Don't prevent module_api from cancelling background processes on shutdown

### Background

As part of Element's plan to support a light form of vhosting (virtual
host) (multiple instances of Synapse in the same Python process), we're
currently diving into the details and implications of running multiple
instances of Synapse in the same Python process.

"Per-tenant logging" tracked internally by
https://github.com/element-hq/synapse-small-hosts/issues/48

### Prior art

Previously, we exposed `server_name` by providing a static logging
`MetadataFilter` that injected the values:


205d9e4fc4/synapse/config/logger.py (L216)

While this can work fine for the normal case of one Synapse instance per
Python process, this configures things globally and isn't compatible
when we try to start multiple Synapse instances because each subsequent
tenant will overwrite the previous tenant.


### What does this PR do?

We remove the `MetadataFilter` and replace it by tracking the
`server_name` in the `LoggingContext` and expose it with our existing
[`LoggingContextFilter`](205d9e4fc4/synapse/logging/context.py (L584-L622))
that we already use to expose information about the `request`.

This means that the `server_name` value follows wherever we log as
expected even when we have multiple Synapse instances running in the
same process.


### A note on logcontext

Anywhere, Synapse mistakenly uses the `sentinel` logcontext to log
something, we won't know which server sent the log. We've been fixing up
`sentinel` logcontext usage as tracked by
https://github.com/element-hq/synapse/issues/18905

Any further `sentinel` logcontext usage we find in the future can be
fixed piecemeal as normal.


d2a966f922/docs/log_contexts.md (L71-L81)


### Testing strategy

1. Adjust your logging config to include `%(server_name)s` in the format
    ```yaml
    formatters:
        precise:
format: '%(asctime)s - %(server_name)s - %(name)s - %(lineno)d -
%(levelname)s - %(request)s - %(message)s'
    ```
1. Start Synapse: `poetry run synapse_homeserver --config-path
homeserver.yaml`
1. Make some requests (`curl
http://localhost:8008/_matrix/client/versions`, etc)
1. Open the homeserver logs and notice the `server_name` in the logs as
expected. `unknown_server_from_sentinel_context` is expected for the
`sentinel` logcontext (things outside of Synapse).

.ci/scripts/triage_labelled_issue.sh

@@ -0,0 +1,29 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# 1) Resolve project ID.
+PROJECT_ID=$(gh project view "$PROJECT_NUMBER" --owner "$PROJECT_OWNER" --format json | jq -r '.id')
+
+# 2) Find existing item (project card) for this issue.
+ITEM_ID=$(
+  gh project item-list "$PROJECT_NUMBER" --owner "$PROJECT_OWNER" --format json \
+  | jq -r --arg url "$ISSUE_URL" '.items[] | select(.content.url==$url) | .id' | head -n1
+)
+
+# 3) If one doesn't exist, add this issue to the project.
+if [ -z "${ITEM_ID:-}" ]; then
+  ITEM_ID=$(gh project item-add "$PROJECT_NUMBER" --owner "$PROJECT_OWNER" --url "$ISSUE_URL" --format json | jq -r '.id')
+fi
+
+# 4) Get Status field id + the option id for TARGET_STATUS.
+FIELDS_JSON=$(gh project field-list "$PROJECT_NUMBER" --owner "$PROJECT_OWNER" --format json)
+STATUS_FIELD=$(echo "$FIELDS_JSON" | jq -r '.fields[] | select(.name=="Status")')
+STATUS_FIELD_ID=$(echo "$STATUS_FIELD" | jq -r '.id')
+OPTION_ID=$(echo "$STATUS_FIELD" | jq -r --arg name "$TARGET_STATUS" '.options[] | select(.name==$name) | .id')
+
+if [ -z "${OPTION_ID:-}" ]; then
+  echo "No Status option named \"$TARGET_STATUS\" found"; exit 1
+fi
+
+# 5) Set Status (moves item to the matching column in the board view).
+gh project item-edit --id "$ITEM_ID" --project-id "$PROJECT_ID" --field-id "$STATUS_FIELD_ID" --single-select-option-id "$OPTION_ID"

.github/workflows/fix_lint.yaml

@@ -25,7 +25,7 @@ jobs:
         with:
           toolchain: ${{ env.RUST_VERSION }}
           components: clippy, rustfmt
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - name: Setup Poetry
         uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0

.github/workflows/latest_deps.yml

@@ -47,7 +47,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       # The dev dependencies aren't exposed in the wheel metadata (at least with current
       # poetry-core versions), so we install with poetry.
@@ -83,7 +83,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - run: sudo apt-get -qq install xmlsec1
       - name: Set up PostgreSQL ${{ matrix.postgres-version }}
@@ -158,7 +158,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - name: Ensure sytest runs `pip install`
         # Delete the lockfile so sytest will `pip install` rather than `poetry install`

.github/workflows/tests.yml

@@ -91,7 +91,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
       - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
         with:
           python-version: "3.x"
@@ -157,7 +157,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - name: Setup Poetry
         uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
@@ -220,7 +220,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
       - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
         with:
           poetry-version: "2.1.1"
@@ -240,7 +240,7 @@ jobs:
         with:
             components: clippy
             toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - run: cargo clippy -- -D warnings
 
@@ -259,7 +259,7 @@ jobs:
         with:
             toolchain: nightly-2025-04-23
             components: clippy
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - run: cargo clippy --all-features -- -D warnings
 
@@ -276,7 +276,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - name: Setup Poetry
         uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
@@ -315,7 +315,7 @@ jobs:
           # `.rustfmt.toml`.
           toolchain: nightly-2025-04-23
           components: rustfmt
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - run: cargo fmt --check
 
@@ -415,7 +415,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
         with:
@@ -459,7 +459,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       # There aren't wheels for some of the older deps, so we need to install
       # their build dependencies
@@ -576,7 +576,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - name: Run SyTest
         run: /bootstrap.sh synapse
@@ -722,7 +722,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - name: Prepare Complement's Prerequisites
         run: synapse/.ci/scripts/setup_complement_prerequisites.sh
@@ -756,7 +756,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - run: cargo test
 
@@ -776,7 +776,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
             toolchain: nightly-2022-12-01
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - run: cargo bench --no-run
 

.github/workflows/triage_labelled.yml

@@ -6,43 +6,26 @@ on:
 
 jobs:
   move_needs_info:
-    name: Move X-Needs-Info on the triage board
     runs-on: ubuntu-latest
     if: >
       contains(github.event.issue.labels.*.name, 'X-Needs-Info')
+    permissions:
+      contents: read
+    env:
+      # This token must have the following scopes: ["repo:public_repo", "admin:org->read:org", "user->read:user", "project"]
+      GITHUB_TOKEN: ${{ secrets.ELEMENT_BOT_TOKEN }}
+      PROJECT_OWNER: matrix-org
+      # Backend issue triage board.
+      # https://github.com/orgs/matrix-org/projects/67/views/1
+      PROJECT_NUMBER: 67
+      ISSUE_URL: ${{ github.event.issue.html_url }}
+      # This field is case-sensitive.
+      TARGET_STATUS: Needs info
     steps:
-      - uses: actions/add-to-project@4515659e2b458b27365e167605ac44f219494b66 # v1.0.2
-        id: add_project
+      - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
         with:
-          project-url: "https://github.com/orgs/matrix-org/projects/67"
-          github-token: ${{ secrets.ELEMENT_BOT_TOKEN }}
-        # This action will error if the issue already exists on the project. Which is
-        # common as `X-Needs-Info` will often be added to issues that are already in
-        # the triage queue. Prevent the whole job from failing in this case.
-        continue-on-error: true
-      - name: Set status
-        env:
-          GITHUB_TOKEN: ${{ secrets.ELEMENT_BOT_TOKEN }}
-        run: |
-          gh api graphql -f query='
-          mutation(
-              $project: ID!
-              $item: ID!
-              $fieldid: ID!
-              $columnid: String!
-            )  {
-            updateProjectV2ItemFieldValue(
-              input: {
-               projectId: $project
-                itemId: $item
-                fieldId: $fieldid
-                value: { 
-                  singleSelectOptionId: $columnid
-                  }
-              }
-            ) {
-              projectV2Item {
-                id
-              }
-            }
-          }' -f project="PVT_kwDOAIB0Bs4AFDdZ" -f item=${{ steps.add_project.outputs.itemId }} -f fieldid="PVTSSF_lADOAIB0Bs4AFDdZzgC6ZA4" -f columnid=ba22e43c --silent
+          # Only clone the script file we care about, instead of the whole repo.
+          sparse-checkout: .ci/scripts/triage_labelled_issue.sh
+
+      - name: Ensure issue exists on the board, then set Status
+        run: .ci/scripts/triage_labelled_issue.sh

.github/workflows/twisted_trunk.yml

@@ -49,7 +49,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
         with:
@@ -77,7 +77,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
         with:
@@ -123,7 +123,7 @@ jobs:
         uses: dtolnay/rust-toolchain@e97e2d8cc328f1b50210efc529dca0028893a2d9 # master
         with:
           toolchain: ${{ env.RUST_VERSION }}
-      - uses: Swatinem/rust-cache@98c8021b550208e191a6a3145459bfc9fb29c4c0 # v2.8.0
+      - uses: Swatinem/rust-cache@f13886b937689c021905a6b90929199931d60db1 # v2.8.1
 
       - name: Patch dependencies
         # Note: The poetry commands want to create a virtualenv in /src/.venv/,

CHANGES.md

@@ -11,8 +11,7 @@
 
 ## Internal Changes
 
-- Drop support for Ubuntu 24.10 Oracular Oriole, and add support for Ubuntu 25.04 Plucky Puffin. ([\#18962](https://github.com/element-hq/synapse/issues/18962))
-
+- Drop support for Ubuntu 24.10 Oracular Oriole, and add support for Ubuntu 25.04 Plucky Puffin. This change was applied on top of 1.139.0rc1. ([\#18962](https://github.com/element-hq/synapse/issues/18962))
 
 
 
@@ -83,6 +82,23 @@
 * Bump types-requests from 2.32.4.20250611 to 2.32.4.20250809. ([\#18895](https://github.com/element-hq/synapse/issues/18895))
 * Bump types-setuptools from 80.9.0.20250809 to 80.9.0.20250822. ([\#18924](https://github.com/element-hq/synapse/issues/18924))
 
+# Synapse 1.138.2 (2025-09-24)
+
+## Internal Changes
+
+- Drop support for Ubuntu 24.10 Oracular Oriole, and add support for Ubuntu 25.04 Plucky Puffin. This change was applied on top of 1.138.1. ([\#18962](https://github.com/element-hq/synapse/issues/18962))
+
+
+
+# Synapse 1.138.1 (2025-09-24)
+
+## Bugfixes
+
+- Fix a performance regression related to the experimental Delayed Events ([MSC4140](https://github.com/matrix-org/matrix-spec-proposals/pull/4140)) feature. ([\#18926](https://github.com/element-hq/synapse/issues/18926))
+
+
+
+
 # Synapse 1.138.0 (2025-09-09)
 
 No significant changes since 1.138.0rc1.

Cargo.lock

@@ -28,9 +28,9 @@ dependencies = [
 
 [[package]]
 name = "anyhow"
-version = "1.0.99"
+version = "1.0.100"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b0674a1ddeecb70197781e945de4b3b8ffb61fa939a5597bcf48503737663100"
+checksum = "a23eb6b1614318a8071c9b2521f36b424b2c83db5eb3a0fead4a6c0809af6e61"
 
 [[package]]
 name = "arc-swap"
@@ -1250,9 +1250,9 @@ dependencies = [
 
 [[package]]
 name = "serde"
-version = "1.0.224"
+version = "1.0.226"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6aaeb1e94f53b16384af593c71e20b095e958dab1d26939c1b70645c5cfbcc0b"
+checksum = "0dca6411025b24b60bfa7ec1fe1f8e710ac09782dca409ee8237ba74b51295fd"
 dependencies = [
  "serde_core",
  "serde_derive",
@@ -1260,18 +1260,18 @@ dependencies = [
 
 [[package]]
 name = "serde_core"
-version = "1.0.224"
+version = "1.0.226"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "32f39390fa6346e24defbcdd3d9544ba8a19985d0af74df8501fbfe9a64341ab"
+checksum = "ba2ba63999edb9dac981fb34b3e5c0d111a69b0924e253ed29d83f7c99e966a4"
 dependencies = [
  "serde_derive",
 ]
 
 [[package]]
 name = "serde_derive"
-version = "1.0.224"
+version = "1.0.226"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "87ff78ab5e8561c9a675bfc1785cb07ae721f0ee53329a595cefd8c04c2ac4e0"
+checksum = "8db53ae22f34573731bafa1db20f04027b2d25e02d8205921b569171699cdb33"
 dependencies = [
  "proc-macro2",
  "quote",

README.rst

@@ -265,6 +265,8 @@ This software is dual-licensed by New Vector Ltd (Element). It can be used eithe
 
 Unless required by applicable law or agreed to in writing, software distributed under the Licenses is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the Licenses for the specific language governing permissions and limitations under the Licenses.
 
+Please contact `licensing@element.io <mailto:licensing@element.io>`_ to purchase an Element commercial license for this software.
+
 
 .. |support| image:: https://img.shields.io/badge/matrix-community%20support-success
   :alt: (get community support in #synapse:matrix.org)

changelog.d/18721.bugfix

@@ -0,0 +1 @@
+Fix room upgrade `room_config` argument and documentation for `user_may_create_room` spam-checker callback.

changelog.d/18767.misc

@@ -0,0 +1 @@
+Update OEmbed providers to use 'X' instead of 'Twitter' in URL previews, following a rebrand. Contributed by @HammyHavoc.

changelog.d/18828.feature

@@ -0,0 +1 @@
+Cleanly shutdown `SynapseHomeServer` object.

changelog.d/18868.misc

@@ -0,0 +1 @@
+Fix `server_name` in logging context for multiple Synapse instances in one process.

changelog.d/18911.feature

@@ -0,0 +1,2 @@
+Add an Admin API that allows server admins to to query and investigate the metadata of local or cached remote media via
+the `origin/media_id` identifier found in a [Matrix Content URI](https://spec.matrix.org/v1.14/client-server-api/#matrix-content-mxc-uris).

changelog.d/18913.misc

@@ -0,0 +1 @@
+Fix the GitHub Actions workflow that moves issues labeled "X-Needs-Info" to the "Needs info" column on the team's internal triage board.

changelog.d/18914.doc

@@ -0,0 +1 @@
+Explain how Deferred callbacks interact with logcontexts.

changelog.d/18932.misc

@@ -0,0 +1 @@
+Disconnect background process work from request trace.

changelog.d/18934.feature

@@ -0,0 +1 @@
+Update [MSC4284: Policy Servers](https://github.com/matrix-org/matrix-spec-proposals/pull/4284) implementation to support signatures when available.

changelog.d/18939.misc

@@ -0,0 +1 @@
+Reduce overall number of calls to `_get_e2e_cross_signing_signatures_for_devices` by increasing the batch size of devices the query is called with, reducing DB load.

changelog.d/18947.misc

@@ -0,0 +1 @@
+Update error code used when an appservice tries to masquerade as an unknown device using [MSC4326](https://github.com/matrix-org/matrix-spec-proposals/pull/4326). Contributed by @tulir @ Beeper.

changelog.d/18948.bugfix

@@ -0,0 +1 @@
+Compute a user's last seen timestamp from their devices' last seen timestamps instead of IPs, because the latter are automatically cleared according to `user_ips_max_age`.

changelog.d/18959.misc

@@ -0,0 +1 @@
+Fix `no active span when trying to log` tracing error on startup (when OpenTracing is enabled).

changelog.d/18964.misc

@@ -0,0 +1 @@
+Fix `run_coroutine_in_background(...)` incorrectly handling logcontext.

changelog.d/18971.misc

@@ -0,0 +1 @@
+Update dockerfile metadata to fix broken link; point to documentation website.

changelog.d/18973.misc

@@ -0,0 +1 @@
+Note that the code is additionally licensed under the [Element Commercial license](https://github.com/element-hq/synapse/blob/develop/LICENSE-COMMERCIAL) in SPDX expression field configs.

changelog.d/18974.misc

@@ -0,0 +1 @@
+Fix logcontext handling in `timeout_deferred` tests.

debian/changelog

@@ -1,3 +1,11 @@
+matrix-synapse-py3 (1.139.0~rc3+nmu1) UNRELEASED; urgency=medium
+
+  * The licensing specifier has been updated to add an optional
+    `LicenseRef-Element-Commercial` license. The code was already licensed in
+    this manner - the debian metadata was just not updated to reflect it.
+
+ -- Synapse Packaging team <packages@matrix.org>  Thu, 25 Sep 2025 12:17:17 +0100
+
 matrix-synapse-py3 (1.139.0~rc3) stable; urgency=medium
 
   * New Synapse release 1.139.0rc3.
@@ -16,6 +24,18 @@ matrix-synapse-py3 (1.139.0~rc1) stable; urgency=medium
 
  -- Synapse Packaging team <packages@matrix.org>  Tue, 23 Sep 2025 13:24:50 +0100
 
+matrix-synapse-py3 (1.138.2) stable; urgency=medium
+
+  * New Synapse release 1.138.2.
+
+ -- Synapse Packaging team <packages@matrix.org>  Wed, 24 Sep 2025 12:26:16 +0100
+
+matrix-synapse-py3 (1.138.1) stable; urgency=medium
+
+  * New Synapse release 1.138.1.
+
+ -- Synapse Packaging team <packages@matrix.org>  Wed, 24 Sep 2025 11:32:38 +0100
+
 matrix-synapse-py3 (1.138.0) stable; urgency=medium
 
   * New Synapse release 1.138.0.

debian/copyright

@@ -8,7 +8,7 @@ License: Apache-2.0
 
 Files: *
 Copyright: 2023 New Vector Ltd
-License: AGPL-3.0-or-later
+License: AGPL-3.0-or-later or LicenseRef-Element-Commercial
 
 Files: synapse/config/saml2.py
 Copyright: 2015, Ericsson

docker/Dockerfile

@@ -171,10 +171,10 @@ FROM docker.io/library/python:${PYTHON_VERSION}-slim-${DEBIAN_VERSION}
 
 ARG TARGETARCH
 
-LABEL org.opencontainers.image.url='https://matrix.org/docs/projects/server/synapse'
-LABEL org.opencontainers.image.documentation='https://github.com/element-hq/synapse/blob/master/docker/README.md'
+LABEL org.opencontainers.image.url='https://github.com/element-hq/synapse'
+LABEL org.opencontainers.image.documentation='https://element-hq.github.io/synapse/latest/'
 LABEL org.opencontainers.image.source='https://github.com/element-hq/synapse.git'
-LABEL org.opencontainers.image.licenses='AGPL-3.0-or-later'
+LABEL org.opencontainers.image.licenses='AGPL-3.0-or-later OR LicenseRef-Element-Commercial'
 
 # On the runtime image, /lib is a symlink to /usr/lib, so we need to copy the
 # libraries to the right place, else the `COPY` won't work.

docs/admin_api/media_admin_api.md

@@ -39,6 +39,40 @@ the use of the
 [List media uploaded by a user](user_admin_api.md#list-media-uploaded-by-a-user)
 Admin API.
 
+## Query a piece of media by ID
+
+This API returns information about a piece of local or cached remote media given the origin server name and media id. If
+information is requested for remote media which is not cached the endpoint will return 404. 
+
+Request:
+```http
+GET /_synapse/admin/v1/media/<origin>/<media_id>
+```
+
+The API returns a JSON body with media info like the following:
+
+Response:
+```json
+{
+  "media_info": {
+    "media_origin": "remote.com",
+    "user_id": null,
+    "media_id": "sdginwegWEG",
+    "media_type": "img/png",
+    "media_length": 67,
+    "upload_name":  "test.png",
+    "created_ts":  300,
+    "filesystem_id":  "wgeweg",
+    "url_cache": null,
+    "last_access_ts": 400,
+    "quarantined_by":  null,
+    "authenticated":  false,
+    "safe_from_quarantine": null,
+    "sha256": "ebf4f635a17d10d6eb46ba680b70142419aa3220f228001a036d311a22ee9d2a"
+  }
+}
+```
+
 # Quarantine media
 
 Quarantining media means that it is marked as inaccessible by users. It applies

docs/log_contexts.md

@@ -143,8 +143,7 @@ cares about.
 The following sections describe pitfalls and helpful patterns when
 implementing these rules.
 
-Always await your awaitables
-----------------------------
+## Always await your awaitables
 
 Whenever you get an awaitable back from a function, you should `await` on
 it as soon as possible. Do not pass go; do not do any logging; do not
@@ -203,6 +202,171 @@ async def sleep(seconds):
     return await context.make_deferred_yieldable(get_sleep_deferred(seconds))
 ```
 
+## Deferred callbacks
+
+When a deferred callback is called, it inherits the current logcontext. The deferred
+callback chain can resume a coroutine, which if following our logcontext rules, will
+restore its own logcontext, then run:
+
+ - until it yields control back to the reactor, setting the sentinel logcontext
+ - or until it finishes, restoring the logcontext it was started with (calling context)
+
+This behavior creates two specific issues:
+
+**Issue 1:** The first issue is that the callback may have reset the logcontext to the
+sentinel before returning. This means our calling function will continue with the
+sentinel logcontext instead of the logcontext it was started with (bad).
+
+**Issue 2:** The second issue is that the current logcontext that called the deferred
+callback could finish before the callback finishes (bad).
+
+In the following example, the deferred callback is called with the "main" logcontext and
+runs until we yield control back to the reactor in the `await` inside `clock.sleep(0)`.
+Since `clock.sleep(0)` follows our logcontext rules, it sets the logcontext to the
+sentinel before yielding control back to the reactor. Our `main` function continues with
+the sentinel logcontext (first bad thing) instead of the "main" logcontext. Then the
+`with LoggingContext("main")` block exits, finishing the "main" logcontext and yielding
+control back to the reactor again. Finally, later on when `clock.sleep(0)` completes,
+our `with LoggingContext("competing")` block exits, and restores the previous "main"
+logcontext which has already finished, resulting in `WARNING: Re-starting finished log
+context main` and leaking the `main` logcontext into the reactor which will then
+erronously be associated with the next task the reactor picks up.
+
+```python
+async def competing_callback():
+    # Since this is run with the "main" logcontext, when the "competing"
+    # logcontext exits, it will restore the previous "main" logcontext which has
+    # already finished and results in "WARNING: Re-starting finished log context main"
+    # and leaking the `main` logcontext into the reactor.
+    with LoggingContext("competing"):
+        await clock.sleep(0)
+
+def main():
+    with LoggingContext("main"):
+        d = defer.Deferred()
+        d.addCallback(lambda _: defer.ensureDeferred(competing_callback()))
+        # Call the callback within the "main" logcontext.
+        d.callback(None)
+        # Bad: This will be logged against sentinel logcontext
+        logger.debug("ugh")
+
+main()
+```
+
+**Solution 1:** We could of course fix this by following the general rule of "always
+await your awaitables":
+
+```python
+async def main():
+    with LoggingContext("main"):
+        d = defer.Deferred()
+        d.addCallback(lambda _: defer.ensureDeferred(competing_callback()))
+        d.callback(None)
+        # Wait for `d` to finish before continuing so the "main" logcontext is
+        # still active. This works because `d` already follows our logcontext
+        # rules. If not, we would also have to use `make_deferred_yieldable(d)`.
+        await d
+        # Good: This will be logged against the "main" logcontext
+        logger.debug("phew")
+``` 
+
+**Solution 2:** We could also fix this by surrounding the call to `d.callback` with a
+`PreserveLoggingContext`, which will reset the logcontext to the sentinel before calling
+the callback, and restore the "main" logcontext afterwards before continuing the `main`
+function. This solves the problem because when the "competing" logcontext exits, it will
+restore the sentinel logcontext which is never finished by its nature, so there is no
+warning and no leakage into the reactor.
+
+```python
+async def main():
+    with LoggingContext("main"):
+        d = defer.Deferred()
+        d.addCallback(lambda _: defer.ensureDeferred(competing_callback()))
+        d.callback(None)
+        with PreserveLoggingContext():
+            # Call the callback with the sentinel logcontext.
+            d.callback(None)
+        # Good: This will be logged against the "main" logcontext
+        logger.debug("phew")
+```
+
+**Solution 3:** But let's say you *do* want to run (fire-and-forget) the deferred
+callback in the current context without running into issues:
+
+We can solve the first issue by using `run_in_background(...)` to run the callback in
+the current logcontext and it handles the magic behind the scenes of a) restoring the
+calling logcontext before returning to the caller and b) resetting the logcontext to the
+sentinel after the deferred completes and we yield control back to the reactor to avoid
+leaking the logcontext into the reactor.
+
+To solve the second issue, we can extend the lifetime of the "main" logcontext by
+avoiding the `LoggingContext`'s context manager lifetime methods
+(`__enter__`/`__exit__`). We can still set "main" as the current logcontext by using
+`PreserveLoggingContext` and passing in the "main" logcontext.
+
+
+```python
+async def main():
+    main_context = LoggingContext("main")
+    with PreserveLoggingContext(main_context):
+        d = defer.Deferred()
+        d.addCallback(lambda _: defer.ensureDeferred(competing_callback()))
+        # The whole lambda will be run in the "main" logcontext. But we're using
+        # a trick to return the deferred `d` itself so that `run_in_background`
+        # will wait on that to complete and reset the logcontext to the sentinel
+        # when it does to avoid leaking the "main" logcontext into the reactor.
+        run_in_background(lambda: (d.callback(None), d)[1])
+        # Good: This will be logged against the "main" logcontext
+        logger.debug("phew")
+
+...
+
+# Wherever possible, it's best to finish the logcontext by calling `__exit__` at some
+# point. This allows us to catch bugs if we later try to erroneously restart a finished
+# logcontext.
+#
+# Since the "main" logcontext stores the `LoggingContext.previous_context` when it is
+# created, we can wrap this call in `PreserveLoggingContext()` to restore the correct
+# previous logcontext. Our goal is to have the calling context remain unchanged after
+# finishing the "main" logcontext.
+with PreserveLoggingContext():
+    # Finish the "main" logcontext
+    with main_context:
+        # Empty block - We're just trying to call `__exit__` on the "main" context
+        # manager to finish it. We can't call `__exit__` directly as the code expects us
+        # to `__enter__` before calling `__exit__` to `start`/`stop` things
+        # appropriately. And in any case, it's probably best not to call the internal
+        # methods directly.
+        pass
+```
+
+The same thing applies if you have some deferreds stored somewhere which you want to
+callback in the current logcontext.
+
+
+### Deferred errbacks and cancellations
+
+The same care should be taken when calling errbacks on deferreds. An errback and
+callback act the same in this regard (see section above).
+
+```python
+d = defer.Deferred()
+d.addErrback(some_other_function)
+d.errback(failure)
+```
+
+Additionally, cancellation is the same as directly calling the errback with a
+`twisted.internet.defer.CancelledError`:
+
+```python
+d = defer.Deferred()
+d.addErrback(some_other_function)
+d.cancel()
+```
+
+
+
+
 ## Fire-and-forget
 
 Sometimes you want to fire off a chain of execution, but not wait for

docs/modules/spam_checker_callbacks.md

@@ -195,12 +195,15 @@ _Changed in Synapse v1.132.0: Added the `room_config` argument. Callbacks that o
 async def user_may_create_room(user_id: str, room_config: synapse.module_api.JsonDict) -> Union["synapse.module_api.NOT_SPAM", "synapse.module_api.errors.Codes", bool]
 ```
 
-Called when processing a room creation request.
+Called when processing a room creation or room upgrade request.
 
 The arguments passed to this callback are:
 
 * `user_id`: The Matrix user ID of the user (e.g. `@alice:example.com`).
-* `room_config`: The contents of the body of a [/createRoom request](https://spec.matrix.org/latest/client-server-api/#post_matrixclientv3createroom) as a dictionary.
+* `room_config`: The contents of the body of the [`/createRoom` request](https://spec.matrix.org/v1.15/client-server-api/#post_matrixclientv3createroom) as a dictionary.
+  For a [room upgrade request](https://spec.matrix.org/v1.15/client-server-api/#post_matrixclientv3roomsroomidupgrade) it is a synthesised subset of what an equivalent
+  `/createRoom` request would have looked like. Specifically, it contains the `creation_content` (linking to the previous room) and `initial_state` (containing a
+  subset of the state of the previous room).
 
 The callback must return one of:
   - `synapse.module_api.NOT_SPAM`, to allow the operation. Other callbacks may still 

docs/upgrade.md

@@ -119,12 +119,6 @@ stacking them up. You can monitor the currently running background updates with
 
 # Upgrading to v1.139.0
 
-## Drop support for Ubuntu 24.10 Oracular Oriole, and add support for Ubuntu 25.04 Plucky Puffin
-
-Ubuntu 24.10 Oracular Oriole [has been end-of-life since 10 Jul
-2025](https://endoflife.date/ubuntu). This release drops support for Ubuntu
-24.10, and in its place adds support for Ubuntu 25.04 Plucky Puffin.
-
 ## `/register` requests from old application service implementations may break when using MAS
 
 Application Services that do not set `inhibit_login=true` when calling `POST
@@ -140,6 +134,16 @@ ensure it is up to date. If it is, then kindly let the author know that they
 need to update their implementation to call `/register` with
 `inhibit_login=true`.
 
+# Upgrading to v1.138.2
+
+## Drop support for Ubuntu 24.10 Oracular Oriole, and add support for Ubuntu 25.04 Plucky Puffin
+
+Ubuntu 24.10 Oracular Oriole [has been end-of-life since 10 Jul
+2025](https://endoflife.date/ubuntu). This release drops support for Ubuntu
+24.10, and in its place adds support for Ubuntu 25.04 Plucky Puffin.
+
+This notice also applies to the v1.139.0 release.
+
 # Upgrading to v1.136.0
 
 ## Deprecate `run_as_background_process` exported as part of the module API interface in favor of `ModuleApi.run_as_background_process`

poetry.lock

@@ -1,4 +1,4 @@
-# This file is automatically @generated by Poetry 2.1.1 and should not be changed by hand.
+# This file is automatically @generated by Poetry 2.2.0 and should not be changed by hand.
 
 [[package]]
 name = "annotated-types"
@@ -34,15 +34,15 @@ tests-mypy = ["mypy (>=1.11.1) ; platform_python_implementation == \"CPython\" a
 
 [[package]]
 name = "authlib"
-version = "1.6.3"
+version = "1.6.4"
 description = "The ultimate Python library in building OAuth and OpenID Connect servers and clients."
 optional = true
 python-versions = ">=3.9"
 groups = ["main"]
 markers = "extra == \"all\" or extra == \"jwt\" or extra == \"oidc\""
 files = [
-    {file = "authlib-1.6.3-py2.py3-none-any.whl", hash = "sha256:7ea0f082edd95a03b7b72edac65ec7f8f68d703017d7e37573aee4fc603f2a48"},
-    {file = "authlib-1.6.3.tar.gz", hash = "sha256:9f7a982cc395de719e4c2215c5707e7ea690ecf84f1ab126f28c053f4219e610"},
+    {file = "authlib-1.6.4-py2.py3-none-any.whl", hash = "sha256:39313d2a2caac3ecf6d8f95fbebdfd30ae6ea6ae6a6db794d976405fdd9aa796"},
+    {file = "authlib-1.6.4.tar.gz", hash = "sha256:104b0442a43061dc8bc23b133d1d06a2b0a9c2e3e33f34c4338929e816287649"},
 ]
 
 [package.dependencies]
@@ -1531,14 +1531,14 @@ files = [
 
 [[package]]
 name = "phonenumbers"
-version = "9.0.13"
+version = "9.0.14"
 description = "Python version of Google's common library for parsing, formatting, storing and validating international phone numbers."
 optional = false
 python-versions = "*"
 groups = ["main"]
 files = [
-    {file = "phonenumbers-9.0.13-py2.py3-none-any.whl", hash = "sha256:b97661e177773e7509c6d503e0f537cd0af22aa3746231654590876eb9430915"},
-    {file = "phonenumbers-9.0.13.tar.gz", hash = "sha256:eca06e01382412c45316868f86a44bb217c02f9ee7196589041556a2f54a7639"},
+    {file = "phonenumbers-9.0.14-py2.py3-none-any.whl", hash = "sha256:6bdf5c46dbfefa1d941d122432d1958418d1dfe3f8c8c81d4c8e80f5442ea41f"},
+    {file = "phonenumbers-9.0.14.tar.gz", hash = "sha256:98afb3e86bf9ae02cc7c98ca44fa8827babb72842f90da9884c5d998937572ae"},
 ]
 
 [[package]]
@@ -1908,14 +1908,14 @@ typing-extensions = ">=4.6.0,<4.7.0 || >4.7.0"
 
 [[package]]
 name = "pygithub"
-version = "2.7.0"
+version = "2.8.1"
 description = "Use the full Github API v3"
 optional = false
 python-versions = ">=3.8"
 groups = ["dev"]
 files = [
-    {file = "pygithub-2.7.0-py3-none-any.whl", hash = "sha256:40ecbfe26dc55cc34ab4b0ffa1d455e6f816ef9a2bc8d6f5ad18ce572f163700"},
-    {file = "pygithub-2.7.0.tar.gz", hash = "sha256:7cd6eafabb09b5369afba3586d86b1f1ad6f1326d2ff01bc47bb26615dce4cbb"},
+    {file = "pygithub-2.8.1-py3-none-any.whl", hash = "sha256:23a0a5bca93baef082e03411bf0ce27204c32be8bfa7abc92fe4a3e132936df0"},
+    {file = "pygithub-2.8.1.tar.gz", hash = "sha256:341b7c78521cb07324ff670afd1baa2bf5c286f8d9fd302c1798ba594a5400c9"},
 ]
 
 [package.dependencies]
@@ -3011,14 +3011,14 @@ files = [
 
 [[package]]
 name = "types-requests"
-version = "2.32.4.20250809"
+version = "2.32.4.20250913"
 description = "Typing stubs for requests"
 optional = false
 python-versions = ">=3.9"
 groups = ["dev"]
 files = [
-    {file = "types_requests-2.32.4.20250809-py3-none-any.whl", hash = "sha256:f73d1832fb519ece02c85b1f09d5f0dd3108938e7d47e7f94bbfa18a6782b163"},
-    {file = "types_requests-2.32.4.20250809.tar.gz", hash = "sha256:d8060de1c8ee599311f56ff58010fb4902f462a1470802cf9f6ed27bc46c4df3"},
+    {file = "types_requests-2.32.4.20250913-py3-none-any.whl", hash = "sha256:78c9c1fffebbe0fa487a418e0fa5252017e9c60d1a2da394077f1780f655d7e1"},
+    {file = "types_requests-2.32.4.20250913.tar.gz", hash = "sha256:abd6d4f9ce3a9383f269775a9835a4c24e5cd6b9f647d64f88aa4613c33def5d"},
 ]
 
 [package.dependencies]
@@ -3038,14 +3038,14 @@ files = [
 
 [[package]]
 name = "typing-extensions"
-version = "4.14.1"
+version = "4.15.0"
 description = "Backported and Experimental Type Hints for Python 3.9+"
 optional = false
 python-versions = ">=3.9"
 groups = ["main", "dev"]
 files = [
-    {file = "typing_extensions-4.14.1-py3-none-any.whl", hash = "sha256:d1e1e3b58374dc93031d6eda2420a48ea44a36c2b4766a4fdeb3710755731d76"},
-    {file = "typing_extensions-4.14.1.tar.gz", hash = "sha256:38b39f4aeeab64884ce9f74c94263ef78f3c22467c8724005483154c26648d36"},
+    {file = "typing_extensions-4.15.0-py3-none-any.whl", hash = "sha256:f0fa19c6845758ab08074a0cfa8b7aecb71c999ca73d62883bc25cc018c4e548"},
+    {file = "typing_extensions-4.15.0.tar.gz", hash = "sha256:0cea48d173cc12fa28ecabc3b837ea3cf6f38c6d1136f85cbaaf598984861466"},
 ]
 
 [[package]]

pyproject.toml

@@ -104,7 +104,7 @@ name = "matrix-synapse"
 version = "1.139.0rc3"
 description = "Homeserver for the Matrix decentralised comms protocol"
 authors = ["Matrix.org Team and Contributors <packages@matrix.org>"]
-license = "AGPL-3.0-or-later"
+license = "AGPL-3.0-or-later OR LicenseRef-Element-Commercial"
 readme = "README.rst"
 repository = "https://github.com/element-hq/synapse"
 packages = [

scripts-dev/mypy_synapse_plugin.py

@@ -68,18 +68,42 @@ PROMETHEUS_METRIC_MISSING_FROM_LIST_TO_CHECK = ErrorCode(
     category="per-homeserver-tenant-metrics",
 )
 
+PREFER_SYNAPSE_CLOCK_CALL_LATER = ErrorCode(
+    "call-later-not-tracked",
+    "Prefer using `synapse.util.Clock.call_later` instead of `reactor.callLater`",
+    category="synapse-reactor-clock",
+)
+
+PREFER_SYNAPSE_CLOCK_LOOPING_CALL = ErrorCode(
+    "prefer-synapse-clock-looping-call",
+    "Prefer using `synapse.util.Clock.looping_call` instead of `task.LoopingCall`",
+    category="synapse-reactor-clock",
+)
+
 PREFER_SYNAPSE_CLOCK_CALL_WHEN_RUNNING = ErrorCode(
     "prefer-synapse-clock-call-when-running",
-    "`synapse.util.Clock.call_when_running` should be used instead of `reactor.callWhenRunning`",
+    "Prefer using `synapse.util.Clock.call_when_running` instead of `reactor.callWhenRunning`",
     category="synapse-reactor-clock",
 )
 
 PREFER_SYNAPSE_CLOCK_ADD_SYSTEM_EVENT_TRIGGER = ErrorCode(
     "prefer-synapse-clock-add-system-event-trigger",
-    "`synapse.util.Clock.add_system_event_trigger` should be used instead of `reactor.addSystemEventTrigger`",
+    "Prefer using `synapse.util.Clock.add_system_event_trigger` instead of `reactor.addSystemEventTrigger`",
     category="synapse-reactor-clock",
 )
 
+MULTIPLE_INTERNAL_CLOCKS_CREATED = ErrorCode(
+    "multiple-internal-clocks",
+    "Only one instance of `clock.Clock` should be created",
+    category="synapse-reactor-clock",
+)
+
+UNTRACKED_BACKGROUND_PROCESS = ErrorCode(
+    "untracked-background-process",
+    "Prefer using `HomeServer.run_as_background_process` method over the bare `run_as_background_process`",
+    category="synapse-tracked-calls",
+)
+
 
 class Sentinel(enum.Enum):
     # defining a sentinel in this way allows mypy to correctly handle the
@@ -222,6 +246,18 @@ class SynapsePlugin(Plugin):
             # callback, let's just pass it in while we have it.
             return lambda ctx: check_prometheus_metric_instantiation(ctx, fullname)
 
+        if fullname == "twisted.internet.task.LoopingCall":
+            return check_looping_call
+
+        if fullname == "synapse.util.clock.Clock":
+            return check_clock_creation
+
+        if (
+            fullname
+            == "synapse.metrics.background_process_metrics.run_as_background_process"
+        ):
+            return check_background_process
+
         return None
 
     def get_method_signature_hook(
@@ -241,6 +277,13 @@ class SynapsePlugin(Plugin):
         ):
             return check_is_cacheable_wrapper
 
+        if fullname in (
+            "twisted.internet.interfaces.IReactorTime.callLater",
+            "synapse.types.ISynapseThreadlessReactor.callLater",
+            "synapse.types.ISynapseReactor.callLater",
+        ):
+            return check_call_later
+
         if fullname in (
             "twisted.internet.interfaces.IReactorCore.callWhenRunning",
             "synapse.types.ISynapseThreadlessReactor.callWhenRunning",
@@ -258,6 +301,78 @@ class SynapsePlugin(Plugin):
         return None
 
 
+def check_clock_creation(ctx: FunctionSigContext) -> CallableType:
+    """
+    Ensure that the only `clock.Clock` instance is the one used by the `HomeServer`.
+    This is so that the `HomeServer` can cancel any tracked delayed or looping calls
+    during server shutdown.
+
+    Args:
+        ctx: The `FunctionSigContext` from mypy.
+    """
+    signature: CallableType = ctx.default_signature
+    ctx.api.fail(
+        "Expected the only `clock.Clock` instance to be the one used by the `HomeServer`. "
+        "This is so that the `HomeServer` can cancel any tracked delayed or looping calls "
+        "during server shutdown",
+        ctx.context,
+        code=MULTIPLE_INTERNAL_CLOCKS_CREATED,
+    )
+
+    return signature
+
+
+def check_call_later(ctx: MethodSigContext) -> CallableType:
+    """
+    Ensure that the `reactor.callLater` callsites aren't used.
+
+    `synapse.util.Clock.call_later` should always be used instead of `reactor.callLater`.
+    This is because the `synapse.util.Clock` tracks delayed calls in order to cancel any
+    outstanding calls during server shutdown. Delayed calls which are either short lived
+    (<~60s) or frequently called and can be tracked via other means could be candidates for
+    using `synapse.util.Clock.call_later` with `call_later_cancel_on_shutdown` set to
+    `False`. There shouldn't be a need to use `reactor.callLater` outside of tests or the
+    `Clock` class itself. If a need arises, you can use a type ignore comment to disable the
+    check, e.g. `# type: ignore[call-later-not-tracked]`.
+
+    Args:
+        ctx: The `FunctionSigContext` from mypy.
+    """
+    signature: CallableType = ctx.default_signature
+    ctx.api.fail(
+        "Expected all `reactor.callLater` calls to use `synapse.util.Clock.call_later` "
+        "instead. This is so that long lived calls can be tracked for cancellation during "
+        "server shutdown",
+        ctx.context,
+        code=PREFER_SYNAPSE_CLOCK_CALL_LATER,
+    )
+
+    return signature
+
+
+def check_looping_call(ctx: FunctionSigContext) -> CallableType:
+    """
+    Ensure that the `task.LoopingCall` callsites aren't used.
+
+    `synapse.util.Clock.looping_call` should always be used instead of `task.LoopingCall`.
+    `synapse.util.Clock` tracks looping calls in order to cancel any outstanding calls
+    during server shutdown.
+
+    Args:
+        ctx: The `FunctionSigContext` from mypy.
+    """
+    signature: CallableType = ctx.default_signature
+    ctx.api.fail(
+        "Expected all `task.LoopingCall` instances to use `synapse.util.Clock.looping_call` "
+        "instead. This is so that long lived calls can be tracked for cancellation during "
+        "server shutdown",
+        ctx.context,
+        code=PREFER_SYNAPSE_CLOCK_LOOPING_CALL,
+    )
+
+    return signature
+
+
 def check_call_when_running(ctx: MethodSigContext) -> CallableType:
     """
     Ensure that the `reactor.callWhenRunning` callsites aren't used.
@@ -312,6 +427,27 @@ def check_add_system_event_trigger(ctx: MethodSigContext) -> CallableType:
     return signature
 
 
+def check_background_process(ctx: FunctionSigContext) -> CallableType:
+    """
+    Ensure that calls to `run_as_background_process` use the `HomeServer` method.
+    This is so that the `HomeServer` can cancel any running background processes during
+    server shutdown.
+
+    Args:
+        ctx: The `FunctionSigContext` from mypy.
+    """
+    signature: CallableType = ctx.default_signature
+    ctx.api.fail(
+        "Prefer using `HomeServer.run_as_background_process` method over the bare "
+        "`run_as_background_process`. This is so that the `HomeServer` can cancel "
+        "any background processes during server shutdown",
+        ctx.context,
+        code=UNTRACKED_BACKGROUND_PROCESS,
+    )
+
+    return signature
+
+
 def analyze_prometheus_metric_classes(ctx: ClassDefContext) -> None:
     """
     Cross-check the list of Prometheus metric classes against the

synapse/_scripts/generate_workers_map.py

@@ -157,7 +157,12 @@ def get_registered_paths_for_default(
     # TODO We only do this to avoid an error, but don't need the database etc
     hs.setup()
     registered_paths = get_registered_paths_for_hs(hs)
-    hs.cleanup()
+    # NOTE: a more robust implementation would properly shutdown/cleanup each server
+    # to avoid resource buildup.
+    # However, the call to `shutdown` is `async` so it would require additional complexity here.
+    # We are intentionally skipping this cleanup because this is a short-lived, one-off
+    # utility script where the simpler approach is sufficient and we shouldn't run into
+    # any resource buildup issues.
 
     return registered_paths
 

synapse/_scripts/update_synapse_database.py

@@ -28,7 +28,6 @@ import yaml
 from twisted.internet import defer, reactor as reactor_
 
 from synapse.config.homeserver import HomeServerConfig
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.server import HomeServer
 from synapse.storage import DataStore
 from synapse.types import ISynapseReactor
@@ -53,7 +52,6 @@ class MockHomeserver(HomeServer):
 
 
 def run_background_updates(hs: HomeServer) -> None:
-    server_name = hs.hostname
     main = hs.get_datastores().main
     state = hs.get_datastores().state
 
@@ -67,9 +65,8 @@ def run_background_updates(hs: HomeServer) -> None:
     def run() -> None:
         # Apply all background updates on the database.
         defer.ensureDeferred(
-            run_as_background_process(
+            hs.run_as_background_process(
                 "background_updates",
-                server_name,
                 run_background_updates,
             )
         )

synapse/api/auth/base.py

@@ -354,12 +354,10 @@ class BaseAuth:
                 effective_user_id, effective_device_id
             )
             if device_opt is None:
-                # For now, use 400 M_EXCLUSIVE if the device doesn't exist.
-                # This is an open thread of discussion on MSC3202 as of 2021-12-09.
                 raise AuthError(
                     400,
                     f"Application service trying to use a device that doesn't exist ('{effective_device_id}' for {effective_user_id})",
-                    Codes.EXCLUSIVE,
+                    Codes.UNKNOWN_DEVICE,
                 )
 
         return create_requester(

synapse/api/errors.py

@@ -149,6 +149,9 @@ class Codes(str, Enum):
     )
     MSC4306_NOT_IN_THREAD = "IO.ELEMENT.MSC4306.M_NOT_IN_THREAD"
 
+    # Part of MSC4326
+    UNKNOWN_DEVICE = "ORG.MATRIX.MSC4326.M_UNKNOWN_DEVICE"
+
 
 class CodeMessageException(RuntimeError):
     """An exception with integer code, a message string attributes and optional headers.

synapse/app/_base.py

@@ -28,6 +28,7 @@ import sys
 import traceback
 import warnings
 from textwrap import indent
+from threading import Thread
 from typing import (
     TYPE_CHECKING,
     Any,
@@ -40,6 +41,7 @@ from typing import (
     Tuple,
     cast,
 )
+from wsgiref.simple_server import WSGIServer
 
 from cryptography.utils import CryptographyDeprecationWarning
 from typing_extensions import ParamSpec
@@ -73,7 +75,6 @@ from synapse.events.presence_router import load_legacy_presence_router
 from synapse.handlers.auth import load_legacy_password_auth_providers
 from synapse.http.site import SynapseSite
 from synapse.logging.context import LoggingContext, PreserveLoggingContext
-from synapse.logging.opentracing import init_tracer
 from synapse.metrics import install_gc_manager, register_threadpool
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.metrics.jemalloc import setup_jemalloc_stats
@@ -98,22 +99,47 @@ reactor = cast(ISynapseReactor, _reactor)
 
 logger = logging.getLogger(__name__)
 
-# list of tuples of function, args list, kwargs dict
-_sighup_callbacks: List[
-    Tuple[Callable[..., None], Tuple[object, ...], Dict[str, object]]
-] = []
+_instance_id_to_sighup_callbacks_map: Dict[
+    str, List[Tuple[Callable[..., None], Tuple[object, ...], Dict[str, object]]]
+] = {}
+"""
+Map from homeserver instance_id to a list of callbacks.
+
+We use `instance_id` instead of `server_name` because it's possible to have multiple
+workers running in the same process with the same `server_name`.
+"""
 P = ParamSpec("P")
 
 
-def register_sighup(func: Callable[P, None], *args: P.args, **kwargs: P.kwargs) -> None:
+def register_sighup(
+    homeserver_instance_id: str,
+    func: Callable[P, None],
+    *args: P.args,
+    **kwargs: P.kwargs,
+) -> None:
     """
     Register a function to be called when a SIGHUP occurs.
 
     Args:
+        homeserver_instance_id: The unique ID for this Synapse process instance
+            (`hs.get_instance_id()`) that this hook is associated with.
         func: Function to be called when sent a SIGHUP signal.
         *args, **kwargs: args and kwargs to be passed to the target function.
     """
-    _sighup_callbacks.append((func, args, kwargs))
+
+    _instance_id_to_sighup_callbacks_map.setdefault(homeserver_instance_id, []).append(
+        (func, args, kwargs)
+    )
+
+
+def unregister_sighups(instance_id: str) -> None:
+    """
+    Unregister all sighup functions associated with this Synapse instance.
+
+    Args:
+        instance_id: Unique ID for this Synapse process instance.
+    """
+    _instance_id_to_sighup_callbacks_map.pop(instance_id, [])
 
 
 def start_worker_reactor(
@@ -282,7 +308,9 @@ def register_start(
     clock.call_when_running(lambda: defer.ensureDeferred(wrapper()))
 
 
-def listen_metrics(bind_addresses: StrCollection, port: int) -> None:
+def listen_metrics(
+    bind_addresses: StrCollection, port: int
+) -> List[Tuple[WSGIServer, Thread]]:
     """
     Start Prometheus metrics server.
 
@@ -295,14 +323,22 @@ def listen_metrics(bind_addresses: StrCollection, port: int) -> None:
     bytecode at a time), this still works because the metrics thread can preempt the
     Twisted reactor thread between bytecode boundaries and the metrics thread gets
     scheduled with roughly equal priority to the Twisted reactor thread.
+
+    Returns:
+        List of WSGIServer with the thread they are running on.
     """
     from prometheus_client import start_http_server as start_http_server_prometheus
 
     from synapse.metrics import RegistryProxy
 
+    servers: List[Tuple[WSGIServer, Thread]] = []
     for host in bind_addresses:
         logger.info("Starting metrics listener on %s:%d", host, port)
-        start_http_server_prometheus(port, addr=host, registry=RegistryProxy)
+        server, thread = start_http_server_prometheus(
+            port, addr=host, registry=RegistryProxy
+        )
+        servers.append((server, thread))
+    return servers
 
 
 def listen_manhole(
@@ -310,7 +346,7 @@ def listen_manhole(
     port: int,
     manhole_settings: ManholeConfig,
     manhole_globals: dict,
-) -> None:
+) -> List[Port]:
     # twisted.conch.manhole 21.1.0 uses "int_from_bytes", which produces a confusing
     # warning. It's fixed by https://github.com/twisted/twisted/pull/1522), so
     # suppress the warning for now.
@@ -322,7 +358,7 @@ def listen_manhole(
 
     from synapse.util.manhole import manhole
 
-    listen_tcp(
+    return listen_tcp(
         bind_addresses,
         port,
         manhole(settings=manhole_settings, globals=manhole_globals),
@@ -499,7 +535,7 @@ def refresh_certificate(hs: "HomeServer") -> None:
         logger.info("Context factories updated.")
 
 
-async def start(hs: "HomeServer") -> None:
+async def start(hs: "HomeServer", freeze: bool = True) -> None:
     """
     Start a Synapse server or worker.
 
@@ -510,6 +546,11 @@ async def start(hs: "HomeServer") -> None:
 
     Args:
         hs: homeserver instance
+        freeze: whether to freeze the homeserver base objects in the garbage collector.
+            May improve garbage collection performance by marking objects with an effectively
+            static lifetime as frozen so they don't need to be considered for cleanup.
+            If you ever want to `shutdown` the homeserver, this needs to be
+            False otherwise the homeserver cannot be garbage collected after `shutdown`.
     """
     server_name = hs.hostname
     reactor = hs.get_reactor()
@@ -542,12 +583,17 @@ async def start(hs: "HomeServer") -> None:
                 # we're not using systemd.
                 sdnotify(b"RELOADING=1")
 
-                for i, args, kwargs in _sighup_callbacks:
-                    i(*args, **kwargs)
+            for sighup_callbacks in _instance_id_to_sighup_callbacks_map.values():
+                for func, args, kwargs in sighup_callbacks:
+                    func(*args, **kwargs)
 
                 sdnotify(b"READY=1")
 
-            return run_as_background_process(
+            # It's okay to ignore the linter error here and call
+            # `run_as_background_process` directly because `_handle_sighup` operates
+            # outside of the scope of a specific `HomeServer` instance and holds no
+            # references to it which would prevent a clean shutdown.
+            return run_as_background_process(  # type: ignore[untracked-background-process]
                 "sighup",
                 server_name,
                 _handle_sighup,
@@ -565,8 +611,8 @@ async def start(hs: "HomeServer") -> None:
 
         signal.signal(signal.SIGHUP, run_sighup)
 
-        register_sighup(refresh_certificate, hs)
-        register_sighup(reload_cache_config, hs.config)
+        register_sighup(hs.get_instance_id(), refresh_certificate, hs)
+        register_sighup(hs.get_instance_id(), reload_cache_config, hs.config)
 
     # Apply the cache config.
     hs.config.caches.resize_all_caches()
@@ -574,9 +620,6 @@ async def start(hs: "HomeServer") -> None:
     # Load the certificate from disk.
     refresh_certificate(hs)
 
-    # Start the tracer
-    init_tracer(hs)  # noqa
-
     # Instantiate the modules so they can register their web resources to the module API
     # before we start the listeners.
     module_api = hs.get_module_api()
@@ -603,11 +646,15 @@ async def start(hs: "HomeServer") -> None:
     hs.get_pusherpool().start()
 
     def log_shutdown() -> None:
-        with LoggingContext("log_shutdown"):
+        with LoggingContext(name="log_shutdown", server_name=server_name):
             logger.info("Shutting down...")
 
     # Log when we start the shut down process.
-    hs.get_clock().add_system_event_trigger("before", "shutdown", log_shutdown)
+    hs.register_sync_shutdown_handler(
+        phase="before",
+        eventType="shutdown",
+        shutdown_func=log_shutdown,
+    )
 
     setup_sentry(hs)
     setup_sdnotify(hs)
@@ -636,18 +683,24 @@ async def start(hs: "HomeServer") -> None:
         # `REQUIRED_ON_BACKGROUND_TASK_STARTUP`
         start_phone_stats_home(hs)
 
-    # We now freeze all allocated objects in the hopes that (almost)
-    # everything currently allocated are things that will be used for the
-    # rest of time. Doing so means less work each GC (hopefully).
-    #
-    # PyPy does not (yet?) implement gc.freeze()
-    if hasattr(gc, "freeze"):
-        gc.collect()
-        gc.freeze()
+    if freeze:
+        # We now freeze all allocated objects in the hopes that (almost)
+        # everything currently allocated are things that will be used for the
+        # rest of time. Doing so means less work each GC (hopefully).
+        #
+        # Note that freezing the homeserver object means that it won't be able to be
+        # garbage collected in the case of attempting an in-memory `shutdown`. This only
+        # needs to be considered if such a case is desirable. Exiting the entire Python
+        # process will function expectedly either way.
+        #
+        # PyPy does not (yet?) implement gc.freeze()
+        if hasattr(gc, "freeze"):
+            gc.collect()
+            gc.freeze()
 
-        # Speed up shutdowns by freezing all allocated objects. This moves everything
-        # into the permanent generation and excludes them from the final GC.
-        atexit.register(gc.freeze)
+            # Speed up process exit by freezing all allocated objects. This moves everything
+            # into the permanent generation and excludes them from the final GC.
+            atexit.register(gc.freeze)
 
 
 def reload_cache_config(config: HomeServerConfig) -> None:

synapse/app/admin_cmd.py

@@ -329,7 +329,7 @@ def start(config: HomeServerConfig, args: argparse.Namespace) -> None:
     # command.
 
     async def run() -> None:
-        with LoggingContext(name="command"):
+        with LoggingContext(name="command", server_name=config.server.server_name):
             await _base.start(ss)
             await args.func(ss, args)
 
@@ -342,5 +342,5 @@ def start(config: HomeServerConfig, args: argparse.Namespace) -> None:
 
 if __name__ == "__main__":
     homeserver_config, args = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config, args)

synapse/app/appservice.py

@@ -27,7 +27,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/client_reader.py

@@ -27,7 +27,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/event_creator.py

@@ -26,7 +26,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/federation_reader.py

@@ -27,7 +27,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/federation_sender.py

@@ -27,7 +27,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/frontend_proxy.py

@@ -27,7 +27,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/generic_worker.py

@@ -49,6 +49,7 @@ from synapse.config.server import ListenerConfig, TCPListenerConfig
 from synapse.federation.transport.server import TransportLayerServer
 from synapse.http.server import JsonResource, OptionsResource
 from synapse.logging.context import LoggingContext
+from synapse.logging.opentracing import init_tracer
 from synapse.metrics import METRICS_PREFIX, MetricsResource, RegistryProxy
 from synapse.replication.http import REPLICATION_PREFIX, ReplicationRestResource
 from synapse.rest import ClientRestResource, admin
@@ -277,11 +278,13 @@ class GenericWorkerServer(HomeServer):
                 self._listen_http(listener)
             elif listener.type == "manhole":
                 if isinstance(listener, TCPListenerConfig):
-                    _base.listen_manhole(
-                        listener.bind_addresses,
-                        listener.port,
-                        manhole_settings=self.config.server.manhole_settings,
-                        manhole_globals={"hs": self},
+                    self._listening_services.extend(
+                        _base.listen_manhole(
+                            listener.bind_addresses,
+                            listener.port,
+                            manhole_settings=self.config.server.manhole_settings,
+                            manhole_globals={"hs": self},
+                        )
                     )
                 else:
                     raise ConfigError(
@@ -295,9 +298,11 @@ class GenericWorkerServer(HomeServer):
                     )
                 else:
                     if isinstance(listener, TCPListenerConfig):
-                        _base.listen_metrics(
-                            listener.bind_addresses,
-                            listener.port,
+                        self._metrics_listeners.extend(
+                            _base.listen_metrics(
+                                listener.bind_addresses,
+                                listener.port,
+                            )
                         )
                     else:
                         raise ConfigError(
@@ -359,6 +364,9 @@ def start(config: HomeServerConfig) -> None:
 
     setup_logging(hs, config, use_worker_options=True)
 
+    # Start the tracer
+    init_tracer(hs)  # noqa
+
     try:
         hs.setup()
 
@@ -382,7 +390,7 @@ def start(config: HomeServerConfig) -> None:
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/homeserver.py

@@ -22,14 +22,13 @@
 import logging
 import os
 import sys
-from typing import Dict, Iterable, List
+from typing import Dict, Iterable, List, Optional
 
 from twisted.internet.tcp import Port
 from twisted.web.resource import EncodingResourceWrapper, Resource
 from twisted.web.server import GzipEncoderFactory
 
 import synapse
-import synapse.config.logger
 from synapse import events
 from synapse.api.urls import (
     CLIENT_API_PREFIX,
@@ -50,6 +49,7 @@ from synapse.app._base import (
 )
 from synapse.config._base import ConfigError, format_config_error
 from synapse.config.homeserver import HomeServerConfig
+from synapse.config.logger import setup_logging
 from synapse.config.server import ListenerConfig, TCPListenerConfig
 from synapse.federation.transport.server import TransportLayerServer
 from synapse.http.additional_resource import AdditionalResource
@@ -60,6 +60,7 @@ from synapse.http.server import (
     StaticResource,
 )
 from synapse.logging.context import LoggingContext
+from synapse.logging.opentracing import init_tracer
 from synapse.metrics import METRICS_PREFIX, MetricsResource, RegistryProxy
 from synapse.replication.http import REPLICATION_PREFIX, ReplicationRestResource
 from synapse.rest import ClientRestResource, admin
@@ -69,6 +70,7 @@ from synapse.rest.synapse.client import build_synapse_client_resource_tree
 from synapse.rest.well_known import well_known_resource
 from synapse.server import HomeServer
 from synapse.storage import DataStore
+from synapse.types import ISynapseReactor
 from synapse.util.check_dependencies import VERSION, check_requirements
 from synapse.util.httpresourcetree import create_resource_tree
 from synapse.util.module_loader import load_module
@@ -276,11 +278,13 @@ class SynapseHomeServer(HomeServer):
                 )
             elif listener.type == "manhole":
                 if isinstance(listener, TCPListenerConfig):
-                    _base.listen_manhole(
-                        listener.bind_addresses,
-                        listener.port,
-                        manhole_settings=self.config.server.manhole_settings,
-                        manhole_globals={"hs": self},
+                    self._listening_services.extend(
+                        _base.listen_manhole(
+                            listener.bind_addresses,
+                            listener.port,
+                            manhole_settings=self.config.server.manhole_settings,
+                            manhole_globals={"hs": self},
+                        )
                     )
                 else:
                     raise ConfigError(
@@ -293,9 +297,11 @@ class SynapseHomeServer(HomeServer):
                     )
                 else:
                     if isinstance(listener, TCPListenerConfig):
-                        _base.listen_metrics(
-                            listener.bind_addresses,
-                            listener.port,
+                        self._metrics_listeners.extend(
+                            _base.listen_metrics(
+                                listener.bind_addresses,
+                                listener.port,
+                            )
                         )
                     else:
                         raise ConfigError(
@@ -339,12 +345,23 @@ def load_or_generate_config(argv_options: List[str]) -> HomeServerConfig:
     return config
 
 
-def setup(config: HomeServerConfig) -> SynapseHomeServer:
+def setup(
+    config: HomeServerConfig,
+    reactor: Optional[ISynapseReactor] = None,
+    freeze: bool = True,
+) -> SynapseHomeServer:
     """
     Create and setup a Synapse homeserver instance given a configuration.
 
     Args:
         config: The configuration for the homeserver.
+        reactor: Optionally provide a reactor to use. Can be useful in different
+            scenarios that you want control over the reactor, such as tests.
+        freeze: whether to freeze the homeserver base objects in the garbage collector.
+            May improve garbage collection performance by marking objects with an effectively
+            static lifetime as frozen so they don't need to be considered for cleanup.
+            If you ever want to `shutdown` the homeserver, this needs to be
+            False otherwise the homeserver cannot be garbage collected after `shutdown`.
 
     Returns:
         A homeserver instance.
@@ -383,9 +400,13 @@ def setup(config: HomeServerConfig) -> SynapseHomeServer:
         config.server.server_name,
         config=config,
         version_string=f"Synapse/{VERSION}",
+        reactor=reactor,
     )
 
-    synapse.config.logger.setup_logging(hs, config, use_worker_options=False)
+    setup_logging(hs, config, use_worker_options=False)
+
+    # Start the tracer
+    init_tracer(hs)  # noqa
 
     logger.info("Setting up server")
 
@@ -401,7 +422,7 @@ def setup(config: HomeServerConfig) -> SynapseHomeServer:
             # Loading the provider metadata also ensures the provider config is valid.
             await oidc.load_metadata()
 
-        await _base.start(hs)
+        await _base.start(hs, freeze)
 
         hs.get_datastores().main.db_pool.updates.start_doing_background_updates()
 
@@ -425,7 +446,7 @@ def run(hs: HomeServer) -> None:
 def main() -> None:
     homeserver_config = load_or_generate_config(sys.argv[1:])
 
-    with LoggingContext("main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         # check base requirements
         check_requirements()
         hs = setup(homeserver_config)

synapse/app/media_repository.py

@@ -27,7 +27,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/phone_stats_home.py

@@ -29,9 +29,6 @@ from prometheus_client import Gauge
 from twisted.internet import defer
 
 from synapse.metrics import SERVER_NAME_LABEL
-from synapse.metrics.background_process_metrics import (
-    run_as_background_process,
-)
 from synapse.types import JsonDict
 from synapse.util.constants import ONE_HOUR_SECONDS, ONE_MINUTE_SECONDS
 
@@ -85,8 +82,6 @@ def phone_stats_home(
     stats: JsonDict,
     stats_process: List[Tuple[int, "resource.struct_rusage"]] = _stats_process,
 ) -> "defer.Deferred[None]":
-    server_name = hs.hostname
-
     async def _phone_stats_home(
         hs: "HomeServer",
         stats: JsonDict,
@@ -200,8 +195,8 @@ def phone_stats_home(
         except Exception as e:
             logger.warning("Error reporting stats: %s", e)
 
-    return run_as_background_process(
-        "phone_stats_home", server_name, _phone_stats_home, hs, stats, stats_process
+    return hs.run_as_background_process(
+        "phone_stats_home", _phone_stats_home, hs, stats, stats_process
     )
 
 
@@ -263,9 +258,8 @@ def start_phone_stats_home(hs: "HomeServer") -> None:
                 float(hs.config.server.max_mau_value)
             )
 
-        return run_as_background_process(
+        return hs.run_as_background_process(
             "generate_monthly_active_users",
-            server_name,
             _generate_monthly_active_users,
         )
 
@@ -285,10 +279,16 @@ def start_phone_stats_home(hs: "HomeServer") -> None:
 
         # We need to defer this init for the cases that we daemonize
         # otherwise the process ID we get is that of the non-daemon process
-        clock.call_later(0, performance_stats_init)
+        clock.call_later(
+            0,
+            performance_stats_init,
+        )
 
         # We wait 5 minutes to send the first set of stats as the server can
         # be quite busy the first few minutes
         clock.call_later(
-            INITIAL_DELAY_BEFORE_FIRST_PHONE_HOME_SECONDS, phone_stats_home, hs, stats
+            INITIAL_DELAY_BEFORE_FIRST_PHONE_HOME_SECONDS,
+            phone_stats_home,
+            hs,
+            stats,
         )

synapse/app/pusher.py

@@ -27,7 +27,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/synchrotron.py

@@ -27,7 +27,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/app/user_dir.py

@@ -27,7 +27,7 @@ from synapse.util.logcontext import LoggingContext
 
 def main() -> None:
     homeserver_config = load_config(sys.argv[1:])
-    with LoggingContext(name="main"):
+    with LoggingContext(name="main", server_name=homeserver_config.server.server_name):
         start(homeserver_config)
 
 

synapse/appservice/__init__.py

@@ -23,15 +23,33 @@
 import logging
 import re
 from enum import Enum
-from typing import TYPE_CHECKING, Dict, Iterable, List, Optional, Pattern, Sequence
+from typing import (
+    TYPE_CHECKING,
+    Dict,
+    Iterable,
+    List,
+    Optional,
+    Pattern,
+    Sequence,
+    cast,
+)
 
 import attr
 from netaddr import IPSet
 
+from twisted.internet import reactor
+
 from synapse.api.constants import EventTypes
 from synapse.events import EventBase
-from synapse.types import DeviceListUpdates, JsonDict, JsonMapping, UserID
+from synapse.types import (
+    DeviceListUpdates,
+    ISynapseThreadlessReactor,
+    JsonDict,
+    JsonMapping,
+    UserID,
+)
 from synapse.util.caches.descriptors import _CacheContext, cached
+from synapse.util.clock import Clock
 
 if TYPE_CHECKING:
     from synapse.appservice.api import ApplicationServiceApi
@@ -98,6 +116,15 @@ class ApplicationService:
         self.sender = sender
         # The application service user should be part of the server's domain.
         self.server_name = sender.domain  # nb must be called this for @cached
+
+        # Ideally we would require passing in the `HomeServer` `Clock` instance.
+        # However this is not currently possible as there are places which use
+        # `@cached` that aren't aware of the `HomeServer` instance.
+        # nb must be called this for @cached
+        self.clock = Clock(
+            cast(ISynapseThreadlessReactor, reactor), server_name=self.server_name
+        )  # type: ignore[multiple-internal-clocks]
+
         self.namespaces = self._check_namespaces(namespaces)
         self.id = id
         self.ip_range_whitelist = ip_range_whitelist

synapse/appservice/scheduler.py

@@ -81,7 +81,6 @@ from synapse.appservice import (
 from synapse.appservice.api import ApplicationServiceApi
 from synapse.events import EventBase
 from synapse.logging.context import run_in_background
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.storage.databases.main import DataStore
 from synapse.types import DeviceListUpdates, JsonMapping
 from synapse.util.clock import Clock
@@ -200,6 +199,7 @@ class _ServiceQueuer:
         )
         self.server_name = hs.hostname
         self.clock = hs.get_clock()
+        self.hs = hs
         self._store = hs.get_datastores().main
 
     def start_background_request(self, service: ApplicationService) -> None:
@@ -207,9 +207,7 @@ class _ServiceQueuer:
         if service.id in self.requests_in_flight:
             return
 
-        run_as_background_process(
-            "as-sender", self.server_name, self._send_request, service
-        )
+        self.hs.run_as_background_process("as-sender", self._send_request, service)
 
     async def _send_request(self, service: ApplicationService) -> None:
         # sanity-check: we shouldn't get here if this service already has a sender
@@ -361,6 +359,7 @@ class _TransactionController:
     def __init__(self, hs: "HomeServer"):
         self.server_name = hs.hostname
         self.clock = hs.get_clock()
+        self.hs = hs
         self.store = hs.get_datastores().main
         self.as_api = hs.get_application_service_api()
 
@@ -448,6 +447,7 @@ class _TransactionController:
         recoverer = self.RECOVERER_CLASS(
             self.server_name,
             self.clock,
+            self.hs,
             self.store,
             self.as_api,
             service,
@@ -494,6 +494,7 @@ class _Recoverer:
         self,
         server_name: str,
         clock: Clock,
+        hs: "HomeServer",
         store: DataStore,
         as_api: ApplicationServiceApi,
         service: ApplicationService,
@@ -501,6 +502,7 @@ class _Recoverer:
     ):
         self.server_name = server_name
         self.clock = clock
+        self.hs = hs
         self.store = store
         self.as_api = as_api
         self.service = service
@@ -513,9 +515,8 @@ class _Recoverer:
         logger.info("Scheduling retries on %s in %fs", self.service.id, delay)
         self.scheduled_recovery = self.clock.call_later(
             delay,
-            run_as_background_process,
+            self.hs.run_as_background_process,
             "as-recoverer",
-            self.server_name,
             self.retry,
         )
 
@@ -535,9 +536,8 @@ class _Recoverer:
         if self.scheduled_recovery:
             self.clock.cancel_call_later(self.scheduled_recovery)
         # Run a retry, which will resechedule a recovery if it fails.
-        run_as_background_process(
+        self.hs.run_as_background_process(
             "retry",
-            self.server_name,
             self.retry,
         )
 

synapse/config/_base.py

@@ -601,7 +601,7 @@ class RootConfig:
 
     @classmethod
     def load_config_with_parser(
-        cls: Type[TRootConfig], parser: argparse.ArgumentParser, argv: List[str]
+        cls: Type[TRootConfig], parser: argparse.ArgumentParser, argv_options: List[str]
     ) -> Tuple[TRootConfig, argparse.Namespace]:
         """Parse the commandline and config files with the given parser
 
@@ -611,14 +611,14 @@ class RootConfig:
 
         Args:
             parser
-            argv
+            argv_options: The options passed to Synapse. Usually `sys.argv[1:]`.
 
         Returns:
             Returns the parsed config object and the parsed argparse.Namespace
             object from parser.parse_args(..)`
         """
 
-        config_args = parser.parse_args(argv)
+        config_args = parser.parse_args(argv_options)
 
         config_files = find_config_files(search_paths=config_args.config_path)
         obj = cls(config_files)

synapse/config/logger.py

@@ -40,7 +40,6 @@ from twisted.logger import (
 )
 
 from synapse.logging.context import LoggingContextFilter
-from synapse.logging.filter import MetadataFilter
 from synapse.synapse_rust import reset_logging_config
 from synapse.types import JsonDict
 
@@ -213,13 +212,11 @@ def _setup_stdlib_logging(
     # writes.
 
     log_context_filter = LoggingContextFilter()
-    log_metadata_filter = MetadataFilter({"server_name": config.server.server_name})
     old_factory = logging.getLogRecordFactory()
 
     def factory(*args: Any, **kwargs: Any) -> logging.LogRecord:
         record = old_factory(*args, **kwargs)
         log_context_filter.filter(record)
-        log_metadata_filter.filter(record)
         return record
 
     logging.setLogRecordFactory(factory)
@@ -348,7 +345,9 @@ def setup_logging(
     # Add a SIGHUP handler to reload the logging configuration, if one is available.
     from synapse.app import _base as appbase
 
-    appbase.register_sighup(_reload_logging_config, log_config_path)
+    appbase.register_sighup(
+        hs.get_instance_id(), _reload_logging_config, log_config_path
+    )
 
     # Log immediately so we can grep backwards.
     logger.warning("***** STARTING SERVER *****")

synapse/crypto/keyring.py

@@ -172,7 +172,7 @@ class Keyring:
             _FetchKeyRequest, Dict[str, Dict[str, FetchKeyResult]]
         ] = BatchingQueue(
             name="keyring_server",
-            server_name=self.server_name,
+            hs=hs,
             clock=hs.get_clock(),
             # The method called to fetch each key
             process_batch_callback=self._inner_fetch_key_requests,
@@ -194,6 +194,14 @@ class Keyring:
             valid_until_ts=2**63,  # fake future timestamp
         )
 
+    def shutdown(self) -> None:
+        """
+        Prepares the KeyRing for garbage collection by shutting down it's queues.
+        """
+        self._fetch_keys_queue.shutdown()
+        for key_fetcher in self._key_fetchers:
+            key_fetcher.shutdown()
+
     async def verify_json_for_server(
         self,
         server_name: str,
@@ -316,7 +324,7 @@ class Keyring:
             if key_result.valid_until_ts < verify_request.minimum_valid_until_ts:
                 continue
 
-            await self._process_json(key_result.verify_key, verify_request)
+            await self.process_json(key_result.verify_key, verify_request)
             verified = True
 
         if not verified:
@@ -326,7 +334,7 @@ class Keyring:
                 Codes.UNAUTHORIZED,
             )
 
-    async def _process_json(
+    async def process_json(
         self, verify_key: VerifyKey, verify_request: VerifyJsonRequest
     ) -> None:
         """Processes the `VerifyJsonRequest`. Raises if the signature can't be
@@ -479,11 +487,17 @@ class KeyFetcher(metaclass=abc.ABCMeta):
         self.server_name = hs.hostname
         self._queue = BatchingQueue(
             name=self.__class__.__name__,
-            server_name=self.server_name,
+            hs=hs,
             clock=hs.get_clock(),
             process_batch_callback=self._fetch_keys,
         )
 
+    def shutdown(self) -> None:
+        """
+        Prepares the KeyFetcher for garbage collection by shutting down it's queue.
+        """
+        self._queue.shutdown()
+
     async def get_keys(
         self, server_name: str, key_ids: List[str], minimum_valid_until_ts: int
     ) -> Dict[str, FetchKeyResult]:

synapse/events/auto_accept_invites.py

@@ -119,7 +119,6 @@ class InviteAutoAccepter:
             event.state_key,
             event.room_id,
             "join",
-            bg_start_span=False,
         )
 
         if is_direct_message:

synapse/federation/federation_client.py

@@ -148,6 +148,7 @@ class FederationClient(FederationBase):
         self._get_pdu_cache: ExpiringCache[str, Tuple[EventBase, str]] = ExpiringCache(
             cache_name="get_pdu_cache",
             server_name=self.server_name,
+            hs=self.hs,
             clock=self._clock,
             max_len=1000,
             expiry_ms=120 * 1000,
@@ -167,6 +168,7 @@ class FederationClient(FederationBase):
         ] = ExpiringCache(
             cache_name="get_room_hierarchy_cache",
             server_name=self.server_name,
+            hs=self.hs,
             clock=self._clock,
             max_len=1000,
             expiry_ms=5 * 60 * 1000,
@@ -495,6 +497,43 @@ class FederationClient(FederationBase):
             )
             return RECOMMENDATION_OK
 
+    @trace
+    @tag_args
+    async def ask_policy_server_to_sign_event(
+        self, destination: str, pdu: EventBase, timeout: Optional[int] = None
+    ) -> Optional[JsonDict]:
+        """Requests that the destination server (typically a policy server)
+        sign the event as not spam.
+
+        If the policy server could not be contacted or the policy server
+        returned an error, this returns no signature.
+
+        Args:
+            destination: The remote homeserver to ask (a policy server)
+            pdu: The event to sign
+            timeout: How long to try (in ms) the destination for before
+                giving up. None indicates no timeout.
+        Returns:
+            The signature from the policy server, structured in the same was as the 'signatures'
+            JSON in the event e.g { "$policy_server_via_domain" : { "ed25519:policy_server": "signature_base64" }}
+        """
+        logger.debug(
+            "ask_policy_server_to_sign_event for event_id=%s from %s",
+            pdu.event_id,
+            destination,
+        )
+        try:
+            return await self.transport_layer.ask_policy_server_to_sign_event(
+                destination, pdu, timeout=timeout
+            )
+        except Exception as e:
+            logger.warning(
+                "ask_policy_server_to_sign_event: server %s responded with error: %s",
+                destination,
+                e,
+            )
+        return None
+
     @trace
     @tag_args
     async def get_pdu(

synapse/federation/federation_server.py

@@ -159,7 +159,7 @@ class FederationServer(FederationBase):
         # with FederationHandlerRegistry.
         hs.get_directory_handler()
 
-        self._server_linearizer = Linearizer("fed_server")
+        self._server_linearizer = Linearizer(name="fed_server", clock=hs.get_clock())
 
         # origins that we are currently processing a transaction from.
         # a dict from origin to txn id.

synapse/federation/send_queue.py

@@ -144,6 +144,9 @@ class FederationRemoteSendQueue(AbstractFederationSender):
 
         self.clock.looping_call(self._clear_queue, 30 * 1000)
 
+    def shutdown(self) -> None:
+        """Stops this federation sender instance from sending further transactions."""
+
     def _next_pos(self) -> int:
         pos = self.pos
         self.pos += 1

synapse/federation/sender/__init__.py

@@ -168,7 +168,6 @@ from synapse.metrics import (
     events_processed_counter,
 )
 from synapse.metrics.background_process_metrics import (
-    run_as_background_process,
     wrap_as_background_process,
 )
 from synapse.types import (
@@ -232,6 +231,11 @@ WAKEUP_INTERVAL_BETWEEN_DESTINATIONS_SEC = 5
 
 
 class AbstractFederationSender(metaclass=abc.ABCMeta):
+    @abc.abstractmethod
+    def shutdown(self) -> None:
+        """Stops this federation sender instance from sending further transactions."""
+        raise NotImplementedError()
+
     @abc.abstractmethod
     def notify_new_events(self, max_token: RoomStreamToken) -> None:
         """This gets called when we have some new events we might want to
@@ -326,6 +330,7 @@ class _DestinationWakeupQueue:
     _MAX_TIME_IN_QUEUE = 30.0
 
     sender: "FederationSender" = attr.ib()
+    hs: "HomeServer" = attr.ib()
     server_name: str = attr.ib()
     """
     Our homeserver name (used to label metrics) (`hs.hostname`).
@@ -453,18 +458,30 @@ class FederationSender(AbstractFederationSender):
             1.0 / hs.config.ratelimiting.federation_rr_transactions_per_room_per_second
         )
         self._destination_wakeup_queue = _DestinationWakeupQueue(
-            self, self.server_name, self.clock, max_delay_s=rr_txn_interval_per_room_s
+            self,
+            hs,
+            self.server_name,
+            self.clock,
+            max_delay_s=rr_txn_interval_per_room_s,
         )
 
+        # It is important for `_is_shutdown` to be instantiated before the looping call
+        # for `wake_destinations_needing_catchup`.
+        self._is_shutdown = False
+
         # Regularly wake up destinations that have outstanding PDUs to be caught up
         self.clock.looping_call_now(
-            run_as_background_process,
+            self.hs.run_as_background_process,
             WAKEUP_RETRY_PERIOD_SEC * 1000.0,
             "wake_destinations_needing_catchup",
-            self.server_name,
             self._wake_destinations_needing_catchup,
         )
 
+    def shutdown(self) -> None:
+        self._is_shutdown = True
+        for queue in self._per_destination_queues.values():
+            queue.shutdown()
+
     def _get_per_destination_queue(
         self, destination: str
     ) -> Optional[PerDestinationQueue]:
@@ -503,16 +520,15 @@ class FederationSender(AbstractFederationSender):
             return
 
         # fire off a processing loop in the background
-        run_as_background_process(
+        self.hs.run_as_background_process(
             "process_event_queue_for_federation",
-            self.server_name,
             self._process_event_queue_loop,
         )
 
     async def _process_event_queue_loop(self) -> None:
         try:
             self._is_processing = True
-            while True:
+            while not self._is_shutdown:
                 last_token = await self.store.get_federation_out_pos("events")
                 (
                     next_token,
@@ -1123,7 +1139,7 @@ class FederationSender(AbstractFederationSender):
 
         last_processed: Optional[str] = None
 
-        while True:
+        while not self._is_shutdown:
             destinations_to_wake = (
                 await self.store.get_catch_up_outstanding_destinations(last_processed)
             )

synapse/federation/sender/per_destination_queue.py

@@ -28,6 +28,8 @@ from typing import TYPE_CHECKING, Dict, Hashable, Iterable, List, Optional, Tupl
 import attr
 from prometheus_client import Counter
 
+from twisted.internet import defer
+
 from synapse.api.constants import EduTypes
 from synapse.api.errors import (
     FederationDeniedError,
@@ -41,7 +43,6 @@ from synapse.handlers.presence import format_user_presence_state
 from synapse.logging import issue9533_logger
 from synapse.logging.opentracing import SynapseTags, set_tag
 from synapse.metrics import SERVER_NAME_LABEL, sent_transactions_counter
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.types import JsonDict, ReadReceipt
 from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter
 from synapse.visibility import filter_events_for_server
@@ -79,6 +80,7 @@ MAX_PRESENCE_STATES_PER_EDU = 50
 class PerDestinationQueue:
     """
     Manages the per-destination transmission queues.
+    Runs until `shutdown()` is called on the queue.
 
     Args:
         hs
@@ -94,6 +96,7 @@ class PerDestinationQueue:
         destination: str,
     ):
         self.server_name = hs.hostname
+        self._hs = hs
         self._clock = hs.get_clock()
         self._storage_controllers = hs.get_storage_controllers()
         self._store = hs.get_datastores().main
@@ -117,6 +120,8 @@ class PerDestinationQueue:
 
         self._destination = destination
         self.transmission_loop_running = False
+        self._transmission_loop_enabled = True
+        self.active_transmission_loop: Optional[defer.Deferred] = None
 
         # Flag to signal to any running transmission loop that there is new data
         # queued up to be sent.
@@ -171,6 +176,20 @@ class PerDestinationQueue:
     def __str__(self) -> str:
         return "PerDestinationQueue[%s]" % self._destination
 
+    def shutdown(self) -> None:
+        """Instruct the queue to stop processing any further requests"""
+        self._transmission_loop_enabled = False
+        # The transaction manager must be shutdown before cancelling the active
+        # transmission loop. Otherwise the transmission loop can enter a new cycle of
+        # sleeping before retrying since the shutdown flag of the _transaction_manager
+        # hasn't been set yet.
+        self._transaction_manager.shutdown()
+        try:
+            if self.active_transmission_loop is not None:
+                self.active_transmission_loop.cancel()
+        except Exception:
+            pass
+
     def pending_pdu_count(self) -> int:
         return len(self._pending_pdus)
 
@@ -309,11 +328,14 @@ class PerDestinationQueue:
             )
             return
 
+        if not self._transmission_loop_enabled:
+            logger.warning("Shutdown has been requested. Not sending transaction")
+            return
+
         logger.debug("TX [%s] Starting transaction loop", self._destination)
 
-        run_as_background_process(
+        self.active_transmission_loop = self._hs.run_as_background_process(
             "federation_transaction_transmission_loop",
-            self.server_name,
             self._transaction_transmission_loop,
         )
 
@@ -321,13 +343,13 @@ class PerDestinationQueue:
         pending_pdus: List[EventBase] = []
         try:
             self.transmission_loop_running = True
-
             # This will throw if we wouldn't retry. We do this here so we fail
             # quickly, but we will later check this again in the http client,
             # hence why we throw the result away.
             await get_retry_limiter(
                 destination=self._destination,
                 our_server_name=self.server_name,
+                hs=self._hs,
                 clock=self._clock,
                 store=self._store,
             )
@@ -339,7 +361,7 @@ class PerDestinationQueue:
                     # not caught up yet
                     return
 
-            while True:
+            while self._transmission_loop_enabled:
                 self._new_data_to_send = False
 
                 async with _TransactionQueueManager(self) as (
@@ -352,8 +374,8 @@ class PerDestinationQueue:
                         # If we've gotten told about new things to send during
                         # checking for things to send, we try looking again.
                         # Otherwise new PDUs or EDUs might arrive in the meantime,
-                        # but not get sent because we hold the
-                        # `transmission_loop_running` flag.
+                        # but not get sent because we currently have an
+                        # `_active_transmission_loop` running.
                         if self._new_data_to_send:
                             continue
                         else:
@@ -442,6 +464,7 @@ class PerDestinationQueue:
                 )
         finally:
             # We want to be *very* sure we clear this after we stop processing
+            self.active_transmission_loop = None
             self.transmission_loop_running = False
 
     async def _catch_up_transmission_loop(self) -> None:
@@ -469,7 +492,7 @@ class PerDestinationQueue:
         last_successful_stream_ordering: int = _tmp_last_successful_stream_ordering
 
         # get at most 50 catchup room/PDUs
-        while True:
+        while self._transmission_loop_enabled:
             event_ids = await self._store.get_catch_up_room_event_ids(
                 self._destination, last_successful_stream_ordering
             )

synapse/federation/sender/transaction_manager.py

@@ -72,6 +72,12 @@ class TransactionManager:
         # HACK to get unique tx id
         self._next_txn_id = int(self.clock.time_msec())
 
+        self._is_shutdown = False
+
+    def shutdown(self) -> None:
+        self._is_shutdown = True
+        self._transport_layer.shutdown()
+
     @measure_func("_send_new_transaction")
     async def send_new_transaction(
         self,
@@ -86,6 +92,12 @@ class TransactionManager:
             edus: List of EDUs to send
         """
 
+        if self._is_shutdown:
+            logger.warning(
+                "TransactionManager has been shutdown, not sending transaction"
+            )
+            return
+
         # Make a transaction-sending opentracing span. This span follows on from
         # all the edus in that transaction. This needs to be done since there is
         # no active span here, so if the edus were not received by the remote the

synapse/federation/transport/client.py

@@ -70,6 +70,9 @@ class TransportLayerClient:
         self.client = hs.get_federation_http_client()
         self._is_mine_server_name = hs.is_mine_server_name
 
+    def shutdown(self) -> None:
+        self.client.shutdown()
+
     async def get_room_state_ids(
         self, destination: str, room_id: str, event_id: str
     ) -> JsonDict:
@@ -170,6 +173,32 @@ class TransportLayerClient:
             timeout=timeout,
         )
 
+    async def ask_policy_server_to_sign_event(
+        self, destination: str, event: EventBase, timeout: Optional[int] = None
+    ) -> JsonDict:
+        """Requests that the destination server (typically a policy server)
+        sign the event as not spam.
+
+        If the policy server could not be contacted or the policy server
+        returned an error, this raises that error.
+
+        Args:
+            destination: The host name of the policy server / homeserver.
+            event: The event to sign.
+            timeout: How long to try (in ms) the destination for before giving up.
+                None indicates no timeout.
+        Returns:
+            The signature from the policy server, structured in the same was as the 'signatures'
+            JSON in the event e.g { "$policy_server_via_domain" : { "ed25519:policy_server": "signature_base64" }}
+        """
+        return await self.client.post_json(
+            destination=destination,
+            path="/_matrix/policy/unstable/org.matrix.msc4284/sign",
+            data=event.get_pdu_json(),
+            ignore_backoff=True,
+            timeout=timeout,
+        )
+
     async def backfill(
         self, destination: str, room_id: str, event_tuples: Collection[str], limit: int
     ) -> Optional[Union[JsonDict, list]]:

synapse/handlers/account_validity.py

@@ -37,10 +37,8 @@ logger = logging.getLogger(__name__)
 
 class AccountValidityHandler:
     def __init__(self, hs: "HomeServer"):
-        self.hs = hs
-        self.server_name = (
-            hs.hostname
-        )  # nb must be called this for @wrap_as_background_process
+        self.hs = hs  # nb must be called this for @wrap_as_background_process
+        self.server_name = hs.hostname
         self.config = hs.config
         self.store = hs.get_datastores().main
         self.send_email_handler = hs.get_send_email_handler()

synapse/handlers/appservice.py

@@ -47,7 +47,6 @@ from synapse.metrics import (
     event_processing_loop_room_count,
 )
 from synapse.metrics.background_process_metrics import (
-    run_as_background_process,
     wrap_as_background_process,
 )
 from synapse.storage.databases.main.directory import RoomAliasMapping
@@ -76,9 +75,8 @@ events_processed_counter = Counter(
 
 class ApplicationServicesHandler:
     def __init__(self, hs: "HomeServer"):
-        self.server_name = (
-            hs.hostname
-        )  # nb must be called this for @wrap_as_background_process
+        self.server_name = hs.hostname
+        self.hs = hs  # nb must be called this for @wrap_as_background_process
         self.store = hs.get_datastores().main
         self.is_mine_id = hs.is_mine_id
         self.appservice_api = hs.get_application_service_api()
@@ -98,7 +96,7 @@ class ApplicationServicesHandler:
         self.is_processing = False
 
         self._ephemeral_events_linearizer = Linearizer(
-            name="appservice_ephemeral_events"
+            name="appservice_ephemeral_events", clock=hs.get_clock()
         )
 
     def notify_interested_services(self, max_token: RoomStreamToken) -> None:
@@ -171,8 +169,8 @@ class ApplicationServicesHandler:
                                 except Exception:
                                     logger.error("Application Services Failure")
 
-                            run_as_background_process(
-                                "as_scheduler", self.server_name, start_scheduler
+                            self.hs.run_as_background_process(
+                                "as_scheduler", start_scheduler
                             )
                             self.started_scheduler = True
 

synapse/handlers/deactivate_account.py

@@ -24,7 +24,6 @@ from typing import TYPE_CHECKING, Optional
 
 from synapse.api.constants import Membership
 from synapse.api.errors import SynapseError
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.replication.http.deactivate_account import (
     ReplicationNotifyAccountDeactivatedServlet,
 )
@@ -272,8 +271,8 @@ class DeactivateAccountHandler:
         pending deactivation, if it isn't already running.
         """
         if not self._user_parter_running:
-            run_as_background_process(
-                "user_parter_loop", self.server_name, self._user_parter_loop
+            self.hs.run_as_background_process(
+                "user_parter_loop", self._user_parter_loop
             )
 
     async def _user_parter_loop(self) -> None:

synapse/handlers/delayed_events.py

@@ -24,9 +24,6 @@ from synapse.config.workers import MAIN_PROCESS_INSTANCE_NAME
 from synapse.logging.context import make_deferred_yieldable
 from synapse.logging.opentracing import set_tag
 from synapse.metrics import SERVER_NAME_LABEL, event_processing_positions
-from synapse.metrics.background_process_metrics import (
-    run_as_background_process,
-)
 from synapse.replication.http.delayed_events import (
     ReplicationAddedDelayedEventRestServlet,
 )
@@ -58,6 +55,7 @@ logger = logging.getLogger(__name__)
 
 class DelayedEventsHandler:
     def __init__(self, hs: "HomeServer"):
+        self.hs = hs
         self.server_name = hs.hostname
         self._store = hs.get_datastores().main
         self._storage_controllers = hs.get_storage_controllers()
@@ -94,7 +92,10 @@ class DelayedEventsHandler:
                 hs.get_notifier().add_replication_callback(self.notify_new_event)
                 # Kick off again (without blocking) to catch any missed notifications
                 # that may have fired before the callback was added.
-                self._clock.call_later(0, self.notify_new_event)
+                self._clock.call_later(
+                    0,
+                    self.notify_new_event,
+                )
 
                 # Delayed events that are already marked as processed on startup might not have been
                 # sent properly on the last run of the server, so unmark them to send them again.
@@ -112,15 +113,14 @@ class DelayedEventsHandler:
                     self._schedule_next_at(next_send_ts)
 
                 # Can send the events in background after having awaited on marking them as processed
-                run_as_background_process(
+                self.hs.run_as_background_process(
                     "_send_events",
-                    self.server_name,
                     self._send_events,
                     events,
                 )
 
-            self._initialized_from_db = run_as_background_process(
-                "_schedule_db_events", self.server_name, _schedule_db_events
+            self._initialized_from_db = self.hs.run_as_background_process(
+                "_schedule_db_events", _schedule_db_events
             )
         else:
             self._repl_client = ReplicationAddedDelayedEventRestServlet.make_client(hs)
@@ -145,9 +145,7 @@ class DelayedEventsHandler:
             finally:
                 self._event_processing = False
 
-        run_as_background_process(
-            "delayed_events.notify_new_event", self.server_name, process
-        )
+        self.hs.run_as_background_process("delayed_events.notify_new_event", process)
 
     async def _unsafe_process_new_event(self) -> None:
         # We purposefully fetch the current max room stream ordering before
@@ -542,9 +540,8 @@ class DelayedEventsHandler:
         if self._next_delayed_event_call is None:
             self._next_delayed_event_call = self._clock.call_later(
                 delay_sec,
-                run_as_background_process,
+                self.hs.run_as_background_process,
                 "_send_on_timeout",
-                self.server_name,
                 self._send_on_timeout,
             )
         else:

synapse/handlers/device.py

@@ -47,7 +47,6 @@ from synapse.api.errors import (
 )
 from synapse.logging.opentracing import log_kv, set_tag, trace
 from synapse.metrics.background_process_metrics import (
-    run_as_background_process,
     wrap_as_background_process,
 )
 from synapse.replication.http.devices import (
@@ -191,10 +190,9 @@ class DeviceHandler:
             and self._delete_stale_devices_after is not None
         ):
             self.clock.looping_call(
-                run_as_background_process,
+                self.hs.run_as_background_process,
                 DELETE_STALE_DEVICES_INTERVAL_MS,
                 desc="delete_stale_devices",
-                server_name=self.server_name,
                 func=self._delete_stale_devices,
             )
 
@@ -1444,14 +1442,19 @@ class DeviceListUpdater(DeviceListWorkerUpdater):
     def __init__(self, hs: "HomeServer", device_handler: DeviceWriterHandler):
         super().__init__(hs)
 
+        self.hs = hs
         self.server_name = hs.hostname
         self.federation = hs.get_federation_client()
         self.server_name = hs.hostname  # nb must be called this for @measure_func
         self.clock = hs.get_clock()  # nb must be called this for @measure_func
         self.device_handler = device_handler
 
-        self._remote_edu_linearizer = Linearizer(name="remote_device_list")
-        self._resync_linearizer = Linearizer(name="remote_device_resync")
+        self._remote_edu_linearizer = Linearizer(
+            name="remote_device_list", clock=self.clock
+        )
+        self._resync_linearizer = Linearizer(
+            name="remote_device_resync", clock=self.clock
+        )
 
         # user_id -> list of updates waiting to be handled.
         self._pending_updates: Dict[
@@ -1464,6 +1467,7 @@ class DeviceListUpdater(DeviceListWorkerUpdater):
         self._seen_updates: ExpiringCache[str, Set[str]] = ExpiringCache(
             cache_name="device_update_edu",
             server_name=self.server_name,
+            hs=self.hs,
             clock=self.clock,
             max_len=10000,
             expiry_ms=30 * 60 * 1000,
@@ -1473,9 +1477,8 @@ class DeviceListUpdater(DeviceListWorkerUpdater):
         # Attempt to resync out of sync device lists every 30s.
         self._resync_retry_lock = Lock()
         self.clock.looping_call(
-            run_as_background_process,
+            self.hs.run_as_background_process,
             30 * 1000,
-            server_name=self.server_name,
             func=self._maybe_retry_device_resync,
             desc="_maybe_retry_device_resync",
         )
@@ -1595,9 +1598,8 @@ class DeviceListUpdater(DeviceListWorkerUpdater):
             if resync:
                 # We mark as stale up front in case we get restarted.
                 await self.store.mark_remote_users_device_caches_as_stale([user_id])
-                run_as_background_process(
+                self.hs.run_as_background_process(
                     "_maybe_retry_device_resync",
-                    self.server_name,
                     self.multi_user_device_resync,
                     [user_id],
                     False,

synapse/handlers/e2e_keys.py

@@ -112,8 +112,7 @@ class E2eKeysHandler:
 
         # Limit the number of in-flight requests from a single device.
         self._query_devices_linearizer = Linearizer(
-            name="query_devices",
-            max_count=10,
+            name="query_devices", max_count=10, clock=hs.get_clock()
         )
 
         self._query_appservices_for_otks = (
@@ -1765,7 +1764,9 @@ class SigningKeyEduUpdater:
         assert isinstance(device_handler, DeviceWriterHandler)
         self._device_handler = device_handler
 
-        self._remote_edu_linearizer = Linearizer(name="remote_signing_key")
+        self._remote_edu_linearizer = Linearizer(
+            name="remote_signing_key", clock=self.clock
+        )
 
         # user_id -> list of updates waiting to be handled.
         self._pending_updates: Dict[str, List[Tuple[JsonDict, JsonDict]]] = {}

synapse/handlers/federation.py

@@ -72,7 +72,6 @@ from synapse.http.servlet import assert_params_in_dict
 from synapse.logging.context import nested_logging_context
 from synapse.logging.opentracing import SynapseTags, set_tag, tag_args, trace
 from synapse.metrics import SERVER_NAME_LABEL
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.module_api import NOT_SPAM
 from synapse.storage.databases.main.events_worker import EventRedactBehaviour
 from synapse.storage.invite_rule import InviteRule
@@ -160,7 +159,7 @@ class FederationHandler:
         self._notifier = hs.get_notifier()
         self._worker_locks = hs.get_worker_locks_handler()
 
-        self._room_backfill = Linearizer("room_backfill")
+        self._room_backfill = Linearizer(name="room_backfill", clock=self.clock)
 
         self._third_party_event_rules = (
             hs.get_module_api_callbacks().third_party_event_rules
@@ -180,16 +179,16 @@ class FederationHandler:
         # When the lock is held for a given room, no other concurrent code may
         # partial state or un-partial state the room.
         self._is_partial_state_room_linearizer = Linearizer(
-            name="_is_partial_state_room_linearizer"
+            name="_is_partial_state_room_linearizer",
+            clock=self.clock,
         )
 
         # if this is the main process, fire off a background process to resume
         # any partial-state-resync operations which were in flight when we
         # were shut down.
         if not hs.config.worker.worker_app:
-            run_as_background_process(
+            self.hs.run_as_background_process(
                 "resume_sync_partial_state_room",
-                self.server_name,
                 self._resume_partial_state_room_sync,
             )
 
@@ -317,9 +316,8 @@ class FederationHandler:
             logger.debug(
                 "_maybe_backfill_inner: all backfill points are *after* current depth. Trying again with later backfill points."
             )
-            run_as_background_process(
+            self.hs.run_as_background_process(
                 "_maybe_backfill_inner_anyway_with_max_depth",
-                self.server_name,
                 self.maybe_backfill,
                 room_id=room_id,
                 # We use `MAX_DEPTH` so that we find all backfill points next
@@ -801,9 +799,8 @@ class FederationHandler:
             # lots of requests for missing prev_events which we do actually
             # have. Hence we fire off the background task, but don't wait for it.
 
-            run_as_background_process(
+            self.hs.run_as_background_process(
                 "handle_queued_pdus",
-                self.server_name,
                 self._handle_queued_pdus,
                 room_queue,
             )
@@ -1876,9 +1873,8 @@ class FederationHandler:
                             room_id=room_id,
                         )
 
-        run_as_background_process(
+        self.hs.run_as_background_process(
             desc="sync_partial_state_room",
-            server_name=self.server_name,
             func=_sync_partial_state_room_wrapper,
         )
 

synapse/handlers/federation_event.py

@@ -81,7 +81,6 @@ from synapse.logging.opentracing import (
     trace,
 )
 from synapse.metrics import SERVER_NAME_LABEL
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.replication.http.federation import (
     ReplicationFederationSendEventsRestServlet,
 )
@@ -153,6 +152,7 @@ class FederationEventHandler:
 
     def __init__(self, hs: "HomeServer"):
         self.server_name = hs.hostname
+        self.hs = hs
         self._clock = hs.get_clock()
         self._store = hs.get_datastores().main
         self._state_store = hs.get_datastores().state
@@ -175,6 +175,7 @@ class FederationEventHandler:
         )
         self._notifier = hs.get_notifier()
 
+        self._server_name = hs.hostname
         self._is_mine_id = hs.is_mine_id
         self._is_mine_server_name = hs.is_mine_server_name
         self._instance_name = hs.get_instance_name()
@@ -191,7 +192,7 @@ class FederationEventHandler:
         # federation event staging area.
         self.room_queues: Dict[str, List[Tuple[EventBase, str]]] = {}
 
-        self._room_pdu_linearizer = Linearizer("fed_room_pdu")
+        self._room_pdu_linearizer = Linearizer(name="fed_room_pdu", clock=self._clock)
 
     async def on_receive_pdu(self, origin: str, pdu: EventBase) -> None:
         """Process a PDU received via a federation /send/ transaction
@@ -974,9 +975,8 @@ class FederationEventHandler:
         # Process previously failed backfill events in the background to not waste
         # time on something that is likely to fail again.
         if len(events_with_failed_pull_attempts) > 0:
-            run_as_background_process(
+            self.hs.run_as_background_process(
                 "_process_new_pulled_events_with_failed_pull_attempts",
-                self.server_name,
                 _process_new_pulled_events,
                 events_with_failed_pull_attempts,
             )
@@ -1568,9 +1568,8 @@ class FederationEventHandler:
                     resync = True
 
             if resync:
-                run_as_background_process(
+                self.hs.run_as_background_process(
                     "resync_device_due_to_pdu",
-                    self.server_name,
                     self._resync_device,
                     event.sender,
                 )

synapse/handlers/message.py

@@ -67,7 +67,6 @@ from synapse.handlers.directory import DirectoryHandler
 from synapse.handlers.worker_lock import NEW_EVENT_DURING_PURGE_LOCK_NAME
 from synapse.logging import opentracing
 from synapse.logging.context import make_deferred_yieldable, run_in_background
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.replication.http.send_events import ReplicationSendEventsRestServlet
 from synapse.storage.databases.main.events_worker import EventRedactBehaviour
 from synapse.types import (
@@ -99,6 +98,7 @@ class MessageHandler:
 
     def __init__(self, hs: "HomeServer"):
         self.server_name = hs.hostname
+        self.hs = hs
         self.auth = hs.get_auth()
         self.clock = hs.get_clock()
         self.state = hs.get_state_handler()
@@ -113,8 +113,8 @@ class MessageHandler:
         self._scheduled_expiry: Optional[IDelayedCall] = None
 
         if not hs.config.worker.worker_app:
-            run_as_background_process(
-                "_schedule_next_expiry", self.server_name, self._schedule_next_expiry
+            self.hs.run_as_background_process(
+                "_schedule_next_expiry", self._schedule_next_expiry
             )
 
     async def get_room_data(
@@ -444,9 +444,8 @@ class MessageHandler:
 
         self._scheduled_expiry = self.clock.call_later(
             delay,
-            run_as_background_process,
+            self.hs.run_as_background_process,
             "_expire_event",
-            self.server_name,
             self._expire_event,
             event_id,
         )
@@ -513,7 +512,9 @@ class EventCreationHandler:
 
         # We limit concurrent event creation for a room to 1. This prevents state resolution
         # from occurring when sending bursts of events to a local room
-        self.limiter = Linearizer(max_count=1, name="room_event_creation_limit")
+        self.limiter = Linearizer(
+            max_count=1, name="room_event_creation_limit", clock=self.clock
+        )
 
         self._bulk_push_rule_evaluator = hs.get_bulk_push_rule_evaluator()
 
@@ -546,9 +547,8 @@ class EventCreationHandler:
             and self.config.server.cleanup_extremities_with_dummy_events
         ):
             self.clock.looping_call(
-                lambda: run_as_background_process(
+                lambda: self.hs.run_as_background_process(
                     "send_dummy_events_to_fill_extremities",
-                    self.server_name,
                     self._send_dummy_events_to_fill_extremities,
                 ),
                 5 * 60 * 1000,
@@ -568,6 +568,7 @@ class EventCreationHandler:
             self._external_cache_joined_hosts_updates = ExpiringCache(
                 cache_name="_external_cache_joined_hosts_updates",
                 server_name=self.server_name,
+                hs=self.hs,
                 clock=self.clock,
                 expiry_ms=30 * 60 * 1000,
             )
@@ -1138,6 +1139,12 @@ class EventCreationHandler:
                 assert self.hs.is_mine_id(event.sender), "User must be our own: %s" % (
                     event.sender,
                 )
+                # if this room uses a policy server, try to get a signature now.
+                # We use verify=False here as we are about to call is_event_allowed on the same event
+                # which will do sig checks.
+                await self._policy_handler.ask_policy_server_to_sign_event(
+                    event, verify=False
+                )
 
                 policy_allowed = await self._policy_handler.is_event_allowed(event)
                 if not policy_allowed:
@@ -2105,9 +2112,8 @@ class EventCreationHandler:
             if event.type == EventTypes.Message:
                 # We don't want to block sending messages on any presence code. This
                 # matters as sometimes presence code can take a while.
-                run_as_background_process(
+                self.hs.run_as_background_process(
                     "bump_presence_active_time",
-                    self.server_name,
                     self._bump_active_time,
                     requester.user,
                     requester.device_id,

synapse/handlers/pagination.py

@@ -29,7 +29,6 @@ from synapse.api.filtering import Filter
 from synapse.events.utils import SerializeEventConfig
 from synapse.handlers.worker_lock import NEW_EVENT_DURING_PURGE_LOCK_NAME
 from synapse.logging.opentracing import trace
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.rest.admin._base import assert_user_is_admin
 from synapse.streams.config import PaginationConfig
 from synapse.types import (
@@ -116,10 +115,9 @@ class PaginationHandler:
                 logger.info("Setting up purge job with config: %s", job)
 
                 self.clock.looping_call(
-                    run_as_background_process,
+                    self.hs.run_as_background_process,
                     job.interval,
                     "purge_history_for_rooms_in_range",
-                    self.server_name,
                     self.purge_history_for_rooms_in_range,
                     job.shortest_max_lifetime,
                     job.longest_max_lifetime,
@@ -244,9 +242,8 @@ class PaginationHandler:
             # We want to purge everything, including local events, and to run the purge in
             # the background so that it's not blocking any other operation apart from
             # other purges in the same room.
-            run_as_background_process(
+            self.hs.run_as_background_process(
                 PURGE_HISTORY_ACTION_NAME,
-                self.server_name,
                 self.purge_history,
                 room_id,
                 token,
@@ -604,9 +601,8 @@ class PaginationHandler:
                 # Otherwise, we can backfill in the background for eventual
                 # consistency's sake but we don't need to block the client waiting
                 # for a costly federation call and processing.
-                run_as_background_process(
+                self.hs.run_as_background_process(
                     "maybe_backfill_in_the_background",
-                    self.server_name,
                     self.hs.get_federation_handler().maybe_backfill,
                     room_id,
                     curr_topo,

synapse/handlers/presence.py

@@ -107,7 +107,6 @@ from synapse.events.presence_router import PresenceRouter
 from synapse.logging.context import run_in_background
 from synapse.metrics import SERVER_NAME_LABEL, LaterGauge
 from synapse.metrics.background_process_metrics import (
-    run_as_background_process,
     wrap_as_background_process,
 )
 from synapse.replication.http.presence import (
@@ -537,19 +536,15 @@ class WorkerPresenceHandler(BasePresenceHandler):
         self._bump_active_client = ReplicationBumpPresenceActiveTime.make_client(hs)
         self._set_state_client = ReplicationPresenceSetState.make_client(hs)
 
-        self._send_stop_syncing_loop = self.clock.looping_call(
-            self.send_stop_syncing, UPDATE_SYNCING_USERS_MS
-        )
-
-        hs.get_clock().add_system_event_trigger(
-            "before",
-            "shutdown",
-            run_as_background_process,
-            "generic_presence.on_shutdown",
-            self.server_name,
-            self._on_shutdown,
+        self.clock.looping_call(self.send_stop_syncing, UPDATE_SYNCING_USERS_MS)
+
+        hs.register_async_shutdown_handler(
+            phase="before",
+            eventType="shutdown",
+            shutdown_func=self._on_shutdown,
         )
 
+    @wrap_as_background_process("WorkerPresenceHandler._on_shutdown")
     async def _on_shutdown(self) -> None:
         if self._track_presence:
             self.hs.get_replication_command_handler().send_command(
@@ -779,9 +774,7 @@ class WorkerPresenceHandler(BasePresenceHandler):
 class PresenceHandler(BasePresenceHandler):
     def __init__(self, hs: "HomeServer"):
         super().__init__(hs)
-        self.server_name = (
-            hs.hostname
-        )  # nb must be called this for @wrap_as_background_process
+        self.server_name = hs.hostname
         self.wheel_timer: WheelTimer[str] = WheelTimer()
         self.notifier = hs.get_notifier()
 
@@ -842,13 +835,10 @@ class PresenceHandler(BasePresenceHandler):
         # have not yet been persisted
         self.unpersisted_users_changes: Set[str] = set()
 
-        hs.get_clock().add_system_event_trigger(
-            "before",
-            "shutdown",
-            run_as_background_process,
-            "presence.on_shutdown",
-            self.server_name,
-            self._on_shutdown,
+        hs.register_async_shutdown_handler(
+            phase="before",
+            eventType="shutdown",
+            shutdown_func=self._on_shutdown,
         )
 
         # Keeps track of the number of *ongoing* syncs on this process. While
@@ -872,14 +862,19 @@ class PresenceHandler(BasePresenceHandler):
         ] = {}
         self.external_process_last_updated_ms: Dict[str, int] = {}
 
-        self.external_sync_linearizer = Linearizer(name="external_sync_linearizer")
+        self.external_sync_linearizer = Linearizer(
+            name="external_sync_linearizer", clock=self.clock
+        )
 
         if self._track_presence:
             # Start a LoopingCall in 30s that fires every 5s.
             # The initial delay is to allow disconnected clients a chance to
             # reconnect before we treat them as offline.
             self.clock.call_later(
-                30, self.clock.looping_call, self._handle_timeouts, 5000
+                30,
+                self.clock.looping_call,
+                self._handle_timeouts,
+                5000,
             )
 
         # Presence information is persisted, whether or not it is being tracked
@@ -906,6 +901,7 @@ class PresenceHandler(BasePresenceHandler):
         self._event_pos = self.store.get_room_max_stream_ordering()
         self._event_processing = False
 
+    @wrap_as_background_process("PresenceHandler._on_shutdown")
     async def _on_shutdown(self) -> None:
         """Gets called when shutting down. This lets us persist any updates that
         we haven't yet persisted, e.g. updates that only changes some internal
@@ -1537,8 +1533,8 @@ class PresenceHandler(BasePresenceHandler):
             finally:
                 self._event_processing = False
 
-        run_as_background_process(
-            "presence.notify_new_event", self.server_name, _process_presence
+        self.hs.run_as_background_process(
+            "presence.notify_new_event", _process_presence
         )
 
     async def _unsafe_process(self) -> None:

synapse/handlers/profile.py

@@ -56,8 +56,8 @@ class ProfileHandler:
 
     def __init__(self, hs: "HomeServer"):
         self.server_name = hs.hostname  # nb must be called this for @cached
+        self.clock = hs.get_clock()  # nb must be called this for @cached
         self.store = hs.get_datastores().main
-        self.clock = hs.get_clock()
         self.hs = hs
 
         self.federation = hs.get_federation_client()

synapse/handlers/read_marker.py

@@ -36,7 +36,9 @@ class ReadMarkerHandler:
     def __init__(self, hs: "HomeServer"):
         self.store = hs.get_datastores().main
         self.account_data_handler = hs.get_account_data_handler()
-        self.read_marker_linearizer = Linearizer(name="read_marker")
+        self.read_marker_linearizer = Linearizer(
+            name="read_marker", clock=hs.get_clock()
+        )
 
     async def received_client_read_marker(
         self, room_id: str, user_id: str, event_id: str

synapse/handlers/register.py

@@ -23,7 +23,14 @@
 """Contains functions for registering clients."""
 
 import logging
-from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple, TypedDict
+from typing import (
+    TYPE_CHECKING,
+    Iterable,
+    List,
+    Optional,
+    Tuple,
+    TypedDict,
+)
 
 from prometheus_client import Counter
 

synapse/handlers/room.py

@@ -597,7 +597,7 @@ class RoomCreationHandler:
                 new_room_version,
                 additional_creators=additional_creators,
             )
-        initial_state = {}
+        initial_state: MutableStateMap = {}
 
         # Replicate relevant room events
         types_to_copy: List[Tuple[str, Optional[str]]] = [
@@ -693,14 +693,23 @@ class RoomCreationHandler:
                 additional_creators,
             )
 
-        # We construct what the body of a call to /createRoom would look like for passing
-        # to the spam checker. We don't include a preset here, as we expect the
+        # We construct a subset of what the body of a call to /createRoom would look like
+        # for passing to the spam checker. We don't include a preset here, as we expect the
         # initial state to contain everything we need.
+        # TODO: given we are upgrading, it would make sense to pass the room_version
+        # TODO: the preset might be useful too
         spam_check = await self._spam_checker_module_callbacks.user_may_create_room(
             user_id,
             {
                 "creation_content": creation_content,
-                "initial_state": list(initial_state.items()),
+                "initial_state": [
+                    {
+                        "type": state_key[0],
+                        "state_key": state_key[1],
+                        "content": event_content,
+                    }
+                    for state_key, event_content in initial_state.items()
+                ],
             },
         )
         if spam_check != self._spam_checker_module_callbacks.NOT_SPAM:

synapse/handlers/room_member.py

@@ -50,7 +50,6 @@ from synapse.handlers.state_deltas import MatchChange, StateDeltasHandler
 from synapse.handlers.worker_lock import NEW_EVENT_DURING_PURGE_LOCK_NAME
 from synapse.logging import opentracing
 from synapse.metrics import SERVER_NAME_LABEL, event_processing_positions
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.replication.http.push import ReplicationCopyPusherRestServlet
 from synapse.storage.databases.main.state_deltas import StateDelta
 from synapse.storage.invite_rule import InviteRule
@@ -114,8 +113,12 @@ class RoomMemberHandler(metaclass=abc.ABCMeta):
         if self.hs.config.server.include_profile_data_on_invite:
             self._membership_types_to_include_profile_data_in.add(Membership.INVITE)
 
-        self.member_linearizer: Linearizer = Linearizer(name="member")
-        self.member_as_limiter = Linearizer(max_count=10, name="member_as_limiter")
+        self.member_linearizer: Linearizer = Linearizer(
+            name="member", clock=hs.get_clock()
+        )
+        self.member_as_limiter = Linearizer(
+            max_count=10, name="member_as_limiter", clock=hs.get_clock()
+        )
 
         self.clock = hs.get_clock()
         self._spam_checker_module_callbacks = hs.get_module_api_callbacks().spam_checker
@@ -2186,7 +2189,10 @@ class RoomForgetterHandler(StateDeltasHandler):
             self._notifier.add_replication_callback(self.notify_new_event)
 
             # We kick this off to pick up outstanding work from before the last restart.
-            self._clock.call_later(0, self.notify_new_event)
+            self._clock.call_later(
+                0,
+                self.notify_new_event,
+            )
 
     def notify_new_event(self) -> None:
         """Called when there may be more deltas to process"""
@@ -2201,9 +2207,7 @@ class RoomForgetterHandler(StateDeltasHandler):
             finally:
                 self._is_processing = False
 
-        run_as_background_process(
-            "room_forgetter.notify_new_event", self.server_name, process
-        )
+        self._hs.run_as_background_process("room_forgetter.notify_new_event", process)
 
     async def _unsafe_process(self) -> None:
         # If self.pos is None then means we haven't fetched it from DB

synapse/handlers/room_policy.py

@@ -17,6 +17,11 @@
 import logging
 from typing import TYPE_CHECKING
 
+from signedjson.key import decode_verify_key_bytes
+from unpaddedbase64 import decode_base64
+
+from synapse.api.errors import SynapseError
+from synapse.crypto.keyring import VerifyJsonRequest
 from synapse.events import EventBase
 from synapse.types.handlers.policy_server import RECOMMENDATION_OK
 from synapse.util.stringutils import parse_and_validate_server_name
@@ -26,6 +31,9 @@ if TYPE_CHECKING:
 
 logger = logging.getLogger(__name__)
 
+POLICY_SERVER_EVENT_TYPE = "org.matrix.msc4284.policy"
+POLICY_SERVER_KEY_ID = "ed25519:policy_server"
+
 
 class RoomPolicyHandler:
     def __init__(self, hs: "HomeServer"):
@@ -54,11 +62,11 @@ class RoomPolicyHandler:
         Returns:
             bool: True if the event is allowed in the room, False otherwise.
         """
-        if event.type == "org.matrix.msc4284.policy" and event.state_key is not None:
+        if event.type == POLICY_SERVER_EVENT_TYPE and event.state_key is not None:
             return True  # always allow policy server change events
 
         policy_event = await self._storage_controllers.state.get_current_state_event(
-            event.room_id, "org.matrix.msc4284.policy", ""
+            event.room_id, POLICY_SERVER_EVENT_TYPE, ""
         )
         if not policy_event:
             return True  # no policy server == default allow
@@ -81,6 +89,22 @@ class RoomPolicyHandler:
         if not is_in_room:
             return True  # policy server not in room == default allow
 
+        # Check if the event has been signed with the public key in the policy server state event.
+        # If it is, we can save an HTTP hit.
+        # We actually want to get the policy server state event BEFORE THE EVENT rather than
+        # the current state value, else changing the public key will cause all of these checks to fail.
+        # However, if we are checking outlier events (which we will due to is_event_allowed being called
+        # near the edges at _check_sigs_and_hash) we won't know the state before the event, so the
+        # only safe option is to use the current state
+        public_key = policy_event.content.get("public_key", None)
+        if public_key is not None and isinstance(public_key, str):
+            valid = await self._verify_policy_server_signature(
+                event, policy_server, public_key
+            )
+            if valid:
+                return True
+            # fallthrough to hit /check manually
+
         # At this point, the server appears valid and is in the room, so ask it to check
         # the event.
         recommendation = await self._federation_client.get_pdu_policy_recommendation(
@@ -90,3 +114,73 @@ class RoomPolicyHandler:
             return False
 
         return True  # default allow
+
+    async def _verify_policy_server_signature(
+        self, event: EventBase, policy_server: str, public_key: str
+    ) -> bool:
+        # check the event is signed with this (via, public_key).
+        verify_json_req = VerifyJsonRequest.from_event(policy_server, event, 0)
+        try:
+            key_bytes = decode_base64(public_key)
+            verify_key = decode_verify_key_bytes(POLICY_SERVER_KEY_ID, key_bytes)
+            # We would normally use KeyRing.verify_event_for_server but we can't here as we don't
+            # want to fetch the server key, and instead want to use the public key in the state event.
+            await self._hs.get_keyring().process_json(verify_key, verify_json_req)
+            # if the event is correctly signed by the public key in the policy server state event = Allow
+            return True
+        except Exception as ex:
+            logger.warning(
+                "failed to verify event using public key in policy server event: %s", ex
+            )
+        return False
+
+    async def ask_policy_server_to_sign_event(
+        self, event: EventBase, verify: bool = False
+    ) -> None:
+        """Ask the policy server to sign this event. The signature is added to the event signatures block.
+
+        Does nothing if there is no policy server state event in the room. If the policy server
+        refuses to sign the event (as it's marked as spam) does nothing.
+
+        Args:
+            event: The event to sign
+            verify: If True, verify that the signature is correctly signed by the public_key in the
+            policy server state event.
+        Raises:
+            if verify=True and the policy server signed the event with an invalid signature. Does
+            not raise if the policy server refuses to sign the event.
+        """
+        policy_event = await self._storage_controllers.state.get_current_state_event(
+            event.room_id, POLICY_SERVER_EVENT_TYPE, ""
+        )
+        if not policy_event:
+            return
+        policy_server = policy_event.content.get("via", None)
+        if policy_server is None or not isinstance(policy_server, str):
+            return
+        # Only ask to sign events if the policy state event has a public_key (so they can be subsequently verified)
+        public_key = policy_event.content.get("public_key", None)
+        if public_key is None or not isinstance(public_key, str):
+            return
+
+        # Ask the policy server to sign this event.
+        # We set a smallish timeout here as we don't want to block event sending too long.
+        signature = await self._federation_client.ask_policy_server_to_sign_event(
+            policy_server,
+            event,
+            timeout=3000,
+        )
+        if (
+            # the policy server returns {} if it refuses to sign the event.
+            signature and len(signature) > 0
+        ):
+            event.signatures.update(signature)
+            if verify:
+                is_valid = await self._verify_policy_server_signature(
+                    event, policy_server, public_key
+                )
+                if not is_valid:
+                    raise SynapseError(
+                        500,
+                        f"policy server {policy_server} failed to sign event correctly",
+                    )

synapse/handlers/sso.py

@@ -224,7 +224,7 @@ class SsoHandler:
         )
 
         # a lock on the mappings
-        self._mapping_lock = Linearizer(name="sso_user_mapping", clock=hs.get_clock())
+        self._mapping_lock = Linearizer(clock=hs.get_clock(), name="sso_user_mapping")
 
         # a map from session id to session data
         self._username_mapping_sessions: Dict[str, UsernameMappingSession] = {}

synapse/handlers/stats.py

@@ -33,7 +33,6 @@ from typing import (
 
 from synapse.api.constants import EventContentFields, EventTypes, Membership
 from synapse.metrics import SERVER_NAME_LABEL, event_processing_positions
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.storage.databases.main.state_deltas import StateDelta
 from synapse.types import JsonDict
 from synapse.util.events import get_plain_text_topic_from_event_content
@@ -75,7 +74,10 @@ class StatsHandler:
 
             # We kick this off so that we don't have to wait for a change before
             # we start populating stats
-            self.clock.call_later(0, self.notify_new_event)
+            self.clock.call_later(
+                0,
+                self.notify_new_event,
+            )
 
     def notify_new_event(self) -> None:
         """Called when there may be more deltas to process"""
@@ -90,7 +92,7 @@ class StatsHandler:
             finally:
                 self._is_processing = False
 
-        run_as_background_process("stats.notify_new_event", self.server_name, process)
+        self.hs.run_as_background_process("stats.notify_new_event", process)
 
     async def _unsafe_process(self) -> None:
         # If self.pos is None then means we haven't fetched it from DB

synapse/handlers/sync.py

@@ -323,6 +323,7 @@ class SyncHandler:
         ] = ExpiringCache(
             cache_name="lazy_loaded_members_cache",
             server_name=self.server_name,
+            hs=hs,
             clock=self.clock,
             max_len=0,
             expiry_ms=LAZY_LOADED_MEMBERS_CACHE_MAX_AGE,
@@ -980,7 +981,11 @@ class SyncHandler:
         )
         if cache is None:
             logger.debug("creating LruCache for %r", cache_key)
-            cache = LruCache(max_size=LAZY_LOADED_MEMBERS_CACHE_MAX_SIZE)
+            cache = LruCache(
+                max_size=LAZY_LOADED_MEMBERS_CACHE_MAX_SIZE,
+                clock=self.clock,
+                server_name=self.server_name,
+            )
             self.lazy_loaded_members_cache[cache_key] = cache
         else:
             logger.debug("found LruCache for %r", cache_key)

synapse/handlers/typing.py

@@ -28,7 +28,6 @@ from synapse.api.constants import EduTypes
 from synapse.api.errors import AuthError, ShadowBanError, SynapseError
 from synapse.appservice import ApplicationService
 from synapse.metrics.background_process_metrics import (
-    run_as_background_process,
     wrap_as_background_process,
 )
 from synapse.replication.tcp.streams import TypingStream
@@ -78,11 +77,10 @@ class FollowerTypingHandler:
     """
 
     def __init__(self, hs: "HomeServer"):
+        self.hs = hs  # nb must be called this for @wrap_as_background_process
         self.store = hs.get_datastores().main
         self._storage_controllers = hs.get_storage_controllers()
-        self.server_name = (
-            hs.hostname
-        )  # nb must be called this for @wrap_as_background_process
+        self.server_name = hs.hostname
         self.clock = hs.get_clock()
         self.is_mine_id = hs.is_mine_id
         self.is_mine_server_name = hs.is_mine_server_name
@@ -144,9 +142,8 @@ class FollowerTypingHandler:
         if self.federation and self.is_mine_id(member.user_id):
             last_fed_poke = self._member_last_federation_poke.get(member, None)
             if not last_fed_poke or last_fed_poke + FEDERATION_PING_INTERVAL <= now:
-                run_as_background_process(
+                self.hs.run_as_background_process(
                     "typing._push_remote",
-                    self.server_name,
                     self._push_remote,
                     member=member,
                     typing=True,
@@ -220,9 +217,8 @@ class FollowerTypingHandler:
             self._rooms_updated.add(row.room_id)
 
             if self.federation:
-                run_as_background_process(
+                self.hs.run_as_background_process(
                     "_send_changes_in_typing_to_remotes",
-                    self.server_name,
                     self._send_changes_in_typing_to_remotes,
                     row.room_id,
                     prev_typing,
@@ -384,9 +380,8 @@ class TypingWriterHandler(FollowerTypingHandler):
     def _push_update(self, member: RoomMember, typing: bool) -> None:
         if self.hs.is_mine_id(member.user_id):
             # Only send updates for changes to our own users.
-            run_as_background_process(
+            self.hs.run_as_background_process(
                 "typing._push_remote",
-                self.server_name,
                 self._push_remote,
                 member,
                 typing,

synapse/handlers/user_directory.py

@@ -36,7 +36,6 @@ from synapse.api.constants import (
 from synapse.api.errors import Codes, SynapseError
 from synapse.handlers.state_deltas import MatchChange, StateDeltasHandler
 from synapse.metrics import SERVER_NAME_LABEL
-from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.storage.databases.main.state_deltas import StateDelta
 from synapse.storage.databases.main.user_directory import SearchResult
 from synapse.storage.roommember import ProfileInfo
@@ -137,11 +136,15 @@ class UserDirectoryHandler(StateDeltasHandler):
 
             # We kick this off so that we don't have to wait for a change before
             # we start populating the user directory
-            self.clock.call_later(0, self.notify_new_event)
+            self.clock.call_later(
+                0,
+                self.notify_new_event,
+            )
 
             # Kick off the profile refresh process on startup
             self._refresh_remote_profiles_call_later = self.clock.call_later(
-                10, self.kick_off_remote_profile_refresh_process
+                10,
+                self.kick_off_remote_profile_refresh_process,
             )
 
     async def search_users(
@@ -193,9 +196,7 @@ class UserDirectoryHandler(StateDeltasHandler):
                 self._is_processing = False
 
         self._is_processing = True
-        run_as_background_process(
-            "user_directory.notify_new_event", self.server_name, process
-        )
+        self._hs.run_as_background_process("user_directory.notify_new_event", process)
 
     async def handle_local_profile_change(
         self, user_id: str, profile: ProfileInfo
@@ -609,8 +610,8 @@ class UserDirectoryHandler(StateDeltasHandler):
                 self._is_refreshing_remote_profiles = False
 
         self._is_refreshing_remote_profiles = True
-        run_as_background_process(
-            "user_directory.refresh_remote_profiles", self.server_name, process
+        self._hs.run_as_background_process(
+            "user_directory.refresh_remote_profiles", process
         )
 
     async def _unsafe_refresh_remote_profiles(self) -> None:
@@ -655,8 +656,9 @@ class UserDirectoryHandler(StateDeltasHandler):
                 if not users:
                     return
                 _, _, next_try_at_ts = users[0]
+                delay = ((next_try_at_ts - self.clock.time_msec()) // 1000) + 2
                 self._refresh_remote_profiles_call_later = self.clock.call_later(
-                    ((next_try_at_ts - self.clock.time_msec()) // 1000) + 2,
+                    delay,
                     self.kick_off_remote_profile_refresh_process,
                 )
 
@@ -692,9 +694,8 @@ class UserDirectoryHandler(StateDeltasHandler):
                 self._is_refreshing_remote_profiles_for_servers.remove(server_name)
 
         self._is_refreshing_remote_profiles_for_servers.add(server_name)
-        run_as_background_process(
+        self._hs.run_as_background_process(
             "user_directory.refresh_remote_profiles_for_remote_server",
-            self.server_name,
             process,
         )
 

synapse/handlers/worker_lock.py

@@ -37,13 +37,13 @@ from weakref import WeakSet
 import attr
 
 from twisted.internet import defer
-from twisted.internet.interfaces import IReactorTime
 
 from synapse.logging.context import PreserveLoggingContext
 from synapse.logging.opentracing import start_active_span
 from synapse.metrics.background_process_metrics import wrap_as_background_process
 from synapse.storage.databases.main.lock import Lock, LockStore
 from synapse.util.async_helpers import timeout_deferred
+from synapse.util.clock import Clock
 from synapse.util.constants import ONE_MINUTE_SECONDS
 
 if TYPE_CHECKING:
@@ -66,10 +66,8 @@ class WorkerLocksHandler:
     """
 
     def __init__(self, hs: "HomeServer") -> None:
-        self.server_name = (
-            hs.hostname
-        )  # nb must be called this for @wrap_as_background_process
-        self._reactor = hs.get_reactor()
+        self.hs = hs  # nb must be called this for @wrap_as_background_process
+        self._clock = hs.get_clock()
         self._store = hs.get_datastores().main
         self._clock = hs.get_clock()
         self._notifier = hs.get_notifier()
@@ -98,7 +96,7 @@ class WorkerLocksHandler:
         """
 
         lock = WaitingLock(
-            reactor=self._reactor,
+            clock=self._clock,
             store=self._store,
             handler=self,
             lock_name=lock_name,
@@ -129,7 +127,7 @@ class WorkerLocksHandler:
         """
 
         lock = WaitingLock(
-            reactor=self._reactor,
+            clock=self._clock,
             store=self._store,
             handler=self,
             lock_name=lock_name,
@@ -160,7 +158,7 @@ class WorkerLocksHandler:
         lock = WaitingMultiLock(
             lock_names=lock_names,
             write=write,
-            reactor=self._reactor,
+            clock=self._clock,
             store=self._store,
             handler=self,
         )
@@ -197,7 +195,11 @@ class WorkerLocksHandler:
                 if not deferred.called:
                     deferred.callback(None)
 
-        self._clock.call_later(0, _wake_all_locks, locks)
+        self._clock.call_later(
+            0,
+            _wake_all_locks,
+            locks,
+        )
 
     @wrap_as_background_process("_cleanup_locks")
     async def _cleanup_locks(self) -> None:
@@ -207,7 +209,7 @@ class WorkerLocksHandler:
 
 @attr.s(auto_attribs=True, eq=False)
 class WaitingLock:
-    reactor: IReactorTime
+    clock: Clock
     store: LockStore
     handler: WorkerLocksHandler
     lock_name: str
@@ -246,10 +248,11 @@ class WaitingLock:
                     # periodically wake up in case the lock was released but we
                     # weren't notified.
                     with PreserveLoggingContext():
+                        timeout = self._get_next_retry_interval()
                         await timeout_deferred(
                             deferred=self.deferred,
-                            timeout=self._get_next_retry_interval(),
-                            reactor=self.reactor,
+                            timeout=timeout,
+                            clock=self.clock,
                         )
                 except Exception:
                     pass
@@ -290,7 +293,7 @@ class WaitingMultiLock:
 
     write: bool
 
-    reactor: IReactorTime
+    clock: Clock
     store: LockStore
     handler: WorkerLocksHandler
 
@@ -323,10 +326,11 @@ class WaitingMultiLock:
                     # periodically wake up in case the lock was released but we
                     # weren't notified.
                     with PreserveLoggingContext():
+                        timeout = self._get_next_retry_interval()
                         await timeout_deferred(
                             deferred=self.deferred,
-                            timeout=self._get_next_retry_interval(),
-                            reactor=self.reactor,
+                            timeout=timeout,
+                            clock=self.clock,
                         )
                 except Exception:
                     pass

synapse/http/client.py

@@ -54,7 +54,6 @@ from twisted.internet.interfaces import (
     IOpenSSLContextFactory,
     IReactorCore,
     IReactorPluggableNameResolver,
-    IReactorTime,
     IResolutionReceiver,
     ITCPTransport,
 )
@@ -88,6 +87,7 @@ from synapse.logging.opentracing import set_tag, start_active_span, tags
 from synapse.metrics import SERVER_NAME_LABEL
 from synapse.types import ISynapseReactor, StrSequence
 from synapse.util.async_helpers import timeout_deferred
+from synapse.util.clock import Clock
 from synapse.util.json import json_decoder
 
 if TYPE_CHECKING:
@@ -165,16 +165,17 @@ def _is_ip_blocked(
 _EPSILON = 0.00000001
 
 
-def _make_scheduler(
-    reactor: IReactorTime,
-) -> Callable[[Callable[[], object]], IDelayedCall]:
+def _make_scheduler(clock: Clock) -> Callable[[Callable[[], object]], IDelayedCall]:
     """Makes a schedular suitable for a Cooperator using the given reactor.
 
     (This is effectively just a copy from `twisted.internet.task`)
     """
 
     def _scheduler(x: Callable[[], object]) -> IDelayedCall:
-        return reactor.callLater(_EPSILON, x)
+        return clock.call_later(
+            _EPSILON,
+            x,
+        )
 
     return _scheduler
 
@@ -367,7 +368,7 @@ class BaseHttpClient:
 
         # We use this for our body producers to ensure that they use the correct
         # reactor.
-        self._cooperator = Cooperator(scheduler=_make_scheduler(hs.get_reactor()))
+        self._cooperator = Cooperator(scheduler=_make_scheduler(hs.get_clock()))
 
     async def request(
         self,
@@ -436,9 +437,9 @@ class BaseHttpClient:
                 # we use our own timeout mechanism rather than treq's as a workaround
                 # for https://twistedmatrix.com/trac/ticket/9534.
                 request_deferred = timeout_deferred(
-                    request_deferred,
-                    60,
-                    self.hs.get_reactor(),
+                    deferred=request_deferred,
+                    timeout=60,
+                    clock=self.hs.get_clock(),
                 )
 
                 # turn timeouts into RequestTimedOutErrors
@@ -763,7 +764,11 @@ class BaseHttpClient:
             d = read_body_with_max_size(response, output_stream, max_size)
 
             # Ensure that the body is not read forever.
-            d = timeout_deferred(d, 30, self.hs.get_reactor())
+            d = timeout_deferred(
+                deferred=d,
+                timeout=30,
+                clock=self.hs.get_clock(),
+            )
 
             length = await make_deferred_yieldable(d)
         except BodyExceededMaxSize:
@@ -957,9 +962,9 @@ class ReplicationClient(BaseHttpClient):
                 # for https://twistedmatrix.com/trac/ticket/9534.
                 # (Updated url https://github.com/twisted/twisted/issues/9534)
                 request_deferred = timeout_deferred(
-                    request_deferred,
-                    60,
-                    self.hs.get_reactor(),
+                    deferred=request_deferred,
+                    timeout=60,
+                    clock=self.hs.get_clock(),
                 )
 
                 # turn timeouts into RequestTimedOutErrors

synapse/http/federation/matrix_federation_agent.py

@@ -67,6 +67,9 @@ class MatrixFederationAgent:
     Args:
         reactor: twisted reactor to use for underlying requests
 
+        clock: Internal `HomeServer` clock used to track delayed and looping calls.
+            Should be obtained from `hs.get_clock()`.
+
         tls_client_options_factory:
             factory to use for fetching client tls options, or none to disable TLS.
 
@@ -97,6 +100,7 @@ class MatrixFederationAgent:
         *,
         server_name: str,
         reactor: ISynapseReactor,
+        clock: Clock,
         tls_client_options_factory: Optional[FederationPolicyForHTTPS],
         user_agent: bytes,
         ip_allowlist: Optional[IPSet],
@@ -109,6 +113,7 @@ class MatrixFederationAgent:
         Args:
             server_name: Our homeserver name (used to label metrics) (`hs.hostname`).
             reactor
+            clock: Should be the `hs` clock from `hs.get_clock()`
             tls_client_options_factory
             user_agent
             ip_allowlist
@@ -124,7 +129,6 @@ class MatrixFederationAgent:
         # addresses, to prevent DNS rebinding.
         reactor = BlocklistingReactorWrapper(reactor, ip_allowlist, ip_blocklist)
 
-        self._clock = Clock(reactor)
         self._pool = HTTPConnectionPool(reactor)
         self._pool.retryAutomatically = False
         self._pool.maxPersistentPerHost = 5
@@ -147,6 +151,7 @@ class MatrixFederationAgent:
             _well_known_resolver = WellKnownResolver(
                 server_name=server_name,
                 reactor=reactor,
+                clock=clock,
                 agent=BlocklistingAgentWrapper(
                     ProxyAgent(
                         reactor=reactor,

synapse/http/federation/well_known_resolver.py

@@ -90,6 +90,7 @@ class WellKnownResolver:
         self,
         server_name: str,
         reactor: ISynapseThreadlessReactor,
+        clock: Clock,
         agent: IAgent,
         user_agent: bytes,
         well_known_cache: Optional[TTLCache[bytes, Optional[bytes]]] = None,
@@ -99,6 +100,7 @@ class WellKnownResolver:
         Args:
             server_name: Our homeserver name (used to label metrics) (`hs.hostname`).
             reactor
+            clock: Should be the `hs` clock from `hs.get_clock()`
             agent
             user_agent
             well_known_cache
@@ -107,7 +109,7 @@ class WellKnownResolver:
 
         self.server_name = server_name
         self._reactor = reactor
-        self._clock = Clock(reactor)
+        self._clock = clock
 
         if well_known_cache is None:
             well_known_cache = TTLCache(

synapse/http/matrixfederationclient.py

@@ -90,6 +90,7 @@ from synapse.logging.opentracing import set_tag, start_active_span, tags
 from synapse.metrics import SERVER_NAME_LABEL
 from synapse.types import JsonDict
 from synapse.util.async_helpers import AwakenableSleeper, Linearizer, timeout_deferred
+from synapse.util.clock import Clock
 from synapse.util.json import json_decoder
 from synapse.util.metrics import Measure
 from synapse.util.stringutils import parse_and_validate_server_name
@@ -270,6 +271,7 @@ class LegacyJsonSendParser(_BaseJsonParser[Tuple[int, JsonDict]]):
 
 
 async def _handle_response(
+    clock: Clock,
     reactor: IReactorTime,
     timeout_sec: float,
     request: MatrixFederationRequest,
@@ -299,7 +301,12 @@ async def _handle_response(
         check_content_type_is(response.headers, parser.CONTENT_TYPE)
 
         d = read_body_with_max_size(response, parser, max_response_size)
-        d = timeout_deferred(d, timeout=timeout_sec, reactor=reactor)
+        d = timeout_deferred(
+            deferred=d,
+            timeout=timeout_sec,
+            cancel_on_shutdown=False,  # We don't track this call since it's short
+            clock=clock,
+        )
 
         length = await make_deferred_yieldable(d)
 
@@ -411,6 +418,7 @@ class MatrixFederationHttpClient:
         self.server_name = hs.hostname
 
         self.reactor = hs.get_reactor()
+        self.clock = hs.get_clock()
 
         user_agent = hs.version_string
         if hs.config.server.user_agent_suffix:
@@ -424,6 +432,7 @@ class MatrixFederationHttpClient:
             federation_agent: IAgent = MatrixFederationAgent(
                 server_name=self.server_name,
                 reactor=self.reactor,
+                clock=self.clock,
                 tls_client_options_factory=tls_client_options_factory,
                 user_agent=user_agent.encode("ascii"),
                 ip_allowlist=hs.config.server.federation_ip_range_allowlist,
@@ -457,7 +466,6 @@ class MatrixFederationHttpClient:
             ip_blocklist=hs.config.server.federation_ip_range_blocklist,
         )
 
-        self.clock = hs.get_clock()
         self._store = hs.get_datastores().main
         self.version_string_bytes = hs.version_string.encode("ascii")
         self.default_timeout_seconds = hs.config.federation.client_timeout_ms / 1000
@@ -470,9 +478,9 @@ class MatrixFederationHttpClient:
         self.max_long_retries = hs.config.federation.max_long_retries
         self.max_short_retries = hs.config.federation.max_short_retries
 
-        self._cooperator = Cooperator(scheduler=_make_scheduler(self.reactor))
+        self._cooperator = Cooperator(scheduler=_make_scheduler(self.clock))
 
-        self._sleeper = AwakenableSleeper(self.reactor)
+        self._sleeper = AwakenableSleeper(self.clock)
 
         self._simple_http_client = SimpleHttpClient(
             hs,
@@ -481,7 +489,13 @@ class MatrixFederationHttpClient:
             use_proxy=True,
         )
 
-        self.remote_download_linearizer = Linearizer("remote_download_linearizer", 6)
+        self.remote_download_linearizer = Linearizer(
+            name="remote_download_linearizer", max_count=6, clock=self.clock
+        )
+        self._is_shutdown = False
+
+    def shutdown(self) -> None:
+        self._is_shutdown = True
 
     def wake_destination(self, destination: str) -> None:
         """Called when the remote server may have come back online."""
@@ -627,6 +641,7 @@ class MatrixFederationHttpClient:
         limiter = await synapse.util.retryutils.get_retry_limiter(
             destination=request.destination,
             our_server_name=self.server_name,
+            hs=self.hs,
             clock=self.clock,
             store=self._store,
             backoff_on_404=backoff_on_404,
@@ -673,7 +688,7 @@ class MatrixFederationHttpClient:
                 (b"", b"", path_bytes, None, query_bytes, b"")
             )
 
-            while True:
+            while not self._is_shutdown:
                 try:
                     json = request.get_json()
                     if json:
@@ -731,9 +746,10 @@ class MatrixFederationHttpClient:
                                 bodyProducer=producer,
                             )
                             request_deferred = timeout_deferred(
-                                request_deferred,
+                                deferred=request_deferred,
                                 timeout=_sec_timeout,
-                                reactor=self.reactor,
+                                cancel_on_shutdown=False,  # We don't track this call since it will typically be short
+                                clock=self.clock,
                             )
 
                             response = await make_deferred_yieldable(request_deferred)
@@ -791,7 +807,10 @@ class MatrixFederationHttpClient:
                         # Update transactions table?
                         d = treq.content(response)
                         d = timeout_deferred(
-                            d, timeout=_sec_timeout, reactor=self.reactor
+                            deferred=d,
+                            timeout=_sec_timeout,
+                            cancel_on_shutdown=False,  # We don't track this call since it will typically be short
+                            clock=self.clock,
                         )
 
                         try:
@@ -860,6 +879,15 @@ class MatrixFederationHttpClient:
                             delay_seconds,
                         )
 
+                        if self._is_shutdown:
+                            # Immediately fail sending the request instead of starting a
+                            # potentially long sleep after the server has requested
+                            # shutdown.
+                            # This is the code path followed when the
+                            # `federation_transaction_transmission_loop` has been
+                            # cancelled.
+                            raise
+
                         # Sleep for the calculated delay, or wake up immediately
                         # if we get notified that the server is back up.
                         await self._sleeper.sleep(
@@ -1072,6 +1100,7 @@ class MatrixFederationHttpClient:
             parser = cast(ByteParser[T], JsonParser())
 
         body = await _handle_response(
+            self.clock,
             self.reactor,
             _sec_timeout,
             request,
@@ -1150,7 +1179,13 @@ class MatrixFederationHttpClient:
             _sec_timeout = self.default_timeout_seconds
 
         body = await _handle_response(
-            self.reactor, _sec_timeout, request, response, start_ms, parser=JsonParser()
+            self.clock,
+            self.reactor,
+            _sec_timeout,
+            request,
+            response,
+            start_ms,
+            parser=JsonParser(),
         )
         return body
 
@@ -1356,6 +1391,7 @@ class MatrixFederationHttpClient:
             parser = cast(ByteParser[T], JsonParser())
 
         body = await _handle_response(
+            self.clock,
             self.reactor,
             _sec_timeout,
             request,
@@ -1429,7 +1465,13 @@ class MatrixFederationHttpClient:
             _sec_timeout = self.default_timeout_seconds
 
         body = await _handle_response(
-            self.reactor, _sec_timeout, request, response, start_ms, parser=JsonParser()
+            self.clock,
+            self.reactor,
+            _sec_timeout,
+            request,
+            response,
+            start_ms,
+            parser=JsonParser(),
         )
         return body
 

synapse/http/proxy.py

@@ -161,12 +161,13 @@ class ProxyResource(_AsyncResource):
             bodyProducer=QuieterFileBodyProducer(request.content),
         )
         request_deferred = timeout_deferred(
-            request_deferred,
+            deferred=request_deferred,
             # This should be set longer than the timeout in `MatrixFederationHttpClient`
             # so that it has enough time to complete and pass us the data before we give
             # up.
             timeout=90,
-            reactor=self.reactor,
+            cancel_on_shutdown=False,  # We don't track this call since it will typically be short
+            clock=self._clock,
         )
 
         response = await make_deferred_yieldable(request_deferred)

synapse/http/server.py

@@ -411,8 +411,26 @@ class DirectServeJsonResource(_AsyncResource):
         # Clock is optional as this class is exposed to the module API.
         clock: Optional[Clock] = None,
     ):
+        """
+        Args:
+            canonical_json: TODO
+            extract_context: TODO
+            clock: This is expected to be passed in by any Synapse code.
+                Only optional for the Module API.
+        """
+
         if clock is None:
-            clock = Clock(cast(ISynapseThreadlessReactor, reactor))
+            # Ideally we wouldn't ignore the linter error here and instead enforce a
+            # required `Clock` be passed into the `__init__` function.
+            # However, this would change the function signature which is currently being
+            # exported to the module api. Since we don't want to break that api, we have
+            # to settle with ignoring the linter error here.
+            # As of the time of writing this, all Synapse internal usages of
+            # `DirectServeJsonResource` pass in the existing homeserver clock instance.
+            clock = Clock(  # type: ignore[multiple-internal-clocks]
+                cast(ISynapseThreadlessReactor, reactor),
+                server_name="synapse_module_running_from_unknown_server",
+            )
 
         super().__init__(clock, extract_context)
         self.canonical_json = canonical_json
@@ -590,8 +608,24 @@ class DirectServeHtmlResource(_AsyncResource):
         # Clock is optional as this class is exposed to the module API.
         clock: Optional[Clock] = None,
     ):
+        """
+        Args:
+            extract_context: TODO
+            clock: This is expected to be passed in by any Synapse code.
+                Only optional for the Module API.
+        """
         if clock is None:
-            clock = Clock(cast(ISynapseThreadlessReactor, reactor))
+            # Ideally we wouldn't ignore the linter error here and instead enforce a
+            # required `Clock` be passed into the `__init__` function.
+            # However, this would change the function signature which is currently being
+            # exported to the module api. Since we don't want to break that api, we have
+            # to settle with ignoring the linter error here.
+            # As of the time of writing this, all Synapse internal usages of
+            # `DirectServeHtmlResource` pass in the existing homeserver clock instance.
+            clock = Clock(  # type: ignore[multiple-internal-clocks]
+                cast(ISynapseThreadlessReactor, reactor),
+                server_name="synapse_module_running_from_unknown_server",
+            )
 
         super().__init__(clock, extract_context)
 

synapse/http/site.py

@@ -22,7 +22,7 @@ import contextlib
 import logging
 import time
 from http import HTTPStatus
-from typing import TYPE_CHECKING, Any, Generator, Optional, Tuple, Union
+from typing import TYPE_CHECKING, Any, Generator, List, Optional, Tuple, Union
 
 import attr
 from zope.interface import implementer
@@ -30,6 +30,7 @@ from zope.interface import implementer
 from twisted.internet.address import UNIXAddress
 from twisted.internet.defer import Deferred
 from twisted.internet.interfaces import IAddress
+from twisted.internet.protocol import Protocol
 from twisted.python.failure import Failure
 from twisted.web.http import HTTPChannel
 from twisted.web.resource import IResource, Resource
@@ -302,10 +303,15 @@ class SynapseRequest(Request):
         # this is called once a Resource has been found to serve the request; in our
         # case the Resource in question will normally be a JsonResource.
 
-        # create a LogContext for this request
+        # Create a LogContext for this request
+        #
+        # We only care about associating logs and tallying up metrics at the per-request
+        # level so we don't worry about setting the `parent_context`; preventing us from
+        # unnecessarily piling up metrics on the main process's context.
         request_id = self.get_request_id()
         self.logcontext = LoggingContext(
-            request_id,
+            name=request_id,
+            server_name=self.our_server_name,
             request=ContextRequest(
                 request_id=request_id,
                 ip_address=self.get_client_ip_if_available(),
@@ -655,6 +661,70 @@ class _XForwardedForAddress:
     host: str
 
 
+class SynapseProtocol(HTTPChannel):
+    """
+    Synapse-specific twisted http Protocol.
+
+    This is a small wrapper around the twisted HTTPChannel so we can track active
+    connections in order to close any outstanding connections on shutdown.
+    """
+
+    def __init__(
+        self,
+        site: "SynapseSite",
+        our_server_name: str,
+        max_request_body_size: int,
+        request_id_header: Optional[str],
+        request_class: type,
+    ):
+        super().__init__()
+        self.factory: SynapseSite = site
+        self.site = site
+        self.our_server_name = our_server_name
+        self.max_request_body_size = max_request_body_size
+        self.request_id_header = request_id_header
+        self.request_class = request_class
+
+    def connectionMade(self) -> None:
+        """
+        Called when a connection is made.
+
+        This may be considered the initializer of the protocol, because
+        it is called when the connection is completed.
+
+        Add the connection to the factory's connection list when it's established.
+        """
+        super().connectionMade()
+        self.factory.addConnection(self)
+
+    def connectionLost(self, reason: Failure) -> None:  # type: ignore[override]
+        """
+        Called when the connection is shut down.
+
+        Clear any circular references here, and any external references to this
+        Protocol. The connection has been closed. In our case, we need to remove the
+        connection from the factory's connection list, when it's lost.
+        """
+        super().connectionLost(reason)
+        self.factory.removeConnection(self)
+
+    def requestFactory(self, http_channel: HTTPChannel, queued: bool) -> SynapseRequest:  # type: ignore[override]
+        """
+        A callable used to build `twisted.web.iweb.IRequest` objects.
+
+        Use our own custom SynapseRequest type instead of the regular
+        twisted.web.server.Request.
+        """
+        return self.request_class(
+            self,
+            self.factory,
+            our_server_name=self.our_server_name,
+            max_request_body_size=self.max_request_body_size,
+            queued=queued,
+            request_id_header=self.request_id_header,
+        )
+
+
 class SynapseSite(ProxySite):
     """
     Synapse-specific twisted http Site
@@ -705,23 +775,44 @@ class SynapseSite(ProxySite):
 
         assert config.http_options is not None
         proxied = config.http_options.x_forwarded
-        request_class = XForwardedForRequest if proxied else SynapseRequest
+        self.request_class = XForwardedForRequest if proxied else SynapseRequest
 
-        request_id_header = config.http_options.request_id_header
+        self.request_id_header = config.http_options.request_id_header
+        self.max_request_body_size = max_request_body_size
 
-        def request_factory(channel: HTTPChannel, queued: bool) -> Request:
-            return request_class(
-                channel,
-                self,
-                our_server_name=self.server_name,
-                max_request_body_size=max_request_body_size,
-                queued=queued,
-                request_id_header=request_id_header,
-            )
-
-        self.requestFactory = request_factory  # type: ignore
         self.access_logger = logging.getLogger(logger_name)
         self.server_version_string = server_version_string.encode("ascii")
+        self.connections: List[Protocol] = []
+
+    def buildProtocol(self, addr: IAddress) -> SynapseProtocol:
+        protocol = SynapseProtocol(
+            self,
+            self.server_name,
+            self.max_request_body_size,
+            self.request_id_header,
+            self.request_class,
+        )
+        return protocol
+
+    def addConnection(self, protocol: Protocol) -> None:
+        self.connections.append(protocol)
+
+    def removeConnection(self, protocol: Protocol) -> None:
+        if protocol in self.connections:
+            self.connections.remove(protocol)
+
+    def stopFactory(self) -> None:
+        super().stopFactory()
+
+        # Shutdown any connections which are still active.
+        # These can be long lived HTTP connections which wouldn't normally be closed
+        # when calling `shutdown` on the respective `Port`.
+        # Closing the connections here is required for us to fully shutdown the
+        # `SynapseHomeServer` in order for it to be garbage collected.
+        for protocol in self.connections[:]:
+            if protocol.transport is not None:
+                protocol.transport.loseConnection()
+        self.connections.clear()
 
     def log(self, request: SynapseRequest) -> None:  # type: ignore[override]
         pass

synapse/logging/context.py

@@ -238,12 +238,13 @@ class _Sentinel:
     we should always know which server the logs are coming from.
     """
 
-    __slots__ = ["previous_context", "finished", "request", "tag"]
+    __slots__ = ["previous_context", "finished", "server_name", "request", "tag"]
 
     def __init__(self) -> None:
         # Minimal set for compatibility with LoggingContext
         self.previous_context = None
         self.finished = False
+        self.server_name = "unknown_server_from_sentinel_context"
         self.request = None
         self.tag = None
 
@@ -282,14 +283,19 @@ class LoggingContext:
           child to the parent
 
     Args:
-        name: Name for the context for logging. If this is omitted, it is
-           inherited from the parent context.
+        name: Name for the context for logging.
+        server_name: The name of the server this context is associated with
+            (`config.server.server_name` or `hs.hostname`)
         parent_context (LoggingContext|None): The parent of the new context
+        request: Synapse Request Context object. Useful to associate all the logs
+            happening to a given request.
+
     """
 
     __slots__ = [
         "previous_context",
         "name",
+        "server_name",
         "parent_context",
         "_resource_usage",
         "usage_start",
@@ -301,7 +307,9 @@ class LoggingContext:
 
     def __init__(
         self,
-        name: Optional[str] = None,
+        *,
+        name: str,
+        server_name: str,
         parent_context: "Optional[LoggingContext]" = None,
         request: Optional[ContextRequest] = None,
     ) -> None:
@@ -314,6 +322,8 @@ class LoggingContext:
         # if the context is not currently active.
         self.usage_start: Optional[resource.struct_rusage] = None
 
+        self.name = name
+        self.server_name = server_name
         self.main_thread = get_thread_id()
         self.request = None
         self.tag = ""
@@ -325,23 +335,15 @@ class LoggingContext:
 
         self.parent_context = parent_context
 
+        # Inherit some fields from the parent context
         if self.parent_context is not None:
-            # we track the current request_id
+            # which request this corresponds to
             self.request = self.parent_context.request
 
         if request is not None:
             # the request param overrides the request from the parent context
             self.request = request
 
-        # if we don't have a `name`, but do have a parent context, use its name.
-        if self.parent_context and name is None:
-            name = str(self.parent_context)
-        if name is None:
-            raise ValueError(
-                "LoggingContext must be given either a name or a parent context"
-            )
-        self.name = name
-
     def __str__(self) -> str:
         return self.name
 
@@ -588,7 +590,26 @@ class LoggingContextFilter(logging.Filter):
     record.
     """
 
-    def __init__(self, request: str = ""):
+    def __init__(
+        self,
+        # `request` is here for backwards compatibility since we previously recommended
+        # people manually configure `LoggingContextFilter` like the following.
+        #
+        # ```yaml
+        # filters:
+        #   context:
+        #       (): synapse.logging.context.LoggingContextFilter
+        #       request: ""
+        # ```
+        #
+        # TODO: Since we now configure `LoggingContextFilter` automatically since #8051
+        # (2020-08-11), we could consider removing this useless parameter. This would
+        # require people to remove their own manual configuration of
+        # `LoggingContextFilter` as it would cause `TypeError: Filter.__init__() got an
+        # unexpected keyword argument 'request'` -> `ValueError: Unable to configure
+        # filter 'context'`
+        request: str = "",
+    ):
         self._default_request = request
 
     def filter(self, record: logging.LogRecord) -> Literal[True]:
@@ -598,11 +619,13 @@ class LoggingContextFilter(logging.Filter):
         """
         context = current_context()
         record.request = self._default_request
+        record.server_name = "unknown_server_from_no_context"
 
         # context should never be None, but if it somehow ends up being, then
         # we end up in a death spiral of infinite loops, so let's check, for
         # robustness' sake.
         if context is not None:
+            record.server_name = context.server_name
             # Logging is interested in the request ID. Note that for backwards
             # compatibility this is stored as the "request" on the record.
             record.request = str(context)
@@ -728,12 +751,15 @@ def nested_logging_context(suffix: str) -> LoggingContext:
             "Starting nested logging context from sentinel context: metrics will be lost"
         )
         parent_context = None
+        server_name = "unknown_server_from_sentinel_context"
     else:
         assert isinstance(curr_context, LoggingContext)
         parent_context = curr_context
+        server_name = parent_context.server_name
     prefix = str(curr_context)
     return LoggingContext(
-        prefix + "-" + suffix,
+        name=prefix + "-" + suffix,
+        server_name=server_name,
         parent_context=parent_context,
     )
 
@@ -1058,12 +1084,18 @@ def defer_to_threadpool(
             "Calling defer_to_threadpool from sentinel context: metrics will be lost"
         )
         parent_context = None
+        server_name = "unknown_server_from_sentinel_context"
     else:
         assert isinstance(curr_context, LoggingContext)
         parent_context = curr_context
+        server_name = parent_context.server_name
 
     def g() -> R:
-        with LoggingContext(str(curr_context), parent_context=parent_context):
+        with LoggingContext(
+            name=str(curr_context),
+            server_name=server_name,
+            parent_context=parent_context,
+        ):
             return f(*args, **kwargs)
 
     return make_deferred_yieldable(threads.deferToThreadPool(reactor, threadpool, g))

synapse/logging/filter.py

@@ -1,38 +0,0 @@
-#
-# This file is licensed under the Affero General Public License (AGPL) version 3.
-#
-# Copyright 2020 The Matrix.org Foundation C.I.C.
-# Copyright (C) 2023 New Vector, Ltd
-#
-# This program is free software: you can redistribute it and/or modify
-# it under the terms of the GNU Affero General Public License as
-# published by the Free Software Foundation, either version 3 of the
-# License, or (at your option) any later version.
-#
-# See the GNU Affero General Public License for more details:
-# <https://www.gnu.org/licenses/agpl-3.0.html>.
-#
-# Originally licensed under the Apache License, Version 2.0:
-# <http://www.apache.org/licenses/LICENSE-2.0>.
-#
-# [This file includes modifications made by New Vector Limited]
-#
-#
-import logging
-from typing import Literal
-
-
-class MetadataFilter(logging.Filter):
-    """Logging filter that adds constant values to each record.
-
-    Args:
-        metadata: Key-value pairs to add to each record.
-    """
-
-    def __init__(self, metadata: dict):
-        self._metadata = metadata
-
-    def filter(self, record: logging.LogRecord) -> Literal[True]:
-        for key, value in self._metadata.items():
-            setattr(record, key, value)
-        return True