
+proc sendHSMessage(
+    sconn: Connection,
+    buf: openArray[byte]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
   writeFrame(sconn, buf)
 
 proc handshakeXXOutbound(
     p: Noise, conn: Connection,
-    p2pSecret: seq[byte]): Future[HandshakeResult] {.async.} =
+    p2pSecret: seq[byte]
+): Future[HandshakeResult] {.async: (raises: [
+    CancelledError, LPStreamError]).} =
   const initiator = true
   var
     hs = HandshakeState.init()
@@ -372,13 +391,16 @@ proc handshakeXXOutbound(
     await conn.sendHSMessage(msg.data)
 
     let (cs1, cs2) = hs.ss.split()
-    return HandshakeResult(cs1: cs1, cs2: cs2, remoteP2psecret: remoteP2psecret, rs: hs.rs)
+    return HandshakeResult(
+      cs1: cs1, cs2: cs2, remoteP2psecret: remoteP2psecret, rs: hs.rs)
   finally:
     burnMem(hs)
 
 proc handshakeXXInbound(
     p: Noise, conn: Connection,
-    p2pSecret: seq[byte]): Future[HandshakeResult] {.async.} =
+    p2pSecret: seq[byte]
+): Future[HandshakeResult] {.async: (raises: [
+    CancelledError, LPStreamError]).} =
   const initiator = false
 
   var
@@ -422,11 +444,14 @@ proc handshakeXXInbound(
     let
       remoteP2psecret = hs.ss.decryptAndHash(msg.data)
       (cs1, cs2) = hs.ss.split()
-    return HandshakeResult(cs1: cs1, cs2: cs2, remoteP2psecret: remoteP2psecret, rs: hs.rs)
+    return HandshakeResult(
+      cs1: cs1, cs2: cs2, remoteP2psecret: remoteP2psecret, rs: hs.rs)
   finally:
     burnMem(hs)
 
-method readMessage*(sconn: NoiseConnection): Future[seq[byte]] {.async.} =
+method readMessage*(
+    sconn: NoiseConnection
+): Future[seq[byte]] {.async: (raises: [CancelledError, LPStreamError]).} =
   while true: # Discard 0-length payloads
     let frame = await sconn.stream.readFrame()
     sconn.activity = true
@@ -458,7 +483,11 @@ proc encryptFrame(
 
   cipherFrame[2 + src.len()..<cipherFrame.len] = tag
 
-method write*(sconn: NoiseConnection, message: seq[byte]): Future[void] =
+method write*(
+    sconn: NoiseConnection,
+    message: seq[byte]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
   # Fast path: `{.async.}` would introduce a copy of `message`
   const FramingSize = 2 + sizeof(ChaChaPolyTag)
 
@@ -478,7 +507,8 @@ method write*(sconn: NoiseConnection, message: seq[byte]): Future[void] =
     try:
       encryptFrame(
         sconn,
-        cipherFrames.toOpenArray(woffset, woffset + chunkSize + FramingSize - 1),
+        cipherFrames.toOpenArray(
+          woffset, woffset + chunkSize + FramingSize - 1),
         message.toOpenArray(offset, offset + chunkSize - 1))
     except NoiseNonceMaxError as exc:
       debug "Noise nonce exceeded"
@@ -501,21 +531,28 @@ method write*(sconn: NoiseConnection, message: seq[byte]): Future[void] =
   # sequencing issues
   sconn.stream.write(cipherFrames)
 
-method handshake*(p: Noise, conn: Connection, initiator: bool, peerId: Opt[PeerId]): Future[SecureConn] {.async.} =
+method handshake*(
+    p: Noise,
+    conn: Connection,
+    initiator: bool,
+    peerId: Opt[PeerId]
+): Future[SecureConn] {.async: (raises: [CancelledError, LPStreamError]).} =
   trace "Starting Noise handshake", conn, initiator
 
   let timeout = conn.timeout
   conn.timeout = HandshakeTimeout
 
   # https://github.com/libp2p/specs/tree/master/noise#libp2p-data-in-handshake-messages
-  let
-    signedPayload = p.localPrivateKey.sign(
-      PayloadString & p.noiseKeys.publicKey.getBytes).tryGet()
+  let signedPayload = p.localPrivateKey.sign(
+    PayloadString & p.noiseKeys.publicKey.getBytes)
+  if signedPayload.isErr():
+    raise (ref NoiseHandshakeError)(msg:
+      "Failed to sign public key: " & $signedPayload.error())
 
   var
     libp2pProof = initProtoBuffer()
   libp2pProof.write(1, p.localPublicKey)
-  libp2pProof.write(2, signedPayload.getBytes())
+  libp2pProof.write(2, signedPayload.get().getBytes())
   # data field also there but not used!
   libp2pProof.finish()
 
@@ -534,29 +571,38 @@ method handshake*(p: Noise, conn: Connection, initiator: bool, peerId: Opt[PeerI
       remoteSigBytes: seq[byte]
 
     if not remoteProof.getField(1, remotePubKeyBytes).valueOr(false):
-      raise newException(NoiseHandshakeError, "Failed to deserialize remote public key bytes. (initiator: " & $initiator & ")")
+      raise (ref NoiseHandshakeError)(msg:
+        "Failed to deserialize remote public key bytes. (initiator: " &
+        $initiator & ")")
     if not remoteProof.getField(2, remoteSigBytes).valueOr(false):
-      raise newException(NoiseHandshakeError, "Failed to deserialize remote signature bytes. (initiator: " & $initiator & ")")
+      raise (ref NoiseHandshakeError)(msg:
+        "Failed to deserialize remote signature bytes. (initiator: " &
+        $initiator & ")")
 
     if not remotePubKey.init(remotePubKeyBytes):
-      raise newException(NoiseHandshakeError, "Failed to decode remote public key. (initiator: " & $initiator & ")")
+      raise (ref NoiseHandshakeError)(msg:
+        "Failed to decode remote public key. (initiator: " & $initiator & ")")
     if not remoteSig.init(remoteSigBytes):
-      raise newException(NoiseHandshakeError, "Failed to decode remote signature. (initiator: " & $initiator & ")")
+      raise (ref NoiseHandshakeError)(msg:
+        "Failed to decode remote signature. (initiator: " & $initiator & ")")
 
     let verifyPayload = PayloadString & handshakeRes.rs.getBytes
     if not remoteSig.verify(verifyPayload, remotePubKey):
-      raise newException(NoiseHandshakeError, "Noise handshake signature verify failed.")
+      raise (ref NoiseHandshakeError)(msg:
+        "Noise handshake signature verify failed.")
     else:
       trace "Remote signature verified", conn
 
     let pid = PeerId.init(remotePubKey).valueOr:
-      raise newException(NoiseHandshakeError, "Invalid remote peer id: " & $error)
+      raise (ref NoiseHandshakeError)(msg:
+        "Invalid remote peer id: " & $error)
 
     trace "Remote peer id", pid = $pid
 
     peerId.withValue(targetPid):
       if not targetPid.validate():
-        raise newException(NoiseHandshakeError, "Failed to validate expected peerId.")
+        raise (ref NoiseHandshakeError)(msg:
+          "Failed to validate expected peerId.")
 
       if pid != targetPid:
         var
@@ -566,7 +612,8 @@ method handshake*(p: Noise, conn: Connection, initiator: bool, peerId: Opt[PeerI
           initiator, dealt_peer = conn,
           dealt_key = $failedKey, received_peer = $pid,
           received_key = $remotePubKey
-        raise newException(NoiseHandshakeError, "Noise handshake, peer id don't match! " & $pid & " != " & $targetPid)
+        raise (ref NoiseHandshakeError)(msg:
+          "Noise handshake, peer id don't match! " & $pid & " != " & $targetPid)
     conn.peerId = pid
 
     var tmp = NoiseConnection.new(conn, conn.peerId, conn.observedAddr)
@@ -586,7 +633,7 @@ method handshake*(p: Noise, conn: Connection, initiator: bool, peerId: Opt[PeerI
 
   return secure
 
-method closeImpl*(s: NoiseConnection) {.async.} =
+method closeImpl*(s: NoiseConnection) {.async: (raises: []).} =
   await procCall SecureConn(s).closeImpl()
 
   burnMem(s.readCs)
@@ -597,15 +644,14 @@ method init*(p: Noise) {.gcsafe.} =
   p.codec = NoiseCodec
 
 proc new*(
-  T: typedesc[Noise],
-  rng: ref HmacDrbgContext,
-  privateKey: PrivateKey,
-  outgoing: bool = true,
-  commonPrologue: seq[byte] = @[]): T =
-
+    T: typedesc[Noise],
+    rng: ref HmacDrbgContext,
+    privateKey: PrivateKey,
+    outgoing: bool = true,
+    commonPrologue: seq[byte] = @[]): T =
   let pkBytes = privateKey.getPublicKey()
-  .expect("Expected valid Private Key")
-  .getBytes().expect("Couldn't get public Key bytes")
+    .expect("Expected valid Private Key")
+    .getBytes().expect("Couldn't get public Key bytes")
 
   var noise = Noise(
     rng: rng,

libp2p/protocols/secure/plaintext.nim

@@ -19,7 +19,7 @@ type
 
 method init(p: PlainText) {.gcsafe.} =
   proc handle(conn: Connection, proto: string)
-    {.async, gcsafe.} = discard
+    {.async.} = discard
     ## plain text doesn't do anything
 
   p.codec = PlainTextCodec

libp2p/protocols/secure/secio.nim

@@ -1,444 +0,0 @@
-# Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
-# 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.
-
-{.push raises: [].}
-
-import std/[oids, strformat]
-import bearssl/rand
-import chronos, chronicles, stew/endians2
-import nimcrypto/[hmac, sha2, sha, hash, rijndael, twofish, bcmode]
-import secure,
-       ../../stream/connection,
-       ../../peerinfo,
-       ../../crypto/crypto,
-       ../../crypto/ecnist,
-       ../../peerid,
-       ../../utility,
-       ../../errors
-
-export hmac, sha2, sha, hash, rijndael, bcmode
-
-logScope:
-  topics = "libp2p secio"
-
-const
-  SecioCodec* = "/secio/1.0.0"
-  SecioMaxMessageSize = 8 * 1024 * 1024 ## 8mb
-  SecioMaxMacSize = sha512.sizeDigest
-  SecioNonceSize = 16
-  SecioExchanges = "P-256,P-384,P-521"
-  SecioCiphers = "TwofishCTR,AES-256,AES-128"
-  SecioHashes = "SHA256,SHA512"
-
-type
-  Secio* = ref object of Secure
-    rng: ref HmacDrbgContext
-    localPrivateKey: PrivateKey
-    localPublicKey: PublicKey
-    remotePublicKey: PublicKey
-
-  SecureCipherType {.pure.} = enum
-    Aes128, Aes256, Twofish
-
-  SecureMacType {.pure.} = enum
-    Sha1, Sha256, Sha512
-
-  SecureCipher = object
-    case kind: SecureCipherType
-    of SecureCipherType.Aes128:
-      ctxaes128: CTR[aes128]
-    of SecureCipherType.Aes256:
-      ctxaes256: CTR[aes256]
-    of SecureCipherType.Twofish:
-      ctxtwofish256: CTR[twofish256]
-
-  SecureMac = object
-    case kind: SecureMacType
-    of SecureMacType.Sha256:
-      ctxsha256: HMAC[sha256]
-    of SecureMacType.Sha512:
-      ctxsha512: HMAC[sha512]
-    of SecureMacType.Sha1:
-      ctxsha1: HMAC[sha1]
-
-  SecioConn = ref object of SecureConn
-    writerMac: SecureMac
-    readerMac: SecureMac
-    writerCoder: SecureCipher
-    readerCoder: SecureCipher
-
-  SecioError* = object of LPError
-
-func shortLog*(conn: SecioConn): auto =
-  try:
-    if conn.isNil: "SecioConn(nil)"
-    else: &"{shortLog(conn.peerId)}:{conn.oid}"
-  except ValueError as exc:
-    raise newException(Defect, exc.msg)
-
-chronicles.formatIt(SecioConn): shortLog(it)
-
-proc init(mac: var SecureMac, hash: string, key: openArray[byte]) =
-  if hash == "SHA256":
-    mac = SecureMac(kind: SecureMacType.Sha256)
-    mac.ctxsha256.init(key)
-  elif hash == "SHA512":
-    mac = SecureMac(kind: SecureMacType.Sha512)
-    mac.ctxsha512.init(key)
-  elif hash == "SHA1":
-    mac = SecureMac(kind: SecureMacType.Sha1)
-    mac.ctxsha1.init(key)
-
-proc update(mac: var SecureMac, data: openArray[byte]) =
-  case mac.kind
-  of SecureMacType.Sha256:
-    update(mac.ctxsha256, data)
-  of SecureMacType.Sha512:
-    update(mac.ctxsha512, data)
-  of SecureMacType.Sha1:
-    update(mac.ctxsha1, data)
-
-proc sizeDigest(mac: SecureMac): int {.inline.} =
-  case mac.kind
-  of SecureMacType.Sha256:
-    result = int(mac.ctxsha256.sizeDigest())
-  of SecureMacType.Sha512:
-    result = int(mac.ctxsha512.sizeDigest())
-  of SecureMacType.Sha1:
-    result = int(mac.ctxsha1.sizeDigest())
-
-proc finish(mac: var SecureMac, data: var openArray[byte]) =
-  case mac.kind
-  of SecureMacType.Sha256:
-    discard finish(mac.ctxsha256, data)
-  of SecureMacType.Sha512:
-    discard finish(mac.ctxsha512, data)
-  of SecureMacType.Sha1:
-    discard finish(mac.ctxsha1, data)
-
-proc reset(mac: var SecureMac) =
-  case mac.kind
-  of SecureMacType.Sha256:
-    reset(mac.ctxsha256)
-  of SecureMacType.Sha512:
-    reset(mac.ctxsha512)
-  of SecureMacType.Sha1:
-    reset(mac.ctxsha1)
-
-proc init(sc: var SecureCipher, cipher: string, key: openArray[byte],
-          iv: openArray[byte]) {.inline.} =
-  if cipher == "AES-128":
-    sc = SecureCipher(kind: SecureCipherType.Aes128)
-    sc.ctxaes128.init(key, iv)
-  elif cipher == "AES-256":
-    sc = SecureCipher(kind: SecureCipherType.Aes256)
-    sc.ctxaes256.init(key, iv)
-  elif cipher == "TwofishCTR":
-    sc = SecureCipher(kind: SecureCipherType.Twofish)
-    sc.ctxtwofish256.init(key, iv)
-
-proc encrypt(cipher: var SecureCipher, input: openArray[byte],
-             output: var openArray[byte]) {.inline.} =
-  case cipher.kind
-  of SecureCipherType.Aes128:
-    cipher.ctxaes128.encrypt(input, output)
-  of SecureCipherType.Aes256:
-    cipher.ctxaes256.encrypt(input, output)
-  of SecureCipherType.Twofish:
-    cipher.ctxtwofish256.encrypt(input, output)
-
-proc decrypt(cipher: var SecureCipher, input: openArray[byte],
-             output: var openArray[byte]) {.inline.} =
-  case cipher.kind
-  of SecureCipherType.Aes128:
-    cipher.ctxaes128.decrypt(input, output)
-  of SecureCipherType.Aes256:
-    cipher.ctxaes256.decrypt(input, output)
-  of SecureCipherType.Twofish:
-    cipher.ctxtwofish256.decrypt(input, output)
-
-proc macCheckAndDecode(sconn: SecioConn, data: var seq[byte]): bool =
-  ## This procedure checks MAC of recieved message ``data`` and if message is
-  ## authenticated, then decrypt message.
-  ##
-  ## Procedure returns ``false`` if message is too short or MAC verification
-  ## failed.
-  var macData: array[SecioMaxMacSize, byte]
-  let macsize = sconn.readerMac.sizeDigest()
-  if len(data) < macsize:
-    trace "Message is shorter then MAC size", message_length = len(data),
-                                              mac_size = macsize
-    return false
-  let mark = len(data) - macsize
-  sconn.readerMac.update(data.toOpenArray(0, mark - 1))
-  sconn.readerMac.finish(macData)
-  sconn.readerMac.reset()
-  if not equalMem(addr data[mark], addr macData[0], macsize):
-    trace "Invalid MAC",
-          calculated = toHex(macData.toOpenArray(0, macsize - 1)),
-          stored = toHex(data.toOpenArray(mark, data.high))
-    return false
-
-  sconn.readerCoder.decrypt(data.toOpenArray(0, mark - 1),
-                            data.toOpenArray(0, mark - 1))
-  data.setLen(mark)
-  result = true
-
-proc readRawMessage(conn: Connection): Future[seq[byte]] {.async.} =
-  while true: # Discard 0-length payloads
-    var lengthBuf: array[4, byte]
-    await conn.readExactly(addr lengthBuf[0], lengthBuf.len)
-    let length = uint32.fromBytesBE(lengthBuf)
-
-    trace "Recieved message header", header = lengthBuf.shortLog, length = length
-
-    if length > SecioMaxMessageSize: # Verify length before casting!
-      trace "Received size of message exceed limits", conn, length = length
-      raise (ref SecioError)(msg: "Message exceeds maximum length")
-
-    if length > 0:
-      var buf = newSeq[byte](int(length))
-      await conn.readExactly(addr buf[0], buf.len)
-      trace "Received message body",
-        conn, length = buf.len, buff = buf.shortLog
-      return buf
-
-    trace "Discarding 0-length payload", conn
-
-method readMessage*(sconn: SecioConn): Future[seq[byte]] {.async.} =
-  ## Read message from channel secure connection ``sconn``.
-  when chronicles.enabledLogLevel == LogLevel.TRACE:
-    logScope:
-      stream_oid = $sconn.stream.oid
-  var buf = await sconn.stream.readRawMessage()
-  if sconn.macCheckAndDecode(buf):
-    result = buf
-  else:
-    trace "Message MAC verification failed", buf = buf.shortLog
-    raise (ref SecioError)(msg: "message failed MAC verification")
-
-method write*(sconn: SecioConn, message: seq[byte]) {.async.} =
-  ## Write message ``message`` to secure connection ``sconn``.
-  if message.len == 0:
-    return
-
-  var
-    left = message.len
-    offset = 0
-  while left > 0:
-    let
-      chunkSize = if left > SecioMaxMessageSize - 64: SecioMaxMessageSize - 64 else: left
-      macsize = sconn.writerMac.sizeDigest()
-      length = chunkSize + macsize
-
-    var msg = newSeq[byte](chunkSize + 4 + macsize)
-    msg[0..<4] = uint32(length).toBytesBE()
-
-    sconn.writerCoder.encrypt(message.toOpenArray(offset, offset + chunkSize - 1),
-                              msg.toOpenArray(4, 4 + chunkSize - 1))
-    left = left - chunkSize
-    offset = offset + chunkSize
-    let mo = 4 + chunkSize
-    sconn.writerMac.update(msg.toOpenArray(4, 4 + chunkSize - 1))
-    sconn.writerMac.finish(msg.toOpenArray(mo, mo + macsize - 1))
-    sconn.writerMac.reset()
-
-    trace "Writing message", message = msg.shortLog, left, offset
-    await sconn.stream.write(msg)
-    sconn.activity = true
-
-proc newSecioConn(conn: Connection,
-                  hash: string,
-                  cipher: string,
-                  secrets: Secret,
-                  order: int,
-                  remotePubKey: PublicKey): SecioConn
-                  {.raises: [LPError].} =
-  ## Create new secure stream/lpstream, using specified hash algorithm ``hash``,
-  ## cipher algorithm ``cipher``, stretched keys ``secrets`` and order
-  ## ``order``.
-
-  result = SecioConn.new(conn, conn.peerId, conn.observedAddr)
-
-  let i0 = if order < 0: 1 else: 0
-  let i1 = if order < 0: 0 else: 1
-
-  trace "Writer credentials", mackey = secrets.macOpenArray(i0).shortLog,
-                              enckey = secrets.keyOpenArray(i0).shortLog,
-                              iv = secrets.ivOpenArray(i0).shortLog
-  trace "Reader credentials", mackey = secrets.macOpenArray(i1).shortLog,
-                              enckey = secrets.keyOpenArray(i1).shortLog,
-                              iv = secrets.ivOpenArray(i1).shortLog
-  result.writerMac.init(hash, secrets.macOpenArray(i0))
-  result.readerMac.init(hash, secrets.macOpenArray(i1))
-  result.writerCoder.init(cipher, secrets.keyOpenArray(i0),
-                          secrets.ivOpenArray(i0))
-  result.readerCoder.init(cipher, secrets.keyOpenArray(i1),
-                          secrets.ivOpenArray(i1))
-
-proc transactMessage(conn: Connection,
-                     msg: seq[byte]): Future[seq[byte]] {.async.} =
-  trace "Sending message", message = msg.shortLog, length = len(msg)
-  await conn.write(msg)
-  return await conn.readRawMessage()
-
-method handshake*(s: Secio, conn: Connection, initiator: bool, peerId: Opt[PeerId]): Future[SecureConn] {.async.} =
-  var
-    localNonce: array[SecioNonceSize, byte]
-    remoteNonce: seq[byte]
-    remoteBytesPubkey: seq[byte]
-    remoteEBytesPubkey: seq[byte]
-    remoteEBytesSig: seq[byte]
-    remotePubkey: PublicKey
-    remoteEPubkey: ecnist.EcPublicKey
-    remoteESignature: Signature
-    remoteExchanges: string
-    remoteCiphers: string
-    remoteHashes: string
-    remotePeerId: PeerId
-    localPeerId: PeerId
-    localBytesPubkey = s.localPublicKey.getBytes().tryGet()
-
-  hmacDrbgGenerate(s.rng[], localNonce)
-
-  var request = createProposal(localNonce,
-                               localBytesPubkey,
-                               SecioExchanges,
-                               SecioCiphers,
-                               SecioHashes)
-
-  localPeerId = PeerId.init(s.localPublicKey).tryGet()
-
-  trace "Local proposal", schemes = SecioExchanges,
-                          ciphers = SecioCiphers,
-                          hashes = SecioHashes,
-                          pubkey = localBytesPubkey.shortLog,
-                          peer = localPeerId
-
-  var answer = await transactMessage(conn, request)
-
-  if len(answer) == 0:
-    trace "Proposal exchange failed", conn
-    raise (ref SecioError)(msg: "Proposal exchange failed")
-
-  if not decodeProposal(answer, remoteNonce, remoteBytesPubkey, remoteExchanges,
-                        remoteCiphers, remoteHashes):
-    trace "Remote proposal decoding failed", conn
-    raise (ref SecioError)(msg: "Remote proposal decoding failed")
-
-  if not remotePubkey.init(remoteBytesPubkey):
-    trace "Remote public key incorrect or corrupted",
-          pubkey = remoteBytesPubkey.shortLog
-    raise (ref SecioError)(msg: "Remote public key incorrect or corrupted")
-
-  remotePeerId = PeerId.init(remotePubkey).tryGet()
-
-  peerId.withValue(targetPid):
-    if not targetPid.validate():
-      raise newException(SecioError, "Failed to validate expected peerId.")
-
-    if remotePeerId != targetPid:
-      raise newException(SecioError, "Peer ids don't match!")
-  conn.peerId = remotePeerId
-  let order = getOrder(remoteBytesPubkey, localNonce, localBytesPubkey,
-                       remoteNonce).tryGet()
-  trace "Remote proposal", schemes = remoteExchanges, ciphers = remoteCiphers,
-                           hashes = remoteHashes,
-                           pubkey = remoteBytesPubkey.shortLog, order = order,
-                           peer = remotePeerId
-
-  let scheme = selectBest(order, SecioExchanges, remoteExchanges)
-  let cipher = selectBest(order, SecioCiphers, remoteCiphers)
-  let hash = selectBest(order, SecioHashes, remoteHashes)
-  if len(scheme) == 0 or len(cipher) == 0 or len(hash) == 0:
-    trace "No algorithms in common", peer = remotePeerId
-    raise (ref SecioError)(msg: "No algorithms in common")
-
-  trace "Encryption scheme selected", scheme = scheme, cipher = cipher,
-                                      hash = hash
-
-  var ekeypair = ephemeral(scheme, s.rng[]).tryGet()
-  # We need EC public key in raw binary form
-  var epubkey = ekeypair.pubkey.getRawBytes().tryGet()
-  var localCorpus = request[4..^1] & answer & epubkey
-  var signature = s.localPrivateKey.sign(localCorpus).tryGet()
-
-  var localExchange = createExchange(epubkey, signature.getBytes())
-  var remoteExchange = await transactMessage(conn, localExchange)
-  if len(remoteExchange) == 0:
-    trace "Corpus exchange failed", conn
-    raise (ref SecioError)(msg: "Corpus exchange failed")
-
-  if not decodeExchange(remoteExchange, remoteEBytesPubkey, remoteEBytesSig):
-    trace "Remote exchange decoding failed", conn
-    raise (ref SecioError)(msg: "Remote exchange decoding failed")
-
-  if not remoteESignature.init(remoteEBytesSig):
-    trace "Remote signature incorrect or corrupted", signature = remoteEBytesSig.shortLog
-    raise (ref SecioError)(msg: "Remote signature incorrect or corrupted")
-
-  var remoteCorpus = answer & request[4..^1] & remoteEBytesPubkey
-  if not remoteESignature.verify(remoteCorpus, remotePubkey):
-    trace "Signature verification failed", scheme = $remotePubkey.scheme,
-                                           signature = $remoteESignature,
-                                           pubkey = $remotePubkey,
-                                           corpus = $remoteCorpus
-    raise (ref SecioError)(msg: "Signature verification failed")
-
-  trace "Signature verified", scheme = remotePubkey.scheme
-
-  if not remoteEPubkey.initRaw(remoteEBytesPubkey):
-    trace "Remote ephemeral public key incorrect or corrupted",
-          pubkey = toHex(remoteEBytesPubkey)
-    raise (ref SecioError)(msg: "Remote ephemeral public key incorrect or corrupted")
-
-  var secret = getSecret(remoteEPubkey, ekeypair.seckey)
-  if len(secret) == 0:
-    trace "Shared secret could not be created"
-    raise (ref SecioError)(msg: "Shared secret could not be created")
-
-  trace "Shared secret calculated", secret = secret.shortLog
-
-  var keys = stretchKeys(cipher, hash, secret)
-
-  trace "Authenticated encryption parameters",
-        iv0 = toHex(keys.ivOpenArray(0)), key0 = keys.keyOpenArray(0).shortLog,
-        mac0 = keys.macOpenArray(0).shortLog,
-        iv1 = keys.ivOpenArray(1).shortLog, key1 = keys.keyOpenArray(1).shortLog,
-        mac1 = keys.macOpenArray(1).shortLog
-
-  # Perform Nonce exchange over encrypted channel.
-
-  var secioConn = newSecioConn(conn, hash, cipher, keys, order, remotePubkey)
-  result = secioConn
-  await secioConn.write(remoteNonce)
-  var res = await secioConn.readMessage()
-
-  if res != @localNonce:
-    trace "Nonce verification failed", receivedNonce = res.shortLog,
-                                       localNonce = localNonce.shortLog
-    raise (ref SecioError)(msg: "Nonce verification failed")
-  else:
-    trace "Secure handshake succeeded"
-
-method init(s: Secio) {.gcsafe.} =
-  procCall Secure(s).init()
-  s.codec = SecioCodec
-
-proc new*(
-  T: typedesc[Secio],
-  rng: ref HmacDrbgContext,
-  localPrivateKey: PrivateKey): T =
-  let secio = Secio(
-    rng: rng,
-    localPrivateKey: localPrivateKey,
-    localPublicKey: localPrivateKey.getPublicKey().expect("Invalid private key"),
-  )
-  secio.init()
-  secio

libp2p/protocols/secure/secure.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -37,18 +37,19 @@ type
 
 func shortLog*(conn: SecureConn): auto =
   try:
-    if conn.isNil: "SecureConn(nil)"
+    if conn == nil: "SecureConn(nil)"
     else: &"{shortLog(conn.peerId)}:{conn.oid}"
   except ValueError as exc:
-    raise newException(Defect, exc.msg)
+    raiseAssert(exc.msg)
 
 chronicles.formatIt(SecureConn): shortLog(it)
 
-proc new*(T: type SecureConn,
-           conn: Connection,
-           peerId: PeerId,
-           observedAddr: Opt[MultiAddress],
-           timeout: Duration = DefaultConnectionTimeout): T =
+proc new*(
+    T: type SecureConn,
+    conn: Connection,
+    peerId: PeerId,
+    observedAddr: Opt[MultiAddress],
+    timeout: Duration = DefaultConnectionTimeout): T =
   result = T(stream: conn,
              peerId: peerId,
              observedAddr: observedAddr,
@@ -63,55 +64,72 @@ method initStream*(s: SecureConn) =
 
   procCall Connection(s).initStream()
 
-method closeImpl*(s: SecureConn) {.async.} =
+method closeImpl*(s: SecureConn) {.async: (raises: []).} =
   trace "Closing secure conn", s, dir = s.dir
-  if not(isNil(s.stream)):
+  if s.stream != nil:
     await s.stream.close()
 
   await procCall Connection(s).closeImpl()
 
-method readMessage*(c: SecureConn): Future[seq[byte]] {.async, base.} =
-  doAssert(false, "Not implemented!")
+method readMessage*(
+    c: SecureConn
+): Future[seq[byte]] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true), base.} =
+  raiseAssert("Not implemented!")
 
 method getWrapped*(s: SecureConn): Connection = s.stream
 
-method handshake*(s: Secure,
-                  conn: Connection,
-                  initiator: bool,
-                  peerId: Opt[PeerId]): Future[SecureConn] {.async, base.} =
-  doAssert(false, "Not implemented!")
+method handshake*(
+    s: Secure,
+    conn: Connection,
+    initiator: bool,
+    peerId: Opt[PeerId]
+): Future[SecureConn] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true), base.} =
+  raiseAssert("Not implemented!")
 
-proc handleConn(s: Secure,
-                 conn: Connection,
-                 initiator: bool,
-                 peerId: Opt[PeerId]): Future[Connection] {.async.} =
+proc handleConn(
+    s: Secure,
+    conn: Connection,
+    initiator: bool,
+    peerId: Opt[PeerId]
+): Future[Connection] {.async: (raises: [CancelledError, LPStreamError]).} =
   var sconn = await s.handshake(conn, initiator, peerId)
   # mark connection bottom level transport direction
   # this is the safest place to do this
   # we require this information in for example gossipsub
   sconn.transportDir = if initiator: Direction.Out else: Direction.In
 
-  proc cleanup() {.async.} =
+  proc cleanup() {.async: (raises: []).} =
     try:
-      let futs = [conn.join(), sconn.join()]
-      await futs[0] or futs[1]
-      for f in futs:
-        if not f.finished: await f.cancelAndWait() # cancel outstanding join()
+      block:
+        let
+          fut1 = conn.join()
+          fut2 = sconn.join()
+        try:  # https://github.com/status-im/nim-chronos/issues/516
+          discard await race(fut1, fut2)
+        except ValueError: raiseAssert("Futures list is not empty")
+        # at least one join() completed, cancel pending one, if any
+        if not fut1.finished: await fut1.cancelAndWait()
+        if not fut2.finished: await fut2.cancelAndWait()
+      block:
+        let
+          fut1 = sconn.close()
+          fut2 = conn.close()
+        await allFutures(fut1, fut2)
+        static: doAssert typeof(fut1).E is void  # Cannot fail
+        static: doAssert typeof(fut2).E is void  # Cannot fail
 
-      await allFuturesThrowing(
-        sconn.close(), conn.close())
     except CancelledError:
       # This is top-level procedure which will work as separate task, so it
       # do not need to propagate CancelledError.
       discard
-    except CatchableError as exc:
-      debug "error cleaning up secure connection", err = exc.msg, sconn
 
-  if not isNil(sconn):
+  if sconn != nil:
     # All the errors are handled inside `cleanup()` procedure.
     asyncSpawn cleanup()
 
-  return sconn
+  sconn
 
 method init*(s: Secure) =
   procCall LPProtocol(s).init()
@@ -127,23 +145,25 @@ method init*(s: Secure) =
       warn "securing connection canceled", conn
       await conn.close()
       raise exc
-    except CatchableError as exc:
+    except LPStreamError as exc:
       warn "securing connection failed", err = exc.msg, conn
       await conn.close()
 
   s.handler = handle
 
-method secure*(s: Secure,
-               conn: Connection,
-               initiator: bool,
-               peerId: Opt[PeerId]):
-               Future[Connection] {.base.} =
-  s.handleConn(conn, initiator, peerId)
+method secure*(
+    s: Secure,
+    conn: Connection,
+    peerId: Opt[PeerId]
+): Future[Connection] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true), base.} =
+  s.handleConn(conn, conn.dir == Direction.Out, peerId)
 
-method readOnce*(s: SecureConn,
-                 pbytes: pointer,
-                 nbytes: int):
-                 Future[int] {.async.} =
+method readOnce*(
+    s: SecureConn,
+    pbytes: pointer,
+    nbytes: int
+): Future[int] {.async: (raises: [CancelledError, LPStreamError]).} =
   doAssert(nbytes > 0, "nbytes must be positive integer")
 
   if s.isEof:
@@ -160,7 +180,7 @@ method readOnce*(s: SecureConn,
       raise err
     except CancelledError as exc:
       raise exc
-    except CatchableError as err:
+    except LPStreamError as err:
       debug "Error while reading message from secure connection, closing.",
         error = err.name,
         message = err.msg,

libp2p/services/autorelayservice.nim

@@ -37,7 +37,7 @@ proc isRunning*(self: AutoRelayService): bool =
 
 proc addressMapper(
   self: AutoRelayService,
-  listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.gcsafe, async.} =
+  listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.async.} =
   return concat(toSeq(self.relayAddresses.values))
 
 proc reserveAndUpdate(self: AutoRelayService, relayPid: PeerId, switch: Switch) {.async.} =
@@ -58,8 +58,8 @@ proc reserveAndUpdate(self: AutoRelayService, relayPid: PeerId, switch: Switch)
         self.onReservation(concat(toSeq(self.relayAddresses.values)))
     await sleepAsync chronos.seconds(ttl - 30)
 
-method setup*(self: AutoRelayService, switch: Switch): Future[bool] {.async, gcsafe.} =
-  self.addressMapper = proc (listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.gcsafe, async.} =
+method setup*(self: AutoRelayService, switch: Switch): Future[bool] {.async.} =
+  self.addressMapper = proc (listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.async.} =
     return await addressMapper(self, listenAddrs)
 
   let hasBeenSetUp = await procCall Service(self).setup(switch)
@@ -83,7 +83,7 @@ proc manageBackedOff(self: AutoRelayService, pid: PeerId) {.async.} =
   self.backingOff.keepItIf(it != pid)
   self.peerAvailable.fire()
 
-proc innerRun(self: AutoRelayService, switch: Switch) {.async, gcsafe.} =
+proc innerRun(self: AutoRelayService, switch: Switch) {.async.} =
   while true:
     # Remove relayPeers that failed
     let peers = toSeq(self.relayPeers.keys())
@@ -116,14 +116,14 @@ proc innerRun(self: AutoRelayService, switch: Switch) {.async, gcsafe.} =
       await self.peerAvailable.wait()
       await sleepAsync(200.millis)
 
-method run*(self: AutoRelayService, switch: Switch) {.async, gcsafe.} =
+method run*(self: AutoRelayService, switch: Switch) {.async.} =
   if self.running:
     trace "Autorelay is already running"
     return
   self.running = true
   self.runner = self.innerRun(switch)
 
-method stop*(self: AutoRelayService, switch: Switch): Future[bool] {.async, gcsafe.} =
+method stop*(self: AutoRelayService, switch: Switch): Future[bool] {.async.} =
   let hasBeenStopped = await procCall Service(self).stop(switch)
   if hasBeenStopped:
     self.running = false

libp2p/services/hpservice.nim

@@ -94,7 +94,7 @@ method setup*(self: HPService, switch: Switch): Future[bool] {.async.} =
 
     switch.connManager.addPeerEventHandler(self.newConnectedPeerHandler, PeerEventKind.Joined)
 
-    self.onNewStatusHandler = proc (networkReachability: NetworkReachability, confidence: Opt[float]) {.gcsafe, async.} =
+    self.onNewStatusHandler = proc (networkReachability: NetworkReachability, confidence: Opt[float]) {.async.} =
       if networkReachability == NetworkReachability.NotReachable and not self.autoRelayService.isRunning():
         discard await self.autoRelayService.setup(switch)
       elif networkReachability == NetworkReachability.Reachable and self.autoRelayService.isRunning():

libp2p/stream/bufferstream.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -34,10 +34,10 @@ type
 
 func shortLog*(s: BufferStream): auto =
   try:
-    if s.isNil: "BufferStream(nil)"
+    if s == nil: "BufferStream(nil)"
     else: &"{shortLog(s.peerId)}:{s.oid}"
   except ValueError as exc:
-    raise newException(Defect, exc.msg)
+    raiseAssert(exc.msg)
 
 chronicles.formatIt(BufferStream): shortLog(it)
 
@@ -55,14 +55,16 @@ method initStream*(s: BufferStream) =
   trace "BufferStream created", s
 
 proc new*(
-  T: typedesc[BufferStream],
-  timeout: Duration = DefaultConnectionTimeout): T =
-
+    T: typedesc[BufferStream],
+    timeout: Duration = DefaultConnectionTimeout): T =
   let bufferStream = T(timeout: timeout)
   bufferStream.initStream()
   bufferStream
 
-method pushData*(s: BufferStream, data: seq[byte]) {.base, async.} =
+method pushData*(
+    s: BufferStream,
+    data: seq[byte]
+) {.base, async: (raises: [CancelledError, LPStreamError]).} =
   ## Write bytes to internal read buffer, use this to fill up the
   ## buffer with data.
   ##
@@ -70,7 +72,7 @@ method pushData*(s: BufferStream, data: seq[byte]) {.base, async.} =
   ##
 
   doAssert(not s.pushing,
-    &"Only one concurrent push allowed for stream {s.shortLog()}")
+    "Only one concurrent push allowed for stream " & s.shortLog())
 
   if s.isClosed or s.pushedEof:
     raise newLPStreamClosedError()
@@ -87,12 +89,14 @@ method pushData*(s: BufferStream, data: seq[byte]) {.base, async.} =
   finally:
     s.pushing = false
 
-method pushEof*(s: BufferStream) {.base, async.} =
+method pushEof*(
+    s: BufferStream
+) {.base, async: (raises: [CancelledError, LPStreamError]).} =
   if s.pushedEof:
     return
 
   doAssert(not s.pushing,
-    &"Only one concurrent push allowed for stream {s.shortLog()}")
+    "Only one concurrent push allowed for stream " & s.shortLog())
 
   s.pushedEof = true
 
@@ -108,13 +112,14 @@ method pushEof*(s: BufferStream) {.base, async.} =
 method atEof*(s: BufferStream): bool =
   s.isEof and s.readBuf.len == 0
 
-method readOnce*(s: BufferStream,
-                 pbytes: pointer,
-                 nbytes: int):
-                 Future[int] {.async.} =
+method readOnce*(
+    s: BufferStream,
+    pbytes: pointer,
+    nbytes: int
+): Future[int] {.async: (raises: [CancelledError, LPStreamError]).} =
   doAssert(nbytes > 0, "nbytes must be positive integer")
   doAssert(not s.reading,
-    &"Only one concurrent read allowed for stream {s.shortLog()}")
+    "Only one concurrent read allowed for stream " & s.shortLog())
 
   if s.returnedEof:
     raise newLPStreamEOFError()
@@ -135,13 +140,6 @@ method readOnce*(s: BufferStream,
         # Not very efficient, but shouldn't happen often
         s.readBuf.assign(@(p.toOpenArray(0, rbytes - 1)) & @(s.readBuf.data))
         raise exc
-      except CatchableError as exc:
-        # When an exception happens here, the Bufferstream is effectively
-        # broken and no more reads will be valid - for now, return EOF if it's
-        # called again, though this is not completely true - EOF represents an
-        # "orderly" shutdown and that's not what happened here..
-        s.returnedEof = true
-        raise exc
       finally:
         s.reading = false
 
@@ -173,7 +171,8 @@ method readOnce*(s: BufferStream,
 
   return rbytes
 
-method closeImpl*(s: BufferStream): Future[void] =
+method closeImpl*(
+    s: BufferStream): Future[void] {.async: (raises: [], raw: true).} =
   ## close the stream and clear the buffer
   trace "Closing BufferStream", s, len = s.len
 
@@ -209,8 +208,8 @@ method closeImpl*(s: BufferStream): Future[void] =
         if not s.readQueue.empty():
           discard s.readQueue.popFirstNoWait()
   except AsyncQueueFullError, AsyncQueueEmptyError:
-    raise newException(Defect, getCurrentExceptionMsg())
+    raiseAssert(getCurrentExceptionMsg())
 
   trace "Closed BufferStream", s
 
-  procCall Connection(s).closeImpl() # noraises, nocancels
+  procCall Connection(s).closeImpl()

libp2p/stream/chronosstream.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -31,18 +31,22 @@ type
       tracked: bool
 
 when defined(libp2p_agents_metrics):
-  declareGauge(libp2p_peers_identity, "peers identities", labels = ["agent"])
-  declareCounter(libp2p_peers_traffic_read, "incoming traffic", labels = ["agent"])
-  declareCounter(libp2p_peers_traffic_write, "outgoing traffic", labels = ["agent"])
+  declareGauge libp2p_peers_identity,
+    "peers identities", labels = ["agent"]
+  declareCounter libp2p_peers_traffic_read,
+    "incoming traffic", labels = ["agent"]
+  declareCounter libp2p_peers_traffic_write,
+    "outgoing traffic", labels = ["agent"]
 
-declareCounter(libp2p_network_bytes, "total traffic", labels = ["direction"])
+declareCounter libp2p_network_bytes,
+  "total traffic", labels = ["direction"]
 
 func shortLog*(conn: ChronosStream): auto =
   try:
-    if conn.isNil: "ChronosStream(nil)"
+    if conn == nil: "ChronosStream(nil)"
     else: &"{shortLog(conn.peerId)}:{conn.oid}"
   except ValueError as exc:
-    raise newException(Defect, exc.msg)
+    raiseAssert(exc.msg)
 
 chronicles.formatIt(ChronosStream): shortLog(it)
 
@@ -50,17 +54,18 @@ method initStream*(s: ChronosStream) =
   if s.objName.len == 0:
     s.objName = ChronosStreamTrackerName
 
-  s.timeoutHandler = proc() {.async, gcsafe.} =
+  s.timeoutHandler = proc(): Future[void] {.async: (raises: [], raw: true).} =
     trace "Idle timeout expired, closing ChronosStream", s
-    await s.close()
+    s.close()
 
   procCall Connection(s).initStream()
 
-proc init*(C: type ChronosStream,
-           client: StreamTransport,
-           dir: Direction,
-           timeout = DefaultChronosStreamTimeout,
-           observedAddr: Opt[MultiAddress]): ChronosStream =
+proc init*(
+    C: type ChronosStream,
+    client: StreamTransport,
+    dir: Direction,
+    timeout = DefaultChronosStreamTimeout,
+    observedAddr: Opt[MultiAddress]): ChronosStream =
   result = C(client: client,
              timeout: timeout,
              dir: dir,
@@ -94,7 +99,11 @@ when defined(libp2p_agents_metrics):
       libp2p_peers_identity.dec(labelValues = [s.shortAgent])
       s.tracked = false
 
-method readOnce*(s: ChronosStream, pbytes: pointer, nbytes: int): Future[int] {.async.} =
+method readOnce*(
+    s: ChronosStream,
+    pbytes: pointer,
+    nbytes: int
+): Future[int] {.async: (raises: [CancelledError, LPStreamError]).} =
   if s.atEof:
     raise newLPStreamEOFError()
   withExceptions:
@@ -107,7 +116,10 @@ method readOnce*(s: ChronosStream, pbytes: pointer, nbytes: int): Future[int] {.
         libp2p_peers_traffic_read.inc(result.int64, labelValues = [s.shortAgent])
 
 proc completeWrite(
-    s: ChronosStream, fut: Future[int], msgLen: int): Future[void] {.async.} =
+    s: ChronosStream,
+    fut: Future[int].Raising([TransportError, CancelledError]),
+    msgLen: int
+): Future[void] {.async: (raises: [CancelledError, LPStreamError]).} =
   withExceptions:
     # StreamTransport will only return written < msg.len on fatal failures where
     # further writing is not possible - in such cases, we'll raise here,
@@ -124,7 +136,11 @@ proc completeWrite(
       if s.tracked:
         libp2p_peers_traffic_write.inc(msgLen.int64, labelValues = [s.shortAgent])
 
-method write*(s: ChronosStream, msg: seq[byte]): Future[void] =
+method write*(
+    s: ChronosStream,
+    msg: seq[byte]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
   # Avoid a copy of msg being kept in the closure created by `{.async.}` as this
   # drives up memory usage
   if msg.len == 0:
@@ -145,19 +161,14 @@ method closed*(s: ChronosStream): bool =
 method atEof*(s: ChronosStream): bool =
   s.client.atEof()
 
-method closeImpl*(s: ChronosStream) {.async.} =
-  try:
-    trace "Shutting down chronos stream", address = $s.client.remoteAddress(), s
+method closeImpl*(
+    s: ChronosStream) {.async: (raises: []).} =
+  trace "Shutting down chronos stream", address = $s.client.remoteAddress(), s
 
-    if not s.client.closed():
-      await s.client.closeWait()
+  if not s.client.closed():
+    await s.client.closeWait()
 
-    trace "Shutdown chronos stream", address = $s.client.remoteAddress(), s
-
-  except CancelledError as exc:
-    raise exc
-  except CatchableError as exc:
-    trace "Error closing chronosstream", s, msg = exc.msg
+  trace "Shutdown chronos stream", address = $s.client.remoteAddress(), s
 
   when defined(libp2p_agents_metrics):
     # do this after closing!

libp2p/stream/connection.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -27,25 +27,25 @@ const
   DefaultConnectionTimeout* = 5.minutes
 
 type
-  TimeoutHandler* = proc(): Future[void] {.gcsafe, raises: [].}
+  TimeoutHandler* = proc(): Future[void] {.async: (raises: []).}
 
   Connection* = ref object of LPStream
-    activity*: bool                 # reset every time data is sent or received
-    timeout*: Duration              # channel timeout if no activity
-    timerTaskFut: Future[void]      # the current timer instance
+    activity*: bool  # reset every time data is sent or received
+    timeout*: Duration  # channel timeout if no activity
+    timerTaskFut: Future[void].Raising([])  # the current timer instance
     timeoutHandler*: TimeoutHandler # timeout handler
     peerId*: PeerId
     observedAddr*: Opt[MultiAddress]
-    protocol*: string               # protocol used by the connection, used as tag for metrics
-    transportDir*: Direction        # The bottom level transport (generally the socket) direction
+    protocol*: string  # protocol used by the connection, used as metrics tag
+    transportDir*: Direction  # underlying transport (usually socket) direction
     when defined(libp2p_agents_metrics):
       shortAgent*: string
 
-proc timeoutMonitor(s: Connection) {.async, gcsafe.}
+proc timeoutMonitor(s: Connection) {.async: (raises: []).}
 
 func shortLog*(conn: Connection): string =
   try:
-    if conn.isNil: "Connection(nil)"
+    if conn == nil: "Connection(nil)"
     else: &"{shortLog(conn.peerId)}:{conn.oid}"
   except ValueError as exc:
     raiseAssert(exc.msg)
@@ -58,23 +58,28 @@ method initStream*(s: Connection) =
 
   procCall LPStream(s).initStream()
 
-  doAssert(isNil(s.timerTaskFut))
+  doAssert(s.timerTaskFut == nil)
 
   if s.timeout > 0.millis:
     trace "Monitoring for timeout", s, timeout = s.timeout
 
     s.timerTaskFut = s.timeoutMonitor()
-    if isNil(s.timeoutHandler):
-      s.timeoutHandler = proc(): Future[void] =
-        trace "Idle timeout expired, closing connection", s
-        s.close()
+    if s.timeoutHandler == nil:
+      s.timeoutHandler =
+        proc(): Future[void] {.async: (raises: [], raw: true).} =
+          trace "Idle timeout expired, closing connection", s
+          s.close()
 
-method closeImpl*(s: Connection): Future[void] =
+method closeImpl*(s: Connection): Future[void] {.async: (raises: []).} =
   # Cleanup timeout timer
   trace "Closing connection", s
 
-  if not isNil(s.timerTaskFut) and not s.timerTaskFut.finished:
-    s.timerTaskFut.cancel()
+  if s.timerTaskFut != nil and not s.timerTaskFut.finished:
+    # Don't `cancelAndWait` here to avoid risking deadlock in this scenario:
+    # - `pollActivity` is waiting for `s.timeoutHandler` to complete.
+    # - `s.timeoutHandler` may have triggered `closeImpl` and we are now here.
+    # In this situation, we have to return for `s.timerTaskFut` to complete.
+    s.timerTaskFut.cancelSoon()
     s.timerTaskFut = nil
 
   trace "Closed connection", s
@@ -84,7 +89,7 @@ method closeImpl*(s: Connection): Future[void] =
 func hash*(p: Connection): Hash =
   cast[pointer](p).hash
 
-proc pollActivity(s: Connection): Future[bool] {.async.} =
+proc pollActivity(s: Connection): Future[bool] {.async: (raises: []).} =
   if s.closed and s.atEof:
     return false # Done, no more monitoring
 
@@ -95,22 +100,13 @@ proc pollActivity(s: Connection): Future[bool] {.async.} =
   # Inactivity timeout happened, call timeout monitor
 
   trace "Connection timed out", s
-  if not(isNil(s.timeoutHandler)):
+  if s.timeoutHandler != nil:
     trace "Calling timeout handler", s
-
-    try:
-      await s.timeoutHandler()
-    except CancelledError:
-      # timeoutHandler is expected to be fast, but it's still possible that
-      # cancellation will happen here - no need to warn about it - we do want to
-      # stop the polling however
-      debug "Timeout handler cancelled", s
-    except CatchableError as exc: # Shouldn't happen
-      warn "exception in timeout handler", s, exc = exc.msg
+    await s.timeoutHandler()
 
   return false
 
-proc timeoutMonitor(s: Connection) {.async, gcsafe.} =
+proc timeoutMonitor(s: Connection) {.async: (raises: []).} =
   ## monitor the channel for inactivity
   ##
   ## if the timeout was hit, it means that
@@ -129,21 +125,22 @@ proc timeoutMonitor(s: Connection) {.async, gcsafe.} =
       return
 
 method getWrapped*(s: Connection): Connection {.base.} =
-  doAssert(false, "not implemented!")
+  raiseAssert("Not implemented!")
 
 when defined(libp2p_agents_metrics):
   proc setShortAgent*(s: Connection, shortAgent: string) =
     var conn = s
-    while not isNil(conn):
+    while conn != nil:
       conn.shortAgent = shortAgent
       conn = conn.getWrapped()
 
-proc new*(C: type Connection,
-           peerId: PeerId,
-           dir: Direction,
-           observedAddr: Opt[MultiAddress],
-           timeout: Duration = DefaultConnectionTimeout,
-           timeoutHandler: TimeoutHandler = nil): Connection =
+proc new*(
+    C: type Connection,
+    peerId: PeerId,
+    dir: Direction,
+    observedAddr: Opt[MultiAddress],
+    timeout: Duration = DefaultConnectionTimeout,
+    timeoutHandler: TimeoutHandler = nil): Connection =
   result = C(peerId: peerId,
              dir: dir,
              timeout: timeout,

libp2p/stream/lpstream.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -23,8 +23,8 @@ import ../varint,
 
 export errors
 
-declareGauge(libp2p_open_streams,
-  "open stream instances", labels = ["type", "dir"])
+declareGauge libp2p_open_streams,
+  "open stream instances", labels = ["type", "dir"]
 
 export oids
 
@@ -50,12 +50,7 @@ type
 
   LPStreamError* = object of LPError
   LPStreamIncompleteError* = object of LPStreamError
-  LPStreamIncorrectDefect* = object of Defect
   LPStreamLimitError* = object of LPStreamError
-  LPStreamReadError* = object of LPStreamError
-    par*: ref CatchableError
-  LPStreamWriteError* = object of LPStreamError
-    par*: ref CatchableError
   LPStreamEOFError* = object of LPStreamError
 
 #        X        |           Read            |         Write
@@ -77,54 +72,12 @@ type
     opened*: uint64
     closed*: uint64
 
-proc setupStreamTracker*(name: string): StreamTracker =
-  let tracker = new StreamTracker
-
-  proc dumpTracking(): string {.gcsafe.} =
-    return "Opened " & tracker.id & ": " & $tracker.opened & "\n" &
-            "Closed " & tracker.id & ": " & $tracker.closed
-
-  proc leakTransport(): bool {.gcsafe.} =
-    return (tracker.opened != tracker.closed)
-
-  tracker.id = name
-  tracker.opened = 0
-  tracker.closed = 0
-  tracker.dump = dumpTracking
-  tracker.isLeaked = leakTransport
-  addTracker(name, tracker)
-
-  return tracker
-
-proc getStreamTracker(name: string): StreamTracker {.gcsafe.} =
-  result = cast[StreamTracker](getTracker(name))
-  if isNil(result):
-    result = setupStreamTracker(name)
-
-proc newLPStreamReadError*(p: ref CatchableError): ref LPStreamReadError =
-  var w = newException(LPStreamReadError, "Read stream failed")
-  w.msg = w.msg & ", originated from [" & $p.name & "] " & p.msg
-  w.par = p
-  result = w
-
-proc newLPStreamReadError*(msg: string): ref LPStreamReadError =
-  newException(LPStreamReadError, msg)
-
-proc newLPStreamWriteError*(p: ref CatchableError): ref LPStreamWriteError =
-  var w = newException(LPStreamWriteError, "Write stream failed")
-  w.msg = w.msg & ", originated from [" & $p.name & "] " & p.msg
-  w.par = p
-  result = w
-
 proc newLPStreamIncompleteError*(): ref LPStreamIncompleteError =
   result = newException(LPStreamIncompleteError, "Incomplete data received")
 
 proc newLPStreamLimitError*(): ref LPStreamLimitError =
   result = newException(LPStreamLimitError, "Buffer limit reached")
 
-proc newLPStreamIncorrectDefect*(m: string): ref LPStreamIncorrectDefect =
-  result = newException(LPStreamIncorrectDefect, m)
-
 proc newLPStreamEOFError*(): ref LPStreamEOFError =
   result = newException(LPStreamEOFError, "Stream EOF!")
 
@@ -145,8 +98,9 @@ proc newLPStreamConnDownError*(
     parentException)
 
 func shortLog*(s: LPStream): auto =
-  if s.isNil: "LPStream(nil)"
+  if s == nil: "LPStream(nil)"
   else: $s.oid
+
 chronicles.formatIt(LPStream): shortLog(it)
 
 method initStream*(s: LPStream) {.base.} =
@@ -157,10 +111,12 @@ method initStream*(s: LPStream) {.base.} =
   s.oid = genOid()
 
   libp2p_open_streams.inc(labelValues = [s.objName, $s.dir])
-  inc getStreamTracker(s.objName).opened
+  trackCounter(s.objName)
   trace "Stream created", s, objName = s.objName, dir = $s.dir
 
-proc join*(s: LPStream): Future[void] {.public.} =
+proc join*(
+    s: LPStream
+): Future[void] {.async: (raises: [CancelledError], raw: true), public.} =
   ## Wait for the stream to be closed
   s.closeEvent.wait()
 
@@ -171,19 +127,21 @@ method atEof*(s: LPStream): bool {.base, public.} =
   s.isEof
 
 method readOnce*(
-  s: LPStream,
-  pbytes: pointer,
-  nbytes: int):
-  Future[int] {.base, async, public.} =
+    s: LPStream,
+    pbytes: pointer,
+    nbytes: int
+): Future[int] {.base, async: (raises: [
+    CancelledError, LPStreamError], raw: true), public.} =
   ## Reads whatever is available in the stream,
   ## up to `nbytes`. Will block if nothing is
   ## available
-  doAssert(false, "not implemented!")
+  raiseAssert("Not implemented!")
 
-proc readExactly*(s: LPStream,
-                  pbytes: pointer,
-                  nbytes: int):
-                  Future[void] {.async, public.} =
+proc readExactly*(
+    s: LPStream,
+    pbytes: pointer,
+    nbytes: int
+): Future[void] {.async: (raises: [CancelledError, LPStreamError]), public.} =
   ## Waits for `nbytes` to be available, then read
   ## them and return them
   if s.atEof:
@@ -217,10 +175,11 @@ proc readExactly*(s: LPStream,
     trace "couldn't read all bytes, incomplete data", s, nbytes, read
     raise newLPStreamIncompleteError()
 
-proc readLine*(s: LPStream,
-               limit = 0,
-               sep = "\r\n"): Future[string]
-               {.async, public.} =
+proc readLine*(
+    s: LPStream,
+    limit = 0,
+    sep = "\r\n"
+): Future[string] {.async: (raises: [CancelledError, LPStreamError]), public.} =
   ## Reads up to `limit` bytes are read, or a `sep` is found
   # TODO replace with something that exploits buffering better
   var lim = if limit <= 0: -1 else: limit
@@ -246,7 +205,9 @@ proc readLine*(s: LPStream,
       if len(result) == lim:
         break
 
-proc readVarint*(conn: LPStream): Future[uint64] {.async, gcsafe, public.} =
+proc readVarint*(
+    conn: LPStream
+): Future[uint64] {.async: (raises: [CancelledError, LPStreamError]), public.} =
   var
     buffer: array[10, byte]
 
@@ -264,7 +225,11 @@ proc readVarint*(conn: LPStream): Future[uint64] {.async, gcsafe, public.} =
   if true: # can't end with a raise apparently
     raise (ref InvalidVarintError)(msg: "Cannot parse varint")
 
-proc readLp*(s: LPStream, maxSize: int): Future[seq[byte]] {.async, gcsafe, public.} =
+proc readLp*(
+    s: LPStream,
+    maxSize: int
+): Future[seq[byte]] {.async: (raises: [
+    CancelledError, LPStreamError]), public.} =
   ## read length prefixed msg, with the length encoded as a varint
   let
     length = await s.readVarint()
@@ -278,13 +243,21 @@ proc readLp*(s: LPStream, maxSize: int): Future[seq[byte]] {.async, gcsafe, publ
 
   var res = newSeqUninitialized[byte](length)
   await s.readExactly(addr res[0], res.len)
-  return res
+  res
 
-method write*(s: LPStream, msg: seq[byte]): Future[void] {.base, public.} =
+method write*(
+    s: LPStream,
+    msg: seq[byte]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true), base, public.} =
   # Write `msg` to stream, waiting for the write to be finished
-  doAssert(false, "not implemented!")
+  raiseAssert("Not implemented!")
 
-proc writeLp*(s: LPStream, msg: openArray[byte]): Future[void] {.public.} =
+proc writeLp*(
+    s: LPStream,
+    msg: openArray[byte]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true), public.} =
   ## Write `msg` with a varint-encoded length prefix
   let vbytes = PB.toBytes(msg.len().uint64)
   var buf = newSeqUninitialized[byte](msg.len() + vbytes.len)
@@ -292,35 +265,53 @@ proc writeLp*(s: LPStream, msg: openArray[byte]): Future[void] {.public.} =
   buf[vbytes.len..<buf.len] = msg
   s.write(buf)
 
-proc writeLp*(s: LPStream, msg: string): Future[void] {.public.} =
+proc writeLp*(
+    s: LPStream,
+    msg: string
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true), public.} =
   writeLp(s, msg.toOpenArrayByte(0, msg.high))
 
-proc write*(s: LPStream, msg: string): Future[void] {.public.} =
+proc write*(
+    s: LPStream,
+    msg: string
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true), public.} =
   s.write(msg.toBytes())
 
-method closeImpl*(s: LPStream): Future[void] {.async, base.} =
+method closeImpl*(
+    s: LPStream
+): Future[void] {.async: (raises: [], raw: true), base.} =
   ## Implementation of close - called only once
   trace "Closing stream", s, objName = s.objName, dir = $s.dir
   libp2p_open_streams.dec(labelValues = [s.objName, $s.dir])
-  inc getStreamTracker(s.objName).closed
+  untrackCounter(s.objName)
   s.closeEvent.fire()
   trace "Closed stream", s, objName = s.objName, dir = $s.dir
+  let fut = newFuture[void]()
+  fut.complete()
+  fut
 
-method close*(s: LPStream): Future[void] {.base, async, public.} = # {.raises [Defect].}
+method close*(
+    s: LPStream
+): Future[void] {.async: (raises: [], raw: true), base, public.} =
   ## close the stream - this may block, but will not raise exceptions
   ##
   if s.isClosed:
     trace "Already closed", s
-    return
+    let fut = newFuture[void]()
+    fut.complete()
+    return fut
 
   s.isClosed = true # Set flag before performing virtual close
 
-  # An separate implementation method is used so that even when derived types
+  # A separate implementation method is used so that even when derived types
   # override `closeImpl`, it is called only once - anyone overriding `close`
   # itself must implement this - once-only check as well, with their own field
-  await closeImpl(s)
+  closeImpl(s)
 
-proc closeWithEOF*(s: LPStream): Future[void] {.async, public.} =
+proc closeWithEOF*(
+    s: LPStream): Future[void] {.async: (raises: []), public.} =
   ## Close the stream and wait for EOF - use this with half-closed streams where
   ## an EOF is expected to arrive from the other end.
   ##
@@ -349,9 +340,9 @@ proc closeWithEOF*(s: LPStream): Future[void] {.async, public.} =
     var buf: array[8, byte]
     if (await readOnce(s, addr buf[0], buf.len)) != 0:
       debug "Unexpected bytes while waiting for EOF", s
+  except CancelledError:
+    discard
   except LPStreamEOFError:
     trace "Expected EOF came", s
-  except CancelledError as exc:
-    raise exc
-  except CatchableError as exc:
+  except LPStreamError as exc:
     debug "Unexpected error while waiting for EOF", s, msg = exc.msg

libp2p/switch.nim

@@ -71,17 +71,17 @@ type
       inUse: bool
 
 
-method setup*(self: Service, switch: Switch): Future[bool] {.base, async, gcsafe.} =
+method setup*(self: Service, switch: Switch): Future[bool] {.base, async.} =
   if self.inUse:
     warn "service setup has already been called"
     return false
   self.inUse = true
   return true
 
-method run*(self: Service, switch: Switch) {.base, async, gcsafe.} =
+method run*(self: Service, switch: Switch) {.base, async.} =
   doAssert(false, "Not implemented!")
 
-method stop*(self: Service, switch: Switch): Future[bool] {.base, async, gcsafe.} =
+method stop*(self: Service, switch: Switch): Future[bool] {.base, async.} =
   if not self.inUse:
     warn "service is already stopped"
     return false
@@ -141,10 +141,10 @@ method connect*(
   addrs: seq[MultiAddress],
   forceDial = false,
   reuseConnection = true,
-  upgradeDir = Direction.Out): Future[void] {.public.} =
+  dir = Direction.Out): Future[void] {.public.} =
   ## Connects to a peer without opening a stream to it
 
-  s.dialer.connect(peerId, addrs, forceDial, reuseConnection, upgradeDir)
+  s.dialer.connect(peerId, addrs, forceDial, reuseConnection, dir)
 
 method connect*(
   s: Switch,
@@ -213,7 +213,7 @@ proc mount*[T: LPProtocol](s: Switch, proto: T, matcher: Matcher = nil)
   s.peerInfo.protocols.add(proto.codec)
 
 proc upgrader(switch: Switch, trans: Transport, conn: Connection) {.async.} =
-  let muxed = await trans.upgrade(conn, Direction.In, Opt.none(PeerId))
+  let muxed = await trans.upgrade(conn, Opt.none(PeerId))
   switch.connManager.storeMuxer(muxed)
   await switch.peerStore.identify(muxed)
   trace "Connection upgrade succeeded"
@@ -273,6 +273,7 @@ proc accept(s: Switch, transport: Transport) {.async.} = # noraises
     except CancelledError as exc:
       trace "releasing semaphore on cancellation"
       upgrades.release() # always release the slot
+      return
     except CatchableError as exc:
       error "Exception in accept loop, exiting", exc = exc.msg
       upgrades.release() # always release the slot
@@ -288,6 +289,12 @@ proc stop*(s: Switch) {.async, public.} =
 
   s.started = false
 
+  try:
+    # Stop accepting incoming connections
+    await allFutures(s.acceptFuts.mapIt(it.cancelAndWait())).wait(1.seconds)
+  except CatchableError as exc:
+    debug "Cannot cancel accepts", error = exc.msg
+
   for service in s.services:
     discard await service.stop(s)
 
@@ -302,18 +309,6 @@ proc stop*(s: Switch) {.async, public.} =
     except CatchableError as exc:
       warn "error cleaning up transports", msg = exc.msg
 
-  try:
-    await allFutures(s.acceptFuts)
-      .wait(1.seconds)
-  except CatchableError as exc:
-    trace "Exception while stopping accept loops", exc = exc.msg
-
-  # check that all futures were properly
-  # stopped and otherwise cancel them
-  for a in s.acceptFuts:
-    if not a.finished:
-      a.cancel()
-
   for service in s.services:
     discard await service.stop(s)
 
@@ -321,7 +316,7 @@ proc stop*(s: Switch) {.async, public.} =
 
   trace "Switch stopped"
 
-proc start*(s: Switch) {.async, gcsafe, public.} =
+proc start*(s: Switch) {.async, public.} =
   ## Start listening on every transport
 
   if s.started:

libp2p/transports/tcptransport.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -12,274 +12,327 @@
 {.push raises: [].}
 
 import std/[sequtils]
-import stew/results
 import chronos, chronicles
-import transport,
-       ../errors,
-       ../wire,
-       ../multicodec,
-       ../connmanager,
-       ../multiaddress,
-       ../stream/connection,
-       ../stream/chronosstream,
-       ../upgrademngrs/upgrade,
-       ../utility
+import
+  ./transport,
+  ../wire,
+  ../multiaddress,
+  ../stream/connection,
+  ../stream/chronosstream,
+  ../upgrademngrs/upgrade,
+  ../utility
 
 logScope:
   topics = "libp2p tcptransport"
 
-export transport, results
+export transport, connection, upgrade
 
-const
-  TcpTransportTrackerName* = "libp2p.tcptransport"
+const TcpTransportTrackerName* = "libp2p.tcptransport"
 
 type
+  AcceptFuture = typeof(default(StreamServer).accept())
+
   TcpTransport* = ref object of Transport
     servers*: seq[StreamServer]
     clients: array[Direction, seq[StreamTransport]]
     flags: set[ServerFlags]
     clientFlags: set[SocketFlags]
-    acceptFuts: seq[Future[StreamTransport]]
+    acceptFuts: seq[AcceptFuture]
     connectionsTimeout: Duration
-
-  TcpTransportTracker* = ref object of TrackerBase
-    opened*: uint64
-    closed*: uint64
+    stopping: bool
 
   TcpTransportError* = object of transport.TransportError
 
-proc setupTcpTransportTracker(): TcpTransportTracker {.gcsafe, raises: [].}
-
-proc getTcpTransportTracker(): TcpTransportTracker {.gcsafe.} =
-  result = cast[TcpTransportTracker](getTracker(TcpTransportTrackerName))
-  if isNil(result):
-    result = setupTcpTransportTracker()
-
-proc dumpTracking(): string {.gcsafe.} =
-  var tracker = getTcpTransportTracker()
-  result = "Opened tcp transports: " & $tracker.opened & "\n" &
-           "Closed tcp transports: " & $tracker.closed
-
-proc leakTransport(): bool {.gcsafe.} =
-  var tracker = getTcpTransportTracker()
-  result = (tracker.opened != tracker.closed)
-
-proc setupTcpTransportTracker(): TcpTransportTracker =
-  result = new TcpTransportTracker
-  result.opened = 0
-  result.closed = 0
-  result.dump = dumpTracking
-  result.isLeaked = leakTransport
-  addTracker(TcpTransportTrackerName, result)
-
-proc connHandler*(self: TcpTransport,
-                  client: StreamTransport,
-                  observedAddr: Opt[MultiAddress],
-                  dir: Direction): Future[Connection] {.async.} =
-
-  trace "Handling tcp connection", address = $observedAddr,
-                                   dir = $dir,
-                                   clients = self.clients[Direction.In].len +
-                                   self.clients[Direction.Out].len
+proc connHandler*(
+    self: TcpTransport,
+    client: StreamTransport,
+    observedAddr: Opt[MultiAddress],
+    dir: Direction,
+): Connection =
+  trace "Handling tcp connection",
+    address = $observedAddr,
+    dir = $dir,
+    clients = self.clients[Direction.In].len + self.clients[Direction.Out].len
 
   let conn = Connection(
     ChronosStream.init(
       client = client,
       dir = dir,
       observedAddr = observedAddr,
-      timeout = self.connectionsTimeout
-    ))
+      timeout = self.connectionsTimeout,
+    )
+  )
 
-  proc onClose() {.async.} =
-    try:
-      let futs = @[client.join(), conn.join()]
-      await futs[0] or futs[1]
-      for f in futs:
-        if not f.finished: await f.cancelAndWait() # cancel outstanding join()
+  proc onClose() {.async: (raises: []).} =
+    await noCancel client.join()
 
-      trace "Cleaning up client", addrs = $client.remoteAddress,
-                                  conn
+    trace "Cleaning up client", addrs = $client.remoteAddress, conn
 
-      self.clients[dir].keepItIf( it != client )
-      await allFuturesThrowing(
-        conn.close(), client.closeWait())
+    self.clients[dir].keepItIf(it != client)
 
-      trace "Cleaned up client", addrs = $client.remoteAddress,
-                                 conn
+    # Propagate the chronos client being closed to the connection
+    # TODO This is somewhat dubious since it's the connection that owns the
+    #      client, but it allows the transport to close all connections when
+    #      shutting down (also dubious! it would make more sense that the owner
+    #      of all connections closes them, or the next read detects the closed
+    #      socket and does the right thing..)
 
-    except CatchableError as exc:
-      let useExc {.used.} = exc
-      debug "Error cleaning up client", errMsg = exc.msg, conn
+    await conn.close()
+
+    trace "Cleaned up client", addrs = $client.remoteAddress, conn
 
   self.clients[dir].add(client)
+
   asyncSpawn onClose()
 
   return conn
 
 proc new*(
-  T: typedesc[TcpTransport],
-  flags: set[ServerFlags] = {},
-  upgrade: Upgrade,
-  connectionsTimeout = 10.minutes): T {.public.} =
+    T: typedesc[TcpTransport],
+    flags: set[ServerFlags] = {},
+    upgrade: Upgrade,
+    connectionsTimeout = 10.minutes,
+): T {.public.} =
+  T(
+    flags: flags,
+    clientFlags:
+      if ServerFlags.TcpNoDelay in flags:
+        {SocketFlags.TcpNoDelay}
+      else:
+        default(set[SocketFlags])
+    ,
+    upgrader: upgrade,
+    networkReachability: NetworkReachability.Unknown,
+    connectionsTimeout: connectionsTimeout,
+  )
 
-  let
-    transport = T(
-      flags: flags,
-      clientFlags:
-        if ServerFlags.TcpNoDelay in flags:
-          compilesOr:
-            {SocketFlags.TcpNoDelay}
-          do:
-            doAssert(false)
-            default(set[SocketFlags])
-        else:
-          default(set[SocketFlags]),
-      upgrader: upgrade,
-      networkReachability: NetworkReachability.Unknown,
-      connectionsTimeout: connectionsTimeout)
+method start*(self: TcpTransport, addrs: seq[MultiAddress]): Future[void] =
+  ## Start transport listening to the given addresses - for dial-only transports,
+  ## start with an empty list
 
-  return transport
+  # TODO remove `impl` indirection throughout when `raises` is added to base
 
-method start*(
-  self: TcpTransport,
-  addrs: seq[MultiAddress]) {.async.} =
-  ## listen on the transport
-  ##
-
-  if self.running:
-    warn "TCP transport already running"
-    return
-
-  await procCall Transport(self).start(addrs)
-  trace "Starting TCP transport"
-  inc getTcpTransportTracker().opened
-
-  for i, ma in addrs:
-    if not self.handles(ma):
-      trace "Invalid address detected, skipping!", address = ma
-      continue
-
-    self.flags.incl(ServerFlags.ReusePort)
-    let server = createStreamServer(
-      ma = ma,
-      flags = self.flags,
-      udata = self)
-
-    # always get the resolved address in case we're bound to 0.0.0.0:0
-    self.addrs[i] = MultiAddress.init(
-      server.sock.getLocalAddress()
-    ).tryGet()
-
-    self.servers &= server
-
-    trace "Listening on", address = ma
-
-method stop*(self: TcpTransport) {.async, gcsafe.} =
-  ## stop the transport
-  ##
-  try:
-    trace "Stopping TCP transport"
-
-    checkFutures(
-      await allFinished(
-        self.clients[Direction.In].mapIt(it.closeWait()) &
-        self.clients[Direction.Out].mapIt(it.closeWait())))
-
-    if not self.running:
-      warn "TCP transport already stopped"
+  proc impl(
+      self: TcpTransport, addrs: seq[MultiAddress]
+  ): Future[void] {.async: (raises: [transport.TransportError, CancelledError]).} =
+    if self.running:
+      warn "TCP transport already running"
       return
 
-    await procCall Transport(self).stop() # call base
-    var toWait: seq[Future[void]]
-    for fut in self.acceptFuts:
-      if not fut.finished:
-        toWait.add(fut.cancelAndWait())
-      elif fut.done:
-        toWait.add(fut.read().closeWait())
+    trace "Starting TCP transport"
 
-    for server in self.servers:
-      server.stop()
-      toWait.add(server.closeWait())
+    self.flags.incl(ServerFlags.ReusePort)
 
-    await allFutures(toWait)
+    var supported: seq[MultiAddress]
+    var initialized = false
+    try:
+      for i, ma in addrs:
+        if not self.handles(ma):
+          trace "Invalid address detected, skipping!", address = ma
+          continue
 
-    self.servers = @[]
+        let
+          ta = initTAddress(ma).expect("valid address per handles check above")
+          server =
+            try:
+              createStreamServer(ta, flags = self.flags)
+            except common.TransportError as exc:
+              raise (ref TcpTransportError)(msg: exc.msg, parent: exc)
 
-    trace "Transport stopped"
-    inc getTcpTransportTracker().closed
-  except CatchableError as exc:
-    trace "Error shutting down tcp transport", exc = exc.msg
+        self.servers &= server
 
-method accept*(self: TcpTransport): Future[Connection] {.async, gcsafe.} =
-  ## accept a new TCP connection
+        trace "Listening on", address = ma
+        supported.add(
+          MultiAddress.init(server.sock.getLocalAddress()).expect(
+            "Can init from local address"
+          )
+        )
+
+      initialized = true
+    finally:
+      if not initialized:
+        # Clean up partial success on exception
+        await noCancel allFutures(self.servers.mapIt(it.closeWait()))
+        reset(self.servers)
+
+    try:
+      await procCall Transport(self).start(supported)
+    except CatchableError:
+      raiseAssert "Base method does not raise"
+
+    trackCounter(TcpTransportTrackerName)
+
+  impl(self, addrs)
+
+method stop*(self: TcpTransport): Future[void] =
+  ## Stop the transport and close all connections it created
+  proc impl(self: TcpTransport) {.async: (raises: []).} =
+    trace "Stopping TCP transport"
+    self.stopping = true
+    defer:
+      self.stopping = false
+
+    if self.running:
+      # Reset the running flag
+      try:
+        await noCancel procCall Transport(self).stop()
+      except CatchableError: # TODO remove when `accept` is annotated with raises
+        raiseAssert "doesn't actually raise"
+
+      # Stop each server by closing the socket - this will cause all accept loops
+      # to fail - since the running flag has been reset, it's also safe to close
+      # all known clients since no more of them will be added
+      await noCancel allFutures(
+        self.servers.mapIt(it.closeWait()) &
+          self.clients[Direction.In].mapIt(it.closeWait()) &
+          self.clients[Direction.Out].mapIt(it.closeWait())
+      )
+
+      self.servers = @[]
+
+      for acceptFut in self.acceptFuts:
+        if acceptFut.completed():
+          await acceptFut.value().closeWait()
+      self.acceptFuts = @[]
+
+      if self.clients[Direction.In].len != 0 or self.clients[Direction.Out].len != 0:
+        # Future updates could consider turning this warn into an assert since
+        # it should never happen if the shutdown code is correct
+        warn "Couldn't clean up clients",
+          len = self.clients[Direction.In].len + self.clients[Direction.Out].len
+
+      trace "Transport stopped"
+      untrackCounter(TcpTransportTrackerName)
+    else:
+      # For legacy reasons, `stop` on a transpart that wasn't started is
+      # expected to close outgoing connections created by the transport
+      warn "TCP transport already stopped"
+
+      doAssert self.clients[Direction.In].len == 0,
+        "No incoming connections possible without start"
+      await noCancel allFutures(self.clients[Direction.Out].mapIt(it.closeWait()))
+
+  impl(self)
+
+method accept*(self: TcpTransport): Future[Connection] =
+  ## accept a new TCP connection, returning nil on non-fatal errors
   ##
+  ## Raises an exception when the transport is broken and cannot be used for
+  ## accepting further connections
+  # TODO returning nil for non-fatal errors is problematic in that error
+  #      information is lost and must be logged here instead of being
+  #      available to the caller - further refactoring should propagate errors
+  #      to the caller instead
+  proc impl(
+      self: TcpTransport
+  ): Future[Connection] {.async: (raises: [transport.TransportError, CancelledError]).} =
+    if not self.running:
+      raise newTransportClosedError()
 
-  if not self.running:
-    raise newTransportClosedError()
-
-  try:
     if self.acceptFuts.len <= 0:
       self.acceptFuts = self.servers.mapIt(it.accept())
 
-    if self.acceptFuts.len <= 0:
-      return
-
     let
-      finished = await one(self.acceptFuts)
+      finished =
+        try:
+          await one(self.acceptFuts)
+        except ValueError:
+          raise (ref TcpTransportError)(msg: "No listeners configured")
+
       index = self.acceptFuts.find(finished)
+      transp =
+        try:
+          await finished
+        except TransportTooManyError as exc:
+          debug "Too many files opened", exc = exc.msg
+          return nil
+        except TransportAbortedError as exc:
+          debug "Connection aborted", exc = exc.msg
+          return nil
+        except TransportUseClosedError as exc:
+          raise newTransportClosedError(exc)
+        except TransportOsError as exc:
+          raise (ref TcpTransportError)(msg: exc.msg, parent: exc)
+        except common.TransportError as exc: # Needed for chronos 4.0.0 support
+          raise (ref TcpTransportError)(msg: exc.msg, parent: exc)
+        except CancelledError as exc:
+          raise exc
+
+    if not self.running: # Stopped while waiting
+      await transp.closeWait()
+      raise newTransportClosedError()
 
     self.acceptFuts[index] = self.servers[index].accept()
 
-    let transp = await finished
-    try:
-      let observedAddr = MultiAddress.init(transp.remoteAddress).tryGet()
-      return await self.connHandler(transp, Opt.some(observedAddr), Direction.In)
-    except CancelledError as exc:
-      transp.close()
-      raise exc
-    except CatchableError as exc:
-      debug "Failed to handle connection", exc = exc.msg
-      transp.close()
-  except TransportTooManyError as exc:
-    debug "Too many files opened", exc = exc.msg
-  except TransportAbortedError as exc:
-    debug "Connection aborted", exc = exc.msg
-  except TransportUseClosedError as exc:
-    debug "Server was closed", exc = exc.msg
-    raise newTransportClosedError(exc)
-  except CancelledError as exc:
-    raise exc
-  except TransportOsError as exc:
-    info "OS Error", exc = exc.msg
-    raise exc
-  except CatchableError as exc:
-    info "Unexpected error accepting connection", exc = exc.msg
-    raise exc
+    let remote =
+      try:
+        transp.remoteAddress
+      except TransportOsError as exc:
+        # The connection had errors / was closed before `await` returned control
+        await transp.closeWait()
+        debug "Cannot read remote address", exc = exc.msg
+        return nil
+
+    let observedAddr =
+      MultiAddress.init(remote).expect("Can initialize from remote address")
+    self.connHandler(transp, Opt.some(observedAddr), Direction.In)
+
+  impl(self)
 
 method dial*(
-  self: TcpTransport,
-  hostname: string,
-  address: MultiAddress,
-  peerId: Opt[PeerId] = Opt.none(PeerId)): Future[Connection] {.async, gcsafe.} =
+    self: TcpTransport,
+    hostname: string,
+    address: MultiAddress,
+    peerId: Opt[PeerId] = Opt.none(PeerId),
+): Future[Connection] =
   ## dial a peer
-  ##
+  proc impl(
+      self: TcpTransport, hostname: string, address: MultiAddress, peerId: Opt[PeerId]
+  ): Future[Connection] {.async: (raises: [transport.TransportError, CancelledError]).} =
+    if self.stopping:
+      raise newTransportClosedError()
 
-  trace "Dialing remote peer", address = $address
-  let transp =
-    if self.networkReachability == NetworkReachability.NotReachable and self.addrs.len > 0:
-      self.clientFlags.incl(SocketFlags.ReusePort)
-      await connect(address, flags = self.clientFlags, localAddress = Opt.some(self.addrs[0]))
-    else:
-      await connect(address, flags = self.clientFlags)
+    let ta = initTAddress(address).valueOr:
+      raise (ref TcpTransportError)(msg: "Unsupported address: " & $address)
 
-  try:
-    let observedAddr = MultiAddress.init(transp.remoteAddress).tryGet()
-    return await self.connHandler(transp, Opt.some(observedAddr), Direction.Out)
-  except CatchableError as err:
-    await transp.closeWait()
-    raise err
+    trace "Dialing remote peer", address = $address
+    let transp =
+      try:
+        await(
+          if self.networkReachability == NetworkReachability.NotReachable and
+              self.addrs.len > 0:
+            let local = initTAddress(self.addrs[0]).expect("self address is valid")
+            self.clientFlags.incl(SocketFlags.ReusePort)
+            connect(ta, flags = self.clientFlags, localAddress = local)
+          else:
+            connect(ta, flags = self.clientFlags)
+        )
+      except CancelledError as exc:
+        raise exc
+      except CatchableError as exc:
+        raise (ref TcpTransportError)(msg: exc.msg, parent: exc)
 
-method handles*(t: TcpTransport, address: MultiAddress): bool {.gcsafe.} =
+    # If `stop` is called after `connect` but before `await` returns, we might
+    # end up with a race condition where `stop` returns but not all connections
+    # have been closed - we drop connections in this case in order not to leak
+    # them
+    if self.stopping:
+      # Stopped while waiting for new connection
+      await transp.closeWait()
+      raise newTransportClosedError()
+
+    let observedAddr =
+      try:
+        MultiAddress.init(transp.remoteAddress).expect("remote address is valid")
+      except TransportOsError as exc:
+        await transp.closeWait()
+        raise (ref TcpTransportError)(msg: exc.msg)
+
+    self.connHandler(transp, Opt.some(observedAddr), Direction.Out)
+
+  impl(self, hostname, address, peerId)
+
+method handles*(t: TcpTransport, address: MultiAddress): bool =
   if procCall Transport(t).handles(address):
     if address.protocols.isOk:
-      return TCP.match(address)
+      return TCP.match(address)

libp2p/transports/tortransport.nim

@@ -82,7 +82,7 @@ proc handlesStart(address: MultiAddress): bool {.gcsafe.} =
   return TcpOnion3.match(address)
 
 proc connectToTorServer(
-    transportAddress: TransportAddress): Future[StreamTransport] {.async, gcsafe.} =
+    transportAddress: TransportAddress): Future[StreamTransport] {.async.} =
   let transp = await connect(transportAddress)
   try:
     discard await transp.write(@[Socks5ProtocolVersion, NMethods, Socks5AuthMethod.NoAuth.byte])
@@ -99,7 +99,7 @@ proc connectToTorServer(
     await transp.closeWait()
     raise err
 
-proc readServerReply(transp: StreamTransport) {.async, gcsafe.} =
+proc readServerReply(transp: StreamTransport) {.async.} =
   ## The specification for this code is defined on
   ## [link text](https://www.rfc-editor.org/rfc/rfc1928#section-5)
   ## and [link text](https://www.rfc-editor.org/rfc/rfc1928#section-6).
@@ -121,7 +121,7 @@ proc readServerReply(transp: StreamTransport) {.async, gcsafe.} =
   let atyp = firstFourOctets[3]
   case atyp:
     of Socks5AddressType.IPv4.byte:
-      discard await transp.read(ipV4NumOctets + portNumOctets) 
+      discard await transp.read(ipV4NumOctets + portNumOctets)
     of Socks5AddressType.FQDN.byte:
       let fqdnNumOctets = await transp.read(1)
       discard await transp.read(int(uint8.fromBytes(fqdnNumOctets)) + portNumOctets)
@@ -166,7 +166,7 @@ proc parseDnsTcp(address: MultiAddress):
   (Socks5AddressType.FQDN.byte, dstAddr, dstPort)
 
 proc dialPeer(
-    transp: StreamTransport, address: MultiAddress) {.async, gcsafe.} =
+    transp: StreamTransport, address: MultiAddress) {.async.} =
   let (atyp, dstAddr, dstPort) =
     if Onion3.match(address):
       parseOnion3(address)
@@ -190,7 +190,7 @@ method dial*(
   self: TorTransport,
   hostname: string,
   address: MultiAddress,
-  peerId: Opt[PeerId] = Opt.none(PeerId)): Future[Connection] {.async, gcsafe.} =
+  peerId: Opt[PeerId] = Opt.none(PeerId)): Future[Connection] {.async.} =
   ## dial a peer
   ##
   if not handlesDial(address):
@@ -200,7 +200,7 @@ method dial*(
 
   try:
     await dialPeer(transp, address)
-    return await self.tcpTransport.connHandler(transp, Opt.none(MultiAddress), Direction.Out)
+    return self.tcpTransport.connHandler(transp, Opt.none(MultiAddress), Direction.Out)
   except CatchableError as err:
     await transp.closeWait()
     raise err
@@ -229,14 +229,14 @@ method start*(
   else:
     raise newException(TransportStartError, "Tor Transport couldn't start, no supported addr was provided.")
 
-method accept*(self: TorTransport): Future[Connection] {.async, gcsafe.} =
+method accept*(self: TorTransport): Future[Connection] {.async.} =
   ## accept a new Tor connection
   ##
   let conn = await self.tcpTransport.accept()
   conn.observedAddr = Opt.none(MultiAddress)
   return conn
 
-method stop*(self: TorTransport) {.async, gcsafe.} =
+method stop*(self: TorTransport) {.async.} =
   ## stop the transport
   ##
   await procCall Transport(self).stop() # call base

libp2p/transports/transport.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -35,7 +35,7 @@ type
     upgrader*: Upgrade
     networkReachability*: NetworkReachability
 
-proc newTransportClosedError*(parent: ref Exception = nil): ref LPError =
+proc newTransportClosedError*(parent: ref Exception = nil): ref TransportError =
   newException(TransportClosedError,
     "Transport closed, no more connections!", parent)
 
@@ -81,26 +81,25 @@ proc dial*(
   self.dial("", address)
 
 method upgrade*(
-  self: Transport,
-  conn: Connection,
-  direction: Direction,
-  peerId: Opt[PeerId]): Future[Muxer] {.base, gcsafe.} =
+    self: Transport,
+    conn: Connection,
+    peerId: Opt[PeerId]
+): Future[Muxer] {.base, async: (raises: [
+    CancelledError, LPError], raw: true).} =
   ## base upgrade method that the transport uses to perform
   ## transport specific upgrades
   ##
-
-  self.upgrader.upgrade(conn, direction, peerId)
+  self.upgrader.upgrade(conn, peerId)
 
 method handles*(
-  self: Transport,
-  address: MultiAddress): bool {.base, gcsafe.} =
+    self: Transport,
+    address: MultiAddress): bool {.base, gcsafe.} =
   ## check if transport supports the multiaddress
   ##
-
   # by default we skip circuit addresses to avoid
   # having to repeat the check in every transport
   let protocols = address.protocols.valueOr: return false
-  return protocols
+  protocols
     .filterIt(
       it == multiCodec("p2p-circuit")
     ).len == 0

libp2p/transports/wstransport.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -44,11 +44,12 @@ method initStream*(s: WsStream) =
 
   procCall Connection(s).initStream()
 
-proc new*(T: type WsStream,
-           session: WSSession,
-           dir: Direction,
-           observedAddr: Opt[MultiAddress],
-           timeout = 10.minutes): T =
+proc new*(
+    T: type WsStream,
+    session: WSSession,
+    dir: Direction,
+    observedAddr: Opt[MultiAddress],
+    timeout = 10.minutes): T =
 
   let stream = T(
     session: session,
@@ -63,18 +64,23 @@ template mapExceptions(body: untyped) =
   try:
     body
   except AsyncStreamIncompleteError:
-    raise newLPStreamEOFError()
+    raise newLPStreamIncompleteError()
+  except AsyncStreamLimitError:
+    raise newLPStreamLimitError()
   except AsyncStreamUseClosedError:
     raise newLPStreamEOFError()
   except WSClosedError:
     raise newLPStreamEOFError()
-  except AsyncStreamLimitError:
-    raise newLPStreamLimitError()
+  except WebSocketError:
+    raise newLPStreamEOFError()
+  except CatchableError:
+    raise newLPStreamEOFError()
 
 method readOnce*(
-  s: WsStream,
-  pbytes: pointer,
-  nbytes: int): Future[int] {.async.} =
+    s: WsStream,
+    pbytes: pointer,
+    nbytes: int
+): Future[int] {.async: (raises: [CancelledError, LPStreamError]).} =
   let res = mapExceptions(await s.session.recv(pbytes, nbytes))
 
   if res == 0 and s.session.readyState == ReadyState.Closed:
@@ -83,13 +89,17 @@ method readOnce*(
   return res
 
 method write*(
-  s: WsStream,
-  msg: seq[byte]): Future[void] {.async.} =
+    s: WsStream,
+    msg: seq[byte]
+): Future[void] {.async: (raises: [CancelledError, LPStreamError]).} =
   mapExceptions(await s.session.send(msg, Opcode.Binary))
   s.activity = true # reset activity flag
 
-method closeImpl*(s: WsStream): Future[void] {.async.} =
-  await s.session.close()
+method closeImpl*(s: WsStream): Future[void] {.async: (raises: []).} =
+  try:
+    await s.session.close()
+  except CatchableError:
+    discard
   await procCall Connection(s).closeImpl()
 
 method getWrapped*(s: WsStream): Connection = nil
@@ -136,7 +146,7 @@ method start*(
       if WSS.match(ma):
         if self.secure: true
         else:
-          warn "Trying to listen on a WSS address without setting the certificate!"
+          warn "Trying to listen on a WSS address without setting certificate!"
           false
       else: false
 
@@ -173,7 +183,7 @@ method start*(
 
   self.running = true
 
-method stop*(self: WsTransport) {.async, gcsafe.} =
+method stop*(self: WsTransport) {.async.} =
   ## stop the transport
   ##
 
@@ -237,7 +247,7 @@ proc connHandler(self: WsTransport,
   asyncSpawn onClose()
   return conn
 
-method accept*(self: WsTransport): Future[Connection] {.async, gcsafe.} =
+method accept*(self: WsTransport): Future[Connection] {.async.} =
   ## accept a new WS connection
   ##
 
@@ -295,7 +305,7 @@ method dial*(
   self: WsTransport,
   hostname: string,
   address: MultiAddress,
-  peerId: Opt[PeerId] = Opt.none(PeerId)): Future[Connection] {.async, gcsafe.} =
+  peerId: Opt[PeerId] = Opt.none(PeerId)): Future[Connection] {.async.} =
   ## dial a peer
   ##
 

libp2p/upgrademngrs/muxedupgrade.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -25,55 +25,61 @@ type
     muxers*: seq[MuxerProvider]
     streamHandler*: StreamHandler
 
-proc getMuxerByCodec(self: MuxedUpgrade, muxerName: string): MuxerProvider =
+func getMuxerByCodec(
+    self: MuxedUpgrade, muxerName: string): Opt[MuxerProvider] =
+  if muxerName.len == 0 or muxerName == "na":
+    return Opt.none(MuxerProvider)
   for m in self.muxers:
     if muxerName == m.codec:
-      return m
+      return Opt.some(m)
+  Opt.none(MuxerProvider)
 
-proc mux*(
-  self: MuxedUpgrade,
-  conn: Connection,
-  direction: Direction): Future[Muxer] {.async, gcsafe.} =
+proc mux(
+    self: MuxedUpgrade,
+    conn: Connection
+): Future[Opt[Muxer]] {.async: (raises: [
+    CancelledError, LPStreamError, MultiStreamError]).} =
   ## mux connection
-
   trace "Muxing connection", conn
   if self.muxers.len == 0:
     warn "no muxers registered, skipping upgrade flow", conn
-    return
+    return Opt.none(Muxer)
 
-  let muxerName =
-    if direction == Out: await self.ms.select(conn, self.muxers.mapIt(it.codec))
-    else: await MultistreamSelect.handle(conn, self.muxers.mapIt(it.codec))
-
-  if muxerName.len == 0 or muxerName == "na":
-    debug "no muxer available, early exit", conn
-    return
+  let
+    muxerName =
+      case conn.dir
+      of Direction.Out:
+        await self.ms.select(conn, self.muxers.mapIt(it.codec))
+      of Direction.In:
+        await MultistreamSelect.handle(conn, self.muxers.mapIt(it.codec))
+    muxerProvider = self.getMuxerByCodec(muxerName).valueOr:
+      debug "no muxer available, early exit", conn, muxerName
+      return Opt.none(Muxer)
 
   trace "Found a muxer", conn, muxerName
 
   # create new muxer for connection
-  let muxer = self.getMuxerByCodec(muxerName).newMuxer(conn)
+  let muxer = muxerProvider.newMuxer(conn)
 
   # install stream handler
   muxer.streamHandler = self.streamHandler
   muxer.handler = muxer.handle()
-  return muxer
+  Opt.some(muxer)
 
 method upgrade*(
-  self: MuxedUpgrade,
-  conn: Connection,
-  direction: Direction,
-  peerId: Opt[PeerId]): Future[Muxer] {.async.} =
-  trace "Upgrading connection", conn, direction
+    self: MuxedUpgrade,
+    conn: Connection,
+    peerId: Opt[PeerId]
+): Future[Muxer] {.async: (raises: [CancelledError, LPError]).} =
+  trace "Upgrading connection", conn, direction = conn.dir
 
-  let sconn = await self.secure(conn, direction, peerId) # secure the connection
-  if isNil(sconn):
-    raise newException(UpgradeFailedError,
+  let sconn = await self.secure(conn, peerId)  # secure the connection
+  if sconn == nil:
+    raise (ref UpgradeFailedError)(msg:
       "unable to secure connection, stopping upgrade")
 
-  let muxer = await self.mux(sconn, direction) # mux it if possible
-  if muxer == nil:
-    raise newException(UpgradeFailedError,
+  let muxer = (await self.mux(sconn)).valueOr:  # mux it if possible
+    raise (ref UpgradeFailedError)(msg:
       "a muxer is required for outgoing connections")
 
   when defined(libp2p_agents_metrics):
@@ -81,32 +87,28 @@ method upgrade*(
 
   if sconn.closed():
     await sconn.close()
-    raise newException(UpgradeFailedError,
+    raise (ref UpgradeFailedError)(msg:
       "Connection closed or missing peer info, stopping upgrade")
 
-  trace "Upgraded connection", conn, sconn, direction
-  return muxer
+  trace "Upgraded connection", conn, sconn, direction = conn.dir
+  muxer
 
 proc new*(
-  T: type MuxedUpgrade,
-  muxers: seq[MuxerProvider],
-  secureManagers: openArray[Secure] = [],
-  ms: MultistreamSelect): T =
-
+    T: type MuxedUpgrade,
+    muxers: seq[MuxerProvider],
+    secureManagers: openArray[Secure] = [],
+    ms: MultistreamSelect): T =
   let upgrader = T(
     muxers: muxers,
     secureManagers: @secureManagers,
     ms: ms)
 
-  upgrader.streamHandler = proc(conn: Connection)
-    {.async, gcsafe, raises: [].} =
+  upgrader.streamHandler = proc(conn: Connection) {.async: (raises: []).} =
     trace "Starting stream handler", conn
     try:
       await upgrader.ms.handle(conn) # handle incoming connection
     except CancelledError as exc:
-      raise exc
-    except CatchableError as exc:
-      trace "exception in stream handler", conn, msg = exc.msg
+      return
     finally:
       await conn.closeWithEOF()
     trace "Stream handler done", conn

libp2p/upgrademngrs/upgrade.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -24,8 +24,10 @@ import ../stream/connection,
 
 export connmanager, connection, identify, secure, multistream
 
-declarePublicCounter(libp2p_failed_upgrades_incoming, "incoming connections failed upgrades")
-declarePublicCounter(libp2p_failed_upgrades_outgoing, "outgoing connections failed upgrades")
+declarePublicCounter(libp2p_failed_upgrades_incoming,
+  "incoming connections failed upgrades")
+declarePublicCounter(libp2p_failed_upgrades_outgoing,
+  "outgoing connections failed upgrades")
 
 logScope:
   topics = "libp2p upgrade"
@@ -38,25 +40,28 @@ type
     secureManagers*: seq[Secure]
 
 method upgrade*(
-  self: Upgrade,
-  conn: Connection,
-  direction: Direction,
-  peerId: Opt[PeerId]): Future[Muxer] {.base.} =
-  doAssert(false, "Not implemented!")
+    self: Upgrade,
+    conn: Connection,
+    peerId: Opt[PeerId]
+): Future[Muxer] {.async: (raises: [
+    CancelledError, LPError], raw: true), base.} =
+  raiseAssert("Not implemented!")
 
 proc secure*(
-  self: Upgrade,
-  conn: Connection,
-  direction: Direction,
-  peerId: Opt[PeerId]): Future[Connection] {.async, gcsafe.} =
+    self: Upgrade,
+    conn: Connection,
+    peerId: Opt[PeerId]
+): Future[Connection] {.async: (raises: [CancelledError, LPError]).} =
   if self.secureManagers.len <= 0:
-    raise newException(UpgradeFailedError, "No secure managers registered!")
+    raise (ref UpgradeFailedError)(msg: "No secure managers registered!")
 
   let codec =
-    if direction == Out: await self.ms.select(conn, self.secureManagers.mapIt(it.codec))
-    else: await MultistreamSelect.handle(conn, self.secureManagers.mapIt(it.codec))
+    if conn.dir == Out:
+      await self.ms.select(conn, self.secureManagers.mapIt(it.codec))
+    else:
+      await MultistreamSelect.handle(conn, self.secureManagers.mapIt(it.codec))
   if codec.len == 0:
-    raise newException(UpgradeFailedError, "Unable to negotiate a secure channel!")
+    raise (ref UpgradeFailedError)(msg: "Unable to negotiate a secure channel!")
 
   trace "Securing connection", conn, codec
   let secureProtocol = self.secureManagers.filterIt(it.codec == codec)
@@ -65,4 +70,4 @@ proc secure*(
   # let's avoid duplicating checks but detect if it fails to do it properly
   doAssert(secureProtocol.len > 0)
 
-  return await secureProtocol[0].secure(conn, direction == Out, peerId)
+  await secureProtocol[0].secure(conn, peerId)

libp2p/utility.nim

@@ -89,28 +89,50 @@ template exceptionToAssert*(body: untyped): untyped =
     res
 
 template withValue*[T](self: Opt[T] | Option[T], value, body: untyped): untyped =
-  if self.isSome:
-    let value {.inject.} = self.get()
+  ## This template provides a convenient way to work with `Option` types in Nim.
+  ## It allows you to execute a block of code (`body`) only when the `Option` is not empty.
+  ##
+  ## `self` is the `Option` instance being checked.
+  ## `value` is the variable name to be used within the `body` for the unwrapped value.
+  ## `body` is a block of code that is executed only if `self` contains a value.
+  ##
+  ## The `value` within `body` is automatically unwrapped from the `Option`, making it
+  ## simpler to work with without needing explicit checks or unwrapping.
+  ##
+  ## Example:
+  ## ```nim
+  ## let myOpt = Opt.some(5)
+  ## myOpt.withValue(value):
+  ##   echo value # Will print 5
+  ## ```
+  ##
+  ## Note: This is a template, and it will be inlined at the call site, offering good performance.
+  let temp = (self)
+  if temp.isSome:
+    let value {.inject.} = temp.get()
     body
 
-macro withValue*[T](self: Opt[T] | Option[T], value, body, body2: untyped): untyped =
-  let elseBody = body2[0]
+macro withValue*[T](self: Opt[T] | Option[T], value, body, elseStmt: untyped): untyped =
+  let elseBody = elseStmt[0]
   quote do:
-    if `self`.isSome:
-      let `value` {.inject.} = `self`.get()
+    let temp = (`self`)
+    if temp.isSome:
+      let `value` {.inject.} = temp.get()
       `body`
     else:
       `elseBody`
 
 template valueOr*[T](self: Option[T], body: untyped): untyped =
-  if self.isSome:
-    self.get()
+  let temp = (self)
+  if temp.isSome:
+    temp.get()
   else:
     body
 
 template toOpt*[T, E](self: Result[T, E]): Opt[T] =
-  if self.isOk:
+  let temp = (self)
+  if temp.isOk:
     when T is void: Result[void, void].ok()
-    else: Opt.some(self.unsafeGet())
+    else: Opt.some(temp.unsafeGet())
   else:
     Opt.none(type(T))

libp2p/wire.nim

@@ -13,6 +13,8 @@
 import chronos, stew/endians2
 import multiaddress, multicodec, errors, utility
 
+export multiaddress, chronos
+
 when defined(windows):
   import winlean
 else:
@@ -30,7 +32,6 @@ const
     UDP,
   )
 
-
 proc initTAddress*(ma: MultiAddress): MaResult[TransportAddress] =
   ## Initialize ``TransportAddress`` with MultiAddress ``ma``.
   ##
@@ -76,7 +77,7 @@ proc connect*(
   child: StreamTransport = nil,
   flags = default(set[SocketFlags]),
   localAddress: Opt[MultiAddress] = Opt.none(MultiAddress)): Future[StreamTransport]
-  {.raises: [LPError, MaInvalidAddress].} =
+  {.async.} =
   ## Open new connection to remote peer with address ``ma`` and create
   ## new transport object ``StreamTransport`` for established connection.
   ## ``bufferSize`` is size of internal buffer for transport.
@@ -88,12 +89,12 @@ proc connect*(
   let transportAddress = initTAddress(ma).tryGet()
 
   compilesOr:
-    return connect(transportAddress, bufferSize, child,
+    return await connect(transportAddress, bufferSize, child,
       if localAddress.isSome(): initTAddress(localAddress.expect("just checked")).tryGet() else: TransportAddress(),
       flags)
   do:
     # support for older chronos versions
-    return connect(transportAddress, bufferSize, child)
+    return await connect(transportAddress, bufferSize, child)
 
 proc createStreamServer*[T](ma: MultiAddress,
                             cbproc: StreamCallback,

scripts/build_p2pd.sh

@@ -89,6 +89,7 @@ build_target() {
 		mkdir "$CACHE_DIR"
 		cp -a "$TARGET_DIR"/* "$CACHE_DIR"/
 	fi
+	echo "Binary built successfully."
 }
 
 if target_needs_rebuilding; then

tests/asyncunit.nim

@@ -5,21 +5,21 @@ export unittest2, chronos
 template asyncTeardown*(body: untyped): untyped =
   teardown:
     waitFor((
-      proc() {.async, gcsafe.} =
+      proc() {.async.} =
         body
     )())
 
 template asyncSetup*(body: untyped): untyped =
   setup:
     waitFor((
-      proc() {.async, gcsafe.} =
+      proc() {.async.} =
         body
     )())
 
 template asyncTest*(name: string, body: untyped): untyped =
   test name:
     waitFor((
-      proc() {.async, gcsafe.} =
+      proc() {.async.} =
         body
     )())
 
@@ -31,7 +31,7 @@ template flakyAsyncTest*(name: string, attempts: int, body: untyped): untyped =
       inc attemptNumber
       try:
         waitFor((
-          proc() {.async, gcsafe.} =
+          proc() {.async.} =
             body
         )())
       except Exception as e:

tests/commoninterop.nim

@@ -20,7 +20,7 @@ proc writeLp(s: StreamTransport, msg: string | seq[byte]): Future[int] {.gcsafe.
   buf.finish()
   result = s.write(buf.buffer)
 
-proc readLp(s: StreamTransport): Future[seq[byte]] {.async, gcsafe.} =
+proc readLp(s: StreamTransport): Future[seq[byte]] {.async.} =
   ## read length prefixed msg
   var
     size: uint

tests/commontransport.nim

@@ -30,7 +30,7 @@ template commonTransportTest*(prov: TransportProvider, ma1: string, ma2: string
 
       let transport2 = transpProvider()
 
-      proc acceptHandler() {.async, gcsafe.} =
+      proc acceptHandler() {.async.} =
         let conn = await transport1.accept()
         if conn.observedAddr.isSome():
           check transport1.handles(conn.observedAddr.get())
@@ -58,7 +58,7 @@ template commonTransportTest*(prov: TransportProvider, ma1: string, ma2: string
       let transport1 = transpProvider()
       await transport1.start(ma)
 
-      proc acceptHandler() {.async, gcsafe.} =
+      proc acceptHandler() {.async.} =
         let conn = await transport1.accept()
         await conn.write("Hello!")
         await conn.close()
@@ -85,7 +85,7 @@ template commonTransportTest*(prov: TransportProvider, ma1: string, ma2: string
       let transport1 = transpProvider()
       await transport1.start(ma)
 
-      proc acceptHandler() {.async, gcsafe.} =
+      proc acceptHandler() {.async.} =
         let conn = await transport1.accept()
         var msg = newSeq[byte](6)
         await conn.readExactly(addr msg[0], 6)
@@ -147,7 +147,7 @@ template commonTransportTest*(prov: TransportProvider, ma1: string, ma2: string
       let transport1 = transpProvider()
       await transport1.start(addrs)
 
-      proc acceptHandler() {.async, gcsafe.} =
+      proc acceptHandler() {.async.} =
         while true:
           let conn = await transport1.accept()
           await conn.write(newSeq[byte](0))
@@ -214,7 +214,7 @@ template commonTransportTest*(prov: TransportProvider, ma1: string, ma2: string
       let transport1 = transpProvider()
       await transport1.start(ma)
 
-      proc acceptHandler() {.async, gcsafe.} =
+      proc acceptHandler() {.async.} =
         let conn = await transport1.accept()
         await conn.close()
 

tests/errorhelpers.nim

@@ -0,0 +1,29 @@
+import
+  std/sequtils,
+  chronos
+
+proc allFuturesThrowing*(args: varargs[FutureBase]): Future[void] =
+  # This proc is only meant for use in tests / not suitable for general use.
+  # - Swallowing errors arbitrarily instead of aggregating them is bad design
+  # - It raises `CatchableError` instead of the union of the `futs` errors,
+  #   inflating the caller's `raises` list unnecessarily. `macro` could fix it
+  let futs = @args
+  (proc() {.async: (raises: [CatchableError]).} =
+    await allFutures(futs)
+    var firstErr: ref CatchableError
+    for fut in futs:
+      if fut.failed:
+        let err = fut.error()
+        if err of CancelledError:
+          raise err
+        if firstErr == nil:
+          firstErr = err
+    if firstErr != nil:
+      raise firstErr)()
+
+proc allFuturesThrowing*[T](futs: varargs[Future[T]]): Future[void] =
+  allFuturesThrowing(futs.mapIt(FutureBase(it)))
+
+proc allFuturesThrowing*[T, E](  # https://github.com/nim-lang/Nim/issues/23432
+    futs: varargs[InternalRaisesFuture[T, E]]): Future[void] =
+  allFuturesThrowing(futs.mapIt(FutureBase(it)))

tests/helpers.nim

@@ -1,6 +1,7 @@
 {.push raises: [].}
 
 import chronos
+import macros
 import algorithm
 
 import ../libp2p/transports/tcptransport
@@ -13,8 +14,8 @@ import ../libp2p/protocols/secure/secure
 import ../libp2p/switch
 import ../libp2p/nameresolving/[nameresolver, mockresolver]
 
-import ./asyncunit
-export asyncunit, mockresolver
+import "."/[asyncunit, errorhelpers]
+export asyncunit, errorhelpers, mockresolver
 
 const
   StreamTransportTrackerName = "stream.transport"
@@ -34,25 +35,19 @@ const
     ChronosStreamTrackerName
   ]
 
-iterator testTrackers*(extras: openArray[string] = []): TrackerBase =
-  for name in trackerNames:
-    let t = getTracker(name)
-    if not isNil(t): yield t
-  for name in extras:
-    let t = getTracker(name)
-    if not isNil(t): yield t
-
 template checkTracker*(name: string) =
-  var tracker = getTracker(name)
-  if tracker.isLeaked():
-    checkpoint tracker.dump()
+  if isCounterLeaked(name):
+    let
+      tracker = getTrackerCounter(name)
+      trackerDescription =
+        "Opened " & name & ": " & $tracker.opened & "\n" &
+        "Closed " & name & ": " & $tracker.closed
+    checkpoint trackerDescription
     fail()
 
 template checkTrackers*() =
-  for tracker in testTrackers():
-    if tracker.isLeaked():
-      checkpoint tracker.dump()
-      fail()
+  for name in trackerNames:
+    checkTracker(name)
   # Also test the GC is not fooling with us
   when defined(nimHasWarnBareExcept):
     {.push warning[BareExcept]:off.}
@@ -81,11 +76,18 @@ template rng*(): ref HmacDrbgContext =
   getRng()
 
 type
-  WriteHandler* = proc(data: seq[byte]): Future[void] {.gcsafe, raises: [].}
+  WriteHandler* = proc(
+      data: seq[byte]
+  ): Future[void] {.async: (raises: [CancelledError, LPStreamError]).}
+
   TestBufferStream* = ref object of BufferStream
     writeHandler*: WriteHandler
 
-method write*(s: TestBufferStream, msg: seq[byte]): Future[void] =
+method write*(
+    s: TestBufferStream,
+    msg: seq[byte]
+): Future[void] {.async: (raises: [
+    CancelledError, LPStreamError], raw: true).} =
   s.writeHandler(msg)
 
 method getWrapped*(s: TestBufferStream): Connection = nil
@@ -103,26 +105,94 @@ proc bridgedConnections*: (Connection, Connection) =
   connB.dir = Direction.In
   connA.initStream()
   connB.initStream()
-  connA.writeHandler = proc(data: seq[byte]) {.async.} =
-    await connB.pushData(data)
+  connA.writeHandler =
+    proc(data: seq[byte]) {.async: (raises: [
+        CancelledError, LPStreamError], raw: true).} =
+      connB.pushData(data)
 
-  connB.writeHandler = proc(data: seq[byte]) {.async.} =
-    await connA.pushData(data)
+  connB.writeHandler =
+    proc(data: seq[byte]) {.async: (raises: [
+        CancelledError, LPStreamError], raw: true).} =
+      connA.pushData(data)
   return (connA, connB)
 
-
-proc checkExpiringInternal(cond: proc(): bool {.raises: [], gcsafe.} ): Future[bool] {.async, gcsafe.} =
-  let start = Moment.now()
-  while true:
-    if Moment.now() > (start + chronos.seconds(5)):
-      return false
-    elif cond():
-      return true
+macro checkUntilCustomTimeout*(timeout: Duration, code: untyped): untyped =
+  ## Periodically checks a given condition until it is true or a timeout occurs.
+  ##
+  ## `code`: untyped - A condition expression that should eventually evaluate to true.
+  ## `timeout`: Duration - The maximum duration to wait for the condition to be true.
+  ##
+  ## Examples:
+  ##   ```nim
+  ##   # Example 1:
+  ##   asyncTest "checkUntilCustomTimeout should pass if the condition is true":
+  ##     let a = 2
+  ##     let b = 2
+  ##     checkUntilCustomTimeout(2.seconds):
+  ##       a == b
+  ##
+  ##   # Example 2: Multiple conditions
+  ##   asyncTest "checkUntilCustomTimeout should pass if the conditions are true":
+  ##     let a = 2
+  ##     let b = 2
+  ##     checkUntilCustomTimeout(5.seconds)::
+  ##       a == b
+  ##       a == 2
+  ##       b == 1
+  ##   ```
+  # Helper proc to recursively build a combined boolean expression
+  proc buildAndExpr(n: NimNode): NimNode =
+    if n.kind == nnkStmtList and n.len > 0:
+      var combinedExpr = n[0]  # Start with the first expression
+      for i in 1..<n.len:
+        # Combine the current expression with the next using 'and'
+        combinedExpr = newCall("and", combinedExpr, n[i])
+      return combinedExpr
     else:
-      await sleepAsync(1.millis)
+      return n
 
-template checkExpiring*(code: untyped): untyped =
-  check await checkExpiringInternal(proc(): bool = code)
+  # Build the combined expression
+  let combinedBoolExpr = buildAndExpr(code)
+
+  result = quote do:
+    proc checkExpiringInternal(): Future[void] {.gensym, async.} =
+      let start = Moment.now()
+      while true:
+        if Moment.now() > (start + `timeout`):
+          checkpoint("[TIMEOUT] Timeout was reached and the conditions were not true. Check if the code is working as " &
+           "expected or consider increasing the timeout param.")
+          check `code`
+          return
+        else:
+         if `combinedBoolExpr`:
+           return
+         else:
+           await sleepAsync(1.millis)
+    await checkExpiringInternal()
+
+macro checkUntilTimeout*(code: untyped): untyped =
+  ## Same as `checkUntilCustomTimeout` but with a default timeout of 10 seconds.
+  ##
+  ## Examples:
+  ##   ```nim
+  ##   # Example 1:
+  ##   asyncTest "checkUntilTimeout should pass if the condition is true":
+  ##     let a = 2
+  ##     let b = 2
+  ##     checkUntilTimeout:
+  ##       a == b
+  ##
+  ##   # Example 2: Multiple conditions
+  ##   asyncTest "checkUntilTimeout should pass if the conditions are true":
+  ##     let a = 2
+  ##     let b = 2
+  ##     checkUntilTimeout:
+  ##       a == b
+  ##       a == 2
+  ##       b == 1
+  ##   ```
+  result = quote do:
+    checkUntilCustomTimeout(10.seconds, `code`)
 
 proc unorderedCompare*[T](a, b: seq[T]): bool =
   if a == b:
@@ -146,8 +216,8 @@ proc default*(T: typedesc[MockResolver]): T =
   resolver.ipResponses[("localhost", true)] = @["::1"]
   resolver
 
-proc setDNSAddr*(switch: Switch) {.gcsafe, async.} =
-  proc addressMapper(listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.gcsafe, async.} =
+proc setDNSAddr*(switch: Switch) {.async.} =
+  proc addressMapper(listenAddrs: seq[MultiAddress]): Future[seq[MultiAddress]] {.async.} =
       return @[MultiAddress.init("/dns4/localhost/").tryGet() & listenAddrs[0][1].tryGet()]
   switch.peerInfo.addressMappers.add(addressMapper)
   await switch.peerInfo.update()

tests/hole-punching-interop/Dockerfile

@@ -0,0 +1,17 @@
+# syntax=docker/dockerfile:1.5-labs
+FROM nimlang/nim:1.6.16 as builder
+
+WORKDIR /workspace
+
+COPY .pinned libp2p.nimble nim-libp2p/
+
+RUN cd nim-libp2p && nimble install_pinned && nimble install redis -y
+
+COPY . nim-libp2p/
+
+RUN cd nim-libp2p && nim c --skipParentCfg --NimblePath:./nimbledeps/pkgs -d:chronicles_log_level=DEBUG -d:chronicles_default_output_device=stderr -d:release --threads:off --skipProjCfg -o:hole-punching-tests ./tests/hole-punching-interop/hole_punching.nim
+
+FROM --platform=linux/amd64 debian:bookworm-slim
+RUN --mount=type=cache,target=/var/cache/apt apt-get update && apt-get install -y dnsutils jq curl tcpdump iproute2
+COPY --from=builder /workspace/nim-libp2p/hole-punching-tests /usr/bin/hole-punch-client
+ENV RUST_BACKTRACE=1

tests/hole-punching-interop/hole_punching.nim

@@ -0,0 +1,115 @@
+import std/[os, options, strformat]
+import redis
+import chronos, chronicles
+import ../../libp2p/[builders,
+                  switch,
+                  observedaddrmanager,
+                  services/hpservice,
+                  services/autorelayservice,
+                  protocols/connectivity/autonat/client as aclient,
+                  protocols/connectivity/relay/client as rclient,
+                  protocols/connectivity/relay/relay,
+                  protocols/connectivity/autonat/service,
+                  protocols/ping]
+import ../stubs/autonatclientstub
+import ../errorhelpers
+
+proc createSwitch(r: Relay = nil, hpService: Service = nil): Switch =
+  let rng = newRng()
+  var builder = SwitchBuilder.new()
+    .withRng(rng)
+    .withAddresses(@[ MultiAddress.init("/ip4/0.0.0.0/tcp/0").tryGet() ])
+    .withObservedAddrManager(ObservedAddrManager.new(maxSize = 1, minCount = 1))
+    .withTcpTransport({ServerFlags.TcpNoDelay})
+    .withYamux()
+    .withAutonat()
+    .withNoise()
+
+  if hpService != nil:
+    builder = builder.withServices(@[hpService])
+
+  if r != nil:
+    builder = builder.withCircuitRelay(r)
+
+  let s =  builder.build()
+  s.mount(Ping.new(rng=rng))
+  return s
+
+proc main() {.async.} =
+  try:
+    let relayClient = RelayClient.new()
+    let autoRelayService = AutoRelayService.new(1, relayClient, nil, newRng())
+    let autonatClientStub = AutonatClientStub.new(expectedDials = 1)
+    autonatClientStub.answer = NotReachable
+    let autonatService = AutonatService.new(autonatClientStub, newRng(), maxQueueSize = 1)
+    let hpservice = HPService.new(autonatService, autoRelayService)
+
+    let
+      isListener = getEnv("MODE") == "listen"
+      switch = createSwitch(relayClient, hpservice)
+      auxSwitch = createSwitch()
+      redisClient = open("redis", 6379.Port)
+
+    debug "Connected to redis"
+
+    await switch.start()
+    await auxSwitch.start()
+
+    let relayAddr =
+      try:
+        redisClient.bLPop(@["RELAY_TCP_ADDRESS"], 0)
+      except Exception as e:
+        raise newException(CatchableError, e.msg)
+
+    # This is necessary to make the autonat service work. It will ask this peer for our reachability which the autonat
+    # client stub will answer NotReachable.
+    await switch.connect(auxSwitch.peerInfo.peerId, auxSwitch.peerInfo.addrs)
+
+    # Wait for autonat to be NotReachable
+    while autonatService.networkReachability != NetworkReachability.NotReachable:
+      await sleepAsync(100.milliseconds)
+
+    # This will trigger the autonat relay service to make a reservation.
+    let relayMA = MultiAddress.init(relayAddr[1]).tryGet()
+    debug "Got relay address", relayMA
+    let relayId = await switch.connect(relayMA)
+    debug "Connected to relay", relayId
+
+    # Wait for our relay address to be published
+    while switch.peerInfo.addrs.len == 0:
+      await sleepAsync(100.milliseconds)
+
+    if isListener:
+      let listenerPeerId = switch.peerInfo.peerId
+      discard redisClient.rPush("LISTEN_CLIENT_PEER_ID", $listenerPeerId)
+      debug "Pushed listener client peer id to redis", listenerPeerId
+
+      # Nothing to do anymore, wait to be killed
+      await sleepAsync(2.minutes)
+    else:
+      let listenerId =
+        try:
+          PeerId.init(redisClient.bLPop(@["LISTEN_CLIENT_PEER_ID"], 0)[1]).tryGet()
+        except Exception as e:
+          raise newException(CatchableError, e.msg)
+
+      debug "Got listener peer id", listenerId
+      let listenerRelayAddr = MultiAddress.init($relayMA & "/p2p-circuit").tryGet()
+
+      debug "Dialing listener relay address", listenerRelayAddr
+      await switch.connect(listenerId, @[listenerRelayAddr])
+
+      # wait for hole-punching to complete in the background
+      await sleepAsync(5000.milliseconds)
+
+      let conn = switch.connManager.selectMuxer(listenerId).connection
+      let channel = await switch.dial(listenerId, @[listenerRelayAddr], PingCodec)
+      let delay = await Ping.new().ping(channel)
+      await allFuturesThrowing(channel.close(), conn.close(), switch.stop(), auxSwitch.stop())
+      echo &"""{{"rtt_to_holepunched_peer_millis":{delay.millis}}}"""
+      quit(0)
+  except CatchableError as e:
+    error "Unexpected error", msg = e.msg
+
+discard waitFor(main().withTimeout(4.minutes))
+quit(1)

tests/hole-punching-interop/version.json

@@ -0,0 +1,7 @@
+{
+  "id": "nim-libp2p-head",
+  "containerImageID": "nim-libp2p-head",
+  "transports": [
+    "tcp"
+  ]
+}

tests/pubsub/testfloodsub.nim

@@ -26,7 +26,7 @@ import ../../libp2p/protocols/pubsub/errors as pubsub_errors
 
 import ../helpers
 
-proc waitSub(sender, receiver: auto; key: string) {.async, gcsafe.} =
+proc waitSub(sender, receiver: auto; key: string) {.async.} =
   # turn things deterministic
   # this is for testing purposes only
   var ceil = 15
@@ -43,7 +43,7 @@ suite "FloodSub":
 
   asyncTest "FloodSub basic publish/subscribe A -> B":
     var completionFut = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       completionFut.complete(true)
 
@@ -81,7 +81,7 @@ suite "FloodSub":
 
   asyncTest "FloodSub basic publish/subscribe B -> A":
     var completionFut = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       completionFut.complete(true)
 
@@ -113,7 +113,7 @@ suite "FloodSub":
 
   asyncTest "FloodSub validation should succeed":
     var handlerFut = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       handlerFut.complete(true)
 
@@ -151,7 +151,7 @@ suite "FloodSub":
     await allFuturesThrowing(nodesFut)
 
   asyncTest "FloodSub validation should fail":
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check false # if we get here, it should fail
 
     let
@@ -186,7 +186,7 @@ suite "FloodSub":
 
   asyncTest "FloodSub validation one fails and one succeeds":
     var handlerFut = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foo"
       handlerFut.complete(true)
 
@@ -235,7 +235,7 @@ suite "FloodSub":
           counter = new int
         futs[i] = (
           fut,
-          (proc(topic: string, data: seq[byte]) {.async, gcsafe.} =
+          (proc(topic: string, data: seq[byte]) {.async.} =
             check topic == "foobar"
             inc counter[]
             if counter[] == runs - 1:
@@ -283,7 +283,7 @@ suite "FloodSub":
           counter = new int
         futs[i] = (
           fut,
-          (proc(topic: string, data: seq[byte]) {.async, gcsafe.} =
+          (proc(topic: string, data: seq[byte]) {.async.} =
             check topic == "foobar"
             inc counter[]
             if counter[] == runs - 1:
@@ -333,7 +333,7 @@ suite "FloodSub":
 
   asyncTest "FloodSub message size validation":
     var messageReceived = 0
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check data.len < 50
       inc(messageReceived)
 
@@ -361,7 +361,7 @@ suite "FloodSub":
     check (await smallNode[0].publish("foo", smallMessage1)) > 0
     check (await bigNode[0].publish("foo", smallMessage2)) > 0
 
-    checkExpiring: messageReceived == 2
+    checkUntilTimeout: messageReceived == 2
 
     check (await smallNode[0].publish("foo", bigMessage)) > 0
     check (await bigNode[0].publish("foo", bigMessage)) > 0
@@ -375,7 +375,7 @@ suite "FloodSub":
 
   asyncTest "FloodSub message size validation 2":
     var messageReceived = 0
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       inc(messageReceived)
 
     let
@@ -396,7 +396,7 @@ suite "FloodSub":
 
     check (await bigNode1[0].publish("foo", bigMessage)) > 0
 
-    checkExpiring: messageReceived == 1
+    checkUntilTimeout: messageReceived == 1
 
     await allFuturesThrowing(
       bigNode1[0].switch.stop(),

tests/pubsub/testgossipinternal.nim

@@ -1,5 +1,5 @@
 # Nim-LibP2P
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE))
 #  * MIT license ([LICENSE-MIT](LICENSE-MIT))
@@ -24,7 +24,8 @@ import utils
 
 import ../helpers
 
-proc noop(data: seq[byte]) {.async, gcsafe.} = discard
+proc noop(data: seq[byte]) {.async: (raises: [CancelledError, LPStreamError]).} =
+  discard
 
 const MsgIdSuccess = "msg id gen success"
 
@@ -524,6 +525,17 @@ suite "GossipSub internal":
     await conn.close()
     await gossipSub.switch.stop()
 
+  asyncTest "invalid message bytes":
+    let gossipSub = TestGossipSub.init(newStandardSwitch())
+
+    let peerId = randomPeerId()
+    let peer = gossipSub.getPubSubPeer(peerId)
+
+    expect(CatchableError):
+      await gossipSub.rpcHandler(peer, @[byte 1, 2, 3])
+
+    await gossipSub.switch.stop()
+
   asyncTest "rebalanceMesh fail due to backoff":
     let gossipSub = TestGossipSub.init(newStandardSwitch())
     let topic = "foobar"
@@ -680,7 +692,7 @@ suite "GossipSub internal":
       )
       peer.iHaveBudget = 0
       let iwants = gossipSub.handleIHave(peer, @[msg])
-      check: iwants.messageIds.len == 0
+      check: iwants.messageIDs.len == 0
 
     block:
       # given duplicate ihave should generate only one iwant
@@ -695,7 +707,7 @@ suite "GossipSub internal":
         messageIDs: @[id, id, id]
       )
       let iwants = gossipSub.handleIHave(peer, @[msg])
-      check: iwants.messageIds.len == 1
+      check: iwants.messageIDs.len == 1
 
     block:
       # given duplicate iwant should generate only one message
@@ -718,104 +730,6 @@ suite "GossipSub internal":
     await allFuturesThrowing(conns.mapIt(it.close()))
     await gossipSub.switch.stop()
 
-  asyncTest "two IHAVEs should generate only one IWANT":
-    let gossipSub = TestGossipSub.init(newStandardSwitch())
-
-    var iwantCount = 0
-
-    proc handler(peer: PubSubPeer, data: seq[byte]) {.async.} =
-     check false
-
-    proc handler2(topic: string, data: seq[byte]) {.async.} = discard
-
-    let topic = "foobar"
-    var conns = newSeq[Connection]()
-    gossipSub.subscribe(topic, handler2)
-
-    # Setup two connections and two peers
-    var ihaveMessageId: string
-    var firstPeer: PubSubPeer
-    let seqno = @[0'u8, 1, 2, 3]
-    for i in 0..<2:
-      let conn = TestBufferStream.new(noop)
-      conns &= conn
-      let peerId = randomPeerId()
-      conn.peerId = peerId
-      let peer = gossipSub.getPubSubPeer(peerId)
-      if isNil(firstPeer):
-        firstPeer = peer
-        ihaveMessageId = byteutils.toHex(seqno) & $firstPeer.peerId
-      peer.handler = handler
-
-      # Simulate that each peer sends an IHAVE message to our node
-      let msg = ControlIHave(
-        topicID: topic,
-        messageIDs: @[ihaveMessageId.toBytes()]
-      )
-      let iwants =  gossipSub.handleIHave(peer, @[msg])
-      if iwants.messageIds.len > 0:
-        iwantCount += 1
-
-    # Verify that our node responds with only one IWANT message
-    check: iwantCount == 1
-    check: gossipSub.outstandingIWANTs.contains(ihaveMessageId.toBytes())
-
-    # Simulate that our node receives the RPCMsg in response to the IWANT
-    let actualMessageData = "Hello, World!".toBytes
-    let rpcMsg = RPCMsg(
-      messages: @[Message(
-        fromPeer: firstPeer.peerId,
-        seqno: seqno,
-        data: actualMessageData
-      )]
-    )
-    await gossipSub.rpcHandler(firstPeer, encodeRpcMsg(rpcMsg, false))
-
-    check: not gossipSub.outstandingIWANTs.contains(ihaveMessageId.toBytes())
-
-    await allFuturesThrowing(conns.mapIt(it.close()))
-    await gossipSub.switch.stop()
-
-  asyncTest "handle unanswered IWANT messages":
-    let gossipSub = TestGossipSub.init(newStandardSwitch())
-    gossipSub.parameters.heartbeatInterval = 50.milliseconds
-    gossipSub.parameters.iwantTimeout = 10.milliseconds
-    await gossipSub.start()
-
-    proc handler(peer: PubSubPeer, data: seq[byte]) {.async.} = discard
-    proc handler2(topic: string, data: seq[byte]) {.async.} = discard
-
-    let topic = "foobar"
-    var conns = newSeq[Connection]()
-    gossipSub.subscribe(topic, handler2)
-
-    # Setup a connection and a peer
-    let conn = TestBufferStream.new(noop)
-    conns &= conn
-    let peerId = randomPeerId()
-    conn.peerId = peerId
-    let peer = gossipSub.getPubSubPeer(peerId)
-    peer.handler = handler
-
-    # Simulate that the peer sends an IHAVE message to our node
-    let ihaveMessageId = @[0'u8, 1, 2, 3]
-    let ihaveMsg = ControlIHave(
-      topicID: topic,
-      messageIDs: @[ihaveMessageId]
-    )
-    discard gossipSub.handleIHave(peer, @[ihaveMsg])
-
-    check: gossipSub.outstandingIWANTs.contains(ihaveMessageId)
-    check: peer.behaviourPenalty == 0.0
-
-    await sleepAsync(60.milliseconds)
-
-    check: not gossipSub.outstandingIWANTs.contains(ihaveMessageId)
-    check: peer.behaviourPenalty == 0.1
-
-    await allFuturesThrowing(conns.mapIt(it.close()))
-    await gossipSub.switch.stop()
-
   proc setupTest(): Future[tuple[gossip0: GossipSub, gossip1: GossipSub, receivedMessages: ref HashSet[seq[byte]]]] {.async.} =
     let
       nodes = generateNodes(2, gossip = true, verifySignature = false)
@@ -828,10 +742,10 @@ suite "GossipSub internal":
 
     var receivedMessages = new(HashSet[seq[byte]])
 
-    proc handlerA(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handlerA(topic: string, data: seq[byte]) {.async.} =
       receivedMessages[].incl(data)
 
-    proc handlerB(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handlerB(topic: string, data: seq[byte]) {.async.} =
       discard
 
     nodes[0].subscribe("foobar", handlerA)
@@ -876,10 +790,10 @@ suite "GossipSub internal":
     let (iwantMessageIds, sentMessages) = createMessages(gossip0, gossip1, messageSize, messageSize)
 
     gossip1.broadcast(gossip1.mesh["foobar"], RPCMsg(control: some(ControlMessage(
-      ihave: @[ControlIHave(topicId: "foobar", messageIds: iwantMessageIds)]
-    ))))
+      ihave: @[ControlIHave(topicID: "foobar", messageIDs: iwantMessageIds)]
+    ))), isHighPriority = false)
 
-    checkExpiring: receivedMessages[] == sentMessages
+    checkUntilTimeout: receivedMessages[] == sentMessages
     check receivedMessages[].len == 2
 
     await teardownTest(gossip0, gossip1)
@@ -893,11 +807,11 @@ suite "GossipSub internal":
     let (bigIWantMessageIds, sentMessages) = createMessages(gossip0, gossip1, messageSize, messageSize)
 
     gossip1.broadcast(gossip1.mesh["foobar"], RPCMsg(control: some(ControlMessage(
-      ihave: @[ControlIHave(topicId: "foobar", messageIds: bigIWantMessageIds)]
-    ))))
+      ihave: @[ControlIHave(topicID: "foobar", messageIDs: bigIWantMessageIds)]
+    ))), isHighPriority = false)
 
     await sleepAsync(300.milliseconds)
-    checkExpiring: receivedMessages[].len == 0
+    checkUntilTimeout: receivedMessages[].len == 0
 
     await teardownTest(gossip0, gossip1)
 
@@ -910,10 +824,10 @@ suite "GossipSub internal":
     let (bigIWantMessageIds, sentMessages) = createMessages(gossip0, gossip1, size1, size2)
 
     gossip1.broadcast(gossip1.mesh["foobar"], RPCMsg(control: some(ControlMessage(
-      ihave: @[ControlIHave(topicId: "foobar", messageIds: bigIWantMessageIds)]
-    ))))
+      ihave: @[ControlIHave(topicID: "foobar", messageIDs: bigIWantMessageIds)]
+    ))), isHighPriority = false)
 
-    checkExpiring: receivedMessages[] == sentMessages
+    checkUntilTimeout: receivedMessages[] == sentMessages
     check receivedMessages[].len == 2
 
     await teardownTest(gossip0, gossip1)
@@ -928,8 +842,8 @@ suite "GossipSub internal":
     let (bigIWantMessageIds, sentMessages) = createMessages(gossip0, gossip1, size1, size2)
 
     gossip1.broadcast(gossip1.mesh["foobar"], RPCMsg(control: some(ControlMessage(
-      ihave: @[ControlIHave(topicId: "foobar", messageIds: bigIWantMessageIds)]
-    ))))
+      ihave: @[ControlIHave(topicID: "foobar", messageIDs: bigIWantMessageIds)]
+    ))), isHighPriority = false)
 
     var smallestSet: HashSet[seq[byte]]
     let seqs = toSeq(sentMessages)
@@ -938,7 +852,7 @@ suite "GossipSub internal":
     else:
       smallestSet.incl(seqs[1])
 
-    checkExpiring: receivedMessages[] == smallestSet
+    checkUntilTimeout: receivedMessages[] == smallestSet
     check receivedMessages[].len == 1
 
     await teardownTest(gossip0, gossip1)

tests/pubsub/testgossipsub.nim

@@ -47,7 +47,7 @@ suite "GossipSub":
 
   asyncTest "GossipSub validation should succeed":
     var handlerFut = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       handlerFut.complete(true)
 
@@ -92,7 +92,7 @@ suite "GossipSub":
     await allFuturesThrowing(nodesFut.concat())
 
   asyncTest "GossipSub validation should fail (reject)":
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check false # if we get here, it should fail
 
     let
@@ -138,7 +138,7 @@ suite "GossipSub":
     await allFuturesThrowing(nodesFut.concat())
 
   asyncTest "GossipSub validation should fail (ignore)":
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check false # if we get here, it should fail
 
     let
@@ -185,7 +185,7 @@ suite "GossipSub":
 
   asyncTest "GossipSub validation one fails and one succeeds":
     var handlerFut = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foo"
       handlerFut.complete(true)
 
@@ -238,7 +238,7 @@ suite "GossipSub":
 
   asyncTest "GossipSub unsub - resub faster than backoff":
     var handlerFut = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       handlerFut.complete(true)
 
@@ -289,7 +289,7 @@ suite "GossipSub":
     await allFuturesThrowing(nodesFut.concat())
 
   asyncTest "e2e - GossipSub should add remote peer topic subscriptions":
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       discard
 
     let
@@ -310,9 +310,9 @@ suite "GossipSub":
     let gossip1 = GossipSub(nodes[0])
     let gossip2 = GossipSub(nodes[1])
 
-    checkExpiring:
-      "foobar" in gossip2.topics and
-      "foobar" in gossip1.gossipsub and
+    checkUntilTimeout:
+      "foobar" in gossip2.topics
+      "foobar" in gossip1.gossipsub
       gossip1.gossipsub.hasPeerId("foobar", gossip2.peerInfo.peerId)
 
     await allFuturesThrowing(
@@ -323,7 +323,7 @@ suite "GossipSub":
     await allFuturesThrowing(nodesFut.concat())
 
   asyncTest "e2e - GossipSub should add remote peer topic subscriptions if both peers are subscribed":
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       discard
 
     let
@@ -374,7 +374,7 @@ suite "GossipSub":
 
   asyncTest "e2e - GossipSub send over fanout A -> B":
     var passed = newFuture[void]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       passed.complete()
 
@@ -428,7 +428,7 @@ suite "GossipSub":
 
   asyncTest "e2e - GossipSub send over fanout A -> B for subscribed topic":
     var passed = newFuture[void]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       passed.complete()
 
@@ -454,9 +454,9 @@ suite "GossipSub":
     nodes[1].subscribe("foobar", handler)
 
     let gsNode = GossipSub(nodes[1])
-    checkExpiring:
-      gsNode.mesh.getOrDefault("foobar").len == 0 and
-      GossipSub(nodes[0]).mesh.getOrDefault("foobar").len == 0 and
+    checkUntilTimeout:
+      gsNode.mesh.getOrDefault("foobar").len == 0
+      GossipSub(nodes[0]).mesh.getOrDefault("foobar").len == 0
       (
         GossipSub(nodes[0]).gossipsub.getOrDefault("foobar").len == 1 or
         GossipSub(nodes[0]).fanout.getOrDefault("foobar").len == 1
@@ -481,7 +481,7 @@ suite "GossipSub":
 
   asyncTest "e2e - GossipSub send over mesh A -> B":
     var passed: Future[bool] = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       passed.complete(true)
 
@@ -548,11 +548,11 @@ suite "GossipSub":
     var
       aReceived = 0
       cReceived = 0
-    proc handlerA(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handlerA(topic: string, data: seq[byte]) {.async.} =
       inc aReceived
       check aReceived < 2
-    proc handlerB(topic: string, data: seq[byte]) {.async, gcsafe.} = discard
-    proc handlerC(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handlerB(topic: string, data: seq[byte]) {.async.} = discard
+    proc handlerC(topic: string, data: seq[byte]) {.async.} =
       inc cReceived
       check cReceived < 2
       cRelayed.complete()
@@ -569,19 +569,19 @@ suite "GossipSub":
     proc slowValidator(topic: string, message: Message): Future[ValidationResult] {.async.} =
       await cRelayed
       # Empty A & C caches to detect duplicates
-      gossip1.seen = TimedCache[MessageId].init()
-      gossip3.seen = TimedCache[MessageId].init()
+      gossip1.seen = TimedCache[SaltedId].init()
+      gossip3.seen = TimedCache[SaltedId].init()
       let msgId = toSeq(gossip2.validationSeen.keys)[0]
-      checkExpiring(try: gossip2.validationSeen[msgId].len > 0 except: false)
+      checkUntilTimeout(try: gossip2.validationSeen[msgId].len > 0 except: false)
       result = ValidationResult.Accept
       bFinished.complete()
 
     nodes[1].addValidator("foobar", slowValidator)
 
-    checkExpiring(
-      gossip1.mesh.getOrDefault("foobar").len == 2 and
-      gossip2.mesh.getOrDefault("foobar").len == 2 and
-      gossip3.mesh.getOrDefault("foobar").len == 2)
+    checkUntilTimeout:
+      gossip1.mesh.getOrDefault("foobar").len == 2
+      gossip2.mesh.getOrDefault("foobar").len == 2
+      gossip3.mesh.getOrDefault("foobar").len == 2
     tryPublish await nodes[0].publish("foobar", "Hello!".toBytes()), 2
 
     await bFinished
@@ -596,7 +596,7 @@ suite "GossipSub":
 
   asyncTest "e2e - GossipSub send over floodPublish A -> B":
     var passed: Future[bool] = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       passed.complete(true)
 
@@ -653,7 +653,7 @@ suite "GossipSub":
     )
 
   proc connectNodes(nodes: seq[PubSub], target: PubSub) {.async.} =
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
 
     for node in nodes:
@@ -661,7 +661,7 @@ suite "GossipSub":
       await node.switch.connect(target.peerInfo.peerId, target.peerInfo.addrs)
 
   proc baseTestProcedure(nodes: seq[PubSub], gossip1: GossipSub, numPeersFirstMsg: int, numPeersSecondMsg: int) {.async.} =
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
 
     block setup:
@@ -676,7 +676,7 @@ suite "GossipSub":
 
     # Now try with a mesh
     gossip1.subscribe("foobar", handler)
-    checkExpiring: gossip1.mesh.peers("foobar") > 5
+    checkUntilTimeout: gossip1.mesh.peers("foobar") > 5
 
     # use a different length so that the message is not equal to the last
     check (await nodes[0].publish("foobar", newSeq[byte](500_000))) == numPeersSecondMsg
@@ -727,7 +727,7 @@ suite "GossipSub":
       var handler: TopicHandler
       closureScope:
         var peerName = $dialer.peerInfo.peerId
-        handler = proc(topic: string, data: seq[byte]) {.async, gcsafe, closure.} =
+        handler = proc(topic: string, data: seq[byte]) {.async, closure.} =
           if peerName notin seen:
             seen[peerName] = 0
           seen[peerName].inc
@@ -778,7 +778,7 @@ suite "GossipSub":
       var handler: TopicHandler
       capture dialer, i:
         var peerName = $dialer.peerInfo.peerId
-        handler = proc(topic: string, data: seq[byte]) {.async, gcsafe, closure.} =
+        handler = proc(topic: string, data: seq[byte]) {.async, closure.} =
           if peerName notin seen:
             seen[peerName] = 0
           seen[peerName].inc
@@ -819,7 +819,7 @@ suite "GossipSub":
     # PX to A & C
     #
     # C sent his SPR, not A
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       discard # not used in this test
 
     let
@@ -895,9 +895,9 @@ suite "GossipSub":
     await nodes[1].switch.connect(nodes[2].switch.peerInfo.peerId, nodes[2].switch.peerInfo.addrs)
 
     let bFinished = newFuture[void]()
-    proc handlerA(topic: string, data: seq[byte]) {.async, gcsafe.} = discard
-    proc handlerB(topic: string, data: seq[byte]) {.async, gcsafe.} = bFinished.complete()
-    proc handlerC(topic: string, data: seq[byte]) {.async, gcsafe.} = doAssert false
+    proc handlerA(topic: string, data: seq[byte]) {.async.} = discard
+    proc handlerB(topic: string, data: seq[byte]) {.async.} = bFinished.complete()
+    proc handlerC(topic: string, data: seq[byte]) {.async.} = doAssert false
 
     nodes[0].subscribe("foobar", handlerA)
     nodes[1].subscribe("foobar", handlerB)
@@ -911,15 +911,15 @@ suite "GossipSub":
     check: gossip3.mesh.peers("foobar") == 1
 
     gossip3.broadcast(gossip3.mesh["foobar"], RPCMsg(control: some(ControlMessage(
-      idontwant: @[ControlIWant(messageIds: @[newSeq[byte](10)])]
-    ))))
-    checkExpiring: gossip2.mesh.getOrDefault("foobar").anyIt(it.heDontWants[^1].len == 1)
+      idontwant: @[ControlIWant(messageIDs: @[newSeq[byte](10)])]
+    ))), isHighPriority = true)
+    checkUntilTimeout: gossip2.mesh.getOrDefault("foobar").anyIt(it.heDontWants[^1].len == 1)
 
     tryPublish await nodes[0].publish("foobar", newSeq[byte](10000)), 1
 
     await bFinished
 
-    checkExpiring: toSeq(gossip3.mesh.getOrDefault("foobar")).anyIt(it.heDontWants[^1].len == 1)
+    checkUntilTimeout: toSeq(gossip3.mesh.getOrDefault("foobar")).anyIt(it.heDontWants[^1].len == 1)
     check: toSeq(gossip1.mesh.getOrDefault("foobar")).anyIt(it.heDontWants[^1].len == 0)
 
     await allFuturesThrowing(
@@ -943,7 +943,7 @@ suite "GossipSub":
 
     await subscribeNodes(nodes)
 
-    proc handle(topic: string, data: seq[byte]) {.async, gcsafe.} = discard
+    proc handle(topic: string, data: seq[byte]) {.async.} = discard
 
     let gossip0 = GossipSub(nodes[0])
     let gossip1 = GossipSub(nodes[1])
@@ -952,6 +952,10 @@ suite "GossipSub":
     gossip1.subscribe("foobar", handle)
     await waitSubGraph(nodes, "foobar")
 
+    # Avoid being disconnected by failing signature verification
+    gossip0.verifySignature = false
+    gossip1.verifySignature = false
+
     return (nodes, gossip0,  gossip1)
 
   proc currentRateLimitHits(): float64 =
@@ -964,8 +968,11 @@ suite "GossipSub":
     let rateLimitHits = currentRateLimitHits()
     let (nodes, gossip0, gossip1) = await initializeGossipTest()
 
-    let msg = RPCMsg(messages: @[Message(topicIDs: @["foobar"], data: "Valid data".toBytes)])
-    gossip0.broadcast(gossip0.mesh["foobar"], msg)
+    gossip0.broadcast(
+      gossip0.mesh["foobar"],
+      RPCMsg(messages: @[Message(topic: "foobar", data: newSeq[byte](10))]),
+      isHighPriority = true,
+    )
     await sleepAsync(300.millis)
 
     check currentRateLimitHits() == rateLimitHits
@@ -973,9 +980,14 @@ suite "GossipSub":
 
     # Disconnect peer when rate limiting is enabled
     gossip1.parameters.disconnectPeerAboveRateLimit = true
-    gossip0.broadcast(gossip0.mesh["foobar"], msg)
+    gossip0.broadcast(
+      gossip0.mesh["foobar"],
+      RPCMsg(messages: @[Message(topic: "foobar", data: newSeq[byte](12))]),
+      isHighPriority = true,
+    )
+    await sleepAsync(300.millis)
 
-    checkExpiring gossip1.switch.isConnected(gossip0.switch.peerInfo.peerId) == true
+    check gossip1.switch.isConnected(gossip0.switch.peerInfo.peerId) == true
     check currentRateLimitHits() == rateLimitHits
 
     await stopNodes(nodes)
@@ -986,8 +998,7 @@ suite "GossipSub":
     let (nodes, gossip0, gossip1) = await initializeGossipTest()
 
     # Simulate sending an undecodable message
-    let msg = newSeqWith[byte](30, 1.byte)
-    await gossip1.peers[gossip0.switch.peerInfo.peerId].sendEncoded(msg)
+    await gossip1.peers[gossip0.switch.peerInfo.peerId].sendEncoded(newSeqWith[byte](33, 1.byte), isHighPriority = true)
     await sleepAsync(300.millis)
 
     check currentRateLimitHits() == rateLimitHits + 1
@@ -995,9 +1006,9 @@ suite "GossipSub":
 
     # Disconnect peer when rate limiting is enabled
     gossip1.parameters.disconnectPeerAboveRateLimit = true
-    await gossip0.peers[gossip1.switch.peerInfo.peerId].sendEncoded(msg)
+    await gossip0.peers[gossip1.switch.peerInfo.peerId].sendEncoded(newSeqWith[byte](35, 1.byte), isHighPriority = true)
 
-    checkExpiring gossip1.switch.isConnected(gossip0.switch.peerInfo.peerId) == false
+    checkUntilTimeout gossip1.switch.isConnected(gossip0.switch.peerInfo.peerId) == false
     check currentRateLimitHits() == rateLimitHits + 2
 
     await stopNodes(nodes)
@@ -1008,11 +1019,10 @@ suite "GossipSub":
 
     let msg = RPCMsg(control: some(ControlMessage(prune: @[
         ControlPrune(topicID: "foobar", peers: @[
-            PeerInfoMsg(peerId: PeerId(data: newSeq[byte](30)))
+            PeerInfoMsg(peerId: PeerId(data: newSeq[byte](33)))
         ], backoff: 123'u64)
     ])))
-
-    gossip0.broadcast(gossip0.mesh["foobar"], msg)
+    gossip0.broadcast(gossip0.mesh["foobar"], msg, isHighPriority = true)
     await sleepAsync(300.millis)
 
     check currentRateLimitHits() == rateLimitHits + 1
@@ -1020,9 +1030,48 @@ suite "GossipSub":
 
     # Disconnect peer when rate limiting is enabled
     gossip1.parameters.disconnectPeerAboveRateLimit = true
-    gossip0.broadcast(gossip0.mesh["foobar"], msg)
+    let msg2 = RPCMsg(control: some(ControlMessage(prune: @[
+        ControlPrune(topicID: "foobar", peers: @[
+            PeerInfoMsg(peerId: PeerId(data: newSeq[byte](35)))
+        ], backoff: 123'u64)
+    ])))
+    gossip0.broadcast(gossip0.mesh["foobar"], msg2, isHighPriority = true)
 
-    checkExpiring gossip1.switch.isConnected(gossip0.switch.peerInfo.peerId) == false
+    checkUntilTimeout gossip1.switch.isConnected(gossip0.switch.peerInfo.peerId) == false
+    check currentRateLimitHits() == rateLimitHits + 2
+
+    await stopNodes(nodes)
+
+  asyncTest "e2e - GossipSub should rate limit invalid messages above the size allowed":
+    let rateLimitHits = currentRateLimitHits()
+    let (nodes, gossip0, gossip1) = await initializeGossipTest()
+
+    let topic = "foobar"
+    proc execValidator(topic: string, message: messages.Message): Future[ValidationResult] {.raises: [].} =
+      let res = newFuture[ValidationResult]()
+      res.complete(ValidationResult.Reject)
+      res
+
+    gossip0.addValidator(topic, execValidator)
+    gossip1.addValidator(topic, execValidator)
+
+    let msg = RPCMsg(messages: @[Message(topic: topic, data: newSeq[byte](40))])
+
+    gossip0.broadcast(gossip0.mesh[topic], msg, isHighPriority = true)
+    await sleepAsync(300.millis)
+
+    check currentRateLimitHits() == rateLimitHits + 1
+    check gossip1.switch.isConnected(gossip0.switch.peerInfo.peerId) == true
+
+    # Disconnect peer when rate limiting is enabled
+    gossip1.parameters.disconnectPeerAboveRateLimit = true
+    gossip0.broadcast(
+      gossip0.mesh[topic],
+      RPCMsg(messages: @[Message(topic: topic, data: newSeq[byte](35))]),
+      isHighPriority = true,
+    )
+
+    checkUntilTimeout gossip1.switch.isConnected(gossip0.switch.peerInfo.peerId) == false
     check currentRateLimitHits() == rateLimitHits + 2
 
     await stopNodes(nodes)

tests/pubsub/testgossipsub2.nim

@@ -59,7 +59,7 @@ suite "GossipSub":
       var handler: TopicHandler
       closureScope:
         var peerName = $dialer.peerInfo.peerId
-        handler = proc(topic: string, data: seq[byte]) {.async, gcsafe, closure.} =
+        handler = proc(topic: string, data: seq[byte]) {.async, closure.} =
           if peerName notin seen:
             seen[peerName] = 0
           seen[peerName].inc
@@ -93,7 +93,7 @@ suite "GossipSub":
 
   asyncTest "GossipSub invalid topic subscription":
     var handlerFut = newFuture[bool]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       handlerFut.complete(true)
 
@@ -155,7 +155,7 @@ suite "GossipSub":
     # DO NOT SUBSCRIBE, CONNECTION SHOULD HAPPEN
     ### await subscribeNodes(nodes)
 
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} = discard
+    proc handler(topic: string, data: seq[byte]) {.async.} = discard
     nodes[1].subscribe("foobar", handler)
 
     await invalidDetected.wait(10.seconds)
@@ -182,10 +182,10 @@ suite "GossipSub":
     await GossipSub(nodes[2]).addDirectPeer(nodes[1].switch.peerInfo.peerId, nodes[1].switch.peerInfo.addrs)
 
     var handlerFut = newFuture[void]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       handlerFut.complete()
-    proc noop(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc noop(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
 
     nodes[0].subscribe("foobar", noop)
@@ -226,7 +226,7 @@ suite "GossipSub":
     GossipSub(nodes[1]).parameters.graylistThreshold = 100000
 
     var handlerFut = newFuture[void]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
       handlerFut.complete()
 
@@ -272,7 +272,7 @@ suite "GossipSub":
       var handler: TopicHandler
       closureScope:
         var peerName = $dialer.peerInfo.peerId
-        handler = proc(topic: string, data: seq[byte]) {.async, gcsafe, closure.} =
+        handler = proc(topic: string, data: seq[byte]) {.async, closure.} =
           if peerName notin seen:
             seen[peerName] = 0
           seen[peerName].inc
@@ -324,7 +324,7 @@ suite "GossipSub":
 
     # Adding again subscriptions
 
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       check topic == "foobar"
 
     for i in 0..<runs:
@@ -368,7 +368,7 @@ suite "GossipSub":
       )
 
     var handlerFut = newFuture[void]()
-    proc handler(topic: string, data: seq[byte]) {.async, gcsafe.} =
+    proc handler(topic: string, data: seq[byte]) {.async.} =
       handlerFut.complete()
 
     await subscribeNodes(nodes)

tests/pubsub/testmcache.nim

@@ -1,12 +1,11 @@
 {.used.}
 
-import unittest2, options, sets, sequtils
+import unittest2, sequtils
 import stew/byteutils
 import ../../libp2p/[peerid,
                      crypto/crypto,
                      protocols/pubsub/mcache,
-                     protocols/pubsub/rpc/messages]
-import ./utils
+                     protocols/pubsub/rpc/message]
 
 var rng = newRng()
 
@@ -27,48 +26,48 @@ suite "MCache":
     var mCache = MCache.init(3, 5)
 
     for i in 0..<3:
-      var msg = Message(fromPeer: randomPeerId(),
-                        seqno: "12345".toBytes(),
-                        topicIDs: @["foo"])
+      var
+        msg =
+          Message(fromPeer: randomPeerId(), seqno: "12345".toBytes(), topic: "foo")
       mCache.put(defaultMsgIdProvider(msg).expect(MsgIdGenSuccess), msg)
 
     for i in 0..<5:
-      var msg = Message(fromPeer: randomPeerId(),
-                        seqno: "12345".toBytes(),
-                        topicIDs: @["bar"])
+      var
+        msg =
+          Message(fromPeer: randomPeerId(), seqno: "12345".toBytes(), topic: "bar")
       mCache.put(defaultMsgIdProvider(msg).expect(MsgIdGenSuccess), msg)
 
     var mids = mCache.window("foo")
     check mids.len == 3
 
     var id = toSeq(mids)[0]
-    check mCache.get(id).get().topicIds[0] == "foo"
+    check mCache.get(id).get().topic == "foo"
 
   test "shift - shift 1 window at a time":
     var mCache = MCache.init(1, 5)
 
     for i in 0..<3:
-      var msg = Message(fromPeer: randomPeerId(),
-                        seqno: "12345".toBytes(),
-                        topicIDs: @["foo"])
+      var
+        msg =
+          Message(fromPeer: randomPeerId(), seqno: "12345".toBytes(), topic: "foo")
       mCache.put(defaultMsgIdProvider(msg).expect(MsgIdGenSuccess), msg)
 
     mCache.shift()
     check mCache.window("foo").len == 0
 
     for i in 0..<3:
-      var msg = Message(fromPeer: randomPeerId(),
-                        seqno: "12345".toBytes(),
-                        topicIDs: @["bar"])
+      var
+        msg =
+          Message(fromPeer: randomPeerId(), seqno: "12345".toBytes(), topic: "bar")
       mCache.put(defaultMsgIdProvider(msg).expect(MsgIdGenSuccess), msg)
 
     mCache.shift()
     check mCache.window("bar").len == 0
 
     for i in 0..<3:
-      var msg = Message(fromPeer: randomPeerId(),
-                        seqno: "12345".toBytes(),
-                        topicIDs: @["baz"])
+      var
+        msg =
+          Message(fromPeer: randomPeerId(), seqno: "12345".toBytes(), topic: "baz")
       mCache.put(defaultMsgIdProvider(msg).expect(MsgIdGenSuccess), msg)
 
     mCache.shift()
@@ -78,21 +77,21 @@ suite "MCache":
     var mCache = MCache.init(1, 5)
 
     for i in 0..<3:
-      var msg = Message(fromPeer: randomPeerId(),
-                        seqno: "12345".toBytes(),
-                        topicIDs: @["foo"])
+      var
+        msg =
+          Message(fromPeer: randomPeerId(), seqno: "12345".toBytes(), topic: "foo")
       mCache.put(defaultMsgIdProvider(msg).expect(MsgIdGenSuccess), msg)
 
     for i in 0..<3:
-      var msg = Message(fromPeer: randomPeerId(),
-                        seqno: "12345".toBytes(),
-                        topicIDs: @["bar"])
+      var
+        msg =
+          Message(fromPeer: randomPeerId(), seqno: "12345".toBytes(), topic: "bar")
       mCache.put(defaultMsgIdProvider(msg).expect(MsgIdGenSuccess), msg)
 
     for i in 0..<3:
-      var msg = Message(fromPeer: randomPeerId(),
-                        seqno: "12345".toBytes(),
-                        topicIDs: @["baz"])
+      var
+        msg =
+          Message(fromPeer: randomPeerId(), seqno: "12345".toBytes(), topic: "baz")
       mCache.put(defaultMsgIdProvider(msg).expect(MsgIdGenSuccess), msg)
 
     mCache.shift()

tests/pubsub/testmessage.nim

@@ -75,14 +75,17 @@ suite "Message":
       msgIdResult.error == ValidationResult.Reject
 
   test "byteSize for RPCMsg":
-    var msg = Message(
-      fromPeer: PeerId(data: @['a'.byte, 'b'.byte]),  # 2 bytes
-      data: @[1'u8, 2, 3],  # 3 bytes
-      seqno: @[4'u8, 5],  # 2 bytes
-      signature: @['c'.byte, 'd'.byte],  # 2 bytes
-      key: @[6'u8, 7],  # 2 bytes
-      topicIds: @["abc", "defgh"]  # 3 + 5 = 8 bytes
-    )
+    var
+      msg =
+        Message(
+          fromPeer: PeerId(data: @['a'.byte, 'b'.byte]), # 2 bytes
+          data: @[1'u8, 2, 3], # 3 bytes
+          seqno: @[4'u8, 5], # 2 bytes
+          signature: @['c'.byte, 'd'.byte], # 2 bytes
+          key: @[6'u8, 7], # 2 bytes
+          topic: "abcde" # 5 bytes
+          ,
+        )
 
     var peerInfo = PeerInfoMsg(
       peerId: PeerId(data: @['e'.byte]),  # 1 byte
@@ -90,20 +93,20 @@ suite "Message":
     )
 
     var controlIHave = ControlIHave(
-      topicId: "ijk",  # 3 bytes
-      messageIds: @[ @['l'.byte], @['m'.byte, 'n'.byte] ]  # 1 + 2 = 3 bytes
+      topicID: "ijk",  # 3 bytes
+      messageIDs: @[ @['l'.byte], @['m'.byte, 'n'.byte] ]  # 1 + 2 = 3 bytes
     )
 
     var controlIWant = ControlIWant(
-      messageIds: @[ @['o'.byte, 'p'.byte], @['q'.byte] ]  # 2 + 1 = 3 bytes
+      messageIDs: @[ @['o'.byte, 'p'.byte], @['q'.byte] ]  # 2 + 1 = 3 bytes
     )
 
     var controlGraft = ControlGraft(
-      topicId: "rst"  # 3 bytes
+      topicID: "rst"  # 3 bytes
     )
 
     var controlPrune = ControlPrune(
-      topicId: "uvw",  # 3 bytes
+      topicID: "uvw",  # 3 bytes
       peers: @[peerInfo, peerInfo],  # (1 + 2) * 2 = 6 bytes
       backoff: 12345678  # 8 bytes for uint64
     )
@@ -118,10 +121,10 @@ suite "Message":
 
     var rpcMsg = RPCMsg(
       subscriptions: @[SubOpts(subscribe: true, topic: "a".repeat(12)), SubOpts(subscribe: false, topic: "b".repeat(14))],  # 1 + 12 + 1 + 14 = 28 bytes
-      messages: @[msg, msg],  # 19 * 2 = 38 bytes
+      messages: @[msg, msg],  # 16 * 2 = 32 bytes
       ping: @[1'u8, 2],  # 2 bytes
       pong: @[3'u8, 4],  # 2 bytes
       control: some(control)  # 12 + 3 + 3 + 17 + 3 = 38 bytes
     )
 
-    check byteSize(rpcMsg) == 28 + 38 + 2 + 2 + 38  # Total: 108 bytes
+    check byteSize(rpcMsg) == 28 + 32 + 2 + 2 + 38 # Total: 102 bytes

tests/pubsub/testtimedcache.nim

@@ -24,6 +24,8 @@ suite "TimedCache":
       2 in cache
       3 in cache
 
+      cache.addedAt(2) == now + 3.seconds
+
     check:
       cache.put(2, now + 7.seconds) # refreshes 2
       not cache.put(4, now + 12.seconds) # expires 3
@@ -33,6 +35,23 @@ suite "TimedCache":
       3 notin cache
       4 in cache
 
+    check:
+      cache.del(4).isSome()
+      4 notin cache
+
     check:
       not cache.put(100, now + 100.seconds) # expires everything
       100 in cache
+      2 notin cache
+
+  test "enough items to force cache heap storage growth":
+    var cache = TimedCache[int].init(5.seconds)
+
+    let now = Moment.now()
+    for i in 101..100000:
+      check:
+        not cache.put(i, now)
+
+    for i in 101..100000:
+      check:
+        i in cache