Move classifiers under project

If we're only defining static classifiers, then they should live under `project` instead of `tool.poetry`.
Merge branch 'develop' of github.com:element-hq/synapse into anoa/update_pyproject
2025-05-13 15:40:01 +01:00 · 2025-05-13 15:39:13 +01:00 · 2025-05-13 15:19:30 +01:00 · 2025-05-13 15:14:00 +01:00 · 2025-05-13 15:12:34 +01:00 · 2025-05-13 15:12:22 +01:00
133 changed files with 5947 additions and 1501 deletions
@@ -11,12 +11,12 @@ with open("poetry.lock", "rb") as f:

 try:
    lock_version = lockfile["metadata"]["lock-version"]
-    assert lock_version == "2.0"
+    assert lock_version == "2.1"
 except Exception:
    print(
        """\
-    Lockfile is not version 2.0. You probably need to upgrade poetry on your local box
-    and re-run `poetry lock --no-update`. See the Poetry cheat sheet at
+    Lockfile is not version 2.1. You probably need to upgrade poetry on your local box
+    and re-run `poetry lock`. See the Poetry cheat sheet at
    https://element-hq.github.io/synapse/develop/development/dependencies.html
    """
    )
@@ -25,7 +25,7 @@ export VIRTUALENV_NO_DOWNLOAD=1

 sed -i \
   -e "s/[~>]=/==/g" \
-   -e '/^python = "^/!s/\^/==/g' \
+   -e '/^requires-python = "^/!s/\^/==/g' \
   -e "/psycopg2/d" \
   -e 's/pyOpenSSL = "==16.0.0"/pyOpenSSL = "==17.0.0"/' \
   -e '/systemd/d' \
@@ -9,5 +9,4 @@
  - End with either a period (.) or an exclamation mark (!).
  - Start with a capital letter.
  - Feel free to credit yourself, by adding a sentence "Contributed by @github_username." or "Contributed by [Your Name]." to the end of the entry.
-* [ ] [Code style](https://element-hq.github.io/synapse/latest/code_style.html) is correct
-  (run the [linters](https://element-hq.github.io/synapse/latest/development/contributing_guide.html#run-the-linters))
+* [ ] [Code style](https://element-hq.github.io/synapse/latest/code_style.html) is correct (run the [linters](https://element-hq.github.io/synapse/latest/development/contributing_guide.html#run-the-linters))
@@ -18,22 +18,22 @@ jobs:
    steps:
      - name: Set up QEMU
        id: qemu
-        uses: docker/setup-qemu-action@v3
+        uses: docker/setup-qemu-action@29109295f81e9208d7d86ff1c6c12d2833863392 # v3.6.0
        with:
          platforms: arm64

      - name: Set up Docker Buildx
        id: buildx
-        uses: docker/setup-buildx-action@v3
+        uses: docker/setup-buildx-action@b5ca514318bd6ebac0fb2aedd5d36ec1b5c232a2 # v3.10.0

      - name: Inspect builder
        run: docker buildx inspect

      - name: Install Cosign
-        uses: sigstore/cosign-installer@v3.8.1
+        uses: sigstore/cosign-installer@3454372f43399081ed03b604cb2d021dabca52bb # v3.8.2

      - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Extract version from pyproject.toml
        # Note: explicitly requesting bash will mean bash is invoked with `-eo pipefail`, see
@@ -43,13 +43,13 @@ jobs:
          echo "SYNAPSE_VERSION=$(grep "^version" pyproject.toml | sed -E 's/version\s*=\s*["]([^"]*)["]/\1/')" >> $GITHUB_ENV

      - name: Log in to DockerHub
-        uses: docker/login-action@v3
+        uses: docker/login-action@74a5d142397b4f367a81961eba4e8cd7edddf772 # v3.4.0
        with:
          username: ${{ secrets.DOCKERHUB_USERNAME }}
          password: ${{ secrets.DOCKERHUB_TOKEN }}

      - name: Log in to GHCR
-        uses: docker/login-action@v3
+        uses: docker/login-action@74a5d142397b4f367a81961eba4e8cd7edddf772 # v3.4.0
        with:
          registry: ghcr.io
          username: ${{ github.repository_owner }}
@@ -57,7 +57,7 @@ jobs:

      - name: Calculate docker image tag
        id: set-tag
-        uses: docker/metadata-action@master
+        uses: docker/metadata-action@902fa8ec7d6ecbf8d84d538b9b233a880e428804 # v5.7.0
        with:
          images: |
            docker.io/matrixdotorg/synapse
@@ -72,7 +72,7 @@ jobs:

      - name: Build and push all platforms
        id: build-and-push
-        uses: docker/build-push-action@v6
+        uses: docker/build-push-action@471d1dc4e07e5cdedd4c2171150001c434f0b7a4 # v6.15.0
        with:
          push: true
          labels: |
@@ -14,7 +14,7 @@ jobs:
      # There's a 'download artifact' action, but it hasn't been updated for the workflow_run action
      # (https://github.com/actions/download-artifact/issues/60) so instead we get this mess:
      - name: 📥 Download artifact
-        uses: dawidd6/action-download-artifact@20319c5641d495c8a52e688b7dc5fada6c3a9fbc # v8
+        uses: dawidd6/action-download-artifact@07ab29fd4a977ae4d2b275087cf67563dfdf0295 # v9
        with:
          workflow: docs-pr.yaml
          run_id: ${{ github.event.workflow_run.id }}
@@ -22,7 +22,7 @@ jobs:
          path: book

      - name: 📤 Deploy to Netlify
-        uses: matrix-org/netlify-pr-preview@v3
+        uses: matrix-org/netlify-pr-preview@9805cd123fc9a7e421e35340a05e1ebc5dee46b5 # v3
        with:
          path: book
          owner: ${{ github.event.workflow_run.head_repository.owner.login }}
@@ -13,7 +13,7 @@ jobs:
    name: GitHub Pages
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          # Fetch all history so that the schema_versions script works.
          fetch-depth: 0
@@ -24,7 +24,7 @@ jobs:
          mdbook-version: '0.4.17'

      - name: Setup python
-        uses: actions/setup-python@v5
+        uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: "3.x"

@@ -39,7 +39,7 @@ jobs:
          cp book/welcome_and_overview.html book/index.html

      - name: Upload Artifact
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
        with:
          name: book
          path: book
@@ -50,7 +50,7 @@ jobs:
    name: Check links in documentation
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Setup mdbook
        uses: peaceiris/actions-mdbook@ee69d230fe19748b7abf22df32acaa93833fad08 # v2.0.0
@@ -50,7 +50,7 @@ jobs:
    needs:
      - pre
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          # Fetch all history so that the schema_versions script works.
          fetch-depth: 0
@@ -64,7 +64,7 @@ jobs:
        run: echo 'window.SYNAPSE_VERSION = "${{ needs.pre.outputs.branch-version }}";' > ./docs/website_files/version.js

      - name: Setup python
-        uses: actions/setup-python@v5
+        uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: "3.x"

@@ -13,21 +13,22 @@ jobs:

    steps:
      - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@master
+        uses: dtolnay/rust-toolchain@56f84321dbccf38fb67ce29ab63e4754056677e0 # master (rust 1.85.1)
        with:
          # We use nightly so that `fmt` correctly groups together imports, and
          # clippy correctly fixes up the benchmarks.
          toolchain: nightly-2022-12-01
          components: clippy, rustfmt
-      - uses: Swatinem/rust-cache@v2
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - name: Setup Poetry
-        uses: matrix-org/setup-python-poetry@v1
+        uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
          install-project: "false"
+          poetry-version: "2.1.1"

      - name: Run ruff check
        continue-on-error: true
@@ -43,6 +44,6 @@ jobs:
      - run: cargo fmt
        continue-on-error: true

-      - uses: stefanzweifel/git-auto-commit-action@v5
+      - uses: stefanzweifel/git-auto-commit-action@b863ae1933cb653a53c021fe36dbb774e1fb9403 # v5.2.0
        with:
          commit_message: "Attempt to fix linting"
@@ -39,17 +39,17 @@ jobs:
    if: needs.check_repo.outputs.should_run_workflow == 'true'
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - name: Install Rust
-        uses: dtolnay/rust-toolchain@stable
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@fcf085fcb4b4b8f63f96906cd713eb52181b5ea4 # stable (rust 1.85.1)
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      # The dev dependencies aren't exposed in the wheel metadata (at least with current
      # poetry-core versions), so we install with poetry.
-      - uses: matrix-org/setup-python-poetry@v1
+      - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
          python-version: "3.x"
-          poetry-version: "1.3.2"
+          poetry-version: "2.1.1"
          extras: "all"
      # Dump installed versions for debugging.
      - run: poetry run pip list > before.txt
@@ -72,11 +72,11 @@ jobs:
            postgres-version: "14"

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@stable
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@fcf085fcb4b4b8f63f96906cd713eb52181b5ea4 # stable (rust 1.85.1)
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - run: sudo apt-get -qq install xmlsec1
      - name: Set up PostgreSQL ${{ matrix.postgres-version }}
@@ -86,7 +86,7 @@ jobs:
            -e POSTGRES_PASSWORD=postgres \
            -e POSTGRES_INITDB_ARGS="--lc-collate C --lc-ctype C --encoding UTF8" \
            postgres:${{ matrix.postgres-version }}
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: "3.x"
      - run: pip install .[all,test]
@@ -145,11 +145,11 @@ jobs:
      BLACKLIST: ${{ matrix.workers && 'synapse-blacklist-with-workers' }}

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@stable
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@fcf085fcb4b4b8f63f96906cd713eb52181b5ea4 # stable (rust 1.85.1)
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - name: Ensure sytest runs `pip install`
        # Delete the lockfile so sytest will `pip install` rather than `poetry install`
@@ -164,7 +164,7 @@ jobs:
        if: ${{ always() }}
        run: /sytest/scripts/tap_to_gha.pl /logs/results.tap
      - name: Upload SyTest logs
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
        if: ${{ always() }}
        with:
          name: Sytest Logs - ${{ job.status }} - (${{ join(matrix.*, ', ') }})
@@ -192,15 +192,15 @@ jobs:
            database: Postgres

    steps:
-      - name: Run actions/checkout@v4 for synapse
-        uses: actions/checkout@v4
+      - name: Check out synapse codebase
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          path: synapse

      - name: Prepare Complement's Prerequisites
        run: synapse/.ci/scripts/setup_complement_prerequisites.sh

-      - uses: actions/setup-go@v5
+      - uses: actions/setup-go@d35c59abb061a4a6fb18e82ac0862c26744d6ab5 # v5.5.0
        with:
          cache-dependency-path: complement/go.sum
          go-version-file: complement/go.mod
@@ -225,7 +225,7 @@ jobs:
    runs-on: ubuntu-latest

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - uses: JasonEtco/create-an-issue@1b14a70e4d8dc185e5cc76d3bec9eab20257b2c5 # v2.9.2
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -16,8 +16,8 @@ jobs:
    name: "Check locked dependencies have sdists"
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: '3.x'
      - run: pip install tomli
@@ -33,29 +33,29 @@ jobs:
      packages: write
    steps:
      - name: Checkout specific branch (debug build)
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        if: github.event_name == 'workflow_dispatch'
        with:
          ref: ${{ inputs.branch }}
      - name: Checkout clean copy of develop (scheduled build)
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        if: github.event_name == 'schedule'
        with:
          ref: develop
      - name: Checkout clean copy of master (on-push)
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        if: github.event_name == 'push'
        with:
          ref: master
      - name: Login to registry
-        uses: docker/login-action@v3
+        uses: docker/login-action@74a5d142397b4f367a81961eba4e8cd7edddf772 # v3.4.0
        with:
          registry: ghcr.io
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}
      - name: Work out labels for complement image
        id: meta
-        uses: docker/metadata-action@v5
+        uses: docker/metadata-action@902fa8ec7d6ecbf8d84d538b9b233a880e428804 # v5.7.0
        with:
          images: ghcr.io/${{ github.repository }}/complement-synapse
          tags: |
@@ -27,8 +27,8 @@ jobs:
    name: "Calculate list of debian distros"
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: '3.x'
      - id: set-distros
@@ -55,18 +55,18 @@ jobs:

    steps:
      - name: Checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          path: src

      - name: Set up Docker Buildx
        id: buildx
-        uses: docker/setup-buildx-action@v3
+        uses: docker/setup-buildx-action@b5ca514318bd6ebac0fb2aedd5d36ec1b5c232a2 # v3.10.0
        with:
          install: true

      - name: Set up docker layer caching
-        uses: actions/cache@v4
+        uses: actions/cache@5a3ec84eff668545956fd18022155c47e93e2684 # v4.2.3
        with:
          path: /tmp/.buildx-cache
          key: ${{ runner.os }}-buildx-${{ github.sha }}
@@ -74,7 +74,7 @@ jobs:
            ${{ runner.os }}-buildx-

      - name: Set up python
-        uses: actions/setup-python@v5
+        uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: '3.x'

@@ -101,7 +101,7 @@ jobs:
          echo "ARTIFACT_NAME=${DISTRO#*:}" >> "$GITHUB_OUTPUT"

      - name: Upload debs as artifacts
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
        with:
          name: debs-${{ steps.artifact-name.outputs.ARTIFACT_NAME }}
          path: debs/*
@@ -130,9 +130,9 @@ jobs:
            arch: aarch64

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          # setup-python@v4 doesn't impose a default python version. Need to use 3.x
          # here, because `python` on osx points to Python 2.7.
@@ -143,7 +143,7 @@ jobs:

      - name: Set up QEMU to emulate aarch64
        if: matrix.arch == 'aarch64'
-        uses: docker/setup-qemu-action@v3
+        uses: docker/setup-qemu-action@29109295f81e9208d7d86ff1c6c12d2833863392 # v3.6.0
        with:
          platforms: arm64

@@ -165,7 +165,7 @@ jobs:
          CARGO_NET_GIT_FETCH_WITH_CLI: true
          CIBW_ENVIRONMENT_PASS_LINUX: CARGO_NET_GIT_FETCH_WITH_CLI

-      - uses: actions/upload-artifact@v4
+      - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
        with:
          name: Wheel-${{ matrix.os }}-${{ matrix.arch }}
          path: ./wheelhouse/*.whl
@@ -176,8 +176,8 @@ jobs:
    if: ${{ !startsWith(github.ref, 'refs/pull/') }}

    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: '3.10'

@@ -186,7 +186,7 @@ jobs:
      - name: Build sdist
        run: python -m build --sdist

-      - uses: actions/upload-artifact@v4
+      - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
        with:
          name: Sdist
          path: dist/*.tar.gz
@@ -203,7 +203,7 @@ jobs:
    runs-on: ubuntu-latest
    steps:
      - name: Download all workflow run artifacts
-        uses: actions/download-artifact@v4
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
      - name: Build a tarball for the debs
        # We need to merge all the debs uploads into one folder, then compress
        # that.
@@ -213,7 +213,7 @@ jobs:
          tar -cvJf debs.tar.xz debs
      - name: Attach to release
        # Pinned to work around https://github.com/softprops/action-gh-release/issues/445
-        uses: softprops/action-gh-release@v0.1.15
+        uses: softprops/action-gh-release@c95fe1489396fe8a9eb87c0abf8aa5b2ef267fda # v0.1.15
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
        with:
@@ -23,7 +23,7 @@ jobs:
      linting: ${{ !startsWith(github.ref, 'refs/pull/') || steps.filter.outputs.linting }}
      linting_readme: ${{ !startsWith(github.ref, 'refs/pull/') || steps.filter.outputs.linting_readme }}
    steps:
-    - uses: dorny/paths-filter@v3
+    - uses: dorny/paths-filter@de90cc6fb38fc0963ad72b210f1f284cd68cea36 # v3.0.2
      id: filter
      # We only check on PRs
      if: startsWith(github.ref, 'refs/pull/')
@@ -83,14 +83,14 @@ jobs:
    if: ${{ needs.changes.outputs.linting == 'true' }}

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - name: Install Rust
-        uses: dtolnay/rust-toolchain@1.66.0
-      - uses: Swatinem/rust-cache@v2
-      - uses: matrix-org/setup-python-poetry@v1
+        uses: dtolnay/rust-toolchain@e05ebb0e73db581a4877c6ce762e29fe1e0b5073 # 1.66.0
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8
+      - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
          python-version: "3.x"
-          poetry-version: "1.3.2"
+          poetry-version: "2.1.1"
          extras: "all"
      - run: poetry run scripts-dev/generate_sample_config.sh --check
      - run: poetry run scripts-dev/config-lint.sh
@@ -101,8 +101,8 @@ jobs:
    if: ${{ needs.changes.outputs.linting == 'true' }}

    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: "3.x"
      - run: "pip install 'click==8.1.1' 'GitPython>=3.1.20'"
@@ -111,8 +111,8 @@ jobs:
  check-lockfile:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: "3.x"
      - run: .ci/scripts/check_lockfile.py
@@ -124,11 +124,12 @@ jobs:

    steps:
      - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Setup Poetry
-        uses: matrix-org/setup-python-poetry@v1
+        uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
+          poetry-version: "2.1.1"
          install-project: "false"

      - name: Run ruff check
@@ -145,14 +146,14 @@ jobs:

    steps:
      - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@1.66.0
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@e05ebb0e73db581a4877c6ce762e29fe1e0b5073 # 1.66.0
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - name: Setup Poetry
-        uses: matrix-org/setup-python-poetry@v1
+        uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
          # We want to make use of type hints in optional dependencies too.
          extras: all
@@ -161,11 +162,12 @@ jobs:
          # https://github.com/matrix-org/synapse/pull/15376#issuecomment-1498983775
          # To make CI green, err towards caution and install the project.
          install-project: "true"
+          poetry-version: "2.1.1"

      # Cribbed from
      # https://github.com/AustinScola/mypy-cache-github-action/blob/85ea4f2972abed39b33bd02c36e341b28ca59213/src/restore.ts#L10-L17
      - name: Restore/persist mypy's cache
-        uses: actions/cache@v4
+        uses: actions/cache@5a3ec84eff668545956fd18022155c47e93e2684 # v4.2.3
        with:
          path: |
            .mypy_cache
@@ -178,7 +180,7 @@ jobs:
  lint-crlf:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - name: Check line endings
        run: scripts-dev/check_line_terminators.sh

@@ -186,11 +188,11 @@ jobs:
    if: ${{ (github.base_ref == 'develop'  || contains(github.base_ref, 'release-')) && github.actor != 'dependabot[bot]' }}
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          ref: ${{ github.event.pull_request.head.sha }}
          fetch-depth: 0
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: "3.x"
      - run: "pip install 'towncrier>=18.6.0rc1'"
@@ -204,15 +206,15 @@ jobs:
    if: ${{ needs.changes.outputs.linting == 'true' }}

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          ref: ${{ github.event.pull_request.head.sha }}
      - name: Install Rust
-        uses: dtolnay/rust-toolchain@1.66.0
-      - uses: Swatinem/rust-cache@v2
-      - uses: matrix-org/setup-python-poetry@v1
+        uses: dtolnay/rust-toolchain@e05ebb0e73db581a4877c6ce762e29fe1e0b5073 # 1.66.0
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8
+      - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
-          poetry-version: "1.3.2"
+          poetry-version: "2.1.1"
          extras: "all"
      - run: poetry run scripts-dev/check_pydantic_models.py

@@ -222,13 +224,13 @@ jobs:
    if: ${{ needs.changes.outputs.rust == 'true' }}

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@1.66.0
+        uses: dtolnay/rust-toolchain@e05ebb0e73db581a4877c6ce762e29fe1e0b5073 # 1.66.0
        with:
            components: clippy
-      - uses: Swatinem/rust-cache@v2
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - run: cargo clippy -- -D warnings

@@ -240,14 +242,14 @@ jobs:
    if: ${{ needs.changes.outputs.rust == 'true' }}

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@master
+        uses: dtolnay/rust-toolchain@56f84321dbccf38fb67ce29ab63e4754056677e0 # master (rust 1.85.1)
        with:
            toolchain: nightly-2022-12-01
            components: clippy
-      - uses: Swatinem/rust-cache@v2
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - run: cargo clippy --all-features -- -D warnings

@@ -257,15 +259,15 @@ jobs:
    if: ${{ needs.changes.outputs.rust == 'true' }}

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@master
+        uses: dtolnay/rust-toolchain@56f84321dbccf38fb67ce29ab63e4754056677e0 # master (rust 1.85.1)
        with:
          # We use nightly so that it correctly groups together imports
          toolchain: nightly-2022-12-01
          components: rustfmt
-      - uses: Swatinem/rust-cache@v2
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - run: cargo fmt --check

@@ -276,8 +278,8 @@ jobs:
    needs: changes
    if: ${{ needs.changes.outputs.linting_readme == 'true' }}
    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: "3.x"
      - run: "pip install rstcheck"
@@ -301,7 +303,7 @@ jobs:
      - lint-readme
    runs-on: ubuntu-latest
    steps:
-      - uses: matrix-org/done-action@v3
+      - uses: matrix-org/done-action@3409aa904e8a2aaf2220f09bc954d3d0b0a2ee67 # v3
        with:
          needs: ${{ toJSON(needs) }}

@@ -324,8 +326,8 @@ jobs:
    needs: linting-done
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: "3.x"
      - id: get-matrix
@@ -345,7 +347,7 @@ jobs:
        job:  ${{ fromJson(needs.calculate-test-jobs.outputs.trial_test_matrix) }}

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - run: sudo apt-get -qq install xmlsec1
      - name: Set up PostgreSQL ${{ matrix.job.postgres-version }}
        if: ${{ matrix.job.postgres-version }}
@@ -360,13 +362,13 @@ jobs:
            postgres:${{ matrix.job.postgres-version }}

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@1.66.0
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@e05ebb0e73db581a4877c6ce762e29fe1e0b5073 # 1.66.0
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

-      - uses: matrix-org/setup-python-poetry@v1
+      - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
          python-version: ${{ matrix.job.python-version }}
-          poetry-version: "1.3.2"
+          poetry-version: "2.1.1"
          extras: ${{ matrix.job.extras }}
      - name: Await PostgreSQL
        if: ${{ matrix.job.postgres-version }}
@@ -399,11 +401,11 @@ jobs:
      - changes
    runs-on: ubuntu-22.04
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@1.66.0
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@e05ebb0e73db581a4877c6ce762e29fe1e0b5073 # 1.66.0
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      # There aren't wheels for some of the older deps, so we need to install
      # their build dependencies
@@ -412,7 +414,7 @@ jobs:
          sudo apt-get -qq install build-essential libffi-dev python3-dev \
            libxml2-dev libxslt-dev xmlsec1 zlib1g-dev libjpeg-dev libwebp-dev

-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@8d9ed9ac5c53483de85588cdf95a591a75ab9f55 # v5.5.0
        with:
          python-version: '3.9'

@@ -462,13 +464,13 @@ jobs:
        extras: ["all"]

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      # Install libs necessary for PyPy to build binary wheels for dependencies
      - run: sudo apt-get -qq install xmlsec1 libxml2-dev libxslt-dev
-      - uses: matrix-org/setup-python-poetry@v1
+      - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
          python-version: ${{ matrix.python-version }}
-          poetry-version: "1.3.2"
+          poetry-version: "2.1.1"
          extras: ${{ matrix.extras }}
      - run: poetry run trial --jobs=2 tests
      - name: Dump logs
@@ -512,13 +514,13 @@ jobs:
        job: ${{ fromJson(needs.calculate-test-jobs.outputs.sytest_test_matrix) }}

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - name: Prepare test blacklist
        run: cat sytest-blacklist .ci/worker-blacklist > synapse-blacklist-with-workers

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@1.66.0
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@e05ebb0e73db581a4877c6ce762e29fe1e0b5073 # 1.66.0
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - name: Run SyTest
        run: /bootstrap.sh synapse
@@ -527,7 +529,7 @@ jobs:
        if: ${{ always() }}
        run: /sytest/scripts/tap_to_gha.pl /logs/results.tap
      - name: Upload SyTest logs
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
        if: ${{ always() }}
        with:
          name: Sytest Logs - ${{ job.status }} - (${{ join(matrix.job.*, ', ') }})
@@ -557,11 +559,11 @@ jobs:
          --health-retries 5

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - run: sudo apt-get -qq install xmlsec1 postgresql-client
-      - uses: matrix-org/setup-python-poetry@v1
+      - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
-          poetry-version: "1.3.2"
+          poetry-version: "2.1.1"
          extras: "postgres"
      - run: .ci/scripts/test_export_data_command.sh
        env:
@@ -601,7 +603,7 @@ jobs:
          --health-retries 5

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - name: Add PostgreSQL apt repository
        # We need a version of pg_dump that can handle the version of
        # PostgreSQL being tested against. The Ubuntu package repository lags
@@ -612,10 +614,10 @@ jobs:
          wget --quiet -O - https://www.postgresql.org/media/keys/ACCC4CF8.asc | sudo apt-key add -
          sudo apt-get update
      - run: sudo apt-get -qq install xmlsec1 postgresql-client
-      - uses: matrix-org/setup-python-poetry@v1
+      - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
          python-version: ${{ matrix.python-version }}
-          poetry-version: "1.3.2"
+          poetry-version: "2.1.1"
          extras: "postgres"
      - run: .ci/scripts/test_synapse_port_db.sh
        id: run_tester_script
@@ -625,7 +627,7 @@ jobs:
          PGPASSWORD: postgres
          PGDATABASE: postgres
      - name: "Upload schema differences"
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
        if: ${{ failure() && !cancelled() && steps.run_tester_script.outcome == 'failure' }}
        with:
          name: Schema dumps
@@ -655,19 +657,19 @@ jobs:
            database: Postgres

    steps:
-      - name: Run actions/checkout@v4 for synapse
-        uses: actions/checkout@v4
+      - name: Checkout synapse codebase
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          path: synapse

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@1.66.0
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@e05ebb0e73db581a4877c6ce762e29fe1e0b5073 # 1.66.0
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - name: Prepare Complement's Prerequisites
        run: synapse/.ci/scripts/setup_complement_prerequisites.sh

-      - uses: actions/setup-go@v5
+      - uses: actions/setup-go@d35c59abb061a4a6fb18e82ac0862c26744d6ab5 # v5.5.0
        with:
          cache-dependency-path: complement/go.sum
          go-version-file: complement/go.mod
@@ -690,11 +692,11 @@ jobs:
      - changes

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@1.66.0
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@e05ebb0e73db581a4877c6ce762e29fe1e0b5073 # 1.66.0
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - run: cargo test

@@ -708,13 +710,13 @@ jobs:
      - changes

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@master
+        uses: dtolnay/rust-toolchain@56f84321dbccf38fb67ce29ab63e4754056677e0 # master (rust 1.85.1)
        with:
            toolchain: nightly-2022-12-01
-      - uses: Swatinem/rust-cache@v2
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - run: cargo bench --no-run

@@ -733,7 +735,7 @@ jobs:
      - linting-done
    runs-on: ubuntu-latest
    steps:
-      - uses: matrix-org/done-action@v3
+      - uses: matrix-org/done-action@3409aa904e8a2aaf2220f09bc954d3d0b0a2ee67 # v3
        with:
          needs: ${{ toJSON(needs) }}

@@ -6,7 +6,7 @@ on:

 jobs:
  triage:
-    uses: matrix-org/backend-meta/.github/workflows/triage-incoming.yml@v2
+    uses: matrix-org/backend-meta/.github/workflows/triage-incoming.yml@18beaf3c8e536108bd04d18e6c3dc40ba3931e28 # v2.0.3
    with:
      project_id: 'PVT_kwDOAIB0Bs4AFDdZ'
      content_id: ${{ github.event.issue.node_id }}
@@ -11,7 +11,7 @@ jobs:
    if: >
      contains(github.event.issue.labels.*.name, 'X-Needs-Info')
    steps:
-      - uses: actions/add-to-project@main
+      - uses: actions/add-to-project@5b1a254a3546aef88e0a7724a77a623fa2e47c36 # main (v1.0.2 + 10 commits)
        id: add_project
        with:
          project-url: "https://github.com/orgs/matrix-org/projects/67"
@@ -40,16 +40,17 @@ jobs:
    runs-on: ubuntu-latest

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@stable
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@fcf085fcb4b4b8f63f96906cd713eb52181b5ea4 # stable (rust 1.85.1)
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

-      - uses: matrix-org/setup-python-poetry@v1
+      - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
          python-version: "3.x"
          extras: "all"
+          poetry-version: "2.1.1"
      - run: |
          poetry remove twisted
          poetry add --extras tls git+https://github.com/twisted/twisted.git#${{ inputs.twisted_ref || 'trunk' }}
@@ -64,17 +65,18 @@ jobs:
    runs-on: ubuntu-latest

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - run: sudo apt-get -qq install xmlsec1

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@stable
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@fcf085fcb4b4b8f63f96906cd713eb52181b5ea4 # stable (rust 1.85.1)
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

-      - uses: matrix-org/setup-python-poetry@v1
+      - uses: matrix-org/setup-python-poetry@5bbf6603c5c930615ec8a29f1b5d7d258d905aa4 # v2.0.0
        with:
          python-version: "3.x"
          extras: "all test"
+          poetry-version: "2.1.1"
      - run: |
          poetry remove twisted
          poetry add --extras tls git+https://github.com/twisted/twisted.git#trunk
@@ -108,11 +110,11 @@ jobs:
        - ${{ github.workspace }}:/src

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@stable
-      - uses: Swatinem/rust-cache@v2
+        uses: dtolnay/rust-toolchain@fcf085fcb4b4b8f63f96906cd713eb52181b5ea4 # stable (rust 1.85.1)
+      - uses: Swatinem/rust-cache@9d47c6ad4b02e050fd481d890b2ea34778fd09d6 # v2.7.8

      - name: Patch dependencies
        # Note: The poetry commands want to create a virtualenv in /src/.venv/,
@@ -136,7 +138,7 @@ jobs:
        if: ${{ always() }}
        run: /sytest/scripts/tap_to_gha.pl /logs/results.tap
      - name: Upload SyTest logs
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
        if: ${{ always() }}
        with:
          name: Sytest Logs - ${{ job.status }} - (${{ join(matrix.*, ', ') }})
@@ -164,14 +166,14 @@ jobs:

    steps:
      - name: Run actions/checkout@v4 for synapse
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          path: synapse

      - name: Prepare Complement's Prerequisites
        run: synapse/.ci/scripts/setup_complement_prerequisites.sh

-      - uses: actions/setup-go@v5
+      - uses: actions/setup-go@d35c59abb061a4a6fb18e82ac0862c26744d6ab5 # v5.5.0
        with:
          cache-dependency-path: complement/go.sum
          go-version-file: complement/go.mod
@@ -181,11 +183,11 @@ jobs:
        run: |
          set -x
          DEBIAN_FRONTEND=noninteractive sudo apt-get install -yqq python3 pipx
-          pipx install poetry==1.3.2
+          pipx install poetry==2.1.1

          poetry remove -n twisted
          poetry add -n --extras tls git+https://github.com/twisted/twisted.git#trunk
-          poetry lock --no-update
+          poetry lock
        working-directory: synapse

      - run: |
@@ -206,7 +208,7 @@ jobs:
    runs-on: ubuntu-latest

    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
      - uses: JasonEtco/create-an-issue@1b14a70e4d8dc185e5cc76d3bec9eab20257b2c5 # v2.9.2
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -1,4 +1,217 @@
+# Synapse 1.130.0rc1 (2025-05-13)
+
+### Features
+
+- Add an Admin API endpoint `GET /_synapse/admin/v1/scheduled_tasks`  to fetch scheduled tasks. ([\#18214](https://github.com/element-hq/synapse/issues/18214))
+- Add config option `user_directory.exclude_remote_users` which, when enabled, excludes remote users from user directory search results. ([\#18300](https://github.com/element-hq/synapse/issues/18300))
+- Add support for handling `GET /devices/` on workers. ([\#18355](https://github.com/element-hq/synapse/issues/18355))
+
+### Bugfixes
+
+- Fix a longstanding bug where Synapse would immediately retry a failing push endpoint when a new event is received, ignoring any backoff timers. ([\#18363](https://github.com/element-hq/synapse/issues/18363))
+- Pass leave from remote invite rejection down Sliding Sync. ([\#18375](https://github.com/element-hq/synapse/issues/18375))
+
+### Updates to the Docker image
+
+- In `configure_workers_and_start.py`, use the same absolute path of Python in the interpreter shebang, and invoke child Python processes with `sys.executable`. ([\#18291](https://github.com/element-hq/synapse/issues/18291))
+- Optimize the build of the workers image. ([\#18292](https://github.com/element-hq/synapse/issues/18292))
+- In `start_for_complement.sh`, replace some external program calls with shell builtins. ([\#18293](https://github.com/element-hq/synapse/issues/18293))
+- When generating container scripts from templates, don't add a leading newline so that their shebangs may be handled correctly. ([\#18295](https://github.com/element-hq/synapse/issues/18295))
+
+### Improved Documentation
+
+- Improve formatting of the README file. ([\#18218](https://github.com/element-hq/synapse/issues/18218))
+- Add documentation for configuring [Pocket ID](https://github.com/pocket-id/pocket-id) as an OIDC provider. ([\#18237](https://github.com/element-hq/synapse/issues/18237))
+- Fix typo in docs about the `push` config option. Contributed by @HarHarLinks. ([\#18320](https://github.com/element-hq/synapse/issues/18320))
+- Add `/_matrix/federation/v1/version` to list of federation endpoints that can be handled by workers. ([\#18377](https://github.com/element-hq/synapse/issues/18377))
+- Add an Admin API endpoint `GET /_synapse/admin/v1/scheduled_tasks`  to fetch scheduled tasks. ([\#18384](https://github.com/element-hq/synapse/issues/18384))
+
+### Internal Changes
+
+- Return specific error code when adding an email address / phone number to account is not supported ([MSC4178](https://github.com/matrix-org/matrix-spec-proposals/pull/4178)). ([\#17578](https://github.com/element-hq/synapse/issues/17578))
+- Stop auto-provisionning missing users & devices when delegating auth to Matrix Authentication Service. Requires MAS 0.13.0 or later. ([\#18181](https://github.com/element-hq/synapse/issues/18181))
+- Apply file hashing and existing quarantines to media downloaded for URL previews. ([\#18297](https://github.com/element-hq/synapse/issues/18297))
+- Allow a few admin APIs used by matrix-authentication-service to run on workers. ([\#18313](https://github.com/element-hq/synapse/issues/18313))
+- Apply `should_drop_federated_event` to federation invites. ([\#18330](https://github.com/element-hq/synapse/issues/18330))
+- Allow `/rooms/` admin API to be run on workers. ([\#18360](https://github.com/element-hq/synapse/issues/18360))
+- Minor performance improvements to the notifier. ([\#18367](https://github.com/element-hq/synapse/issues/18367))
+- Slight performance increase when using the ratelimiter. ([\#18369](https://github.com/element-hq/synapse/issues/18369))
+- Don't validate the `at_hash` (access token hash) field in OIDC ID Tokens if we don't end up actually using the OIDC Access Token. ([\#18374](https://github.com/element-hq/synapse/issues/18374), [\#18385](https://github.com/element-hq/synapse/issues/18385))
+- Fixed test failures when using authlib 1.5.2. ([\#18390](https://github.com/element-hq/synapse/issues/18390))
+- Refactor [MSC4186](https://github.com/matrix-org/matrix-spec-proposals/pull/4186) Simplified Sliding Sync room list tests to cover both new and fallback logic paths. ([\#18399](https://github.com/element-hq/synapse/issues/18399))
+
+
+
+### Updates to locked dependencies
+
+* Bump actions/add-to-project from 280af8ae1f83a494cfad2cb10f02f6d13529caa9 to 5b1a254a3546aef88e0a7724a77a623fa2e47c36. ([\#18365](https://github.com/element-hq/synapse/issues/18365))
+* Bump actions/download-artifact from 4.2.1 to 4.3.0. ([\#18364](https://github.com/element-hq/synapse/issues/18364))
+* Bump actions/setup-go from 5.4.0 to 5.5.0. ([\#18426](https://github.com/element-hq/synapse/issues/18426))
+* Bump anyhow from 1.0.97 to 1.0.98. ([\#18336](https://github.com/element-hq/synapse/issues/18336))
+* Bump packaging from 24.2 to 25.0. ([\#18393](https://github.com/element-hq/synapse/issues/18393))
+* Bump pillow from 11.1.0 to 11.2.1. ([\#18429](https://github.com/element-hq/synapse/issues/18429))
+* Bump pydantic from 2.10.3 to 2.11.4. ([\#18394](https://github.com/element-hq/synapse/issues/18394))
+* Bump pyo3-log from 0.12.2 to 0.12.3. ([\#18317](https://github.com/element-hq/synapse/issues/18317))
+* Bump pyopenssl from 24.3.0 to 25.0.0. ([\#18315](https://github.com/element-hq/synapse/issues/18315))
+* Bump sha2 from 0.10.8 to 0.10.9. ([\#18395](https://github.com/element-hq/synapse/issues/18395))
+* Bump sigstore/cosign-installer from 3.8.1 to 3.8.2. ([\#18366](https://github.com/element-hq/synapse/issues/18366))
+* Bump softprops/action-gh-release from 1 to 2. ([\#18264](https://github.com/element-hq/synapse/issues/18264))
+* Bump stefanzweifel/git-auto-commit-action from 5.1.0 to 5.2.0. ([\#18354](https://github.com/element-hq/synapse/issues/18354))
+* Bump txredisapi from 1.4.10 to 1.4.11. ([\#18392](https://github.com/element-hq/synapse/issues/18392))
+* Bump types-jsonschema from 4.23.0.20240813 to 4.23.0.20241208. ([\#18305](https://github.com/element-hq/synapse/issues/18305))
+* Bump types-psycopg2 from 2.9.21.20250121 to 2.9.21.20250318. ([\#18316](https://github.com/element-hq/synapse/issues/18316))
+
+# Synapse 1.129.0 (2025-05-06)
+
+No significant changes since 1.129.0rc2.
+
+
+
+
+# Synapse 1.129.0rc2 (2025-04-30)
+
+Synapse 1.129.0rc1 was never formally released due to regressions discovered during the release process. 1.129.0rc2 fixes those regressions by reverting the affected PRs.
+
+### Internal Changes
+
+- Revert the slow background update introduced by [\#18068](https://github.com/element-hq/synapse/issues/18068) in v1.128.0. ([\#18372](https://github.com/element-hq/synapse/issues/18372))
+- Revert "Add total event, unencrypted message, and e2ee event counts to stats reporting", added in v1.129.0rc1. ([\#18373](https://github.com/element-hq/synapse/issues/18373))
+
+
+
+
+# Synapse 1.129.0rc1 (2025-04-15)
+
+### Features
+
+- Add `passthrough_authorization_parameters` in OIDC configuration to allow passing parameters to the authorization grant URL. ([\#18232](https://github.com/element-hq/synapse/issues/18232))
+- Add `total_event_count`, `total_message_count`, and `total_e2ee_event_count` fields to the homeserver usage statistics. ([\#18260](https://github.com/element-hq/synapse/issues/18260))
+
+### Bugfixes
+
+- Fix `force_tracing_for_users` config when using delegated auth. ([\#18334](https://github.com/element-hq/synapse/issues/18334))
+- Fix the token introspection cache logging access tokens when MAS integration is in use. ([\#18335](https://github.com/element-hq/synapse/issues/18335))
+- Stop caching introspection failures when delegating auth to MAS. ([\#18339](https://github.com/element-hq/synapse/issues/18339))
+- Fix `ExternalIDReuse` exception after migrating to MAS on workers with a high traffic. ([\#18342](https://github.com/element-hq/synapse/issues/18342))
+- Fix minor performance regression caused by tracking of room participation. Regressed in v1.128.0. ([\#18345](https://github.com/element-hq/synapse/issues/18345))
+
+### Updates to the Docker image
+
+- Optimize the build of the complement-synapse image. ([\#18294](https://github.com/element-hq/synapse/issues/18294))
+
+### Internal Changes
+
+- Disable statement timeout during room purge. ([\#18133](https://github.com/element-hq/synapse/issues/18133))
+- Add cache to storage functions used to auth requests when using delegated auth. ([\#18337](https://github.com/element-hq/synapse/issues/18337))
+
+
+
+
+# Synapse 1.128.0 (2025-04-08)
+
+No significant changes since 1.128.0rc1.
+
+
+
+
+# Synapse 1.128.0rc1 (2025-04-01)
+
+### Features
+
+- Add an access token introspection cache to make Matrix Authentication Service integration ([MSC3861](https://github.com/matrix-org/matrix-doc/pull/3861)) more efficient. ([\#18231](https://github.com/element-hq/synapse/issues/18231))
+- Add background job to clear unreferenced state groups. ([\#18254](https://github.com/element-hq/synapse/issues/18254))
+- Hashes of media files are now tracked by Synapse. Media quarantines will now apply to all files with the same hash. ([\#18277](https://github.com/element-hq/synapse/issues/18277), [\#18302](https://github.com/element-hq/synapse/issues/18302), [\#18296](https://github.com/element-hq/synapse/issues/18296))
+
+### Bugfixes
+
+- Add index to sliding sync ([MSC4186](https://github.com/matrix-org/matrix-doc/pull/4186)) membership snapshot table, to fix a performance issue. ([\#18074](https://github.com/element-hq/synapse/issues/18074))
+
+### Updates to the Docker image
+
+- Specify the architecture of installed packages via an APT config option, which is more reliable than appending package names with `:{arch}`. ([\#18271](https://github.com/element-hq/synapse/issues/18271))
+- Always specify base image debian versions with a build argument. ([\#18272](https://github.com/element-hq/synapse/issues/18272))
+- Allow passing arguments to `start_for_complement.sh` (to be sent to `configure_workers_and_start.py`). ([\#18273](https://github.com/element-hq/synapse/issues/18273))
+- Make some improvements to the `prefix-log` script in the workers image. ([\#18274](https://github.com/element-hq/synapse/issues/18274))
+- Use `uv pip` to install `supervisor` in the worker image. ([\#18275](https://github.com/element-hq/synapse/issues/18275))
+- Avoid needing to download & use `rsync` in a build layer. ([\#18287](https://github.com/element-hq/synapse/issues/18287))
+
+### Improved Documentation
+
+- Fix how to obtain access token and change naming from riot to element ([\#18225](https://github.com/element-hq/synapse/issues/18225))
+- Correct a small typo in the SSO mapping providers documentation. ([\#18276](https://github.com/element-hq/synapse/issues/18276))
+- Add docs for how to clear out the Poetry wheel cache. ([\#18283](https://github.com/element-hq/synapse/issues/18283))
+
+### Internal Changes
+
+- Add a column `participant` to `room_memberships` table. ([\#18068](https://github.com/element-hq/synapse/issues/18068))
+- Update Poetry to 2.1.1, including updating the lock file version. ([\#18251](https://github.com/element-hq/synapse/issues/18251))
+- Pin GitHub Actions dependencies by commit hash. ([\#18255](https://github.com/element-hq/synapse/issues/18255))
+- Add DB delta to remove the old state group deletion job. ([\#18284](https://github.com/element-hq/synapse/issues/18284))
+
+
+
+### Updates to locked dependencies
+
+* Bump actions/add-to-project from f5473ace9aeee8b97717b281e26980aa5097023f to 280af8ae1f83a494cfad2cb10f02f6d13529caa9. ([\#18303](https://github.com/element-hq/synapse/issues/18303))
+* Bump actions/cache from 4.2.2 to 4.2.3. ([\#18266](https://github.com/element-hq/synapse/issues/18266))
+* Bump actions/download-artifact from 4.2.0 to 4.2.1. ([\#18268](https://github.com/element-hq/synapse/issues/18268))
+* Bump actions/setup-python from 5.4.0 to 5.5.0. ([\#18298](https://github.com/element-hq/synapse/issues/18298))
+* Bump actions/upload-artifact from 4.6.1 to 4.6.2. ([\#18304](https://github.com/element-hq/synapse/issues/18304))
+* Bump authlib from 1.4.1 to 1.5.1. ([\#18306](https://github.com/element-hq/synapse/issues/18306))
+* Bump dawidd6/action-download-artifact from 8 to 9. ([\#18204](https://github.com/element-hq/synapse/issues/18204))
+* Bump jinja2 from 3.1.5 to 3.1.6. ([\#18223](https://github.com/element-hq/synapse/issues/18223))
+* Bump log from 0.4.26 to 0.4.27. ([\#18267](https://github.com/element-hq/synapse/issues/18267))
+* Bump phonenumbers from 8.13.50 to 9.0.2. ([\#18299](https://github.com/element-hq/synapse/issues/18299))
+* Bump pygithub from 2.5.0 to 2.6.1. ([\#18243](https://github.com/element-hq/synapse/issues/18243))
+* Bump pyo3-log from 0.12.1 to 0.12.2. ([\#18269](https://github.com/element-hq/synapse/issues/18269))
+
+# Synapse 1.127.1 (2025-03-26)
+
+## Security
+- Fix [CVE-2025-30355](https://www.cve.org/CVERecord?id=CVE-2025-30355) / [GHSA-v56r-hwv5-mxg6](https://github.com/element-hq/synapse/security/advisories/GHSA-v56r-hwv5-mxg6). **High severity vulnerability affecting federation. The vulnerability has been exploited in the wild.**
+
+
+
+# Synapse 1.127.0 (2025-03-25)
+
+No significant changes since 1.127.0rc1.
+
+
+
+
+# Synapse 1.127.0rc1 (2025-03-18)
+
+### Features
+
+- Update [MSC4140](https://github.com/matrix-org/matrix-spec-proposals/pull/4140) implementation to no longer cancel a user's own delayed state events with an event type & state key that match a more recent state event sent by that user. ([\#17810](https://github.com/element-hq/synapse/issues/17810))
+
+### Improved Documentation
+
+- Fixed a minor typo in the Synapse documentation. Contributed by @karuto12. ([\#18224](https://github.com/element-hq/synapse/issues/18224))
+
+### Internal Changes
+
+- Remove undocumented `SYNAPSE_USE_FROZEN_DICTS` environment variable. ([\#18123](https://github.com/element-hq/synapse/issues/18123))
+- Fix detection of workflow failures in the release script. ([\#18211](https://github.com/element-hq/synapse/issues/18211))
+- Add caching support to media endpoints. ([\#18235](https://github.com/element-hq/synapse/issues/18235))
+
+
+
+### Updates to locked dependencies
+
+* Bump anyhow from 1.0.96 to 1.0.97. ([\#18201](https://github.com/element-hq/synapse/issues/18201))
+* Bump bcrypt from 4.2.1 to 4.3.0. ([\#18207](https://github.com/element-hq/synapse/issues/18207))
+* Bump bytes from 1.10.0 to 1.10.1. ([\#18227](https://github.com/element-hq/synapse/issues/18227))
+* Bump http from 1.2.0 to 1.3.1. ([\#18245](https://github.com/element-hq/synapse/issues/18245))
+* Bump sentry-sdk from 2.19.2 to 2.22.0. ([\#18205](https://github.com/element-hq/synapse/issues/18205))
+* Bump serde from 1.0.218 to 1.0.219. ([\#18228](https://github.com/element-hq/synapse/issues/18228))
+* Bump serde_json from 1.0.139 to 1.0.140. ([\#18202](https://github.com/element-hq/synapse/issues/18202))
+* Bump ulid from 1.2.0 to 1.2.1. ([\#18246](https://github.com/element-hq/synapse/issues/18246))
+
 # Synapse 1.126.0 (2025-03-11)
+Administrators using the Debian/Ubuntu packages from `packages.matrix.org`, please check
+[the relevant section in the upgrade notes](https://github.com/element-hq/synapse/blob/release-v1.126/docs/upgrade.md#change-of-signing-key-expiry-date-for-the-debianubuntu-package-repository)
+as we have recently updated the expiry date on the repository's GPG signing key. The old version of the key will expire on `2025-03-15`.

 No significant changes since 1.126.0rc3.

@@ -16,9 +229,6 @@ No significant changes since 1.126.0rc3.

 # Synapse 1.126.0rc2 (2025-03-05)

-Administrators using the Debian/Ubuntu packages from `packages.matrix.org`, please check
-[the relevant section in the upgrade notes](https://github.com/element-hq/synapse/blob/release-v1.126/docs/upgrade.md#change-of-signing-key-expiry-date-for-the-debianubuntu-package-repository)
-as we have recently updated the expiry date on the repository's GPG signing key. The old version of the key will expire on `2025-03-15`.

 ### Internal Changes

@@ -13,9 +13,9 @@ dependencies = [

 [[package]]
 name = "anyhow"
-version = "1.0.96"
+version = "1.0.98"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6b964d184e89d9b6b67dd2715bc8e74cf3107fb2b529990c90cf517326150bf4"
+checksum = "e16d2d3311acee920a9eb8d33b8cbc1787ce4a264e85f964c2404b969bdcd487"

 [[package]]
 name = "arc-swap"
@@ -67,9 +67,9 @@ checksum = "79296716171880943b8470b5f8d03aa55eb2e645a4874bdbb28adb49162e012c"

 [[package]]
 name = "bytes"
-version = "1.10.0"
+version = "1.10.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f61dac84819c6588b558454b194026eb1f09c293b9036ae9b159e74e73ab6cf9"
+checksum = "d71b6127be86fdcfddb610f7182ac57211d4b18a3e9c82eb2d17662f2227ad6a"

 [[package]]
 name = "cfg-if"
@@ -173,9 +173,9 @@ checksum = "7f24254aa9a54b5c858eaee2f5bccdb46aaf0e486a595ed5fd8f86ba55232a70"

 [[package]]
 name = "http"
-version = "1.2.0"
+version = "1.3.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f16ca2af56261c99fba8bac40a10251ce8188205a4c448fbb745a2e4daa76fea"
+checksum = "f4a85d31aea989eead29a3aaf9e1115a180df8282431156e533de47660892565"
 dependencies = [
 "bytes",
 "fnv",
@@ -223,9 +223,9 @@ checksum = "ae743338b92ff9146ce83992f766a31066a91a8c84a45e0e9f21e7cf6de6d346"

 [[package]]
 name = "log"
-version = "0.4.26"
+version = "0.4.27"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "30bde2b3dc3671ae49d8e2e9f044c7c005836e7a023ee57cffa25ab82764bb9e"
+checksum = "13dc2df351e3202783a1fe0d44375f7295ffb4049267b0f3018346dc122a1d94"

 [[package]]
 name = "memchr"
@@ -316,9 +316,9 @@ dependencies = [

 [[package]]
 name = "pyo3-log"
-version = "0.12.1"
+version = "0.12.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "be5bb22b77965a7b5394e9aae9897a0607b51df5167561ffc3b02643b4200bc7"
+checksum = "7079e412e909af5d6be7c04a7f29f6a2837a080410e1c529c9dee2c367383db4"
 dependencies = [
 "arc-swap",
 "log",
@@ -437,18 +437,18 @@ checksum = "f3cb5ba0dc43242ce17de99c180e96db90b235b8a9fdc9543c96d2209116bd9f"

 [[package]]
 name = "serde"
-version = "1.0.218"
+version = "1.0.219"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e8dfc9d19bdbf6d17e22319da49161d5d0108e4188e8b680aef6299eed22df60"
+checksum = "5f0e2c6ed6606019b4e29e69dbaba95b11854410e5347d525002456dbbb786b6"
 dependencies = [
 "serde_derive",
 ]

 [[package]]
 name = "serde_derive"
-version = "1.0.218"
+version = "1.0.219"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f09503e191f4e797cb8aac08e9a4a4695c5edf6a2e70e376d961ddd5c969f82b"
+checksum = "5b0276cf7f2c73365f7157c8123c21cd9a50fbbd844757af28ca1f5925fc2a00"
 dependencies = [
 "proc-macro2",
 "quote",
@@ -457,9 +457,9 @@ dependencies = [

 [[package]]
 name = "serde_json"
-version = "1.0.139"
+version = "1.0.140"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "44f86c3acccc9c65b153fe1b85a3be07fe5515274ec9f0653b4a0875731c72a6"
+checksum = "20068b6e96dc6c9bd23e01df8827e6c7e1f2fddd43c21810382803c136b99373"
 dependencies = [
 "itoa",
 "memchr",
@@ -480,9 +480,9 @@ dependencies = [

 [[package]]
 name = "sha2"
-version = "0.10.8"
+version = "0.10.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "793db75ad2bcafc3ffa7c68b215fee268f537982cd901d132f89c6343f3a3dc8"
+checksum = "a7507d819769d01a365ab707794a4084392c824f54a7a6a7862f8c3d0892b283"
 dependencies = [
 "cfg-if",
 "cpufeatures",
@@ -544,9 +544,9 @@ checksum = "42ff0bf0c66b8238c6f3b578df37d0b7848e55df8577b3f74f92a69acceeb825"

 [[package]]
 name = "ulid"
-version = "1.2.0"
+version = "1.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ab82fc73182c29b02e2926a6df32f2241dbadb5cfc111fd595515b3598f46bb3"
+checksum = "470dbf6591da1b39d43c14523b2b469c86879a53e8b758c8e090a470fe7b1fbe"
 dependencies = [
 "rand",
 "web-time",
@@ -253,15 +253,17 @@ Alongside all that, join our developer community on Matrix:
 Copyright and Licensing
 =======================

-Copyright 2014-2017 OpenMarket Ltd
-Copyright 2017 Vector Creations Ltd
-Copyright 2017-2025 New Vector Ltd 
+| Copyright 2014-2017 OpenMarket Ltd
+| Copyright 2017 Vector Creations Ltd
+| Copyright 2017-2025 New Vector Ltd
+|

 This software is dual-licensed by New Vector Ltd (Element). It can be used either:
-  
+
 (1) for free 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); OR
-  
+
 (2) under the terms of a paid-for Element Commercial License agreement between you and Element (the terms of which may vary depending on what you and Element have agreed to).
+
 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.


@@ -0,0 +1 @@
+Explicitly enable PyPy builds in `cibuildwheel`s config to avoid it being disabled on a future upgrade to `cibuildwheel` v3.
@@ -0,0 +1 @@
+Update the PR review template to remove an erroneous line break from the final bullet point.
@@ -0,0 +1 @@
+Update `pyproject.toml` to remove deprecated keys.
@@ -35,7 +35,7 @@ TEMP_VENV="$(mktemp -d)"
 python3 -m venv "$TEMP_VENV"
 source "$TEMP_VENV/bin/activate"
 pip install -U pip
-pip install poetry==1.3.2
+pip install poetry==2.1.1 poetry-plugin-export==1.9.0
 poetry export \
    --extras all \
    --extras test \
@@ -1,3 +1,58 @@
+matrix-synapse-py3 (1.130.0~rc1) stable; urgency=medium
+
+  * New Synapse release 1.130.0rc1.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 13 May 2025 10:44:04 +0100
+
+matrix-synapse-py3 (1.129.0) stable; urgency=medium
+
+  * New Synapse release 1.129.0.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 06 May 2025 12:22:11 +0100
+
+matrix-synapse-py3 (1.129.0~rc2) stable; urgency=medium
+
+  * New synapse release 1.129.0rc2.
+
+ -- Synapse Packaging team <packages@matrix.org>  Wed, 30 Apr 2025 13:13:16 +0000
+
+matrix-synapse-py3 (1.129.0~rc1) stable; urgency=medium
+
+  * New Synapse release 1.129.0rc1.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 15 Apr 2025 10:47:43 -0600
+
+matrix-synapse-py3 (1.128.0) stable; urgency=medium
+
+  * New Synapse release 1.128.0.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 08 Apr 2025 14:09:54 +0100
+
+matrix-synapse-py3 (1.128.0~rc1) stable; urgency=medium
+
+  * Update Poetry to 2.1.1.
+  * New synapse release 1.128.0rc1.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 01 Apr 2025 14:35:33 +0000
+
+matrix-synapse-py3 (1.127.1) stable; urgency=medium
+
+  * New Synapse release 1.127.1.
+
+ -- Synapse Packaging team <packages@matrix.org>  Wed, 26 Mar 2025 21:07:31 +0000
+
+matrix-synapse-py3 (1.127.0) stable; urgency=medium
+
+  * New Synapse release 1.127.0.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 25 Mar 2025 12:04:15 +0000
+
+matrix-synapse-py3 (1.127.0~rc1) stable; urgency=medium
+
+  * New Synapse release 1.127.0rc1.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 18 Mar 2025 13:30:05 +0000
+
 matrix-synapse-py3 (1.126.0) stable; urgency=medium

  * New Synapse release 1.126.0.
@@ -22,7 +22,7 @@

 ARG DEBIAN_VERSION=bookworm
 ARG PYTHON_VERSION=3.12
-ARG POETRY_VERSION=1.8.3
+ARG POETRY_VERSION=2.1.1

 ###
 ### Stage 0: generate requirements.txt
@@ -56,7 +56,7 @@ ENV UV_LINK_MODE=copy
 ARG POETRY_VERSION
 RUN --mount=type=cache,target=/root/.cache/uv \
  if [ -z "$TEST_ONLY_IGNORE_POETRY_LOCKFILE" ]; then \
-    uvx --with poetry-plugin-export==1.8.0 \
+    uvx --with poetry-plugin-export==1.9.0 \
        poetry@${POETRY_VERSION} export --extras all -o /synapse/requirements.txt ${TEST_ONLY_SKIP_DEP_HASH_VERIFICATION:+--without-hashes}; \
  else \
    touch /synapse/requirements.txt; \
@@ -134,7 +134,6 @@ RUN \
  --mount=type=cache,target=/var/cache/apt,sharing=locked \
  --mount=type=cache,target=/var/lib/apt,sharing=locked \
  apt-get update -qq && \
-  apt-get install -y --no-install-recommends rsync && \
  apt-cache depends --recurse --no-recommends --no-suggests --no-conflicts --no-breaks --no-replaces --no-enhances --no-pre-depends \
      curl \
      gosu \
@@ -148,14 +147,10 @@ RUN \
  for arch in arm64 amd64; do \
    mkdir -p /tmp/debs-${arch} && \
    cd /tmp/debs-${arch} && \
-    apt-get download $(sed "s/$/:${arch}/" /tmp/pkg-list); \
+    apt-get -o APT::Architecture="${arch}" download $(cat /tmp/pkg-list); \
  done

 # Extract the debs for each architecture
-# 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.
-# On amd64, we'll also have a /lib64 folder with ld-linux-x86-64.so.2, which is
-# already present in the runtime image.
 RUN \
  for arch in arm64 amd64; do \
    mkdir -p /install-${arch}/var/lib/dpkg/status.d/ && \
@@ -165,8 +160,6 @@ RUN \
      dpkg --ctrl-tarfile $deb | tar -Ox ./control > /install-${arch}/var/lib/dpkg/status.d/${package_name}; \
      dpkg --extract $deb /install-${arch}; \
    done; \
-    rsync -avr /install-${arch}/lib/ /install-${arch}/usr/lib; \
-    rm -rf /install-${arch}/lib /install-${arch}/lib64; \
  done


@@ -183,7 +176,14 @@ LABEL org.opencontainers.image.documentation='https://github.com/element-hq/syna
 LABEL org.opencontainers.image.source='https://github.com/element-hq/synapse.git'
 LABEL org.opencontainers.image.licenses='AGPL-3.0-or-later'

-COPY --from=runtime-deps /install-${TARGETARCH} /
+# On the runtime image, /lib is a symlink to /usr/lib, so we need to copy the
+# libraries to the right place, else the `COPY` won't work.
+# On amd64, we'll also have a /lib64 folder with ld-linux-x86-64.so.2, which is
+# already present in the runtime image.
+COPY --from=runtime-deps /install-${TARGETARCH}/lib /usr/lib
+COPY --from=runtime-deps /install-${TARGETARCH}/etc /etc
+COPY --from=runtime-deps /install-${TARGETARCH}/usr /usr
+COPY --from=runtime-deps /install-${TARGETARCH}/var /var
 COPY --from=builder /install /usr/local
 COPY ./docker/start.py /start.py
 COPY ./docker/conf /conf
@@ -2,18 +2,38 @@

 ARG SYNAPSE_VERSION=latest
 ARG FROM=matrixdotorg/synapse:$SYNAPSE_VERSION
+ARG DEBIAN_VERSION=bookworm
+ARG PYTHON_VERSION=3.12

-# first of all, we create a base image with an nginx which we can copy into the
+# first of all, we create a base image with dependencies which we can copy into the
 # target image. For repeated rebuilds, this is much faster than apt installing
 # each time.

-FROM docker.io/library/debian:bookworm-slim AS deps_base
+FROM ghcr.io/astral-sh/uv:python${PYTHON_VERSION}-${DEBIAN_VERSION} AS deps_base
+
+    # Tell apt to keep downloaded package files, as we're using cache mounts.
+    RUN rm -f /etc/apt/apt.conf.d/docker-clean; echo 'Binary::apt::APT::Keep-Downloaded-Packages "true";' > /etc/apt/apt.conf.d/keep-cache
+
    RUN \
       --mount=type=cache,target=/var/cache/apt,sharing=locked \
       --mount=type=cache,target=/var/lib/apt,sharing=locked \
      apt-get update -qq && \
      DEBIAN_FRONTEND=noninteractive apt-get install -yqq --no-install-recommends \
-          redis-server nginx-light
+          nginx-light
+
+    RUN \
+    # remove default page
+      rm /etc/nginx/sites-enabled/default && \
+    # have nginx log to stderr/out
+      ln -sf /dev/stdout /var/log/nginx/access.log && \
+      ln -sf /dev/stderr /var/log/nginx/error.log
+
+    # --link-mode=copy silences a warning as uv isn't able to do hardlinks between its cache
+    # (mounted as --mount=type=cache) and the target directory.
+    RUN --mount=type=cache,target=/root/.cache/uv \
+      uv pip install --link-mode=copy --prefix="/uv/usr/local" supervisor~=4.2
+
+    RUN mkdir -p /uv/etc/supervisor/conf.d

 # Similarly, a base to copy the redis server from.
 #
@@ -21,31 +41,21 @@ FROM docker.io/library/debian:bookworm-slim AS deps_base
 # which makes it much easier to copy (but we need to make sure we use an image
 # based on the same debian version as the synapse image, to make sure we get
 # the expected version of libc.
-FROM docker.io/library/redis:7-bookworm AS redis_base
+FROM docker.io/library/redis:7-${DEBIAN_VERSION} AS redis_base

 # now build the final image, based on the the regular Synapse docker image
 FROM $FROM

-    # Install supervisord with pip instead of apt, to avoid installing a second
-    # copy of python.
-    RUN --mount=type=cache,target=/root/.cache/pip \
-        pip install supervisor~=4.2
-    RUN mkdir -p /etc/supervisor/conf.d
-
-    # Copy over redis and nginx
+    # Copy over dependencies
    COPY --from=redis_base /usr/local/bin/redis-server /usr/local/bin
-
+    COPY --from=deps_base /uv /
    COPY --from=deps_base /usr/sbin/nginx /usr/sbin
    COPY --from=deps_base /usr/share/nginx /usr/share/nginx
    COPY --from=deps_base /usr/lib/nginx /usr/lib/nginx
    COPY --from=deps_base /etc/nginx /etc/nginx
-    RUN rm /etc/nginx/sites-enabled/default
-    RUN mkdir /var/log/nginx /var/lib/nginx
-    RUN chown www-data /var/lib/nginx
-
-    # have nginx log to stderr/out
-    RUN ln -sf /dev/stdout /var/log/nginx/access.log
-    RUN ln -sf /dev/stderr /var/log/nginx/error.log
+    COPY --from=deps_base /var/log/nginx /var/log/nginx
+    # chown to allow non-root user to write to http-*-temp-path dirs
+    COPY --from=deps_base --chown=www-data:root /var/lib/nginx /var/lib/nginx

    # Copy Synapse worker, nginx and supervisord configuration template files
    COPY ./docker/conf-workers/* /conf/
@@ -64,4 +74,4 @@ FROM $FROM
    # Replace the healthcheck with one which checks *all* the workers. The script
    # is generated by configure_workers_and_start.py.
    HEALTHCHECK --start-period=5s --interval=15s --timeout=5s \
-        CMD /bin/sh /healthcheck.sh
+        CMD ["/healthcheck.sh"]
@@ -9,6 +9,9 @@
 ARG SYNAPSE_VERSION=latest
 # This is an intermediate image, to be built locally (not pulled from a registry).
 ARG FROM=matrixdotorg/synapse-workers:$SYNAPSE_VERSION
+ARG DEBIAN_VERSION=bookworm
+
+FROM docker.io/library/postgres:13-${DEBIAN_VERSION} AS postgres_base

 FROM $FROM
 # First of all, we copy postgres server from the official postgres image,
@@ -20,9 +23,9 @@ FROM $FROM
 # the same debian version as Synapse's docker image (so the versions of the
 # shared libraries match).
 RUN adduser --system --uid 999 postgres --home /var/lib/postgresql
-COPY --from=docker.io/library/postgres:13-bookworm /usr/lib/postgresql /usr/lib/postgresql
-COPY --from=docker.io/library/postgres:13-bookworm /usr/share/postgresql /usr/share/postgresql
-RUN mkdir /var/run/postgresql && chown postgres /var/run/postgresql
+COPY --from=postgres_base /usr/lib/postgresql /usr/lib/postgresql
+COPY --from=postgres_base /usr/share/postgresql /usr/share/postgresql
+COPY --from=postgres_base --chown=postgres /var/run/postgresql /var/run/postgresql
 ENV PATH="${PATH}:/usr/lib/postgresql/13/bin"
 ENV PGDATA=/var/lib/postgresql/data

@@ -55,4 +58,4 @@ ENTRYPOINT ["/start_for_complement.sh"]

 # Update the healthcheck to have a shorter check interval
 HEALTHCHECK --start-period=5s --interval=1s --timeout=1s \
-    CMD /bin/sh /healthcheck.sh
+    CMD ["/healthcheck.sh"]
@@ -5,12 +5,12 @@
 set -e

 echo "Complement Synapse launcher"
-echo "  Args: $@"
+echo "  Args: $*"
 echo "  Env: SYNAPSE_COMPLEMENT_DATABASE=$SYNAPSE_COMPLEMENT_DATABASE SYNAPSE_COMPLEMENT_USE_WORKERS=$SYNAPSE_COMPLEMENT_USE_WORKERS SYNAPSE_COMPLEMENT_USE_ASYNCIO_REACTOR=$SYNAPSE_COMPLEMENT_USE_ASYNCIO_REACTOR"

 function log {
-    d=$(date +"%Y-%m-%d %H:%M:%S,%3N")
-    echo "$d $@"
+    d=$(printf '%(%Y-%m-%d %H:%M:%S)T,%.3s\n' ${EPOCHREALTIME/./ })
+    echo "$d $*"
 }

 # Set the server name of the homeserver
@@ -103,12 +103,11 @@ fi
 # Note that both the key and certificate are in PEM format (not DER).

 # First generate a configuration file to set up a Subject Alternative Name.
-cat > /conf/server.tls.conf <<EOF
+echo "\
 .include /etc/ssl/openssl.cnf

 [SAN]
-subjectAltName=DNS:${SERVER_NAME}
-EOF
+subjectAltName=DNS:${SERVER_NAME}" > /conf/server.tls.conf

 # Generate an RSA key
 openssl genrsa -out /conf/server.tls.key 2048
@@ -123,12 +122,12 @@ openssl x509 -req -in /conf/server.tls.csr \
  -out /conf/server.tls.crt -extfile /conf/server.tls.conf -extensions SAN

 # Assert that we have a Subject Alternative Name in the certificate.
-# (grep will exit with 1 here if there isn't a SAN in the certificate.)
-openssl x509 -in /conf/server.tls.crt -noout -text | grep DNS:
+# (the test will exit with 1 here if there isn't a SAN in the certificate.)
+[[ $(openssl x509 -in /conf/server.tls.crt -noout -text) == *DNS:* ]]

 export SYNAPSE_TLS_CERT=/conf/server.tls.crt
 export SYNAPSE_TLS_KEY=/conf/server.tls.key

 # Run the script that writes the necessary config files and starts supervisord, which in turn
 # starts everything else
-exec /configure_workers_and_start.py
+exec /configure_workers_and_start.py "$@"
@@ -1,4 +1,4 @@
-#!/usr/bin/env python
+#!/usr/local/bin/python
 #
 # This file is licensed under the Affero General Public License (AGPL) version 3.
 #
@@ -202,6 +202,7 @@ WORKERS_CONFIG: Dict[str, Dict[str, Any]] = {
        "app": "synapse.app.generic_worker",
        "listener_resources": ["federation"],
        "endpoint_patterns": [
+            "^/_matrix/federation/v1/version$",
            "^/_matrix/federation/(v1|v2)/event/",
            "^/_matrix/federation/(v1|v2)/state/",
            "^/_matrix/federation/(v1|v2)/state_ids/",
@@ -376,9 +377,11 @@ def convert(src: str, dst: str, **template_vars: object) -> None:
    #
    # We use append mode in case the files have already been written to by something else
    # (for instance, as part of the instructions in a dockerfile).
+    exists = os.path.isfile(dst)
    with open(dst, "a") as outfile:
        # In case the existing file doesn't end with a newline
-        outfile.write("\n")
+        if exists:
+            outfile.write("\n")

        outfile.write(rendered)

@@ -604,7 +607,7 @@ def generate_base_homeserver_config() -> None:
    # start.py already does this for us, so just call that.
    # note that this script is copied in in the official, monolith dockerfile
    os.environ["SYNAPSE_HTTP_PORT"] = str(MAIN_PROCESS_HTTP_LISTENER_PORT)
-    subprocess.run(["/usr/local/bin/python", "/start.py", "migrate_config"], check=True)
+    subprocess.run([sys.executable, "/start.py", "migrate_config"], check=True)


 def parse_worker_types(
@@ -998,6 +1001,7 @@ def generate_worker_files(
        "/healthcheck.sh",
        healthcheck_urls=healthcheck_urls,
    )
+    os.chmod("/healthcheck.sh", 0o755)

    # Ensure the logging directory exists
    log_dir = data_dir + "/logs"
@@ -10,6 +10,9 @@
 # '-W interactive' is a `mawk` extension which disables buffering on stdout and sets line-buffered reads on
 # stdin. The effect is that the output is flushed after each line, rather than being batched, which helps reduce
 # confusion due to to interleaving of the different processes.
-exec 1> >(awk -W interactive '{print "'"${SUPERVISOR_PROCESS_NAME}"' | "$0 }' >&1)
-exec 2> >(awk -W interactive '{print "'"${SUPERVISOR_PROCESS_NAME}"' | "$0 }' >&2)
+prefixer() {
+    mawk -W interactive '{printf("%s | %s\n", ENVIRON["SUPERVISOR_PROCESS_NAME"], $0); fflush() }'
+}
+exec 1> >(prefixer)
+exec 2> >(prefixer >&2)
 exec "$@"
@@ -46,6 +46,14 @@ to any local media, and any locally-cached copies of remote media.

 The media file itself (and any thumbnails) is not deleted from the server.

+Since Synapse 1.128.0, hashes of uploaded media are tracked. If this media
+is quarantined, Synapse will:
+
+ - Quarantine any media with a matching hash that has already been uploaded.
+ - Quarantine any future media.
+ - Quarantine any existing cached remote media.
+ - Quarantine any future remote media.
+
 ## Quarantining media by ID

 This API quarantines a single piece of local or remote media.
@@ -0,0 +1,54 @@
+# Show scheduled tasks
+
+This API returns information about scheduled tasks.
+
+To use it, you will need to authenticate by providing an `access_token`
+for a server admin: see [Admin API](../usage/administration/admin_api/).
+
+The api is:
+```
+GET /_synapse/admin/v1/scheduled_tasks
+```
+
+It returns a JSON body like the following:
+
+```json
+{
+    "scheduled_tasks": [
+        {
+            "id": "GSA124oegf1",
+            "action": "shutdown_room",
+            "status": "complete",
+            "timestamp_ms": 23423523,
+            "resource_id": "!roomid",
+            "result": "some result",
+            "error": null
+        }
+      ]
+}
+```
+
+**Query parameters:**
+
+* `action_name`: string - Is optional. Returns only the scheduled tasks with the given action name.
+* `resource_id`: string - Is optional. Returns only the scheduled tasks with the given resource id.
+* `status`: string - Is optional. Returns only the scheduled tasks matching the given status, one of
+    - "scheduled" - Task is scheduled but not active
+    - "active" - Task is active and probably running, and if not will be run on next scheduler loop run
+    - "complete" - Task has completed successfully
+    - "failed" - Task is over and either returned a failed status, or had an exception
+
+* `max_timestamp`: int - Is optional. Returns only the scheduled tasks with a timestamp inferior to the specified one.
+
+**Response**
+
+The following fields are returned in the JSON response body along with a `200` HTTP status code:
+
+* `id`: string - ID of scheduled task.
+* `action`: string - The name of the scheduled task's action.
+* `status`: string - The status of the scheduled task.
+* `timestamp_ms`: integer - The timestamp (in milliseconds since the unix epoch) of the given task - If the status is "scheduled" then this represents when it should be launched.
+  Otherwise it represents the last time this task got a change of state.
+* `resource_id`: Optional string - The resource id of the scheduled task, if it possesses one
+* `result`: Optional Json - Any result of the scheduled task, if given
+* `error`: Optional string - If the task has the status "failed", the error associated with this failure
@@ -162,7 +162,7 @@ by a unique name, the current status (stored in JSON), and some dependency infor
 * Whether the update requires a previous update to be complete.
 * A rough ordering for which to complete updates.

-A new background updates needs to be added to the `background_updates` table:
+A new background update needs to be added to the `background_updates` table:

 ```sql
 INSERT INTO background_updates (ordering, update_name, depends_on, progress_json) VALUES
@@ -150,6 +150,28 @@ $ poetry shell
 $ poetry install --extras all
 ```

+If you want to go even further and remove the Poetry caches:
+
+```shell
+# Find your Poetry cache directory
+# Docs: https://github.com/python-poetry/poetry/blob/main/docs/configuration.md#cache-directory
+$ poetry config cache-dir
+
+# Remove packages from all cached repositories
+$ poetry cache clear --all .
+
+# Go completely nuclear and clear out everything Poetry cache related
+# including the wheel artifacts which is not covered by the above command
+# (see https://github.com/python-poetry/poetry/issues/10304)
+#
+# This is necessary in order to rebuild or fetch new wheels. For example, if you update
+# the `icu` library in on your system, you will need to rebuild the PyICU Python package
+# in order to incorporate the correct dynamically linked library locations otherwise you
+# will run into errors like: `ImportError: libicui18n.so.75: cannot open shared object file: No such file or directory`
+$ rm -rf $(poetry config cache-dir)
+```
+
+
 ## ...run a command in the `poetry` virtualenv?

 Use `poetry run cmd args` when you need the python virtualenv context.
@@ -187,7 +209,7 @@ useful.
 ## ...add a new dependency?

 Either:
- manually update `pyproject.toml`; then `poetry lock --no-update`; or else
+- manually update `pyproject.toml`; then `poetry lock`; or else
 - `poetry add packagename`. See `poetry add --help`; note the `--dev`,
  `--extras` and `--optional` flags in particular.

@@ -202,12 +224,12 @@ poetry remove packagename
 ```

 ought to do the trick. Alternatively, manually update `pyproject.toml` and
-`poetry lock --no-update`. Include the updated `pyproject.toml` and `poetry.lock`
+`poetry lock`. Include the updated `pyproject.toml` and `poetry.lock`
 files in your commit.

 ## ...update the version range for an existing dependency?

-Best done by manually editing `pyproject.toml`, then `poetry lock --no-update`.
+Best done by manually editing `pyproject.toml`, then `poetry lock`.
 Include the updated `pyproject.toml` and `poetry.lock` in your commit.

 ## ...update a dependency in the locked environment?
@@ -233,7 +255,7 @@ poetry add packagename==1.2.3

 # Get poetry to recompute the content-hash of pyproject.toml without changing
 # the locked package versions.
-poetry lock --no-update
+poetry lock
 ```

 Either way, include the updated `poetry.lock` file in your commit.
@@ -353,6 +353,8 @@ callback returns `False`, Synapse falls through to the next one. The value of th
 callback that does not return `False` will be used. If this happens, Synapse will not call
 any of the subsequent implementations of this callback.

+Note that this check is applied to federation invites as of Synapse v1.130.0.
+

 ### `check_login_for_spam`

@@ -23,6 +23,7 @@ such as [Github][github-idp].
 [auth0]: https://auth0.com/
 [authentik]: https://goauthentik.io/
 [lemonldap]: https://lemonldap-ng.org/
+[pocket-id]: https://pocket-id.org/
 [okta]: https://www.okta.com/
 [dex-idp]: https://github.com/dexidp/dex
 [keycloak-idp]: https://www.keycloak.org/docs/latest/server_admin/#sso-protocols
@@ -624,6 +625,32 @@ oidc_providers:

 Note that the fields `client_id` and `client_secret` are taken from the CURL response above.

+### Pocket ID
+
+[Pocket ID][pocket-id] is a simple OIDC provider that allows users to authenticate with their passkeys.
+1. Go to `OIDC Clients`
+2. Click on `Add OIDC Client`
+3. Add a name, for example `Synapse`
+4. Add `"https://auth.example.org/_synapse/client/oidc/callback` to `Callback URLs`  # Replace `auth.example.org` with your domain
+5. Click on `Save`
+6. Note down your `Client ID` and `Client secret`, these will be used later
+
+Synapse config:
+
+```yaml
+oidc_providers:
+  - idp_id: pocket_id
+    idp_name: Pocket ID
+    issuer: "https://auth.example.org/" # Replace with your domain
+    client_id: "your-client-id" # Replace with the "Client ID" you noted down before
+    client_secret: "your-client-secret" # Replace with the "Client secret" you noted down before
+    scopes: ["openid", "profile"]
+    user_mapping_provider:
+      config:
+        localpart_template: "{{ user.preferred_username }}"
+        display_name_template: "{{ user.name }}"
+```
+
 ### Shibboleth with OIDC Plugin

 [Shibboleth](https://www.shibboleth.net/) is an open Standard IdP solution widely used by Universities.
@@ -10,7 +10,7 @@ As an example, a SSO service may return the email address
 to turn that into a displayname when creating a Matrix user for this individual.
 It may choose `John Smith`, or `Smith, John [Example.com]` or any number of
 variations. As each Synapse configuration may want something different, this is
-where SAML mapping providers come into play.
+where SSO mapping providers come into play.

 SSO mapping providers are currently supported for OpenID and SAML SSO
 configurations. Please see the details below for how to implement your own.
@@ -117,6 +117,16 @@ each upgrade are complete before moving on to the next upgrade, to avoid
 stacking them up. You can monitor the currently running background updates with
 [the Admin API](usage/administration/admin_api/background_updates.html#status).

+# Upgrading to v1.130.0
+
+## Documented endpoint which can be delegated to a federation worker
+
+The endpoint `^/_matrix/federation/v1/version$` can be delegated to a federation
+worker. This is not new behaviour, but had not been documented yet. The 
+[list of delegatable endpoints](workers.md#synapseappgeneric_worker) has 
+been updated to include it. Make sure to check your reverse proxy rules if you
+are using workers. 
+
 # Upgrading to v1.126.0

 ## Room list publication rules change
@@ -160,7 +160,7 @@ Using the following curl command:
 ```console
 curl -H 'Authorization: Bearer <access-token>' -X DELETE https://matrix.org/_matrix/client/r0/directory/room/<room-alias>
 ```
-`<access-token>` - can be obtained in riot by looking in the riot settings, down the bottom is:
+`<access-token>` - can be obtained in element by looking in All settings, clicking Help & About and down the bottom is:
 Access Token:\<click to reveal\>

 `<room-alias>` - the room alias, eg. #my_room:matrix.org this possibly needs to be URL encoded also, for example  %23my_room%3Amatrix.org
@@ -3672,6 +3672,9 @@ Options for each entry include:
 * `additional_authorization_parameters`: String to string dictionary that will be passed as
   additional parameters to the authorization grant URL.

+* `passthrough_authorization_parameters`: List of parameters that will be passed through from the redirect endpoint 
+   to the authorization grant URL.
+
 * `allow_existing_users`: set to true to allow a user logging in via OIDC to
   match a pre-existing account instead of failing. This could be used if
   switching from password logins to OIDC. Defaults to false.
@@ -3798,6 +3801,7 @@ oidc_providers:
    jwks_uri: "https://accounts.example.com/.well-known/jwks.json"
    additional_authorization_parameters:
      acr_values: 2fa
+    passthrough_authorization_parameters: ["login_hint"]
    skip_verification: true
    enable_registration: true
    user_mapping_provider:
@@ -4014,7 +4018,7 @@ This option has a number of sub-options. They are as follows:
 * `include_content`: Clients requesting push notifications can either have the body of
   the message sent in the notification poke along with other details
   like the sender, or just the event ID and room ID (`event_id_only`).
-   If clients choose the to have the body sent, this option controls whether the
+   If clients choose to have the body sent, this option controls whether the
   notification request includes the content of the event (other details
   like the sender are still included). If `event_id_only` is enabled, it
   has no effect.
@@ -4091,6 +4095,7 @@ This option has the following sub-options:
 * `prefer_local_users`: Defines whether to prefer local users in search query results.
   If set to true, local users are more likely to appear above remote users when searching the
   user directory. Defaults to false.
+* `exclude_remote_users`: If set to true, the search will only return local users. Defaults to false.
 * `show_locked_users`: Defines whether to show locked users in search query results. Defaults to false.

 Example configuration:
@@ -4099,6 +4104,7 @@ user_directory:
    enabled: false
    search_all_users: true
    prefer_local_users: true
+    exclude_remote_users: false
    show_locked_users: true
 ```
 ---
@@ -200,6 +200,7 @@ information.
    ^/_matrix/client/(api/v1|r0|v3)/rooms/[^/]+/initialSync$

    # Federation requests
+    ^/_matrix/federation/v1/version$
    ^/_matrix/federation/v1/event/
    ^/_matrix/federation/v1/state/
    ^/_matrix/federation/v1/state_ids/
@@ -249,13 +250,14 @@ information.
    ^/_matrix/client/(api/v1|r0|v3|unstable)/directory/room/.*$
    ^/_matrix/client/(r0|v3|unstable)/capabilities$
    ^/_matrix/client/(r0|v3|unstable)/notifications$
+    ^/_synapse/admin/v1/rooms/

    # Encryption requests
    ^/_matrix/client/(r0|v3|unstable)/keys/query$
    ^/_matrix/client/(r0|v3|unstable)/keys/changes$
    ^/_matrix/client/(r0|v3|unstable)/keys/claim$
    ^/_matrix/client/(r0|v3|unstable)/room_keys/
-    ^/_matrix/client/(r0|v3|unstable)/keys/upload/
+    ^/_matrix/client/(r0|v3|unstable)/keys/upload$

    # Registration/login requests
    ^/_matrix/client/(api/v1|r0|v3|unstable)/login$
@@ -280,6 +282,7 @@ Additionally, the following REST endpoints can be handled for GET requests:

    ^/_matrix/client/(api/v1|r0|v3|unstable)/pushrules/
    ^/_matrix/client/unstable/org.matrix.msc4140/delayed_events
+    ^/_matrix/client/(api/v1|r0|v3|unstable)/devices/

    # Account data requests
    ^/_matrix/client/(r0|v3|unstable)/.*/tags
@@ -320,6 +323,15 @@ For multiple workers not handling the SSO endpoints properly, see
 [#7530](https://github.com/matrix-org/synapse/issues/7530) and
 [#9427](https://github.com/matrix-org/synapse/issues/9427).

+Additionally, when MSC3861 is enabled (`experimental_features.msc3861.enabled`
+set to `true`), the following endpoints can be handled by the worker:
+
+    ^/_synapse/admin/v2/users/[^/]+$
+    ^/_synapse/admin/v1/username_available$
+    ^/_synapse/admin/v1/users/[^/]+/_allow_cross_signing_replacement_without_uia$
+    # Only the GET method:
+    ^/_synapse/admin/v1/users/[^/]+/devices$
+
 Note that a [HTTP listener](usage/configuration/config_documentation.md#listeners)
 with `client` and `federation` `resources` must be configured in the
 [`worker_listeners`](usage/configuration/config_documentation.md#worker_listeners)
@@ -1,3 +1,45 @@
+[project]
+name = "matrix-synapse"
+version = "1.129.0"
+description = "Homeserver for the Matrix decentralised comms protocol"
+authors = [
+    { name = "Matrix.org Team and Contributors", email = "packages@matrix.org" },
+]
+license = "AGPL-3.0-or-later"
+readme = "README.rst"
+classifiers = [
+    "Development Status :: 5 - Production/Stable",
+    "Topic :: Communications :: Chat",
+]
+dynamic = [
+    # TODO: Transition our `tool.poetry.dependencies` section to
+    # `project.dependencies`.
+    "dependencies",
+]
+requires-python = ">=3.9.0,<4.0"
+
+# https://python-poetry.org/docs/pyproject/#urls
+[project.urls]
+homepage = "https://github.com/element-hq/synapse"
+repository = "https://github.com/element-hq/synapse"
+documentation = "https://element-hq.github.io/synapse/latest"
+"Bug Tracker" = "https://github.com/element-hq/synapse/issues"
+
+# https://python-poetry.org/docs/pyproject/#scripts
+[project.scripts]
+synapse_homeserver = "synapse.app.homeserver:main"
+synapse_worker = "synapse.app.generic_worker:main"
+synctl = "synapse._scripts.synctl:main"
+export_signing_key = "synapse._scripts.export_signing_key:main"
+generate_config = "synapse._scripts.generate_config:main"
+generate_log_config = "synapse._scripts.generate_log_config:main"
+generate_signing_key = "synapse._scripts.generate_signing_key:main"
+hash_password = "synapse._scripts.hash_password:main"
+register_new_matrix_user = "synapse._scripts.register_new_matrix_user:main"
+synapse_port_db = "synapse._scripts.synapse_port_db:main"
+synapse_review_recent_signups = "synapse._scripts.review_recent_signups:main"
+update_synapse_database = "synapse._scripts.update_synapse_database:main"
+
 [tool.towncrier]
    package = "synapse"
    filename = "CHANGES.md"
@@ -96,20 +138,9 @@ manifest-path = "rust/Cargo.toml"
 module-name = "synapse.synapse_rust"

 [tool.poetry]
-name = "matrix-synapse"
-version = "1.126.0"
-description = "Homeserver for the Matrix decentralised comms protocol"
-authors = ["Matrix.org Team and Contributors <packages@matrix.org>"]
-license = "AGPL-3.0-or-later"
-readme = "README.rst"
-repository = "https://github.com/element-hq/synapse"
 packages = [
    { include = "synapse" },
 ]
-classifiers = [
-    "Development Status :: 5 - Production/Stable",
-    "Topic :: Communications :: Chat",
-]
 include = [
    { path = "AUTHORS.rst", format = "sdist" },
    { path = "book.toml", format = "sdist" },
@@ -139,24 +170,7 @@ exclude = [
 script = "build_rust.py"
 generate-setup-file = true

-[tool.poetry.scripts]
-synapse_homeserver = "synapse.app.homeserver:main"
-synapse_worker = "synapse.app.generic_worker:main"
-synctl = "synapse._scripts.synctl:main"
-
-export_signing_key = "synapse._scripts.export_signing_key:main"
-generate_config = "synapse._scripts.generate_config:main"
-generate_log_config = "synapse._scripts.generate_log_config:main"
-generate_signing_key = "synapse._scripts.generate_signing_key:main"
-hash_password = "synapse._scripts.hash_password:main"
-register_new_matrix_user = "synapse._scripts.register_new_matrix_user:main"
-synapse_port_db = "synapse._scripts.synapse_port_db:main"
-synapse_review_recent_signups = "synapse._scripts.review_recent_signups:main"
-update_synapse_database = "synapse._scripts.update_synapse_database:main"
-
 [tool.poetry.dependencies]
-python = "^3.9.0"
-
 # Mandatory Dependencies
 # ----------------------
 # we use the TYPE_CHECKER.redefine method added in jsonschema 3.0.0
@@ -252,7 +266,8 @@ parameterized = { version = ">=0.7.4", optional = true }
 idna = { version = ">=2.5", optional = true }
 pyicu = { version = ">=2.10.2", optional = true }

-[tool.poetry.extras]
+# https://python-poetry.org/docs/pyproject/#optional-dependencies
+[project.optional-dependencies]
 # NB: Packages that should be part of `pip install matrix-synapse[all]` need to be specified
 # twice: once here, and once in the `all` extra.
 matrix-synapse-ldap3 = ["matrix-synapse-ldap3"]
@@ -315,7 +330,7 @@ all = [
    #   - systemd: this is a system-based requirement
 ]

-[tool.poetry.dev-dependencies]
+[tool.poetry.group.dev.dependencies]
 # We pin development dependencies in poetry.lock so that our tests don't start
 # failing on new releases. Keeping lower bounds loose here means that dependabot
 # can bump versions without having to update the content-hash in the lockfile.
@@ -370,7 +385,7 @@ tomli = ">=1.2.3"
 # runtime errors caused by build system changes.
 # We are happy to raise these upper bounds upon request,
 # provided we check that it's safe to do so (i.e. that CI passes).
-requires = ["poetry-core>=1.1.0,<=1.9.1", "setuptools_rust>=1.3,<=1.10.2"]
+requires = ["poetry-core>=2.0.0,<=2.1.3", "setuptools_rust>=1.3,<=1.10.2"]
 build-backend = "poetry.core.masonry.api"


@@ -385,6 +400,9 @@ build-backend = "poetry.core.masonry.api"
 #  - PyPy on Aarch64 and musllinux on aarch64: too slow to build.
 #    c.f. https://github.com/matrix-org/synapse/pull/14259
 skip = "cp36* cp37* cp38* pp37* pp38* *-musllinux_i686 pp*aarch64 *-musllinux_aarch64"
+# Enable non-default builds.
+# "pypy" used to be included by default up until cibuildwheel 3.
+enable = "pypy"

 # We need a rust compiler.
 #
@@ -592,7 +592,7 @@ def _wait_for_actions(gh_token: Optional[str]) -> None:
        if all(
            workflow["status"] != "in_progress" for workflow in resp["workflow_runs"]
        ):
-            success = (
+            success = all(
                workflow["status"] == "completed" for workflow in resp["workflow_runs"]
            )
            if success:
@@ -128,6 +128,7 @@ BOOLEAN_COLUMNS = {
    "pushers": ["enabled"],
    "redactions": ["have_censored"],
    "remote_media_cache": ["authenticated"],
+    "room_memberships": ["participant"],
    "room_stats_state": ["is_federatable"],
    "rooms": ["is_public", "has_auth_chain_index"],
    "sliding_sync_joined_rooms": ["is_encrypted"],
@@ -191,6 +192,11 @@ APPEND_ONLY_TABLES = [


 IGNORED_TABLES = {
+    # Porting the auto generated sequence in this table is non-trivial.
+    # None of the entries in this list are mandatory for Synapse to keep working.
+    # If state group disk space is an issue after the port, the
+    # `mark_unreferenced_state_groups_for_deletion_bg_update` background task can be run again.
+    "state_groups_pending_deletion",
    # We don't port these tables, as they're a faff and we can regenerate
    # them anyway.
    "user_directory",
@@ -216,6 +222,15 @@ IGNORED_TABLES = {
 }


+# These background updates will not be applied upon creation of the postgres database.
+IGNORED_BACKGROUND_UPDATES = {
+    # Reapplying this background update to the postgres database is unnecessary after
+    # already having waited for the SQLite database to complete all running background
+    # updates.
+    "mark_unreferenced_state_groups_for_deletion_bg_update",
+}
+
+
 # Error returned by the run function. Used at the top-level part of the script to
 # handle errors and return codes.
 end_error: Optional[str] = None
@@ -687,6 +702,20 @@ class Porter:
        # 0 means off. 1 means full. 2 means incremental.
        return autovacuum_setting != 0

+    async def remove_ignored_background_updates_from_database(self) -> None:
+        def _remove_delete_unreferenced_state_groups_bg_updates(
+            txn: LoggingTransaction,
+        ) -> None:
+            txn.execute(
+                "DELETE FROM background_updates WHERE update_name = ANY(?)",
+                (list(IGNORED_BACKGROUND_UPDATES),),
+            )
+
+        await self.postgres_store.db_pool.runInteraction(
+            "remove_delete_unreferenced_state_groups_bg_updates",
+            _remove_delete_unreferenced_state_groups_bg_updates,
+        )
+
    async def run(self) -> None:
        """Ports the SQLite database to a PostgreSQL database.

@@ -732,6 +761,8 @@ class Porter:
                self.hs_config.database.get_single_database()
            )

+            await self.remove_ignored_background_updates_from_database()
+
            await self.run_background_updates_on_postgres()

            self.progress.set_state("Creating port tables")
@@ -19,6 +19,7 @@
 #
 #
 import logging
+from dataclasses import dataclass
 from typing import TYPE_CHECKING, Any, Callable, Dict, List, Optional
 from urllib.parse import urlencode

@@ -38,15 +39,16 @@ from synapse.api.errors import (
    HttpResponseException,
    InvalidClientTokenError,
    OAuthInsufficientScopeError,
-    StoreError,
    SynapseError,
    UnrecognizedRequestError,
 )
 from synapse.http.site import SynapseRequest
 from synapse.logging.context import make_deferred_yieldable
+from synapse.logging.opentracing import active_span, force_tracing, start_active_span
 from synapse.types import Requester, UserID, create_requester
 from synapse.util import json_decoder
 from synapse.util.caches.cached_call import RetryOnExceptionCachedCall
+from synapse.util.caches.response_cache import ResponseCache, ResponseCacheContext

 if TYPE_CHECKING:
    from synapse.rest.admin.experimental_features import ExperimentalFeature
@@ -76,6 +78,61 @@ def scope_to_list(scope: str) -> List[str]:
    return scope.strip().split(" ")


+@dataclass
+class IntrospectionResult:
+    _inner: IntrospectionToken
+
+    # when we retrieved this token,
+    # in milliseconds since the Unix epoch
+    retrieved_at_ms: int
+
+    def is_active(self, now_ms: int) -> bool:
+        if not self._inner.get("active"):
+            return False
+
+        expires_in = self._inner.get("expires_in")
+        if expires_in is None:
+            return True
+        if not isinstance(expires_in, int):
+            raise InvalidClientTokenError("token `expires_in` is not an int")
+
+        absolute_expiry_ms = expires_in * 1000 + self.retrieved_at_ms
+        return now_ms < absolute_expiry_ms
+
+    def get_scope_list(self) -> List[str]:
+        value = self._inner.get("scope")
+        if not isinstance(value, str):
+            return []
+        return scope_to_list(value)
+
+    def get_sub(self) -> Optional[str]:
+        value = self._inner.get("sub")
+        if not isinstance(value, str):
+            return None
+        return value
+
+    def get_username(self) -> Optional[str]:
+        value = self._inner.get("username")
+        if not isinstance(value, str):
+            return None
+        return value
+
+    def get_name(self) -> Optional[str]:
+        value = self._inner.get("name")
+        if not isinstance(value, str):
+            return None
+        return value
+
+    def get_device_id(self) -> Optional[str]:
+        value = self._inner.get("device_id")
+        if value is not None and not isinstance(value, str):
+            raise AuthError(
+                500,
+                "Invalid device ID in introspection result",
+            )
+        return value
+
+
 class PrivateKeyJWTWithKid(PrivateKeyJWT):  # type: ignore[misc]
    """An implementation of the private_key_jwt client auth method that includes a kid header.

@@ -120,6 +177,34 @@ class MSC3861DelegatedAuth(BaseAuth):
        self._http_client = hs.get_proxied_http_client()
        self._hostname = hs.hostname
        self._admin_token: Callable[[], Optional[str]] = self._config.admin_token
+        self._force_tracing_for_users = hs.config.tracing.force_tracing_for_users
+
+        # # Token Introspection Cache
+        # This remembers what users/devices are represented by which access tokens,
+        # in order to reduce overall system load:
+        # - on Synapse (as requests are relatively expensive)
+        # - on the network
+        # - on MAS
+        #
+        # Since there is no invalidation mechanism currently,
+        # the entries expire after 2 minutes.
+        # This does mean tokens can be treated as valid by Synapse
+        # for longer than reality.
+        #
+        # Ideally, tokens should logically be invalidated in the following circumstances:
+        # - If a session logout happens.
+        #   In this case, MAS will delete the device within Synapse
+        #   anyway and this is good enough as an invalidation.
+        # - If the client refreshes their token in MAS.
+        #   In this case, the device still exists and it's not the end of the world for
+        #   the old access token to continue working for a short time.
+        self._introspection_cache: ResponseCache[str] = ResponseCache(
+            self._clock,
+            "token_introspection",
+            timeout_ms=120_000,
+            # don't log because the keys are access tokens
+            enable_logging=False,
+        )

        self._issuer_metadata = RetryOnExceptionCachedCall[OpenIDProviderMetadata](
            self._load_metadata
@@ -193,7 +278,9 @@ class MSC3861DelegatedAuth(BaseAuth):
        metadata = await self._issuer_metadata.get()
        return metadata.get("introspection_endpoint")

-    async def _introspect_token(self, token: str) -> IntrospectionToken:
+    async def _introspect_token(
+        self, token: str, cache_context: ResponseCacheContext[str]
+    ) -> IntrospectionResult:
        """
        Send a token to the introspection endpoint and returns the introspection response

@@ -209,6 +296,8 @@ class MSC3861DelegatedAuth(BaseAuth):
        Returns:
            The introspection response
        """
+        # By default, we shouldn't cache the result unless we know it's valid
+        cache_context.should_cache = False
        introspection_endpoint = await self._introspection_endpoint()
        raw_headers: Dict[str, str] = {
            "Content-Type": "application/x-www-form-urlencoded",
@@ -266,7 +355,11 @@ class MSC3861DelegatedAuth(BaseAuth):
                "The introspection endpoint returned an invalid JSON response."
            )

-        return IntrospectionToken(**resp)
+        # We had a valid response, so we can cache it
+        cache_context.should_cache = True
+        return IntrospectionResult(
+            IntrospectionToken(**resp), retrieved_at_ms=self._clock.time_msec()
+        )

    async def is_server_admin(self, requester: Requester) -> bool:
        return "urn:synapse:admin:*" in requester.scope
@@ -277,6 +370,55 @@ class MSC3861DelegatedAuth(BaseAuth):
        allow_guest: bool = False,
        allow_expired: bool = False,
        allow_locked: bool = False,
+    ) -> Requester:
+        """Get a registered user's ID.
+
+        Args:
+            request: An HTTP request with an access_token query parameter.
+            allow_guest: If False, will raise an AuthError if the user making the
+                request is a guest.
+            allow_expired: If True, allow the request through even if the account
+                is expired, or session token lifetime has ended. Note that
+                /login will deliver access tokens regardless of expiration.
+
+        Returns:
+            Resolves to the requester
+        Raises:
+            InvalidClientCredentialsError if no user by that token exists or the token
+                is invalid.
+            AuthError if access is denied for the user in the access token
+        """
+        parent_span = active_span()
+        with start_active_span("get_user_by_req"):
+            requester = await self._wrapped_get_user_by_req(
+                request, allow_guest, allow_expired, allow_locked
+            )
+
+            if parent_span:
+                if requester.authenticated_entity in self._force_tracing_for_users:
+                    # request tracing is enabled for this user, so we need to force it
+                    # tracing on for the parent span (which will be the servlet span).
+                    #
+                    # It's too late for the get_user_by_req span to inherit the setting,
+                    # so we also force it on for that.
+                    force_tracing()
+                    force_tracing(parent_span)
+                parent_span.set_tag(
+                    "authenticated_entity", requester.authenticated_entity
+                )
+                parent_span.set_tag("user_id", requester.user.to_string())
+                if requester.device_id is not None:
+                    parent_span.set_tag("device_id", requester.device_id)
+                if requester.app_service is not None:
+                    parent_span.set_tag("appservice_id", requester.app_service.id)
+            return requester
+
+    async def _wrapped_get_user_by_req(
+        self,
+        request: SynapseRequest,
+        allow_guest: bool = False,
+        allow_expired: bool = False,
+        allow_locked: bool = False,
    ) -> Requester:
        access_token = self.get_access_token_from_request(request)

@@ -344,7 +486,9 @@ class MSC3861DelegatedAuth(BaseAuth):
            )

        try:
-            introspection_result = await self._introspect_token(token)
+            introspection_result = await self._introspection_cache.wrap(
+                token, self._introspect_token, token, cache_context=True
+            )
        except Exception:
            logger.exception("Failed to introspect token")
            raise SynapseError(503, "Unable to introspect the access token")
@@ -353,11 +497,11 @@ class MSC3861DelegatedAuth(BaseAuth):

        # TODO: introspection verification should be more extensive, especially:
        #   - verify the audience
-        if not introspection_result.get("active"):
+        if not introspection_result.is_active(self._clock.time_msec()):
            raise InvalidClientTokenError("Token is not active")

        # Let's look at the scope
-        scope: List[str] = scope_to_list(introspection_result.get("scope", ""))
+        scope: List[str] = introspection_result.get_scope_list()

        # Determine type of user based on presence of particular scopes
        has_user_scope = SCOPE_MATRIX_API in scope
@@ -367,7 +511,7 @@ class MSC3861DelegatedAuth(BaseAuth):
            raise InvalidClientTokenError("No scope in token granting user rights")

        # Match via the sub claim
-        sub: Optional[str] = introspection_result.get("sub")
+        sub = introspection_result.get_sub()
        if sub is None:
            raise InvalidClientTokenError(
                "Invalid sub claim in the introspection result"
@@ -380,29 +524,20 @@ class MSC3861DelegatedAuth(BaseAuth):
            # If we could not find a user via the external_id, it either does not exist,
            # or the external_id was never recorded

-            # TODO: claim mapping should be configurable
-            username: Optional[str] = introspection_result.get("username")
-            if username is None or not isinstance(username, str):
+            username = introspection_result.get_username()
+            if username is None:
                raise AuthError(
                    500,
                    "Invalid username claim in the introspection result",
                )
            user_id = UserID(username, self._hostname)

-            # First try to find a user from the username claim
+            # Try to find a user from the username claim
            user_info = await self.store.get_user_by_id(user_id=user_id.to_string())
            if user_info is None:
-                # If the user does not exist, we should create it on the fly
-                # TODO: we could use SCIM to provision users ahead of time and listen
-                # for SCIM SET events if those ever become standard:
-                # https://datatracker.ietf.org/doc/html/draft-hunt-scim-notify-00
-
-                # TODO: claim mapping should be configurable
-                # If present, use the name claim as the displayname
-                name: Optional[str] = introspection_result.get("name")
-
-                await self.store.register_user(
-                    user_id=user_id.to_string(), create_profile_with_displayname=name
+                raise AuthError(
+                    500,
+                    "User not found",
                )

            # And record the sub as external_id
@@ -414,15 +549,8 @@ class MSC3861DelegatedAuth(BaseAuth):

        # MAS 0.15+ will give us the device ID as an explicit value for compatibility sessions
        # If present, we get it from here, if not we get it in thee scope
-        device_id = introspection_result.get("device_id")
-        if device_id is not None:
-            # We got the device ID explicitly, just sanity check that it's a string
-            if not isinstance(device_id, str):
-                raise AuthError(
-                    500,
-                    "Invalid device ID in introspection result",
-                )
-        else:
+        device_id = introspection_result.get_device_id()
+        if device_id is None:
            # Find device_ids in scope
            # We only allow a single device_id in the scope, so we find them all in the
            # scope list, and raise if there are more than one. The OIDC server should be
@@ -449,17 +577,10 @@ class MSC3861DelegatedAuth(BaseAuth):
                    "Invalid device ID in introspection result",
                )

-            # Create the device on the fly if it does not exist
-            try:
-                await self.store.get_device(
-                    user_id=user_id.to_string(), device_id=device_id
-                )
-            except StoreError:
-                await self.store.store_device(
-                    user_id=user_id.to_string(),
-                    device_id=device_id,
-                    initial_device_display_name="OIDC-native client",
-                )
+            # Make sure the device exists
+            await self.store.get_device(
+                user_id=user_id.to_string(), device_id=device_id
+            )

        # TODO: there is a few things missing in the requester here, which still need
        # to be figured out, like:
@@ -29,8 +29,13 @@ from typing import Final
 # the max size of a (canonical-json-encoded) event
 MAX_PDU_SIZE = 65536

-# the "depth" field on events is limited to 2**63 - 1
-MAX_DEPTH = 2**63 - 1
+# Max/min size of ints in canonical JSON
+CANONICALJSON_MAX_INT = (2**53) - 1
+CANONICALJSON_MIN_INT = -CANONICALJSON_MAX_INT
+
+# the "depth" field on events is limited to the same as what
+# canonicaljson accepts
+MAX_DEPTH = CANONICALJSON_MAX_INT

 # the maximum length for a room alias is 255 characters
 MAX_ALIAS_LENGTH = 255
@@ -70,6 +70,7 @@ class Codes(str, Enum):
    THREEPID_NOT_FOUND = "M_THREEPID_NOT_FOUND"
    THREEPID_DENIED = "M_THREEPID_DENIED"
    INVALID_USERNAME = "M_INVALID_USERNAME"
+    THREEPID_MEDIUM_NOT_SUPPORTED = "M_THREEPID_MEDIUM_NOT_SUPPORTED"
    SERVER_NOT_TRUSTED = "M_SERVER_NOT_TRUSTED"
    CONSENT_NOT_GIVEN = "M_CONSENT_NOT_GIVEN"
    CANNOT_LEAVE_SERVER_NOTICE_ROOM = "M_CANNOT_LEAVE_SERVER_NOTICE_ROOM"
@@ -20,8 +20,7 @@
 #
 #

-from collections import OrderedDict
-from typing import Hashable, Optional, Tuple
+from typing import Dict, Hashable, Optional, Tuple

 from synapse.api.errors import LimitExceededError
 from synapse.config.ratelimiting import RatelimitSettings
@@ -80,12 +79,14 @@ class Ratelimiter:
        self.store = store
        self._limiter_name = cfg.key

-        # An ordered dictionary representing the token buckets tracked by this rate
+        # A dictionary representing the token buckets tracked by this rate
        # limiter. Each entry maps a key of arbitrary type to a tuple representing:
        #   * The number of tokens currently in the bucket,
        #   * The time point when the bucket was last completely empty, and
        #   * The rate_hz (leak rate) of this particular bucket.
-        self.actions: OrderedDict[Hashable, Tuple[float, float, float]] = OrderedDict()
+        self.actions: Dict[Hashable, Tuple[float, float, float]] = {}
+
+        self.clock.looping_call(self._prune_message_counts, 60 * 1000)

    def _get_key(
        self, requester: Optional[Requester], key: Optional[Hashable]
@@ -169,9 +170,6 @@ class Ratelimiter:
        rate_hz = rate_hz if rate_hz is not None else self.rate_hz
        burst_count = burst_count if burst_count is not None else self.burst_count

-        # Remove any expired entries
-        self._prune_message_counts(time_now_s)
-
        # Check if there is an existing count entry for this key
        action_count, time_start, _ = self._get_action_counts(key, time_now_s)

@@ -246,13 +244,12 @@ class Ratelimiter:
        action_count, time_start, rate_hz = self._get_action_counts(key, time_now_s)
        self.actions[key] = (action_count + n_actions, time_start, rate_hz)

-    def _prune_message_counts(self, time_now_s: float) -> None:
+    def _prune_message_counts(self) -> None:
        """Remove message count entries that have not exceeded their defined
        rate_hz limit
-
-        Args:
-            time_now_s: The current time
        """
+        time_now_s = self.clock.time()
+
        # We create a copy of the key list here as the dictionary is modified during
        # the loop
        for key in list(self.actions.keys()):
@@ -51,8 +51,7 @@ from synapse.http.server import JsonResource, OptionsResource
 from synapse.logging.context import LoggingContext
 from synapse.metrics import METRICS_PREFIX, MetricsResource, RegistryProxy
 from synapse.replication.http import REPLICATION_PREFIX, ReplicationRestResource
-from synapse.rest import ClientRestResource
-from synapse.rest.admin import register_servlets_for_media_repo
+from synapse.rest import ClientRestResource, admin
 from synapse.rest.health import HealthResource
 from synapse.rest.key.v2 import KeyResource
 from synapse.rest.synapse.client import build_synapse_client_resource_tree
@@ -176,8 +175,13 @@ class GenericWorkerServer(HomeServer):
    def _listen_http(self, listener_config: ListenerConfig) -> None:
        assert listener_config.http_options is not None

-        # We always include a health resource.
-        resources: Dict[str, Resource] = {"/health": HealthResource()}
+        # We always include an admin resource that we populate with servlets as needed
+        admin_resource = JsonResource(self, canonical_json=False)
+        resources: Dict[str, Resource] = {
+            # We always include a health resource.
+            "/health": HealthResource(),
+            "/_synapse/admin": admin_resource,
+        }

        for res in listener_config.http_options.resources:
            for name in res.names:
@@ -190,6 +194,7 @@ class GenericWorkerServer(HomeServer):

                    resources.update(build_synapse_client_resource_tree(self))
                    resources["/.well-known"] = well_known_resource(self)
+                    admin.register_servlets(self, admin_resource)

                elif name == "federation":
                    resources[FEDERATION_PREFIX] = TransportLayerServer(self)
@@ -199,15 +204,13 @@ class GenericWorkerServer(HomeServer):

                        # We need to serve the admin servlets for media on the
                        # worker.
-                        admin_resource = JsonResource(self, canonical_json=False)
-                        register_servlets_for_media_repo(self, admin_resource)
+                        admin.register_servlets_for_media_repo(self, admin_resource)

                        resources.update(
                            {
                                MEDIA_R0_PREFIX: media_repo,
                                MEDIA_V3_PREFIX: media_repo,
                                LEGACY_MEDIA_PREFIX: media_repo,
-                                "/_synapse/admin": admin_resource,
                            }
                        )

@@ -54,6 +54,7 @@ from synapse.config.server import ListenerConfig, TCPListenerConfig
 from synapse.federation.transport.server import TransportLayerServer
 from synapse.http.additional_resource import AdditionalResource
 from synapse.http.server import (
+    JsonResource,
    OptionsResource,
    RootOptionsRedirectResource,
    StaticResource,
@@ -61,8 +62,7 @@ from synapse.http.server import (
 from synapse.logging.context import LoggingContext
 from synapse.metrics import METRICS_PREFIX, MetricsResource, RegistryProxy
 from synapse.replication.http import REPLICATION_PREFIX, ReplicationRestResource
-from synapse.rest import ClientRestResource
-from synapse.rest.admin import AdminRestResource
+from synapse.rest import ClientRestResource, admin
 from synapse.rest.health import HealthResource
 from synapse.rest.key.v2 import KeyResource
 from synapse.rest.synapse.client import build_synapse_client_resource_tree
@@ -180,11 +180,14 @@ class SynapseHomeServer(HomeServer):
            if compress:
                client_resource = gz_wrap(client_resource)

+            admin_resource = JsonResource(self, canonical_json=False)
+            admin.register_servlets(self, admin_resource)
+
            resources.update(
                {
                    CLIENT_API_PREFIX: client_resource,
                    "/.well-known": well_known_resource(self),
-                    "/_synapse/admin": AdminRestResource(self),
+                    "/_synapse/admin": admin_resource,
                    **build_synapse_client_resource_tree(self),
                }
            )
@@ -356,6 +356,9 @@ def _parse_oidc_config_dict(
        additional_authorization_parameters=oidc_config.get(
            "additional_authorization_parameters", {}
        ),
+        passthrough_authorization_parameters=oidc_config.get(
+            "passthrough_authorization_parameters", []
+        ),
    )


@@ -501,3 +504,6 @@ class OidcProviderConfig:

    # Additional parameters that will be passed to the authorization grant URL
    additional_authorization_parameters: Mapping[str, str]
+
+    # Allow query parameters to the redirect endpoint that will be passed to the authorization grant URL
+    passthrough_authorization_parameters: Collection[str]
@@ -38,6 +38,9 @@ class UserDirectoryConfig(Config):
        self.user_directory_search_all_users = user_directory_config.get(
            "search_all_users", False
        )
+        self.user_directory_exclude_remote_users = user_directory_config.get(
+            "exclude_remote_users", False
+        )
        self.user_directory_search_prefer_local_users = user_directory_config.get(
            "prefer_local_users", False
        )
@@ -22,7 +22,6 @@

 import abc
 import collections.abc
-import os
 from typing import (
    TYPE_CHECKING,
    Any,
@@ -48,21 +47,21 @@ from synapse.synapse_rust.events import EventInternalMetadata
 from synapse.types import JsonDict, StrCollection
 from synapse.util.caches import intern_dict
 from synapse.util.frozenutils import freeze
-from synapse.util.stringutils import strtobool

 if TYPE_CHECKING:
    from synapse.events.builder import EventBuilder

-# Whether we should use frozen_dict in FrozenEvent. Using frozen_dicts prevents
-# bugs where we accidentally share e.g. signature dicts. However, converting a
-# dict to frozen_dicts is expensive.
-#
-# NOTE: This is overridden by the configuration by the Synapse worker apps, but
-# for the sake of tests, it is set here while it cannot be configured on the
-# homeserver object itself.

-USE_FROZEN_DICTS = strtobool(os.environ.get("SYNAPSE_USE_FROZEN_DICTS", "0"))
+USE_FROZEN_DICTS = False
+"""
+Whether we should use frozen_dict in FrozenEvent. Using frozen_dicts prevents
+bugs where we accidentally share e.g. signature dicts. However, converting a
+dict to frozen_dicts is expensive.

+NOTE: This is overridden by the configuration by the Synapse worker apps, but
+for the sake of tests, it is set here because it cannot be configured on the
+homeserver object itself.
+"""

 T = TypeVar("T")

@@ -40,6 +40,8 @@ import attr
 from canonicaljson import encode_canonical_json

 from synapse.api.constants import (
+    CANONICALJSON_MAX_INT,
+    CANONICALJSON_MIN_INT,
    MAX_PDU_SIZE,
    EventContentFields,
    EventTypes,
@@ -61,9 +63,6 @@ SPLIT_FIELD_REGEX = re.compile(r"\\*\.")
 # Find escaped characters, e.g. those with a \ in front of them.
 ESCAPE_SEQUENCE_PATTERN = re.compile(r"\\(.)")

-CANONICALJSON_MAX_INT = (2**53) - 1
-CANONICALJSON_MIN_INT = -CANONICALJSON_MAX_INT
-

 # Module API callback that allows adding fields to the unsigned section of
 # events that are sent to clients.
@@ -86,9 +86,7 @@ class EventValidator:

        # Depending on the room version, ensure the data is spec compliant JSON.
        if event.room_version.strict_canonicaljson:
-            # Note that only the client controlled portion of the event is
-            # checked, since we trust the portions of the event we created.
-            validate_canonicaljson(event.content)
+            validate_canonicaljson(event.get_pdu_json())

        if event.type == EventTypes.Aliases:
            if "aliases" in event.content:
@@ -20,7 +20,7 @@
 #
 #
 import logging
-from typing import TYPE_CHECKING, Awaitable, Callable, Optional
+from typing import TYPE_CHECKING, Awaitable, Callable, List, Optional, Sequence

 from synapse.api.constants import MAX_DEPTH, EventContentFields, EventTypes, Membership
 from synapse.api.errors import Codes, SynapseError
@@ -29,6 +29,7 @@ from synapse.crypto.event_signing import check_event_content_hash
 from synapse.crypto.keyring import Keyring
 from synapse.events import EventBase, make_event_from_dict
 from synapse.events.utils import prune_event, validate_canonicaljson
+from synapse.federation.units import filter_pdus_for_valid_depth
 from synapse.http.servlet import assert_params_in_dict
 from synapse.logging.opentracing import log_kv, trace
 from synapse.types import JsonDict, get_domain_from_id
@@ -267,6 +268,15 @@ def _is_invite_via_3pid(event: EventBase) -> bool:
    )


+def parse_events_from_pdu_json(
+    pdus_json: Sequence[JsonDict], room_version: RoomVersion
+) -> List[EventBase]:
+    return [
+        event_from_pdu_json(pdu_json, room_version)
+        for pdu_json in filter_pdus_for_valid_depth(pdus_json)
+    ]
+
+
 def event_from_pdu_json(pdu_json: JsonDict, room_version: RoomVersion) -> EventBase:
    """Construct an EventBase from an event json received over federation

@@ -68,6 +68,7 @@ from synapse.federation.federation_base import (
    FederationBase,
    InvalidEventSignatureError,
    event_from_pdu_json,
+    parse_events_from_pdu_json,
 )
 from synapse.federation.transport.client import SendJoinResponse
 from synapse.http.client import is_unknown_endpoint
@@ -349,7 +350,7 @@ class FederationClient(FederationBase):

        room_version = await self.store.get_room_version(room_id)

-        pdus = [event_from_pdu_json(p, room_version) for p in transaction_data_pdus]
+        pdus = parse_events_from_pdu_json(transaction_data_pdus, room_version)

        # Check signatures and hash of pdus, removing any from the list that fail checks
        pdus[:] = await self._check_sigs_and_hash_for_pulled_events_and_fetch(
@@ -393,9 +394,7 @@ class FederationClient(FederationBase):
            transaction_data,
        )

-        pdu_list: List[EventBase] = [
-            event_from_pdu_json(p, room_version) for p in transaction_data["pdus"]
-        ]
+        pdu_list = parse_events_from_pdu_json(transaction_data["pdus"], room_version)

        if pdu_list and pdu_list[0]:
            pdu = pdu_list[0]
@@ -809,7 +808,7 @@ class FederationClient(FederationBase):

        room_version = await self.store.get_room_version(room_id)

-        auth_chain = [event_from_pdu_json(p, room_version) for p in res["auth_chain"]]
+        auth_chain = parse_events_from_pdu_json(res["auth_chain"], room_version)

        signed_auth = await self._check_sigs_and_hash_for_pulled_events_and_fetch(
            destination, auth_chain, room_version=room_version
@@ -1529,9 +1528,7 @@ class FederationClient(FederationBase):

            room_version = await self.store.get_room_version(room_id)

-            events = [
-                event_from_pdu_json(e, room_version) for e in content.get("events", [])
-            ]
+            events = parse_events_from_pdu_json(content.get("events", []), room_version)

            signed_events = await self._check_sigs_and_hash_for_pulled_events_and_fetch(
                destination, events, room_version=room_version
@@ -66,7 +66,7 @@ from synapse.federation.federation_base import (
    event_from_pdu_json,
 )
 from synapse.federation.persistence import TransactionActions
-from synapse.federation.units import Edu, Transaction
+from synapse.federation.units import Edu, Transaction, serialize_and_filter_pdus
 from synapse.handlers.worker_lock import NEW_EVENT_DURING_PURGE_LOCK_NAME
 from synapse.http.servlet import assert_params_in_dict
 from synapse.logging.context import (
@@ -469,7 +469,12 @@ class FederationServer(FederationBase):
                logger.info("Ignoring PDU: %s", e)
                continue

-            event = event_from_pdu_json(p, room_version)
+            try:
+                event = event_from_pdu_json(p, room_version)
+            except SynapseError as e:
+                logger.info("Ignoring PDU for failing to deserialize: %s", e)
+                continue
+
            pdus_by_room.setdefault(room_id, []).append(event)

            if event.origin_server_ts > newest_pdu_ts:
@@ -636,8 +641,8 @@ class FederationServer(FederationBase):
        )

        return {
-            "pdus": [pdu.get_pdu_json() for pdu in pdus],
-            "auth_chain": [pdu.get_pdu_json() for pdu in auth_chain],
+            "pdus": serialize_and_filter_pdus(pdus),
+            "auth_chain": serialize_and_filter_pdus(auth_chain),
        }

    async def on_pdu_request(
@@ -696,6 +701,12 @@ class FederationServer(FederationBase):
        pdu = event_from_pdu_json(content, room_version)
        origin_host, _ = parse_server_name(origin)
        await self.check_server_matches_acl(origin_host, pdu.room_id)
+        if await self._spam_checker_module_callbacks.should_drop_federated_event(pdu):
+            logger.info(
+                "Federated event contains spam, dropping %s",
+                pdu.event_id,
+            )
+            raise SynapseError(403, Codes.FORBIDDEN)
        try:
            pdu = await self._check_sigs_and_hash(room_version, pdu)
        except InvalidEventSignatureError as e:
@@ -761,8 +772,8 @@ class FederationServer(FederationBase):
        event_json = event.get_pdu_json(time_now)
        resp = {
            "event": event_json,
-            "state": [p.get_pdu_json(time_now) for p in state_events],
-            "auth_chain": [p.get_pdu_json(time_now) for p in auth_chain_events],
+            "state": serialize_and_filter_pdus(state_events, time_now),
+            "auth_chain": serialize_and_filter_pdus(auth_chain_events, time_now),
            "members_omitted": caller_supports_partial_state,
        }

@@ -1005,7 +1016,7 @@ class FederationServer(FederationBase):

            time_now = self._clock.time_msec()
            auth_pdus = await self.handler.on_event_auth(event_id)
-            res = {"auth_chain": [a.get_pdu_json(time_now) for a in auth_pdus]}
+            res = {"auth_chain": serialize_and_filter_pdus(auth_pdus, time_now)}
        return 200, res

    async def on_query_client_keys(
@@ -1090,7 +1101,7 @@ class FederationServer(FederationBase):

            time_now = self._clock.time_msec()

-        return {"events": [ev.get_pdu_json(time_now) for ev in missing_events]}
+        return {"events": serialize_and_filter_pdus(missing_events, time_now)}

    async def on_openid_userinfo(self, token: str) -> Optional[str]:
        ts_now_ms = self._clock.time_msec()
@@ -24,10 +24,12 @@ server protocol.
 """

 import logging
-from typing import List, Optional
+from typing import List, Optional, Sequence

 import attr

+from synapse.api.constants import CANONICALJSON_MAX_INT, CANONICALJSON_MIN_INT
+from synapse.events import EventBase
 from synapse.types import JsonDict

 logger = logging.getLogger(__name__)
@@ -104,8 +106,28 @@ class Transaction:
        result = {
            "origin": self.origin,
            "origin_server_ts": self.origin_server_ts,
-            "pdus": self.pdus,
+            "pdus": filter_pdus_for_valid_depth(self.pdus),
        }
        if self.edus:
            result["edus"] = self.edus
        return result
+
+
+def filter_pdus_for_valid_depth(pdus: Sequence[JsonDict]) -> List[JsonDict]:
+    filtered_pdus = []
+    for pdu in pdus:
+        # Drop PDUs that have a depth that is outside of the range allowed
+        # by canonical json.
+        if (
+            "depth" in pdu
+            and CANONICALJSON_MIN_INT <= pdu["depth"] <= CANONICALJSON_MAX_INT
+        ):
+            filtered_pdus.append(pdu)
+
+    return filtered_pdus
+
+
+def serialize_and_filter_pdus(
+    pdus: Sequence[EventBase], time_now: Optional[int] = None
+) -> List[JsonDict]:
+    return filter_pdus_for_valid_depth([pdu.get_pdu_json(time_now) for pdu in pdus])
@@ -191,18 +191,36 @@ class DelayedEventsHandler:

    async def _handle_state_deltas(self, deltas: List[StateDelta]) -> None:
        """
-        Process current state deltas to cancel pending delayed events
+        Process current state deltas to cancel other users' pending delayed events
        that target the same state.
        """
        for delta in deltas:
+            if delta.event_id is None:
+                logger.debug(
+                    "Not handling delta for deleted state: %r %r",
+                    delta.event_type,
+                    delta.state_key,
+                )
+                continue
+
            logger.debug(
                "Handling: %r %r, %s", delta.event_type, delta.state_key, delta.event_id
            )

+            event = await self._store.get_event(
+                delta.event_id, check_room_id=delta.room_id
+            )
+            sender = UserID.from_string(event.sender)
+
            next_send_ts = await self._store.cancel_delayed_state_events(
                room_id=delta.room_id,
                event_type=delta.event_type,
                state_key=delta.state_key,
+                not_from_localpart=(
+                    sender.localpart
+                    if sender.domain == self._config.server.server_name
+                    else ""
+                ),
            )

            if self._next_send_ts_changed(next_send_ts):
@@ -163,6 +163,8 @@ class DeviceWorkerHandler:
            raise errors.NotFoundError()

        ips = await self.store.get_last_client_ip_by_device(user_id, device_id)
+
+        device = dict(device)
        _update_device_from_client_ips(device, ips)

        set_tag("device", str(device))
@@ -1462,6 +1462,12 @@ class EventCreationHandler:
                    )
                    return prev_event

+            if not event.is_state() and event.type in [
+                EventTypes.Message,
+                EventTypes.Encrypted,
+            ]:
+                await self.store.set_room_participation(event.user_id, event.room_id)
+
            if event.internal_metadata.is_out_of_band_membership():
                # the only sort of out-of-band-membership events we expect to see here are
                # invite rejections and rescinded knocks that we have generated ourselves.
@@ -467,6 +467,10 @@ class OidcProvider:

        self._sso_handler.register_identity_provider(self)

+        self.passthrough_authorization_parameters = (
+            provider.passthrough_authorization_parameters
+        )
+
    def _validate_metadata(self, m: OpenIDProviderMetadata) -> None:
        """Verifies the provider metadata.

@@ -582,6 +586,24 @@ class OidcProvider:
            or self._user_profile_method == "userinfo_endpoint"
        )

+    @property
+    def _uses_access_token(self) -> bool:
+        """Return True if the `access_token` will be used during the login process.
+
+        This is useful to determine whether the access token
+        returned by the identity provider, and
+        any related metadata (such as the `at_hash` field in
+        the ID token), should be validated.
+        """
+        # Currently, Synapse only uses the access_token to fetch user metadata
+        # from the userinfo endpoint. Therefore we only have a single criteria
+        # to check right now but this may change in the future and this function
+        # should be updated if more usages are introduced.
+        #
+        # For example, if we start to use the access_token given to us by the
+        # IdP for more things, such as accessing Resource Server APIs.
+        return self._uses_userinfo
+
    @property
    def issuer(self) -> str:
        """The issuer identifying this provider."""
@@ -953,9 +975,16 @@ class OidcProvider:
            "nonce": nonce,
            "client_id": self._client_auth.client_id,
        }
-        if "access_token" in token:
+        if self._uses_access_token and "access_token" in token:
            # If we got an `access_token`, there should be an `at_hash` claim
-            # in the `id_token` that we can check against.
+            # in the `id_token` that we can check against. Setting this
+            # instructs authlib to check the value of `at_hash` in the
+            # ID token.
+            #
+            # We only need to verify the access token if we actually make
+            # use of it. Which currently only happens when we need to fetch
+            # the user's information from the userinfo_endpoint. Thus, this
+            # check is also gated on self._uses_userinfo.
            claims_params["access_token"] = token["access_token"]

        claims_options = {"iss": {"values": [metadata["issuer"]]}}
@@ -1005,7 +1034,6 @@ class OidcProvider:
                when everything is done (or None for UI Auth)
            ui_auth_session_id: The session ID of the ongoing UI Auth (or
                None if this is a login).
-
        Returns:
            The redirect URL to the authorization endpoint.

@@ -1078,6 +1106,13 @@ class OidcProvider:
                )
            )

+        # add passthrough additional authorization parameters
+        passthrough_authorization_parameters = self.passthrough_authorization_parameters
+        for parameter in passthrough_authorization_parameters:
+            parameter_value = parse_string(request, parameter)
+            if parameter_value:
+                additional_authorization_parameters.update({parameter: parameter_value})
+
        authorization_endpoint = metadata.get("authorization_endpoint")
        return prepare_grant_uri(
            authorization_endpoint,
@@ -36,10 +36,17 @@ class SetPasswordHandler:
    def __init__(self, hs: "HomeServer"):
        self.store = hs.get_datastores().main
        self._auth_handler = hs.get_auth_handler()
-        # This can only be instantiated on the main process.
-        device_handler = hs.get_device_handler()
-        assert isinstance(device_handler, DeviceHandler)
-        self._device_handler = device_handler
+
+        # We don't need the device handler if password changing is disabled.
+        # This allows us to instantiate the SetPasswordHandler on the workers
+        # that have admin APIs for MAS
+        if self._auth_handler.can_change_password():
+            # This can only be instantiated on the main process.
+            device_handler = hs.get_device_handler()
+            assert isinstance(device_handler, DeviceHandler)
+            self._device_handler: Optional[DeviceHandler] = device_handler
+        else:
+            self._device_handler = None

    async def set_password(
        self,
@@ -51,6 +58,9 @@ class SetPasswordHandler:
        if not self._auth_handler.can_change_password():
            raise SynapseError(403, "Password change disabled", errcode=Codes.FORBIDDEN)

+        # We should have this available only if password changing is enabled.
+        assert self._device_handler is not None
+
        try:
            await self.store.user_set_password_hash(user_id, password_hash)
        except StoreError as e:
@@ -271,6 +271,7 @@ class SlidingSyncHandler:
                from_token=from_token,
                to_token=to_token,
                newly_joined=room_id in interested_rooms.newly_joined_rooms,
+                newly_left=room_id in interested_rooms.newly_left_rooms,
                is_dm=room_id in interested_rooms.dm_room_ids,
            )

@@ -542,6 +543,7 @@ class SlidingSyncHandler:
        from_token: Optional[SlidingSyncStreamToken],
        to_token: StreamToken,
        newly_joined: bool,
+        newly_left: bool,
        is_dm: bool,
    ) -> SlidingSyncResult.RoomResult:
        """
@@ -559,6 +561,7 @@ class SlidingSyncHandler:
            from_token: The point in the stream to sync from.
            to_token: The point in the stream to sync up to.
            newly_joined: If the user has newly joined the room
+            newly_left: If the user has newly left the room
            is_dm: Whether the room is a DM room
        """
        user = sync_config.user
@@ -856,6 +859,26 @@ class SlidingSyncHandler:
            # TODO: Limit the number of state events we're about to send down
            # the room, if its too many we should change this to an
            # `initial=True`?
+
+            # For the case of rejecting remote invites, the leave event won't be
+            # returned by `get_current_state_deltas_for_room`. This is due to the current
+            # state only being filled out for rooms the server is in, and so doesn't pick
+            # up out-of-band leaves (including locally rejected invites) as these events
+            # are outliers and not added to the `current_state_delta_stream`.
+            #
+            # We rely on being explicitly told that the room has been `newly_left` to
+            # ensure we extract the out-of-band leave.
+            if newly_left and room_membership_for_user_at_to_token.event_id is not None:
+                membership_changed = True
+                leave_event = await self.store.get_event(
+                    room_membership_for_user_at_to_token.event_id
+                )
+                state_key = leave_event.get_state_key()
+                if state_key is not None:
+                    room_state_delta_id_map[(leave_event.type, state_key)] = (
+                        room_membership_for_user_at_to_token.event_id
+                    )
+
            deltas = await self.get_current_state_deltas_for_room(
                room_id=room_id,
                room_membership_for_user_at_to_token=room_membership_for_user_at_to_token,
@@ -244,14 +244,47 @@ class SlidingSyncRoomLists:
        # Note: this won't include rooms the user has left themselves. We add back
        # `newly_left` rooms below. This is more efficient than fetching all rooms and
        # then filtering out the old left rooms.
-        room_membership_for_user_map = await self.store.get_sliding_sync_rooms_for_user(
-            user_id
+        room_membership_for_user_map = (
+            await self.store.get_sliding_sync_rooms_for_user_from_membership_snapshots(
+                user_id
+            )
        )
+        # To play nice with the rewind logic below, we need to go fetch the rooms the
+        # user has left themselves but only if it changed after the `to_token`.
+        #
+        # If a leave happens *after* the token range, we may have still been joined (or
+        # any non-self-leave which is relevant to sync) to the room before so we need to
+        # include it in the list of potentially relevant rooms and apply our rewind
+        # logic (outside of this function) to see if it's actually relevant.
+        #
+        # We do this separately from
+        # `get_sliding_sync_rooms_for_user_from_membership_snapshots` as those results
+        # are cached and the `to_token` isn't very cache friendly (people are constantly
+        # requesting with new tokens) so we separate it out here.
+        self_leave_room_membership_for_user_map = (
+            await self.store.get_sliding_sync_self_leave_rooms_after_to_token(
+                user_id, to_token
+            )
+        )
+        if self_leave_room_membership_for_user_map:
+            # FIXME: It would be nice to avoid this copy but since
+            # `get_sliding_sync_rooms_for_user_from_membership_snapshots` is cached, it
+            # can't return a mutable value like a `dict`. We make the copy to get a
+            # mutable dict that we can change. We try to only make a copy when necessary
+            # (if we actually need to change something) as in most cases, the logic
+            # doesn't need to run.
+            room_membership_for_user_map = dict(room_membership_for_user_map)
+            room_membership_for_user_map.update(self_leave_room_membership_for_user_map)

        # Remove invites from ignored users
        ignored_users = await self.store.ignored_users(user_id)
        if ignored_users:
-            # TODO: It would be nice to avoid these copies
+            # FIXME: It would be nice to avoid this copy but since
+            # `get_sliding_sync_rooms_for_user_from_membership_snapshots` is cached, it
+            # can't return a mutable value like a `dict`. We make the copy to get a
+            # mutable dict that we can change. We try to only make a copy when necessary
+            # (if we actually need to change something) as in most cases, the logic
+            # doesn't need to run.
            room_membership_for_user_map = dict(room_membership_for_user_map)
            # Make a copy so we don't run into an error: `dictionary changed size during
            # iteration`, when we remove items
@@ -263,11 +296,23 @@ class SlidingSyncRoomLists:
                ):
                    room_membership_for_user_map.pop(room_id, None)

+        (
+            newly_joined_room_ids,
+            newly_left_room_map,
+        ) = await self._get_newly_joined_and_left_rooms(
+            user_id, from_token=from_token, to_token=to_token
+        )
+
        changes = await self._get_rewind_changes_to_current_membership_to_token(
            sync_config.user, room_membership_for_user_map, to_token=to_token
        )
        if changes:
-            # TODO: It would be nice to avoid these copies
+            # FIXME: It would be nice to avoid this copy but since
+            # `get_sliding_sync_rooms_for_user_from_membership_snapshots` is cached, it
+            # can't return a mutable value like a `dict`. We make the copy to get a
+            # mutable dict that we can change. We try to only make a copy when necessary
+            # (if we actually need to change something) as in most cases, the logic
+            # doesn't need to run.
            room_membership_for_user_map = dict(room_membership_for_user_map)
            for room_id, change in changes.items():
                if change is None:
@@ -278,7 +323,7 @@ class SlidingSyncRoomLists:
                existing_room = room_membership_for_user_map.get(room_id)
                if existing_room is not None:
                    # Update room membership events to the point in time of the `to_token`
-                    room_membership_for_user_map[room_id] = RoomsForUserSlidingSync(
+                    room_for_user = RoomsForUserSlidingSync(
                        room_id=room_id,
                        sender=change.sender,
                        membership=change.membership,
@@ -290,18 +335,18 @@ class SlidingSyncRoomLists:
                        room_type=existing_room.room_type,
                        is_encrypted=existing_room.is_encrypted,
                    )
-
-        (
-            newly_joined_room_ids,
-            newly_left_room_map,
-        ) = await self._get_newly_joined_and_left_rooms(
-            user_id, from_token=from_token, to_token=to_token
-        )
-        dm_room_ids = await self._get_dm_rooms_for_user(user_id)
+                    if filter_membership_for_sync(
+                        user_id=user_id,
+                        room_membership_for_user=room_for_user,
+                        newly_left=room_id in newly_left_room_map,
+                    ):
+                        room_membership_for_user_map[room_id] = room_for_user
+                    else:
+                        room_membership_for_user_map.pop(room_id, None)

        # Add back `newly_left` rooms (rooms left in the from -> to token range).
        #
-        # We do this because `get_sliding_sync_rooms_for_user(...)` doesn't include
+        # We do this because `get_sliding_sync_rooms_for_user_from_membership_snapshots(...)` doesn't include
        # rooms that the user left themselves as it's more efficient to add them back
        # here than to fetch all rooms and then filter out the old left rooms. The user
        # only leaves a room once in a blue moon so this barely needs to run.
@@ -310,7 +355,12 @@ class SlidingSyncRoomLists:
            newly_left_room_map.keys() - room_membership_for_user_map.keys()
        )
        if missing_newly_left_rooms:
-            # TODO: It would be nice to avoid these copies
+            # FIXME: It would be nice to avoid this copy but since
+            # `get_sliding_sync_rooms_for_user_from_membership_snapshots` is cached, it
+            # can't return a mutable value like a `dict`. We make the copy to get a
+            # mutable dict that we can change. We try to only make a copy when necessary
+            # (if we actually need to change something) as in most cases, the logic
+            # doesn't need to run.
            room_membership_for_user_map = dict(room_membership_for_user_map)
            for room_id in missing_newly_left_rooms:
                newly_left_room_for_user = newly_left_room_map[room_id]
@@ -327,14 +377,21 @@ class SlidingSyncRoomLists:
                # If the membership exists, it's just a normal user left the room on
                # their own
                if newly_left_room_for_user_sliding_sync is not None:
-                    room_membership_for_user_map[room_id] = (
-                        newly_left_room_for_user_sliding_sync
-                    )
+                    if filter_membership_for_sync(
+                        user_id=user_id,
+                        room_membership_for_user=newly_left_room_for_user_sliding_sync,
+                        newly_left=room_id in newly_left_room_map,
+                    ):
+                        room_membership_for_user_map[room_id] = (
+                            newly_left_room_for_user_sliding_sync
+                        )
+                    else:
+                        room_membership_for_user_map.pop(room_id, None)

                    change = changes.get(room_id)
                    if change is not None:
                        # Update room membership events to the point in time of the `to_token`
-                        room_membership_for_user_map[room_id] = RoomsForUserSlidingSync(
+                        room_for_user = RoomsForUserSlidingSync(
                            room_id=room_id,
                            sender=change.sender,
                            membership=change.membership,
@@ -346,6 +403,14 @@ class SlidingSyncRoomLists:
                            room_type=newly_left_room_for_user_sliding_sync.room_type,
                            is_encrypted=newly_left_room_for_user_sliding_sync.is_encrypted,
                        )
+                        if filter_membership_for_sync(
+                            user_id=user_id,
+                            room_membership_for_user=room_for_user,
+                            newly_left=room_id in newly_left_room_map,
+                        ):
+                            room_membership_for_user_map[room_id] = room_for_user
+                        else:
+                            room_membership_for_user_map.pop(room_id, None)

                # If we are `newly_left` from the room but can't find any membership,
                # then we have been "state reset" out of the room
@@ -367,7 +432,7 @@ class SlidingSyncRoomLists:
                        newly_left_room_for_user.event_pos.to_room_stream_token(),
                    )

-                    room_membership_for_user_map[room_id] = RoomsForUserSlidingSync(
+                    room_for_user = RoomsForUserSlidingSync(
                        room_id=room_id,
                        sender=newly_left_room_for_user.sender,
                        membership=newly_left_room_for_user.membership,
@@ -378,6 +443,16 @@ class SlidingSyncRoomLists:
                        room_type=room_type,
                        is_encrypted=is_encrypted,
                    )
+                    if filter_membership_for_sync(
+                        user_id=user_id,
+                        room_membership_for_user=room_for_user,
+                        newly_left=room_id in newly_left_room_map,
+                    ):
+                        room_membership_for_user_map[room_id] = room_for_user
+                    else:
+                        room_membership_for_user_map.pop(room_id, None)
+
+        dm_room_ids = await self._get_dm_rooms_for_user(user_id)

        if sync_config.lists:
            sync_room_map = room_membership_for_user_map
@@ -493,7 +568,12 @@ class SlidingSyncRoomLists:

        if sync_config.room_subscriptions:
            with start_active_span("assemble_room_subscriptions"):
-                # TODO: It would be nice to avoid these copies
+                # FIXME: It would be nice to avoid this copy but since
+                # `get_sliding_sync_rooms_for_user_from_membership_snapshots` is cached, it
+                # can't return a mutable value like a `dict`. We make the copy to get a
+                # mutable dict that we can change. We try to only make a copy when necessary
+                # (if we actually need to change something) as in most cases, the logic
+                # doesn't need to run.
                room_membership_for_user_map = dict(room_membership_for_user_map)

                # Find which rooms are partially stated and may need to be filtered out
@@ -1040,7 +1120,7 @@ class SlidingSyncRoomLists:
        (
            newly_joined_room_ids,
            newly_left_room_map,
-        ) = await self._get_newly_joined_and_left_rooms(
+        ) = await self._get_newly_joined_and_left_rooms_fallback(
            user_id, to_token=to_token, from_token=from_token
        )

@@ -1096,6 +1176,53 @@ class SlidingSyncRoomLists:
        "state reset" out of the room, and so that room would not be part of the
        "current memberships" of the user.

+        Returns:
+            A 2-tuple of newly joined room IDs and a map of newly_left room
+            IDs to the `RoomsForUserStateReset` entry.
+
+            We're using `RoomsForUserStateReset` but that doesn't necessarily mean the
+            user was state reset of the rooms. It's just that the `event_id`/`sender`
+            are optional and we can't tell the difference between the server leaving the
+            room when the user was the last person participating in the room and left or
+            was state reset out of the room. To actually check for a state reset, you
+            need to check if a membership still exists in the room.
+        """
+
+        newly_joined_room_ids: Set[str] = set()
+        newly_left_room_map: Dict[str, RoomsForUserStateReset] = {}
+
+        if not from_token:
+            return newly_joined_room_ids, newly_left_room_map
+
+        changes = await self.store.get_sliding_sync_membership_changes(
+            user_id,
+            from_key=from_token.room_key,
+            to_key=to_token.room_key,
+            excluded_room_ids=set(self.rooms_to_exclude_globally),
+        )
+
+        for room_id, entry in changes.items():
+            if entry.membership == Membership.JOIN:
+                newly_joined_room_ids.add(room_id)
+            elif entry.membership == Membership.LEAVE:
+                newly_left_room_map[room_id] = entry
+
+        return newly_joined_room_ids, newly_left_room_map
+
+    @trace
+    async def _get_newly_joined_and_left_rooms_fallback(
+        self,
+        user_id: str,
+        to_token: StreamToken,
+        from_token: Optional[StreamToken],
+    ) -> Tuple[AbstractSet[str], Mapping[str, RoomsForUserStateReset]]:
+        """Fetch the sets of rooms that the user newly joined or left in the
+        given token range.
+
+        Note: there may be rooms in the newly left rooms where the user was
+        "state reset" out of the room, and so that room would not be part of the
+        "current memberships" of the user.
+
        Returns:
            A 2-tuple of newly joined room IDs and a map of newly_left room
            IDs to the `RoomsForUserStateReset` entry.
@@ -108,6 +108,9 @@ class UserDirectoryHandler(StateDeltasHandler):
        self.is_mine_id = hs.is_mine_id
        self.update_user_directory = hs.config.worker.should_update_user_directory
        self.search_all_users = hs.config.userdirectory.user_directory_search_all_users
+        self.exclude_remote_users = (
+            hs.config.userdirectory.user_directory_exclude_remote_users
+        )
        self.show_locked_users = hs.config.userdirectory.show_locked_users
        self._spam_checker_module_callbacks = hs.get_module_api_callbacks().spam_checker
        self._hs = hs
@@ -118,6 +118,9 @@ DEFAULT_MAX_TIMEOUT_MS = 20_000
 # Maximum allowed timeout_ms for download and thumbnail requests
 MAXIMUM_ALLOWED_MAX_TIMEOUT_MS = 60_000

+# The ETag header value to use for immutable media. This can be anything.
+_IMMUTABLE_ETAG = "1"
+

 def respond_404(request: SynapseRequest) -> None:
    assert request.path is not None
@@ -224,12 +227,7 @@ def add_file_headers(

    request.setHeader(b"Content-Disposition", disposition.encode("ascii"))

-    # cache for at least a day.
-    # XXX: we might want to turn this off for data we don't want to
-    # recommend caching as it's sensitive or private - or at least
-    # select private. don't bother setting Expires as all our
-    # clients are smart enough to be happy with Cache-Control
-    request.setHeader(b"Cache-Control", b"public,max-age=86400,s-maxage=86400")
+    _add_cache_headers(request)

    if file_size is not None:
        request.setHeader(b"Content-Length", b"%d" % (file_size,))
@@ -240,6 +238,26 @@ def add_file_headers(
    request.setHeader(b"X-Robots-Tag", "noindex, nofollow, noarchive, noimageindex")


+def _add_cache_headers(request: Request) -> None:
+    """Adds the appropriate cache headers to the response"""
+
+    # Cache on the client for at least a day.
+    #
+    # We set this to "public,s-maxage=0,proxy-revalidate" to allow CDNs to cache
+    # the media, so long as they "revalidate" the media on every request. By
+    # revalidate, we mean send the request to Synapse with a `If-None-Match`
+    # header, to which Synapse can either respond with a 304 if the user is
+    # authenticated/authorized, or a 401/403 if they're not.
+    request.setHeader(
+        b"Cache-Control", b"public,max-age=86400,s-maxage=0,proxy-revalidate"
+    )
+
+    # Set an ETag header to allow requesters to use it in requests to check if
+    # the cache is still valid. Since media is immutable (though may be
+    # deleted), we just set this to a constant.
+    request.setHeader(b"ETag", _IMMUTABLE_ETAG)
+
+
 # separators as defined in RFC2616. SP and HT are handled separately.
 # see _can_encode_filename_as_token.
 _FILENAME_SEPARATOR_CHARS = {
@@ -336,13 +354,15 @@ async def respond_with_multipart_responder(

        from synapse.media.media_storage import MultipartFileConsumer

+        _add_cache_headers(request)
+
        # note that currently the json_object is just {}, this will change when linked media
        # is implemented
        multipart_consumer = MultipartFileConsumer(
            clock,
            request,
            media_type,
-            {},
+            {},  # Note: if we change this we need to change the returned ETag.
            disposition,
            media_length,
        )
@@ -419,6 +439,46 @@ async def respond_with_responder(
    finish_request(request)


+def respond_with_304(request: SynapseRequest) -> None:
+    request.setResponseCode(304)
+
+    # could alternatively use request.notifyFinish() and flip a flag when
+    # the Deferred fires, but since the flag is RIGHT THERE it seems like
+    # a waste.
+    if request._disconnected:
+        logger.warning(
+            "Not sending response to request %s, already disconnected.", request
+        )
+        return None
+
+    _add_cache_headers(request)
+
+    request.finish()
+
+
+def check_for_cached_entry_and_respond(request: SynapseRequest) -> bool:
+    """Check if the request has a conditional header that allows us to return a
+    304 Not Modified response, and if it does, return a 304 response.
+
+    This handles clients and intermediary proxies caching media.
+    This method assumes that the user has already been
+    authorised to request the media.
+
+    Returns True if we have responded."""
+
+    # We've checked the user has access to the media, so we now check if it
+    # is a "conditional request" and we can just return a `304 Not Modified`
+    # response. Since media is immutable (though may be deleted), we just
+    # check this is the expected constant.
+    etag = request.getHeader("If-None-Match")
+    if etag == _IMMUTABLE_ETAG:
+        # Return a `304 Not modified`.
+        respond_with_304(request)
+        return True
+
+    return False
+
+
 class Responder(ABC):
    """Represents a response that can be streamed to the requester.

@@ -52,13 +52,18 @@ from synapse.media._base import (
    FileInfo,
    Responder,
    ThumbnailInfo,
+    check_for_cached_entry_and_respond,
    get_filename_from_headers,
    respond_404,
    respond_with_multipart_responder,
    respond_with_responder,
 )
 from synapse.media.filepath import MediaFilePaths
-from synapse.media.media_storage import MediaStorage
+from synapse.media.media_storage import (
+    MediaStorage,
+    SHA256TransparentIOReader,
+    SHA256TransparentIOWriter,
+)
 from synapse.media.storage_provider import StorageProviderWrapper
 from synapse.media.thumbnailer import Thumbnailer, ThumbnailError
 from synapse.media.url_previewer import UrlPreviewer
@@ -300,15 +305,26 @@ class MediaRepository:
            auth_user: The user_id of the uploader
        """
        file_info = FileInfo(server_name=None, file_id=media_id)
-        fname = await self.media_storage.store_file(content, file_info)
+        sha256reader = SHA256TransparentIOReader(content)
+        # This implements all of IO as it has a passthrough
+        fname = await self.media_storage.store_file(sha256reader.wrap(), file_info)
+        sha256 = sha256reader.hexdigest()
+        should_quarantine = await self.store.get_is_hash_quarantined(sha256)
        logger.info("Stored local media in file %r", fname)

+        if should_quarantine:
+            logger.warn(
+                "Media has been automatically quarantined as it matched existing quarantined media"
+            )
+
        await self.store.update_local_media(
            media_id=media_id,
            media_type=media_type,
            upload_name=upload_name,
            media_length=content_length,
            user_id=auth_user,
+            sha256=sha256,
+            quarantined_by="system" if should_quarantine else None,
        )

        try:
@@ -341,11 +357,19 @@ class MediaRepository:
        media_id = random_string(24)

        file_info = FileInfo(server_name=None, file_id=media_id)
-
-        fname = await self.media_storage.store_file(content, file_info)
+        # This implements all of IO as it has a passthrough
+        sha256reader = SHA256TransparentIOReader(content)
+        fname = await self.media_storage.store_file(sha256reader.wrap(), file_info)
+        sha256 = sha256reader.hexdigest()
+        should_quarantine = await self.store.get_is_hash_quarantined(sha256)

        logger.info("Stored local media in file %r", fname)

+        if should_quarantine:
+            logger.warn(
+                "Media has been automatically quarantined as it matched existing quarantined media"
+            )
+
        await self.store.store_local_media(
            media_id=media_id,
            media_type=media_type,
@@ -353,6 +377,8 @@ class MediaRepository:
            upload_name=upload_name,
            media_length=content_length,
            user_id=auth_user,
+            sha256=sha256,
+            quarantined_by="system" if should_quarantine else None,
        )

        try:
@@ -459,6 +485,11 @@ class MediaRepository:

        self.mark_recently_accessed(None, media_id)

+        # Once we've checked auth we can return early if the media is cached on
+        # the client
+        if check_for_cached_entry_and_respond(request):
+            return
+
        media_type = media_info.media_type
        if not media_type:
            media_type = "application/octet-stream"
@@ -538,6 +569,17 @@ class MediaRepository:
                allow_authenticated,
            )

+        # Check if the media is cached on the client, if so return 304. We need
+        # to do this after we have fetched remote media, as we need it to do the
+        # auth.
+        if check_for_cached_entry_and_respond(request):
+            # We always need to use the responder.
+            if responder:
+                with responder:
+                    pass
+
+            return
+
        # We deliberately stream the file outside the lock
        if responder and media_info:
            upload_name = name if name else media_info.upload_name
@@ -739,11 +781,13 @@ class MediaRepository:
        file_info = FileInfo(server_name=server_name, file_id=file_id)

        async with self.media_storage.store_into_file(file_info) as (f, fname):
+            sha256writer = SHA256TransparentIOWriter(f)
            try:
                length, headers = await self.client.download_media(
                    server_name,
                    media_id,
-                    output_stream=f,
+                    # This implements all of BinaryIO as it has a passthrough
+                    output_stream=sha256writer.wrap(),
                    max_size=self.max_upload_size,
                    max_timeout_ms=max_timeout_ms,
                    download_ratelimiter=download_ratelimiter,
@@ -808,6 +852,7 @@ class MediaRepository:
                upload_name=upload_name,
                media_length=length,
                filesystem_id=file_id,
+                sha256=sha256writer.hexdigest(),
            )

        logger.info("Stored remote media in file %r", fname)
@@ -828,6 +873,7 @@ class MediaRepository:
            last_access_ts=time_now_ms,
            quarantined_by=None,
            authenticated=authenticated,
+            sha256=sha256writer.hexdigest(),
        )

    async def _federation_download_remote_file(
@@ -862,11 +908,13 @@ class MediaRepository:
        file_info = FileInfo(server_name=server_name, file_id=file_id)

        async with self.media_storage.store_into_file(file_info) as (f, fname):
+            sha256writer = SHA256TransparentIOWriter(f)
            try:
                res = await self.client.federation_download_media(
                    server_name,
                    media_id,
-                    output_stream=f,
+                    # This implements all of BinaryIO as it has a passthrough
+                    output_stream=sha256writer.wrap(),
                    max_size=self.max_upload_size,
                    max_timeout_ms=max_timeout_ms,
                    download_ratelimiter=download_ratelimiter,
@@ -937,6 +985,7 @@ class MediaRepository:
                upload_name=upload_name,
                media_length=length,
                filesystem_id=file_id,
+                sha256=sha256writer.hexdigest(),
            )

        logger.debug("Stored remote media in file %r", fname)
@@ -957,6 +1006,7 @@ class MediaRepository:
            last_access_ts=time_now_ms,
            quarantined_by=None,
            authenticated=authenticated,
+            sha256=sha256writer.hexdigest(),
        )

    def _get_thumbnail_requirements(
@@ -19,6 +19,7 @@
 #
 #
 import contextlib
+import hashlib
 import json
 import logging
 import os
@@ -70,6 +71,88 @@ logger = logging.getLogger(__name__)
 CRLF = b"\r\n"


+class SHA256TransparentIOWriter:
+    """Will generate a SHA256 hash from a source stream transparently.
+
+    Args:
+        source: Source stream.
+    """
+
+    def __init__(self, source: BinaryIO):
+        self._hash = hashlib.sha256()
+        self._source = source
+
+    def write(self, buffer: Union[bytes, bytearray]) -> int:
+        """Wrapper for source.write()
+
+        Args:
+            buffer
+
+        Returns:
+            the value of source.write()
+        """
+        res = self._source.write(buffer)
+        self._hash.update(buffer)
+        return res
+
+    def hexdigest(self) -> str:
+        """The digest of the written or read value.
+
+        Returns:
+            The digest in hex formaat.
+        """
+        return self._hash.hexdigest()
+
+    def wrap(self) -> BinaryIO:
+        # This class implements a subset the IO interface and passes through everything else via __getattr__
+        return cast(BinaryIO, self)
+
+    # Passthrough any other calls
+    def __getattr__(self, attr_name: str) -> Any:
+        return getattr(self._source, attr_name)
+
+
+class SHA256TransparentIOReader:
+    """Will generate a SHA256 hash from a source stream transparently.
+
+    Args:
+        source: Source IO stream.
+    """
+
+    def __init__(self, source: IO):
+        self._hash = hashlib.sha256()
+        self._source = source
+
+    def read(self, n: int = -1) -> bytes:
+        """Wrapper for source.read()
+
+        Args:
+            n
+
+        Returns:
+            the value of source.read()
+        """
+        bytes = self._source.read(n)
+        self._hash.update(bytes)
+        return bytes
+
+    def hexdigest(self) -> str:
+        """The digest of the written or read value.
+
+        Returns:
+            The digest in hex formaat.
+        """
+        return self._hash.hexdigest()
+
+    def wrap(self) -> IO:
+        # This class implements a subset the IO interface and passes through everything else via __getattr__
+        return cast(IO, self)
+
+    # Passthrough any other calls
+    def __getattr__(self, attr_name: str) -> Any:
+        return getattr(self._source, attr_name)
+
+
 class MediaStorage:
    """Responsible for storing/fetching files from local sources.

@@ -107,7 +190,6 @@ class MediaStorage:
        Returns:
            the file path written to in the primary media store
        """
-
        async with self.store_into_file(file_info) as (f, fname):
            # Write to the main media repository
            await self.write_to_file(source, f)
@@ -34,6 +34,7 @@ from synapse.logging.opentracing import trace
 from synapse.media._base import (
    FileInfo,
    ThumbnailInfo,
+    check_for_cached_entry_and_respond,
    respond_404,
    respond_with_file,
    respond_with_multipart_responder,
@@ -294,6 +295,11 @@ class ThumbnailProvider:
            if media_info.authenticated:
                raise NotFoundError()

+        # Once we've checked auth we can return early if the media is cached on
+        # the client
+        if check_for_cached_entry_and_respond(request):
+            return
+
        thumbnail_infos = await self.store.get_local_media_thumbnails(media_id)
        await self._select_and_respond_with_thumbnail(
            request,
@@ -334,6 +340,11 @@ class ThumbnailProvider:
            if media_info.authenticated:
                raise NotFoundError()

+        # Once we've checked auth we can return early if the media is cached on
+        # the client
+        if check_for_cached_entry_and_respond(request):
+            return
+
        thumbnail_infos = await self.store.get_local_media_thumbnails(media_id)
        for info in thumbnail_infos:
            t_w = info.width == desired_width
@@ -431,6 +442,10 @@ class ThumbnailProvider:
                respond_404(request)
                return

+        # Check if the media is cached on the client, if so return 304.
+        if check_for_cached_entry_and_respond(request):
+            return
+
        thumbnail_infos = await self.store.get_remote_media_thumbnails(
            server_name, media_id
        )
@@ -510,6 +525,10 @@ class ThumbnailProvider:
            if media_info.authenticated:
                raise NotFoundError()

+        # Check if the media is cached on the client, if so return 304.
+        if check_for_cached_entry_and_respond(request):
+            return
+
        thumbnail_infos = await self.store.get_remote_media_thumbnails(
            server_name, media_id
        )
@@ -41,7 +41,7 @@ from synapse.api.errors import Codes, SynapseError
 from synapse.http.client import SimpleHttpClient
 from synapse.logging.context import make_deferred_yieldable, run_in_background
 from synapse.media._base import FileInfo, get_filename_from_headers
-from synapse.media.media_storage import MediaStorage
+from synapse.media.media_storage import MediaStorage, SHA256TransparentIOWriter
 from synapse.media.oembed import OEmbedProvider
 from synapse.media.preview_html import decode_body, parse_html_to_open_graph
 from synapse.metrics.background_process_metrics import run_as_background_process
@@ -593,17 +593,26 @@ class UrlPreviewer:
        file_info = FileInfo(server_name=None, file_id=file_id, url_cache=True)

        async with self.media_storage.store_into_file(file_info) as (f, fname):
+            sha256writer = SHA256TransparentIOWriter(f)
            if url.startswith("data:"):
                if not allow_data_urls:
                    raise SynapseError(
                        500, "Previewing of data: URLs is forbidden", Codes.UNKNOWN
                    )

-                download_result = await self._parse_data_url(url, f)
+                download_result = await self._parse_data_url(url, sha256writer.wrap())
            else:
-                download_result = await self._download_url(url, f)
+                download_result = await self._download_url(url, sha256writer.wrap())

        try:
+            sha256 = sha256writer.hexdigest()
+            should_quarantine = await self.store.get_is_hash_quarantined(sha256)
+
+            if should_quarantine:
+                logger.warn(
+                    "Media has been automatically quarantined as it matched existing quarantined media"
+                )
+
            time_now_ms = self.clock.time_msec()

            await self.store.store_local_media(
@@ -614,6 +623,8 @@ class UrlPreviewer:
                media_length=download_result.length,
                user_id=user,
                url_cache=url,
+                sha256=sha256,
+                quarantined_by="system" if should_quarantine else None,
            )

        except Exception as e:
@@ -66,7 +66,6 @@ from synapse.types import (
 from synapse.util.async_helpers import (
    timeout_deferred,
 )
-from synapse.util.metrics import Measure
 from synapse.util.stringutils import shortstr
 from synapse.visibility import filter_events_for_client

@@ -520,20 +519,22 @@ class Notifier:
        users = users or []
        rooms = rooms or []

-        with Measure(self.clock, "on_new_event"):
-            user_streams: Set[_NotifierUserStream] = set()
+        user_streams: Set[_NotifierUserStream] = set()

-            log_kv(
-                {
-                    "waking_up_explicit_users": len(users),
-                    "waking_up_explicit_rooms": len(rooms),
-                    "users": shortstr(users),
-                    "rooms": shortstr(rooms),
-                    "stream": stream_key,
-                    "stream_id": new_token,
-                }
-            )
+        log_kv(
+            {
+                "waking_up_explicit_users": len(users),
+                "waking_up_explicit_rooms": len(rooms),
+                "users": shortstr(users),
+                "rooms": shortstr(rooms),
+                "stream": stream_key,
+                "stream_id": new_token,
+            }
+        )

+        # Only calculate which user streams to wake up if there are, in fact,
+        # any user streams registered.
+        if self.user_to_user_stream or self.room_to_user_streams:
            for user in users:
                user_stream = self.user_to_user_stream.get(str(user))
                if user_stream is not None:
@@ -565,25 +566,25 @@ class Notifier:
            # We resolve all these deferreds in one go so that we only need to
            # call `PreserveLoggingContext` once, as it has a bunch of overhead
            # (to calculate performance stats)
-            with PreserveLoggingContext():
-                for listener in listeners:
-                    listener.callback(current_token)
+            if listeners:
+                with PreserveLoggingContext():
+                    for listener in listeners:
+                        listener.callback(current_token)

-            users_woken_by_stream_counter.labels(stream_key).inc(len(user_streams))
+            if user_streams:
+                users_woken_by_stream_counter.labels(stream_key).inc(len(user_streams))

-            self.notify_replication()
+        self.notify_replication()

-            # Notify appservices.
-            try:
-                self.appservice_handler.notify_interested_services_ephemeral(
-                    stream_key,
-                    new_token,
-                    users,
-                )
-            except Exception:
-                logger.exception(
-                    "Error notifying application services of ephemeral events"
-                )
+        # Notify appservices.
+        try:
+            self.appservice_handler.notify_interested_services_ephemeral(
+                stream_key,
+                new_token,
+                users,
+            )
+        except Exception:
+            logger.exception("Error notifying application services of ephemeral events")

    def on_new_replication_data(self) -> None:
        """Used to inform replication listeners that something has happened
@@ -205,6 +205,12 @@ class HttpPusher(Pusher):
        if self._is_processing:
            return

+        # Check if we are trying, but failing, to contact the pusher. If so, we
+        # don't try and start processing immediately and instead wait for the
+        # retry loop to try again later (which is controlled by the timer).
+        if self.failing_since and self.timed_call and self.timed_call.active():
+            return
+
        run_as_background_process("httppush.process", self._process)

    async def _process(self) -> None:
@@ -187,7 +187,6 @@ class ClientRestResource(JsonResource):
                    mutual_rooms.register_servlets,
                    login_token_request.register_servlets,
                    rendezvous.register_servlets,
-                    auth_metadata.register_servlets,
                ]:
                    continue

@@ -39,7 +39,7 @@ from typing import TYPE_CHECKING, Optional, Tuple

 from synapse.api.errors import Codes, NotFoundError, SynapseError
 from synapse.handlers.pagination import PURGE_HISTORY_ACTION_NAME
-from synapse.http.server import HttpServer, JsonResource
+from synapse.http.server import HttpServer
 from synapse.http.servlet import RestServlet, parse_json_object_from_request
 from synapse.http.site import SynapseRequest
 from synapse.rest.admin._base import admin_patterns, assert_requester_is_admin
@@ -51,6 +51,7 @@ from synapse.rest.admin.background_updates import (
 from synapse.rest.admin.devices import (
    DeleteDevicesRestServlet,
    DeviceRestServlet,
+    DevicesGetRestServlet,
    DevicesRestServlet,
 )
 from synapse.rest.admin.event_reports import (
@@ -86,6 +87,7 @@ from synapse.rest.admin.rooms import (
    RoomStateRestServlet,
    RoomTimestampToEventRestServlet,
 )
+from synapse.rest.admin.scheduled_tasks import ScheduledTasksRestServlet
 from synapse.rest.admin.server_notice_servlet import SendServerNoticeServlet
 from synapse.rest.admin.statistics import (
    LargestRoomsStatistics,
@@ -263,27 +265,24 @@ class PurgeHistoryStatusRestServlet(RestServlet):
 ########################################################################################


-class AdminRestResource(JsonResource):
-    """The REST resource which gets mounted at /_synapse/admin"""
-
-    def __init__(self, hs: "HomeServer"):
-        JsonResource.__init__(self, hs, canonical_json=False)
-        register_servlets(hs, self)
-
-
 def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None:
    """
    Register all the admin servlets.
    """
-    # Admin servlets aren't registered on workers.
+    RoomRestServlet(hs).register(http_server)
+
+    # Admin servlets below may not work on workers.
    if hs.config.worker.worker_app is not None:
+        # Some admin servlets can be mounted on workers when MSC3861 is enabled.
+        if hs.config.experimental.msc3861.enabled:
+            register_servlets_for_msc3861_delegation(hs, http_server)
+
        return

    register_servlets_for_client_rest_resource(hs, http_server)
    BlockRoomRestServlet(hs).register(http_server)
    ListRoomRestServlet(hs).register(http_server)
    RoomStateRestServlet(hs).register(http_server)
-    RoomRestServlet(hs).register(http_server)
    RoomRestV2Servlet(hs).register(http_server)
    RoomMembersRestServlet(hs).register(http_server)
    DeleteRoomStatusByDeleteIdRestServlet(hs).register(http_server)
@@ -337,6 +336,7 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None:
    BackgroundUpdateStartJobRestServlet(hs).register(http_server)
    ExperimentalFeaturesRestServlet(hs).register(http_server)
    SuspendAccountRestServlet(hs).register(http_server)
+    ScheduledTasksRestServlet(hs).register(http_server)


 def register_servlets_for_client_rest_resource(
@@ -364,4 +364,16 @@ def register_servlets_for_client_rest_resource(
        ListMediaInRoom(hs).register(http_server)

    # don't add more things here: new servlets should only be exposed on
-    # /_synapse/admin so should not go here. Instead register them in AdminRestResource.
+    # /_synapse/admin so should not go here. Instead register them in register_servlets.
+
+
+def register_servlets_for_msc3861_delegation(
+    hs: "HomeServer", http_server: HttpServer
+) -> None:
+    """Register servlets needed by MAS when MSC3861 is enabled"""
+    assert hs.config.experimental.msc3861.enabled
+
+    UserRestServletV2(hs).register(http_server)
+    UsernameAvailableRestServlet(hs).register(http_server)
+    UserReplaceMasterCrossSigningKeyRestServlet(hs).register(http_server)
+    DevicesGetRestServlet(hs).register(http_server)
@@ -113,18 +113,19 @@ class DeviceRestServlet(RestServlet):
        return HTTPStatus.OK, {}


-class DevicesRestServlet(RestServlet):
+class DevicesGetRestServlet(RestServlet):
    """
    Retrieve the given user's devices
+
+    This can be mounted on workers as it is read-only, as opposed
+    to `DevicesRestServlet`.
    """

    PATTERNS = admin_patterns("/users/(?P<user_id>[^/]*)/devices$", "v2")

    def __init__(self, hs: "HomeServer"):
        self.auth = hs.get_auth()
-        handler = hs.get_device_handler()
-        assert isinstance(handler, DeviceHandler)
-        self.device_handler = handler
+        self.device_worker_handler = hs.get_device_handler()
        self.store = hs.get_datastores().main
        self.is_mine = hs.is_mine

@@ -141,9 +142,24 @@ class DevicesRestServlet(RestServlet):
        if u is None:
            raise NotFoundError("Unknown user")

-        devices = await self.device_handler.get_devices_by_user(target_user.to_string())
+        devices = await self.device_worker_handler.get_devices_by_user(
+            target_user.to_string()
+        )
        return HTTPStatus.OK, {"devices": devices, "total": len(devices)}

+
+class DevicesRestServlet(DevicesGetRestServlet):
+    """
+    Retrieve the given user's devices
+    """
+
+    PATTERNS = admin_patterns("/users/(?P<user_id>[^/]*)/devices$", "v2")
+
+    def __init__(self, hs: "HomeServer"):
+        super().__init__(hs)
+        assert isinstance(self.device_worker_handler, DeviceHandler)
+        self.device_handler = self.device_worker_handler
+
    async def on_POST(
        self, request: SynapseRequest, user_id: str
    ) -> Tuple[int, JsonDict]:
@@ -0,0 +1,70 @@
+#
+# This file is licensed under the Affero General Public License (AGPL) version 3.
+#
+# Copyright (C) 2025 New Vector, Ltd
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Affero General Public License as
+# published by the Free Software Foundation, either version 3 of the
+# License, or (at your option) any later version.
+#
+# See the GNU Affero General Public License for more details:
+# <https://www.gnu.org/licenses/agpl-3.0.html>.
+#
+#
+#
+from typing import TYPE_CHECKING, Tuple
+
+from synapse.http.servlet import RestServlet, parse_integer, parse_string
+from synapse.http.site import SynapseRequest
+from synapse.rest.admin import admin_patterns, assert_requester_is_admin
+from synapse.types import JsonDict, TaskStatus
+
+if TYPE_CHECKING:
+    from synapse.server import HomeServer
+
+
+class ScheduledTasksRestServlet(RestServlet):
+    """Get a list of scheduled tasks and their statuses
+    optionally filtered by action name, resource id, status, and max timestamp
+    """
+
+    PATTERNS = admin_patterns("/scheduled_tasks$")
+
+    def __init__(self, hs: "HomeServer"):
+        self._auth = hs.get_auth()
+        self._store = hs.get_datastores().main
+
+    async def on_GET(self, request: SynapseRequest) -> Tuple[int, JsonDict]:
+        await assert_requester_is_admin(self._auth, request)
+
+        # extract query params
+        action_name = parse_string(request, "action_name")
+        resource_id = parse_string(request, "resource_id")
+        status = parse_string(request, "job_status")
+        max_timestamp = parse_integer(request, "max_timestamp")
+
+        actions = [action_name] if action_name else None
+        statuses = [TaskStatus(status)] if status else None
+
+        tasks = await self._store.get_scheduled_tasks(
+            actions=actions,
+            resource_id=resource_id,
+            statuses=statuses,
+            max_timestamp=max_timestamp,
+        )
+
+        json_tasks = []
+        for task in tasks:
+            result_task = {
+                "id": task.id,
+                "action": task.action,
+                "status": task.status,
+                "timestamp_ms": task.timestamp,
+                "resource_id": task.resource_id,
+                "result": task.result,
+                "error": task.error,
+            }
+            json_tasks.append(result_task)
+
+        return 200, {"scheduled_tasks": json_tasks}
@@ -350,6 +350,7 @@ class EmailThreepidRequestTokenRestServlet(RestServlet):
            raise SynapseError(
                400,
                "Adding an email to your account is disabled on this server",
+                Codes.THREEPID_MEDIUM_NOT_SUPPORTED,
            )

        body = parse_and_validate_json_object_from_request(
@@ -456,6 +457,7 @@ class MsisdnThreepidRequestTokenRestServlet(RestServlet):
            raise SynapseError(
                400,
                "Adding phone numbers to user account is not supported by this homeserver",
+                Codes.THREEPID_MEDIUM_NOT_SUPPORTED,
            )

        ret = await self.identity_handler.requestMsisdnToken(
@@ -498,7 +500,9 @@ class AddThreepidEmailSubmitTokenServlet(RestServlet):
                "Adding emails have been disabled due to lack of an email config"
            )
            raise SynapseError(
-                400, "Adding an email to your account is disabled on this server"
+                400,
+                "Adding an email to your account is disabled on this server",
+                Codes.THREEPID_MEDIUM_NOT_SUPPORTED,
            )

        sid = parse_string(request, "sid", required=True)
@@ -143,11 +143,11 @@ class DeviceRestServlet(RestServlet):
        self.hs = hs
        self.auth = hs.get_auth()
        handler = hs.get_device_handler()
-        assert isinstance(handler, DeviceHandler)
        self.device_handler = handler
        self.auth_handler = hs.get_auth_handler()
        self._msc3852_enabled = hs.config.experimental.msc3852_enabled
        self._msc3861_oauth_delegation_enabled = hs.config.experimental.msc3861.enabled
+        self._is_main_process = hs.config.worker.worker_app is None

    async def on_GET(
        self, request: SynapseRequest, device_id: str
@@ -179,6 +179,14 @@ class DeviceRestServlet(RestServlet):
    async def on_DELETE(
        self, request: SynapseRequest, device_id: str
    ) -> Tuple[int, JsonDict]:
+        # Can only be run on main process, as changes to device lists must
+        # happen on main.
+        if not self._is_main_process:
+            error_message = "DELETE on /devices/ must be routed to main process"
+            logger.error(error_message)
+            raise SynapseError(500, error_message)
+        assert isinstance(self.device_handler, DeviceHandler)
+
        requester = await self.auth.get_user_by_req(request)

        try:
@@ -223,6 +231,14 @@ class DeviceRestServlet(RestServlet):
    async def on_PUT(
        self, request: SynapseRequest, device_id: str
    ) -> Tuple[int, JsonDict]:
+        # Can only be run on main process, as changes to device lists must
+        # happen on main.
+        if not self._is_main_process:
+            error_message = "PUT on /devices/ must be routed to main process"
+            logger.error(error_message)
+            raise SynapseError(500, error_message)
+        assert isinstance(self.device_handler, DeviceHandler)
+
        requester = await self.auth.get_user_by_req(request, allow_guest=True)

        body = parse_and_validate_json_object_from_request(request, self.PutBody)
@@ -585,9 +601,9 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None:
    ):
        DeleteDevicesRestServlet(hs).register(http_server)
    DevicesRestServlet(hs).register(http_server)
+    DeviceRestServlet(hs).register(http_server)

    if hs.config.worker.worker_app is None:
-        DeviceRestServlet(hs).register(http_server)
        if hs.config.experimental.msc2697_enabled:
            DehydratedDeviceServlet(hs).register(http_server)
            ClaimDehydratedDeviceServlet(hs).register(http_server)
@@ -24,7 +24,7 @@ from collections import defaultdict
 from typing import TYPE_CHECKING, Any, Dict, List, Mapping, Optional, Tuple, Union

 from synapse.api.constants import AccountDataTypes, EduTypes, Membership, PresenceState
-from synapse.api.errors import Codes, LimitExceededError, StoreError, SynapseError
+from synapse.api.errors import Codes, StoreError, SynapseError
 from synapse.api.filtering import FilterCollection
 from synapse.api.presence import UserPresenceState
 from synapse.api.ratelimiting import Ratelimiter
@@ -248,9 +248,8 @@ class SyncRestServlet(RestServlet):
        await self._server_notices_sender.on_user_syncing(user.to_string())

        # ignore the presence update if the ratelimit is exceeded but do not pause the request
-        try:
-            await self._presence_per_user_limiter.ratelimit(requester, pause=0.0)
-        except LimitExceededError:
+        allowed, _ = await self._presence_per_user_limiter.can_do_action(requester)
+        if not allowed:
            affect_presence = False
            logger.debug("User set_presence ratelimit exceeded; ignoring it.")
        else:
@@ -130,7 +130,7 @@ class SQLBaseStore(metaclass=ABCMeta):
                "_get_rooms_for_local_user_where_membership_is_inner", (user_id,)
            )
            self._attempt_to_invalidate_cache(
-                "get_sliding_sync_rooms_for_user", (user_id,)
+                "get_sliding_sync_rooms_for_user_from_membership_snapshots", (user_id,)
            )

        # Purge other caches based on room state.
@@ -138,7 +138,9 @@ class SQLBaseStore(metaclass=ABCMeta):
        self._attempt_to_invalidate_cache("get_partial_current_state_ids", (room_id,))
        self._attempt_to_invalidate_cache("get_room_type", (room_id,))
        self._attempt_to_invalidate_cache("get_room_encryption", (room_id,))
-        self._attempt_to_invalidate_cache("get_sliding_sync_rooms_for_user", None)
+        self._attempt_to_invalidate_cache(
+            "get_sliding_sync_rooms_for_user_from_membership_snapshots", None
+        )

    def _invalidate_state_caches_all(self, room_id: str) -> None:
        """Invalidates caches that are based on the current state, but does
@@ -168,7 +170,9 @@ class SQLBaseStore(metaclass=ABCMeta):
        self._attempt_to_invalidate_cache("get_room_summary", (room_id,))
        self._attempt_to_invalidate_cache("get_room_type", (room_id,))
        self._attempt_to_invalidate_cache("get_room_encryption", (room_id,))
-        self._attempt_to_invalidate_cache("get_sliding_sync_rooms_for_user", None)
+        self._attempt_to_invalidate_cache(
+            "get_sliding_sync_rooms_for_user_from_membership_snapshots", None
+        )

    def _attempt_to_invalidate_cache(
        self, cache_name: str, key: Optional[Collection[Any]]
@@ -21,11 +21,19 @@

 import itertools
 import logging
-from typing import TYPE_CHECKING, Collection, Mapping, Set
+from typing import (
+    TYPE_CHECKING,
+    Collection,
+    Mapping,
+    Optional,
+    Set,
+)

 from synapse.logging.context import nested_logging_context
 from synapse.metrics.background_process_metrics import wrap_as_background_process
+from synapse.storage.database import LoggingTransaction
 from synapse.storage.databases import Databases
+from synapse.types.storage import _BackgroundUpdates

 if TYPE_CHECKING:
    from synapse.server import HomeServer
@@ -44,6 +52,11 @@ class PurgeEventsStorageController:
                self._delete_state_groups_loop, 60 * 1000
            )

+        self.stores.state.db_pool.updates.register_background_update_handler(
+            _BackgroundUpdates.MARK_UNREFERENCED_STATE_GROUPS_FOR_DELETION_BG_UPDATE,
+            self._background_delete_unrefereneced_state_groups,
+        )
+
    async def purge_room(self, room_id: str) -> None:
        """Deletes all record of a room"""

@@ -81,7 +94,8 @@ class PurgeEventsStorageController:
            )

    async def _find_unreferenced_groups(
-        self, state_groups: Collection[int]
+        self,
+        state_groups: Collection[int],
    ) -> Set[int]:
        """Used when purging history to figure out which state groups can be
        deleted.
@@ -203,3 +217,232 @@ class PurgeEventsStorageController:
            room_id,
            groups_to_sequences,
        )
+
+    async def _background_delete_unrefereneced_state_groups(
+        self, progress: dict, batch_size: int
+    ) -> int:
+        """This background update will slowly delete any unreferenced state groups"""
+
+        last_checked_state_group = progress.get("last_checked_state_group")
+
+        if last_checked_state_group is None:
+            # This is the first run.
+            last_checked_state_group = (
+                await self.stores.state.db_pool.simple_select_one_onecol(
+                    table="state_groups",
+                    keyvalues={},
+                    retcol="MAX(id)",
+                    allow_none=True,
+                    desc="get_max_state_group",
+                )
+            )
+            if last_checked_state_group is None:
+                # There are no state groups so the background process is finished.
+                await self.stores.state.db_pool.updates._end_background_update(
+                    _BackgroundUpdates.MARK_UNREFERENCED_STATE_GROUPS_FOR_DELETION_BG_UPDATE
+                )
+                return batch_size
+            last_checked_state_group += 1
+
+        (
+            last_checked_state_group,
+            final_batch,
+        ) = await self._delete_unreferenced_state_groups_batch(
+            last_checked_state_group,
+            batch_size,
+        )
+
+        if not final_batch:
+            # There are more state groups to check.
+            progress = {
+                "last_checked_state_group": last_checked_state_group,
+            }
+            await self.stores.state.db_pool.updates._background_update_progress(
+                _BackgroundUpdates.MARK_UNREFERENCED_STATE_GROUPS_FOR_DELETION_BG_UPDATE,
+                progress,
+            )
+        else:
+            # This background process is finished.
+            await self.stores.state.db_pool.updates._end_background_update(
+                _BackgroundUpdates.MARK_UNREFERENCED_STATE_GROUPS_FOR_DELETION_BG_UPDATE
+            )
+
+        return batch_size
+
+    async def _delete_unreferenced_state_groups_batch(
+        self,
+        last_checked_state_group: int,
+        batch_size: int,
+    ) -> tuple[int, bool]:
+        """Looks for unreferenced state groups starting from the last state group
+        checked and marks them for deletion.
+
+        Args:
+            last_checked_state_group: The last state group that was checked.
+            batch_size: How many state groups to process in this iteration.
+
+        Returns:
+            (last_checked_state_group, final_batch)
+        """
+
+        # Find all state groups that can be deleted if any of the original set are deleted.
+        (
+            to_delete,
+            last_checked_state_group,
+            final_batch,
+        ) = await self._find_unreferenced_groups_for_background_deletion(
+            last_checked_state_group, batch_size
+        )
+
+        if len(to_delete) == 0:
+            return last_checked_state_group, final_batch
+
+        await self.stores.state_deletion.mark_state_groups_as_pending_deletion(
+            to_delete
+        )
+
+        return last_checked_state_group, final_batch
+
+    async def _find_unreferenced_groups_for_background_deletion(
+        self,
+        last_checked_state_group: int,
+        batch_size: int,
+    ) -> tuple[Set[int], int, bool]:
+        """Used when deleting unreferenced state groups in the background to figure out
+        which state groups can be deleted.
+        To avoid increased DB usage due to de-deltaing state groups, this returns only
+        state groups which are free standing (ie. no shared edges with referenced groups) or
+        state groups which do not share edges which result in a future referenced group.
+
+        The following scenarios outline the possibilities based on state group data in
+        the DB.
+
+        ie. Free standing -> state groups 1-N would be returned:
+            SG_1
+            |
+            ...
+            |
+            SG_N
+
+        ie. Previous reference -> state groups 2-N would be returned:
+            SG_1 <- referenced by event
+            |
+            SG_2
+            |
+            ...
+            |
+            SG_N
+
+        ie. Future reference -> none of the following state groups would be returned:
+            SG_1
+            |
+            SG_2
+            |
+            ...
+            |
+            SG_N <- referenced by event
+
+        Args:
+            last_checked_state_group: The last state group that was checked.
+            batch_size: How many state groups to process in this iteration.
+
+        Returns:
+            (to_delete, last_checked_state_group, final_batch)
+        """
+
+        # If a state group's next edge is not pending deletion then we don't delete the state group.
+        # If there is no next edge or the next edges are all marked for deletion, then delete
+        # the state group.
+        # This holds since we walk backwards from the latest state groups, ensuring that
+        # we've already checked newer state groups for event references along the way.
+        def get_next_state_groups_marked_for_deletion_txn(
+            txn: LoggingTransaction,
+        ) -> tuple[dict[int, bool], dict[int, int]]:
+            state_group_sql = """
+                SELECT s.id, e.state_group, d.state_group
+                FROM (
+                    SELECT id FROM state_groups
+                    WHERE id < ? ORDER BY id DESC LIMIT ?
+                ) as s
+                LEFT JOIN state_group_edges AS e ON (s.id = e.prev_state_group)
+                LEFT JOIN state_groups_pending_deletion AS d ON (e.state_group = d.state_group)
+            """
+            txn.execute(state_group_sql, (last_checked_state_group, batch_size))
+
+            # Mapping from state group to whether we should delete it.
+            state_groups_to_deletion: dict[int, bool] = {}
+
+            # Mapping from state group to prev state group.
+            state_groups_to_prev: dict[int, int] = {}
+
+            for row in txn:
+                state_group = row[0]
+                next_edge = row[1]
+                pending_deletion = row[2]
+
+                if next_edge is not None:
+                    state_groups_to_prev[next_edge] = state_group
+
+                if next_edge is not None and not pending_deletion:
+                    # We have found an edge not marked for deletion.
+                    # Check previous results to see if this group is part of a chain
+                    # within this batch that qualifies for deletion.
+                    # ie. batch contains:
+                    # SG_1 -> SG_2 -> SG_3
+                    # If SG_3 is a candidate for deletion, then SG_2 & SG_1 should also
+                    # be, even though they have edges which may not be marked for
+                    # deletion.
+                    # This relies on SQL results being sorted in DESC order to work.
+                    next_is_deletion_candidate = state_groups_to_deletion.get(next_edge)
+                    if (
+                        next_is_deletion_candidate is None
+                        or not next_is_deletion_candidate
+                    ):
+                        state_groups_to_deletion[state_group] = False
+                    else:
+                        state_groups_to_deletion.setdefault(state_group, True)
+                else:
+                    # This state group may be a candidate for deletion
+                    state_groups_to_deletion.setdefault(state_group, True)
+
+            return state_groups_to_deletion, state_groups_to_prev
+
+        (
+            state_groups_to_deletion,
+            state_group_edges,
+        ) = await self.stores.state.db_pool.runInteraction(
+            "get_next_state_groups_marked_for_deletion",
+            get_next_state_groups_marked_for_deletion_txn,
+        )
+        deletion_candidates = {
+            state_group
+            for state_group, deletion in state_groups_to_deletion.items()
+            if deletion
+        }
+
+        final_batch = False
+        state_groups = state_groups_to_deletion.keys()
+        if len(state_groups) < batch_size:
+            final_batch = True
+        else:
+            last_checked_state_group = min(state_groups)
+
+        if len(state_groups) == 0:
+            return set(), last_checked_state_group, final_batch
+
+        # Determine if any of the remaining state groups are directly referenced.
+        referenced = await self.stores.main.get_referenced_state_groups(
+            deletion_candidates
+        )
+
+        # Remove state groups from deletion_candidates which are directly referenced or share a
+        # future edge with a referenced state group within this batch.
+        def filter_reference_chains(group: Optional[int]) -> None:
+            while group is not None:
+                deletion_candidates.discard(group)
+                group = state_group_edges.get(group)
+
+        for referenced_group in referenced:
+            filter_reference_chains(referenced_group)
+
+        return deletion_candidates, last_checked_state_group, final_batch
@@ -307,7 +307,7 @@ class CacheInvalidationWorkerStore(SQLBaseStore):
                    "get_rooms_for_user", (data.state_key,)
                )
                self._attempt_to_invalidate_cache(
-                    "get_sliding_sync_rooms_for_user", None
+                    "get_sliding_sync_rooms_for_user_from_membership_snapshots", None
                )
                self._membership_stream_cache.entity_has_changed(data.state_key, token)  # type: ignore[attr-defined]
            elif data.type == EventTypes.RoomEncryption:
@@ -319,7 +319,7 @@ class CacheInvalidationWorkerStore(SQLBaseStore):

            if (data.type, data.state_key) in SLIDING_SYNC_RELEVANT_STATE_SET:
                self._attempt_to_invalidate_cache(
-                    "get_sliding_sync_rooms_for_user", None
+                    "get_sliding_sync_rooms_for_user_from_membership_snapshots", None
                )
        elif row.type == EventsStreamAllStateRow.TypeId:
            assert isinstance(data, EventsStreamAllStateRow)
@@ -330,7 +330,9 @@ class CacheInvalidationWorkerStore(SQLBaseStore):
            self._attempt_to_invalidate_cache("get_rooms_for_user", None)
            self._attempt_to_invalidate_cache("get_room_type", (data.room_id,))
            self._attempt_to_invalidate_cache("get_room_encryption", (data.room_id,))
-            self._attempt_to_invalidate_cache("get_sliding_sync_rooms_for_user", None)
+            self._attempt_to_invalidate_cache(
+                "get_sliding_sync_rooms_for_user_from_membership_snapshots", None
+            )
        else:
            raise Exception("Unknown events stream row type %s" % (row.type,))

@@ -394,7 +396,8 @@ class CacheInvalidationWorkerStore(SQLBaseStore):
                "_get_rooms_for_local_user_where_membership_is_inner", (state_key,)
            )
            self._attempt_to_invalidate_cache(
-                "get_sliding_sync_rooms_for_user", (state_key,)
+                "get_sliding_sync_rooms_for_user_from_membership_snapshots",
+                (state_key,),
            )

            self._attempt_to_invalidate_cache(
@@ -413,7 +416,9 @@ class CacheInvalidationWorkerStore(SQLBaseStore):
            self._attempt_to_invalidate_cache("get_room_encryption", (room_id,))

        if (etype, state_key) in SLIDING_SYNC_RELEVANT_STATE_SET:
-            self._attempt_to_invalidate_cache("get_sliding_sync_rooms_for_user", None)
+            self._attempt_to_invalidate_cache(
+                "get_sliding_sync_rooms_for_user_from_membership_snapshots", None
+            )

        if relates_to:
            self._attempt_to_invalidate_cache(
@@ -470,7 +475,9 @@ class CacheInvalidationWorkerStore(SQLBaseStore):
        self._attempt_to_invalidate_cache(
            "_get_rooms_for_local_user_where_membership_is_inner", None
        )
-        self._attempt_to_invalidate_cache("get_sliding_sync_rooms_for_user", None)
+        self._attempt_to_invalidate_cache(
+            "get_sliding_sync_rooms_for_user_from_membership_snapshots", None
+        )
        self._attempt_to_invalidate_cache("did_forget", None)
        self._attempt_to_invalidate_cache("get_forgotten_rooms_for_user", None)
        self._attempt_to_invalidate_cache("get_references_for_event", None)
@@ -529,7 +536,9 @@ class CacheInvalidationWorkerStore(SQLBaseStore):
        self._attempt_to_invalidate_cache(
            "get_current_hosts_in_room_ordered", (room_id,)
        )
-        self._attempt_to_invalidate_cache("get_sliding_sync_rooms_for_user", None)
+        self._attempt_to_invalidate_cache(
+            "get_sliding_sync_rooms_for_user_from_membership_snapshots", None
+        )
        self._attempt_to_invalidate_cache("did_forget", None)
        self._attempt_to_invalidate_cache("get_forgotten_rooms_for_user", None)
        self._attempt_to_invalidate_cache("_get_membership_from_event_id", None)
@@ -424,25 +424,37 @@ class DelayedEventsStore(SQLBaseStore):
        room_id: str,
        event_type: str,
        state_key: str,
+        not_from_localpart: str,
    ) -> Optional[Timestamp]:
        """
        Cancels all matching delayed state events, i.e. remove them as long as they haven't been processed.

+        Args:
+            room_id: The room ID to match against.
+            event_type: The event type to match against.
+            state_key: The state key to match against.
+            not_from_localpart: The localpart of a user whose delayed events to not cancel.
+                If set to the empty string, any users' delayed events may be cancelled.
+
        Returns: The send time of the next delayed event to be sent, if any.
        """

        def cancel_delayed_state_events_txn(
            txn: LoggingTransaction,
        ) -> Optional[Timestamp]:
-            self.db_pool.simple_delete_txn(
-                txn,
-                table="delayed_events",
-                keyvalues={
-                    "room_id": room_id,
-                    "event_type": event_type,
-                    "state_key": state_key,
-                    "is_processed": False,
-                },
+            txn.execute(
+                """
+                DELETE FROM delayed_events
+                WHERE room_id = ? AND event_type = ? AND state_key = ?
+                    AND user_localpart <> ?
+                    AND NOT is_processed
+                """,
+                (
+                    room_id,
+                    event_type,
+                    state_key,
+                    not_from_localpart,
+                ),
            )
            return self._get_next_delayed_event_send_ts_txn(txn)

@@ -282,9 +282,10 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore):
            "count_devices_by_users", count_devices_by_users_txn, user_ids
        )

+    @cached()
    async def get_device(
        self, user_id: str, device_id: str
-    ) -> Optional[Dict[str, Any]]:
+    ) -> Optional[Mapping[str, Any]]:
        """Retrieve a device. Only returns devices that are not marked as
        hidden.

@@ -1817,6 +1818,8 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore):
                },
                desc="store_device",
            )
+            await self.invalidate_cache_and_stream("get_device", (user_id, device_id))
+
            if not inserted:
                # if the device already exists, check if it's a real device, or
                # if the device ID is reserved by something else
@@ -1882,6 +1885,9 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore):
                values=device_ids,
                keyvalues={"user_id": user_id},
            )
+            self._invalidate_cache_and_stream_bulk(
+                txn, self.get_device, [(user_id, device_id) for device_id in device_ids]
+            )

        for batch in batch_iter(device_ids, 100):
            await self.db_pool.runInteraction(
@@ -1915,6 +1921,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore):
            updatevalues=updates,
            desc="update_device",
        )
+        await self.invalidate_cache_and_stream("get_device", (user_id, device_id))

    async def update_remote_device_list_cache_entry(
        self, user_id: str, device_id: str, content: JsonDict, stream_id: str
@@ -1501,6 +1501,45 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker
            "delete_old_otks_for_next_user_batch", impl
        )

+    async def allow_master_cross_signing_key_replacement_without_uia(
+        self, user_id: str, duration_ms: int
+    ) -> Optional[int]:
+        """Mark this user's latest master key as being replaceable without UIA.
+
+        Said replacement will only be permitted for a short time after calling this
+        function. That time period is controlled by the duration argument.
+
+        Returns:
+            None, if there is no such key.
+            Otherwise, the timestamp before which replacement is allowed without UIA.
+        """
+        timestamp = self._clock.time_msec() + duration_ms
+
+        def impl(txn: LoggingTransaction) -> Optional[int]:
+            txn.execute(
+                """
+                UPDATE e2e_cross_signing_keys
+                SET updatable_without_uia_before_ms = ?
+                WHERE stream_id = (
+                    SELECT stream_id
+                    FROM e2e_cross_signing_keys
+                    WHERE user_id = ? AND keytype = 'master'
+                    ORDER BY stream_id DESC
+                    LIMIT 1
+                )
+            """,
+                (timestamp, user_id),
+            )
+            if txn.rowcount == 0:
+                return None
+
+            return timestamp
+
+        return await self.db_pool.runInteraction(
+            "allow_master_cross_signing_key_replacement_without_uia",
+            impl,
+        )
+

 class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore):
    def __init__(
@@ -1755,42 +1794,3 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore):
            ],
            desc="add_e2e_signing_key",
        )
-
-    async def allow_master_cross_signing_key_replacement_without_uia(
-        self, user_id: str, duration_ms: int
-    ) -> Optional[int]:
-        """Mark this user's latest master key as being replaceable without UIA.
-
-        Said replacement will only be permitted for a short time after calling this
-        function. That time period is controlled by the duration argument.
-
-        Returns:
-            None, if there is no such key.
-            Otherwise, the timestamp before which replacement is allowed without UIA.
-        """
-        timestamp = self._clock.time_msec() + duration_ms
-
-        def impl(txn: LoggingTransaction) -> Optional[int]:
-            txn.execute(
-                """
-                UPDATE e2e_cross_signing_keys
-                SET updatable_without_uia_before_ms = ?
-                WHERE stream_id = (
-                    SELECT stream_id
-                    FROM e2e_cross_signing_keys
-                    WHERE user_id = ? AND keytype = 'master'
-                    ORDER BY stream_id DESC
-                    LIMIT 1
-                )
-            """,
-                (timestamp, user_id),
-            )
-            if txn.rowcount == 0:
-                return None
-
-            return timestamp
-
-        return await self.db_pool.runInteraction(
-            "allow_master_cross_signing_key_replacement_without_uia",
-            impl,
-        )
@@ -19,6 +19,7 @@
 # [This file includes modifications made by New Vector Limited]
 #
 #
+import logging
 from enum import Enum
 from typing import (
    TYPE_CHECKING,
@@ -51,6 +52,8 @@ BG_UPDATE_REMOVE_MEDIA_REPO_INDEX_WITHOUT_METHOD_2 = (
    "media_repository_drop_index_wo_method_2"
 )

+logger = logging.getLogger(__name__)
+

@attr.s(slots=True, frozen=True, auto_attribs=True)
 class LocalMedia:
@@ -65,6 +68,7 @@ class LocalMedia:
    safe_from_quarantine: bool
    user_id: Optional[str]
    authenticated: Optional[bool]
+    sha256: Optional[str]


@attr.s(slots=True, frozen=True, auto_attribs=True)
@@ -79,6 +83,7 @@ class RemoteMedia:
    last_access_ts: int
    quarantined_by: Optional[str]
    authenticated: Optional[bool]
+    sha256: Optional[str]


@attr.s(slots=True, frozen=True, auto_attribs=True)
@@ -154,6 +159,26 @@ class MediaRepositoryBackgroundUpdateStore(SQLBaseStore):
            unique=True,
        )

+        self.db_pool.updates.register_background_index_update(
+            update_name="local_media_repository_sha256_idx",
+            index_name="local_media_repository_sha256",
+            table="local_media_repository",
+            where_clause="sha256 IS NOT NULL",
+            columns=[
+                "sha256",
+            ],
+        )
+
+        self.db_pool.updates.register_background_index_update(
+            update_name="remote_media_cache_sha256_idx",
+            index_name="remote_media_cache_sha256",
+            table="remote_media_cache",
+            where_clause="sha256 IS NOT NULL",
+            columns=[
+                "sha256",
+            ],
+        )
+
        self.db_pool.updates.register_background_update_handler(
            BG_UPDATE_REMOVE_MEDIA_REPO_INDEX_WITHOUT_METHOD_2,
            self._drop_media_index_without_method,
@@ -221,6 +246,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
                "safe_from_quarantine",
                "user_id",
                "authenticated",
+                "sha256",
            ),
            allow_none=True,
            desc="get_local_media",
@@ -239,6 +265,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
            safe_from_quarantine=row[7],
            user_id=row[8],
            authenticated=row[9],
+            sha256=row[10],
        )

    async def get_local_media_by_user_paginate(
@@ -295,7 +322,8 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
                    quarantined_by,
                    safe_from_quarantine,
                    user_id,
-                    authenticated
+                    authenticated,
+                    sha256
                FROM local_media_repository
                WHERE user_id = ?
                ORDER BY {order_by_column} {order}, media_id ASC
@@ -320,6 +348,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
                    safe_from_quarantine=bool(row[8]),
                    user_id=row[9],
                    authenticated=row[10],
+                    sha256=row[11],
                )
                for row in txn
            ]
@@ -449,6 +478,8 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
        media_length: int,
        user_id: UserID,
        url_cache: Optional[str] = None,
+        sha256: Optional[str] = None,
+        quarantined_by: Optional[str] = None,
    ) -> None:
        if self.hs.config.media.enable_authenticated_media:
            authenticated = True
@@ -466,6 +497,8 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
                "user_id": user_id.to_string(),
                "url_cache": url_cache,
                "authenticated": authenticated,
+                "sha256": sha256,
+                "quarantined_by": quarantined_by,
            },
            desc="store_local_media",
        )
@@ -477,20 +510,28 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
        upload_name: Optional[str],
        media_length: int,
        user_id: UserID,
+        sha256: str,
        url_cache: Optional[str] = None,
+        quarantined_by: Optional[str] = None,
    ) -> None:
+        updatevalues = {
+            "media_type": media_type,
+            "upload_name": upload_name,
+            "media_length": media_length,
+            "url_cache": url_cache,
+            "sha256": sha256,
+        }
+
+        # This should never be un-set by this function.
+        if quarantined_by is not None:
+            updatevalues["quarantined_by"] = quarantined_by
+
        await self.db_pool.simple_update_one(
            "local_media_repository",
            keyvalues={
-                "user_id": user_id.to_string(),
                "media_id": media_id,
            },
-            updatevalues={
-                "media_type": media_type,
-                "upload_name": upload_name,
-                "media_length": media_length,
-                "url_cache": url_cache,
-            },
+            updatevalues=updatevalues,
            desc="update_local_media",
        )

@@ -657,6 +698,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
                "last_access_ts",
                "quarantined_by",
                "authenticated",
+                "sha256",
            ),
            allow_none=True,
            desc="get_cached_remote_media",
@@ -674,6 +716,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
            last_access_ts=row[5],
            quarantined_by=row[6],
            authenticated=row[7],
+            sha256=row[8],
        )

    async def store_cached_remote_media(
@@ -685,6 +728,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
        time_now_ms: int,
        upload_name: Optional[str],
        filesystem_id: str,
+        sha256: Optional[str],
    ) -> None:
        if self.hs.config.media.enable_authenticated_media:
            authenticated = True
@@ -703,6 +747,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
                "filesystem_id": filesystem_id,
                "last_access_ts": time_now_ms,
                "authenticated": authenticated,
+                "sha256": sha256,
            },
            desc="store_cached_remote_media",
        )
@@ -946,3 +991,46 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore):
        await self.db_pool.runInteraction(
            "delete_url_cache_media", _delete_url_cache_media_txn
        )
+
+    async def get_is_hash_quarantined(self, sha256: str) -> bool:
+        """Get whether a specific sha256 hash digest matches any quarantined media.
+
+        Returns:
+            None if the media_id doesn't exist.
+        """
+
+        # If we don't have the index yet, performance tanks, so we return False.
+        # In the background updates, remote_media_cache_sha256_idx is created
+        # after local_media_repository_sha256_idx, which is why we only need to
+        # check for the completion of the former.
+        if not await self.db_pool.updates.has_completed_background_update(
+            "remote_media_cache_sha256_idx"
+        ):
+            return False
+
+        def get_matching_media_txn(
+            txn: LoggingTransaction, table: str, sha256: str
+        ) -> bool:
+            # Return on first match
+            sql = """
+            SELECT 1
+            FROM local_media_repository
+            WHERE sha256 = ? AND quarantined_by IS NOT NULL
+
+            UNION ALL
+
+            SELECT 1
+            FROM remote_media_cache
+            WHERE sha256 = ? AND quarantined_by IS NOT NULL
+            LIMIT 1
+            """
+            txn.execute(sql, (sha256, sha256))
+            row = txn.fetchone()
+            return row is not None
+
+        return await self.db_pool.runInteraction(
+            "get_matching_media_txn",
+            get_matching_media_txn,
+            "local_media_repository",
+            sha256,
+        )
@@ -759,17 +759,37 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore):
            external_id: id on that system
            user_id: complete mxid that it is mapped to
        """
+        self._invalidate_cache_and_stream(
+            txn, self.get_user_by_external_id, (auth_provider, external_id)
+        )

-        self.db_pool.simple_insert_txn(
+        # This INSERT ... ON CONFLICT DO NOTHING statement will cause a
+        # 'could not serialize access due to concurrent update'
+        # if the row is added concurrently by another transaction.
+        # This is exactly what we want, as it makes the transaction get retried
+        # in a new snapshot where we can check for a genuine conflict.
+        was_inserted = self.db_pool.simple_upsert_txn(
            txn,
            table="user_external_ids",
-            values={
-                "auth_provider": auth_provider,
-                "external_id": external_id,
-                "user_id": user_id,
-            },
+            keyvalues={"auth_provider": auth_provider, "external_id": external_id},
+            values={},
+            insertion_values={"user_id": user_id},
        )

+        if not was_inserted:
+            existing_id = self.db_pool.simple_select_one_onecol_txn(
+                txn,
+                table="user_external_ids",
+                keyvalues={"auth_provider": auth_provider, "user_id": user_id},
+                retcol="external_id",
+                allow_none=True,
+            )
+
+            if existing_id != external_id:
+                raise ExternalIDReuseException(
+                    f"{user_id!r} has external id {existing_id!r} for {auth_provider} but trying to add {external_id!r}"
+                )
+
    async def remove_user_external_id(
        self, auth_provider: str, external_id: str, user_id: str
    ) -> None:
@@ -789,6 +809,9 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore):
            },
            desc="remove_user_external_id",
        )
+        await self.invalidate_cache_and_stream(
+            "get_user_by_external_id", (auth_provider, external_id)
+        )

    async def replace_user_external_id(
        self,
@@ -809,29 +832,20 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore):
            ExternalIDReuseException if the new external_id could not be mapped.
        """

-        def _remove_user_external_ids_txn(
+        def _replace_user_external_id_txn(
            txn: LoggingTransaction,
-            user_id: str,
        ) -> None:
-            """Remove all mappings from external user ids to a mxid
-            If these mappings are not found, this method does nothing.
-
-            Args:
-                user_id: complete mxid that it is mapped to
-            """
-
            self.db_pool.simple_delete_txn(
                txn,
                table="user_external_ids",
                keyvalues={"user_id": user_id},
            )

-        def _replace_user_external_id_txn(
-            txn: LoggingTransaction,
-        ) -> None:
-            _remove_user_external_ids_txn(txn, user_id)
-
            for auth_provider, external_id in record_external_ids:
+                self._invalidate_cache_and_stream(
+                    txn, self.get_user_by_external_id, (auth_provider, external_id)
+                )
+
                self._record_user_external_id_txn(
                    txn,
                    auth_provider,
@@ -847,6 +861,7 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore):
        except self.database_engine.module.IntegrityError:
            raise ExternalIDReuseException()

+    @cached()
    async def get_user_by_external_id(
        self, auth_provider: str, external_id: str
    ) -> Optional[str]:
@@ -2090,6 +2105,136 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore):
            func=is_user_approved_txn,
        )

+    async def set_user_deactivated_status(
+        self, user_id: str, deactivated: bool
+    ) -> None:
+        """Set the `deactivated` property for the provided user to the provided value.
+
+        Args:
+            user_id: The ID of the user to set the status for.
+            deactivated: The value to set for `deactivated`.
+        """
+
+        await self.db_pool.runInteraction(
+            "set_user_deactivated_status",
+            self.set_user_deactivated_status_txn,
+            user_id,
+            deactivated,
+        )
+
+    def set_user_deactivated_status_txn(
+        self, txn: LoggingTransaction, user_id: str, deactivated: bool
+    ) -> None:
+        self.db_pool.simple_update_one_txn(
+            txn=txn,
+            table="users",
+            keyvalues={"name": user_id},
+            updatevalues={"deactivated": 1 if deactivated else 0},
+        )
+        self._invalidate_cache_and_stream(
+            txn, self.get_user_deactivated_status, (user_id,)
+        )
+        self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
+        self._invalidate_cache_and_stream(txn, self.is_guest, (user_id,))
+
+    async def set_user_suspended_status(self, user_id: str, suspended: bool) -> None:
+        """
+        Set whether the user's account is suspended in the `users` table.
+
+        Args:
+            user_id: The user ID of the user in question
+            suspended: True if the user is suspended, false if not
+        """
+        await self.db_pool.runInteraction(
+            "set_user_suspended_status",
+            self.set_user_suspended_status_txn,
+            user_id,
+            suspended,
+        )
+
+    def set_user_suspended_status_txn(
+        self, txn: LoggingTransaction, user_id: str, suspended: bool
+    ) -> None:
+        self.db_pool.simple_update_one_txn(
+            txn=txn,
+            table="users",
+            keyvalues={"name": user_id},
+            updatevalues={"suspended": suspended},
+        )
+        self._invalidate_cache_and_stream(
+            txn, self.get_user_suspended_status, (user_id,)
+        )
+        self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
+
+    async def set_user_locked_status(self, user_id: str, locked: bool) -> None:
+        """Set the `locked` property for the provided user to the provided value.
+
+        Args:
+            user_id: The ID of the user to set the status for.
+            locked: The value to set for `locked`.
+        """
+
+        await self.db_pool.runInteraction(
+            "set_user_locked_status",
+            self.set_user_locked_status_txn,
+            user_id,
+            locked,
+        )
+
+    def set_user_locked_status_txn(
+        self, txn: LoggingTransaction, user_id: str, locked: bool
+    ) -> None:
+        self.db_pool.simple_update_one_txn(
+            txn=txn,
+            table="users",
+            keyvalues={"name": user_id},
+            updatevalues={"locked": locked},
+        )
+        self._invalidate_cache_and_stream(txn, self.get_user_locked_status, (user_id,))
+        self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
+
+    async def update_user_approval_status(
+        self, user_id: UserID, approved: bool
+    ) -> None:
+        """Set the user's 'approved' flag to the given value.
+
+        The boolean will be turned into an int (in update_user_approval_status_txn)
+        because the column is a smallint.
+
+        Args:
+            user_id: the user to update the flag for.
+            approved: the value to set the flag to.
+        """
+        await self.db_pool.runInteraction(
+            "update_user_approval_status",
+            self.update_user_approval_status_txn,
+            user_id.to_string(),
+            approved,
+        )
+
+    def update_user_approval_status_txn(
+        self, txn: LoggingTransaction, user_id: str, approved: bool
+    ) -> None:
+        """Set the user's 'approved' flag to the given value.
+
+        The boolean is turned into an int because the column is a smallint.
+
+        Args:
+            txn: the current database transaction.
+            user_id: the user to update the flag for.
+            approved: the value to set the flag to.
+        """
+        self.db_pool.simple_update_one_txn(
+            txn=txn,
+            table="users",
+            keyvalues={"name": user_id},
+            updatevalues={"approved": approved},
+        )
+
+        # Invalidate the caches of methods that read the value of the 'approved' flag.
+        self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
+        self._invalidate_cache_and_stream(txn, self.is_user_approved, (user_id,))
+

 class RegistrationBackgroundUpdateStore(RegistrationWorkerStore):
    def __init__(
@@ -2202,117 +2347,6 @@ class RegistrationBackgroundUpdateStore(RegistrationWorkerStore):

        return nb_processed

-    async def set_user_deactivated_status(
-        self, user_id: str, deactivated: bool
-    ) -> None:
-        """Set the `deactivated` property for the provided user to the provided value.
-
-        Args:
-            user_id: The ID of the user to set the status for.
-            deactivated: The value to set for `deactivated`.
-        """
-
-        await self.db_pool.runInteraction(
-            "set_user_deactivated_status",
-            self.set_user_deactivated_status_txn,
-            user_id,
-            deactivated,
-        )
-
-    def set_user_deactivated_status_txn(
-        self, txn: LoggingTransaction, user_id: str, deactivated: bool
-    ) -> None:
-        self.db_pool.simple_update_one_txn(
-            txn=txn,
-            table="users",
-            keyvalues={"name": user_id},
-            updatevalues={"deactivated": 1 if deactivated else 0},
-        )
-        self._invalidate_cache_and_stream(
-            txn, self.get_user_deactivated_status, (user_id,)
-        )
-        self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
-        txn.call_after(self.is_guest.invalidate, (user_id,))
-
-    async def set_user_suspended_status(self, user_id: str, suspended: bool) -> None:
-        """
-        Set whether the user's account is suspended in the `users` table.
-
-        Args:
-            user_id: The user ID of the user in question
-            suspended: True if the user is suspended, false if not
-        """
-        await self.db_pool.runInteraction(
-            "set_user_suspended_status",
-            self.set_user_suspended_status_txn,
-            user_id,
-            suspended,
-        )
-
-    def set_user_suspended_status_txn(
-        self, txn: LoggingTransaction, user_id: str, suspended: bool
-    ) -> None:
-        self.db_pool.simple_update_one_txn(
-            txn=txn,
-            table="users",
-            keyvalues={"name": user_id},
-            updatevalues={"suspended": suspended},
-        )
-        self._invalidate_cache_and_stream(
-            txn, self.get_user_suspended_status, (user_id,)
-        )
-        self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
-
-    async def set_user_locked_status(self, user_id: str, locked: bool) -> None:
-        """Set the `locked` property for the provided user to the provided value.
-
-        Args:
-            user_id: The ID of the user to set the status for.
-            locked: The value to set for `locked`.
-        """
-
-        await self.db_pool.runInteraction(
-            "set_user_locked_status",
-            self.set_user_locked_status_txn,
-            user_id,
-            locked,
-        )
-
-    def set_user_locked_status_txn(
-        self, txn: LoggingTransaction, user_id: str, locked: bool
-    ) -> None:
-        self.db_pool.simple_update_one_txn(
-            txn=txn,
-            table="users",
-            keyvalues={"name": user_id},
-            updatevalues={"locked": locked},
-        )
-        self._invalidate_cache_and_stream(txn, self.get_user_locked_status, (user_id,))
-        self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
-
-    def update_user_approval_status_txn(
-        self, txn: LoggingTransaction, user_id: str, approved: bool
-    ) -> None:
-        """Set the user's 'approved' flag to the given value.
-
-        The boolean is turned into an int because the column is a smallint.
-
-        Args:
-            txn: the current database transaction.
-            user_id: the user to update the flag for.
-            approved: the value to set the flag to.
-        """
-        self.db_pool.simple_update_one_txn(
-            txn=txn,
-            table="users",
-            keyvalues={"name": user_id},
-            updatevalues={"approved": approved},
-        )
-
-        # Invalidate the caches of methods that read the value of the 'approved' flag.
-        self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,))
-        self._invalidate_cache_and_stream(txn, self.is_user_approved, (user_id,))
-

 class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
    def __init__(
@@ -2941,25 +2975,6 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore):
            start_or_continue_validation_session_txn,
        )

-    async def update_user_approval_status(
-        self, user_id: UserID, approved: bool
-    ) -> None:
-        """Set the user's 'approved' flag to the given value.
-
-        The boolean will be turned into an int (in update_user_approval_status_txn)
-        because the column is a smallint.
-
-        Args:
-            user_id: the user to update the flag for.
-            approved: the value to set the flag to.
-        """
-        await self.db_pool.runInteraction(
-            "update_user_approval_status",
-            self.update_user_approval_status_txn,
-            user_id.to_string(),
-            approved,
-        )
-
    @wrap_as_background_process("delete_expired_login_tokens")
    async def _delete_expired_login_tokens(self) -> None:
        """Remove login tokens with expiry dates that have passed."""
@@ -51,11 +51,15 @@ from synapse.api.room_versions import RoomVersion, RoomVersions
 from synapse.config.homeserver import HomeServerConfig
 from synapse.events import EventBase
 from synapse.replication.tcp.streams.partial_state import UnPartialStatedRoomStream
-from synapse.storage._base import db_to_json, make_in_list_sql_clause
+from synapse.storage._base import (
+    db_to_json,
+    make_in_list_sql_clause,
+)
 from synapse.storage.database import (
    DatabasePool,
    LoggingDatabaseConnection,
    LoggingTransaction,
+    make_tuple_in_list_sql_clause,
 )
 from synapse.storage.databases.main.cache import CacheInvalidationWorkerStore
 from synapse.storage.types import Cursor
@@ -1127,6 +1131,109 @@ class RoomWorkerStore(CacheInvalidationWorkerStore):

        return local_media_ids

+    def _quarantine_local_media_txn(
+        self,
+        txn: LoggingTransaction,
+        hashes: Set[str],
+        media_ids: Set[str],
+        quarantined_by: Optional[str],
+    ) -> int:
+        """Quarantine and unquarantine local media items.
+
+        Args:
+            txn (cursor)
+            hashes: A set of sha256 hashes for any media that should be quarantined
+            media_ids: A set of media IDs for any media that should be quarantined
+            quarantined_by: The ID of the user who initiated the quarantine request
+                If it is `None` media will be removed from quarantine
+        Returns:
+            The total number of media items quarantined
+        """
+        total_media_quarantined = 0
+
+        # Effectively a legacy path, update any media that was explicitly named.
+        if media_ids:
+            sql_many_clause_sql, sql_many_clause_args = make_in_list_sql_clause(
+                txn.database_engine, "media_id", media_ids
+            )
+            sql = f"""
+                UPDATE local_media_repository
+                SET quarantined_by = ?
+                WHERE {sql_many_clause_sql}"""
+
+            if quarantined_by is not None:
+                sql += " AND safe_from_quarantine = FALSE"
+
+            txn.execute(sql, [quarantined_by] + sql_many_clause_args)
+            # Note that a rowcount of -1 can be used to indicate no rows were affected.
+            total_media_quarantined += txn.rowcount if txn.rowcount > 0 else 0
+
+        # Update any media that was identified via hash.
+        if hashes:
+            sql_many_clause_sql, sql_many_clause_args = make_in_list_sql_clause(
+                txn.database_engine, "sha256", hashes
+            )
+            sql = f"""
+                UPDATE local_media_repository
+                SET quarantined_by = ?
+                WHERE {sql_many_clause_sql}"""
+
+            if quarantined_by is not None:
+                sql += " AND safe_from_quarantine = FALSE"
+
+            txn.execute(sql, [quarantined_by] + sql_many_clause_args)
+            total_media_quarantined += txn.rowcount if txn.rowcount > 0 else 0
+
+        return total_media_quarantined
+
+    def _quarantine_remote_media_txn(
+        self,
+        txn: LoggingTransaction,
+        hashes: Set[str],
+        media: Set[Tuple[str, str]],
+        quarantined_by: Optional[str],
+    ) -> int:
+        """Quarantine and unquarantine remote items
+
+        Args:
+            txn (cursor)
+            hashes: A set of sha256 hashes for any media that should be quarantined
+            media_ids: A set of tuples (media_origin, media_id) for any media that should be quarantined
+            quarantined_by: The ID of the user who initiated the quarantine request
+                If it is `None` media will be removed from quarantine
+        Returns:
+            The total number of media items quarantined
+        """
+        total_media_quarantined = 0
+
+        if media:
+            sql_in_list_clause, sql_args = make_tuple_in_list_sql_clause(
+                txn.database_engine,
+                ("media_origin", "media_id"),
+                media,
+            )
+            sql = f"""
+                UPDATE remote_media_cache
+                SET quarantined_by = ?
+                WHERE {sql_in_list_clause}"""
+
+            txn.execute(sql, [quarantined_by] + sql_args)
+            total_media_quarantined += txn.rowcount if txn.rowcount > 0 else 0
+
+        total_media_quarantined = 0
+        if hashes:
+            sql_many_clause_sql, sql_many_clause_args = make_in_list_sql_clause(
+                txn.database_engine, "sha256", hashes
+            )
+            sql = f"""
+                UPDATE remote_media_cache
+                SET quarantined_by = ?
+                WHERE {sql_many_clause_sql}"""
+            txn.execute(sql, [quarantined_by] + sql_many_clause_args)
+            total_media_quarantined += txn.rowcount if txn.rowcount > 0 else 0
+
+        return total_media_quarantined
+
    def _quarantine_media_txn(
        self,
        txn: LoggingTransaction,
@@ -1146,40 +1253,49 @@ class RoomWorkerStore(CacheInvalidationWorkerStore):
        Returns:
            The total number of media items quarantined
        """
+        hashes = set()
+        media_ids = set()
+        remote_media = set()

-        # Update all the tables to set the quarantined_by flag
-        sql = """
-            UPDATE local_media_repository
-            SET quarantined_by = ?
-            WHERE media_id = ?
-        """
-
-        # set quarantine
-        if quarantined_by is not None:
-            sql += "AND safe_from_quarantine = FALSE"
-            txn.executemany(
-                sql, [(quarantined_by, media_id) for media_id in local_mxcs]
+        # First, determine the hashes of the media we want to delete.
+        # We also want the media_ids for any media that lacks a hash.
+        if local_mxcs:
+            hash_sql_many_clause_sql, hash_sql_many_clause_args = (
+                make_in_list_sql_clause(txn.database_engine, "media_id", local_mxcs)
            )
-        # remove from quarantine
-        else:
-            txn.executemany(
-                sql, [(quarantined_by, media_id) for media_id in local_mxcs]
+            hash_sql = f"SELECT sha256, media_id FROM local_media_repository WHERE {hash_sql_many_clause_sql}"
+            if quarantined_by is not None:
+                hash_sql += " AND safe_from_quarantine = FALSE"
+
+            txn.execute(hash_sql, hash_sql_many_clause_args)
+            for sha256, media_id in txn:
+                if sha256:
+                    hashes.add(sha256)
+                else:
+                    media_ids.add(media_id)
+
+        # Do the same for remote media
+        if remote_mxcs:
+            hash_sql_in_list_clause, hash_sql_args = make_tuple_in_list_sql_clause(
+                txn.database_engine,
+                ("media_origin", "media_id"),
+                remote_mxcs,
            )

-        # Note that a rowcount of -1 can be used to indicate no rows were affected.
-        total_media_quarantined = txn.rowcount if txn.rowcount > 0 else 0
+            hash_sql = f"SELECT sha256, media_origin, media_id FROM remote_media_cache WHERE {hash_sql_in_list_clause}"
+            txn.execute(hash_sql, hash_sql_args)
+            for sha256, media_origin, media_id in txn:
+                if sha256:
+                    hashes.add(sha256)
+                else:
+                    remote_media.add((media_origin, media_id))

-        txn.executemany(
-            """
-                UPDATE remote_media_cache
-                SET quarantined_by = ?
-                WHERE media_origin = ? AND media_id = ?
-            """,
-            [(quarantined_by, origin, media_id) for origin, media_id in remote_mxcs],
+        count = self._quarantine_local_media_txn(txn, hashes, media_ids, quarantined_by)
+        count += self._quarantine_remote_media_txn(
+            txn, hashes, remote_media, quarantined_by
        )
-        total_media_quarantined += txn.rowcount if txn.rowcount > 0 else 0

-        return total_media_quarantined
+        return count

    async def block_room(self, room_id: str, user_id: str) -> None:
        """Marks the room as blocked.
@@ -53,6 +53,7 @@ from synapse.storage.database import (
 )
 from synapse.storage.databases.main.cache import CacheInvalidationWorkerStore
 from synapse.storage.databases.main.events_worker import EventsWorkerStore
+from synapse.storage.databases.main.stream import _filter_results_by_stream
 from synapse.storage.engines import Sqlite3Engine
 from synapse.storage.roommember import (
    MemberSummary,
@@ -65,6 +66,7 @@ from synapse.types import (
    PersistedEventPosition,
    StateMap,
    StrCollection,
+    StreamToken,
    get_domain_from_id,
 )
 from synapse.util.caches.descriptors import _CacheContext, cached, cachedList
@@ -79,6 +81,7 @@ logger = logging.getLogger(__name__)

 _MEMBERSHIP_PROFILE_UPDATE_NAME = "room_membership_profile_update"
 _CURRENT_STATE_MEMBERSHIP_UPDATE_NAME = "current_state_events_membership"
+_POPULATE_PARTICIPANT_BG_UPDATE_BATCH_SIZE = 1000


@attr.s(frozen=True, slots=True, auto_attribs=True)
@@ -1388,7 +1391,9 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore):
                txn, self.get_forgotten_rooms_for_user, (user_id,)
            )
            self._invalidate_cache_and_stream(
-                txn, self.get_sliding_sync_rooms_for_user, (user_id,)
+                txn,
+                self.get_sliding_sync_rooms_for_user_from_membership_snapshots,
+                (user_id,),
            )

        await self.db_pool.runInteraction("forget_membership", f)
@@ -1420,25 +1425,30 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore):
        )

    @cached(iterable=True, max_entries=10000)
-    async def get_sliding_sync_rooms_for_user(
-        self,
-        user_id: str,
+    async def get_sliding_sync_rooms_for_user_from_membership_snapshots(
+        self, user_id: str
    ) -> Mapping[str, RoomsForUserSlidingSync]:
-        """Get all the rooms for a user to handle a sliding sync request.
+        """
+        Get all the rooms for a user to handle a sliding sync request from the
+        `sliding_sync_membership_snapshots` table. These will be current memberships and
+        need to be rewound to the token range.

        Ignores forgotten rooms and rooms that the user has left themselves.

+        Args:
+            user_id: The user ID to get the rooms for.
+
        Returns:
            Map from room ID to membership info
        """

-        def get_sliding_sync_rooms_for_user_txn(
+        def _txn(
            txn: LoggingTransaction,
        ) -> Dict[str, RoomsForUserSlidingSync]:
            # XXX: If you use any new columns that can change (like from
            # `sliding_sync_joined_rooms` or `forgotten`), make sure to bust the
-            # `get_sliding_sync_rooms_for_user` cache in the appropriate places (and add
-            # tests).
+            # `get_sliding_sync_rooms_for_user_from_membership_snapshots` cache in the
+            # appropriate places (and add tests).
            sql = """
                SELECT m.room_id, m.sender, m.membership, m.membership_event_id,
                    r.room_version,
@@ -1454,6 +1464,7 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore):
                    AND (m.membership != 'leave' OR m.user_id != m.sender)
            """
            txn.execute(sql, (user_id,))
+
            return {
                row[0]: RoomsForUserSlidingSync(
                    room_id=row[0],
@@ -1474,8 +1485,113 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore):
            }

        return await self.db_pool.runInteraction(
-            "get_sliding_sync_rooms_for_user",
-            get_sliding_sync_rooms_for_user_txn,
+            "get_sliding_sync_rooms_for_user_from_membership_snapshots",
+            _txn,
+        )
+
+    async def get_sliding_sync_self_leave_rooms_after_to_token(
+        self,
+        user_id: str,
+        to_token: StreamToken,
+    ) -> Dict[str, RoomsForUserSlidingSync]:
+        """
+        Get all the self-leave rooms for a user after the `to_token` (outside the token
+        range) that are potentially relevant[1] and needed to handle a sliding sync
+        request. The results are from the `sliding_sync_membership_snapshots` table and
+        will be current memberships and need to be rewound to the token range.
+
+        [1] If a leave happens after the token range, we may have still been joined (or
+        any non-self-leave which is relevant to sync) to the room before so we need to
+        include it in the list of potentially relevant rooms and apply
+        our rewind logic (outside of this function) to see if it's actually relevant.
+
+        This is basically a sister-function to
+        `get_sliding_sync_rooms_for_user_from_membership_snapshots`. We could
+        alternatively incorporate this logic into
+        `get_sliding_sync_rooms_for_user_from_membership_snapshots` but those results
+        are cached and the `to_token` isn't very cache friendly (people are constantly
+        requesting with new tokens) so we separate it out here.
+
+        Args:
+            user_id: The user ID to get the rooms for.
+            to_token: Any self-leave memberships after this position will be returned.
+
+        Returns:
+            Map from room ID to membership info
+        """
+        # TODO: Potential to check
+        # `self._membership_stream_cache.has_entity_changed(...)` as an early-return
+        # shortcut.
+
+        def _txn(
+            txn: LoggingTransaction,
+        ) -> Dict[str, RoomsForUserSlidingSync]:
+            sql = """
+                SELECT m.room_id, m.sender, m.membership, m.membership_event_id,
+                    r.room_version,
+                    m.event_instance_name, m.event_stream_ordering,
+                    m.has_known_state,
+                    m.room_type,
+                    m.is_encrypted
+                FROM sliding_sync_membership_snapshots AS m
+                INNER JOIN rooms AS r USING (room_id)
+                WHERE user_id = ?
+                    AND m.forgotten = 0
+                    AND m.membership = 'leave'
+                    AND m.user_id = m.sender
+                    AND (m.event_stream_ordering > ?)
+            """
+            # If a leave happens after the token range, we may have still been joined
+            # (or any non-self-leave which is relevant to sync) to the room before so we
+            # need to include it in the list of potentially relevant rooms and apply our
+            # rewind logic (outside of this function).
+            #
+            # To handle tokens with a non-empty instance_map we fetch more
+            # results than necessary and then filter down
+            min_to_token_position = to_token.room_key.stream
+            txn.execute(sql, (user_id, min_to_token_position))
+
+            # Map from room_id to membership info
+            room_membership_for_user_map: Dict[str, RoomsForUserSlidingSync] = {}
+            for row in txn:
+                room_for_user = RoomsForUserSlidingSync(
+                    room_id=row[0],
+                    sender=row[1],
+                    membership=row[2],
+                    event_id=row[3],
+                    room_version_id=row[4],
+                    event_pos=PersistedEventPosition(row[5], row[6]),
+                    has_known_state=bool(row[7]),
+                    room_type=row[8],
+                    is_encrypted=bool(row[9]),
+                )
+
+                # We filter out unknown room versions proactively. They shouldn't go
+                # down sync and their metadata may be in a broken state (causing
+                # errors).
+                if row[4] not in KNOWN_ROOM_VERSIONS:
+                    continue
+
+                # We only want to include the self-leave membership if it happened after
+                # the token range.
+                #
+                # Since the database pulls out more than necessary, we need to filter it
+                # down here.
+                if _filter_results_by_stream(
+                    lower_token=None,
+                    upper_token=to_token.room_key,
+                    instance_name=room_for_user.event_pos.instance_name,
+                    stream_ordering=room_for_user.event_pos.stream,
+                ):
+                    continue
+
+                room_membership_for_user_map[room_for_user.room_id] = room_for_user
+
+            return room_membership_for_user_map
+
+        return await self.db_pool.runInteraction(
+            "get_sliding_sync_self_leave_rooms_after_to_token",
+            _txn,
        )

    async def get_sliding_sync_room_for_user(
@@ -1606,6 +1722,62 @@ class RoomMemberWorkerStore(EventsWorkerStore, CacheInvalidationWorkerStore):
            from_ts,
        )

+    async def set_room_participation(self, user_id: str, room_id: str) -> None:
+        """
+        Record the provided user as participating in the given room
+
+        Args:
+            user_id: the user ID of the user
+            room_id: ID of the room to set the participant in
+        """
+
+        def _set_room_participation_txn(
+            txn: LoggingTransaction, user_id: str, room_id: str
+        ) -> None:
+            sql = """
+                UPDATE room_memberships
+                SET participant = true
+                WHERE event_id IN (
+                    SELECT event_id FROM local_current_membership
+                    WHERE user_id = ? AND room_id = ?
+                )
+                AND NOT participant
+            """
+            txn.execute(sql, (user_id, room_id))
+
+        await self.db_pool.runInteraction(
+            "_set_room_participation_txn", _set_room_participation_txn, user_id, room_id
+        )
+
+    async def get_room_participation(self, user_id: str, room_id: str) -> bool:
+        """
+        Check whether a user is listed as a participant in a room
+
+        Args:
+            user_id: user ID of the user
+            room_id: ID of the room to check in
+        """
+
+        def _get_room_participation_txn(
+            txn: LoggingTransaction, user_id: str, room_id: str
+        ) -> bool:
+            sql = """
+                SELECT participant
+                FROM local_current_membership AS l
+                INNER JOIN room_memberships AS r USING (event_id)
+                WHERE l.user_id = ?
+                AND l.room_id = ?
+            """
+            txn.execute(sql, (user_id, room_id))
+            res = txn.fetchone()
+            if res:
+                return res[0]
+            return False
+
+        return await self.db_pool.runInteraction(
+            "_get_room_participation_txn", _get_room_participation_txn, user_id, room_id
+        )
+

 class RoomMemberBackgroundUpdateStore(SQLBaseStore):
    def __init__(
@@ -1,7 +1,7 @@
 #
 # This file is licensed under the Affero General Public License (AGPL) version 3.
 #
-# Copyright (C) 2023 New Vector, Ltd
+# Copyright (C) 2023, 2025 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
@@ -61,6 +61,13 @@ class SlidingSyncStore(SQLBaseStore):
            columns=("required_state_id",),
        )

+        self.db_pool.updates.register_background_index_update(
+            update_name="sliding_sync_membership_snapshots_membership_event_id_idx",
+            index_name="sliding_sync_membership_snapshots_membership_event_id_idx",
+            table="sliding_sync_membership_snapshots",
+            columns=("membership_event_id",),
+        )
+
    async def get_latest_bump_stamp_for_room(
        self,
        room_id: str,
@@ -80,6 +80,7 @@ from synapse.storage.database import (
 )
 from synapse.storage.databases.main.events_worker import EventsWorkerStore
 from synapse.storage.engines import BaseDatabaseEngine, PostgresEngine, Sqlite3Engine
+from synapse.storage.roommember import RoomsForUserStateReset
 from synapse.storage.util.id_generators import MultiWriterIdGenerator
 from synapse.types import PersistedEventPosition, RoomStreamToken, StrCollection
 from synapse.util.caches.descriptors import cached, cachedList
@@ -453,6 +454,8 @@ def _filter_results_by_stream(
    stream_ordering falls between the two tokens (taking a None
    token to mean unbounded).

+    The token range is defined by > `lower_token` and <= `upper_token`.
+
    Used to filter results from fetching events in the DB against the given
    tokens. This is necessary to handle the case where the tokens include
    position maps, which we handle by fetching more than necessary from the DB
@@ -991,6 +994,10 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore):
            available in the `current_state_delta_stream` table. To actually check for a
            state reset, you need to check if a membership still exists in the room.
        """
+
+        assert from_key.topological is None
+        assert to_key.topological is None
+
        # Start by ruling out cases where a DB query is not necessary.
        if from_key == to_key:
            return []
@@ -1136,6 +1143,203 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore):
            if membership_change.room_id not in room_ids_to_exclude
        ]

+    @trace
+    async def get_sliding_sync_membership_changes(
+        self,
+        user_id: str,
+        from_key: RoomStreamToken,
+        to_key: RoomStreamToken,
+        excluded_room_ids: Optional[AbstractSet[str]] = None,
+    ) -> Dict[str, RoomsForUserStateReset]:
+        """
+        Fetch membership events that result in a meaningful membership change for a
+        given user.
+
+        A meaningful membership changes is one where the `membership` value actually
+        changes. This means memberships changes from `join` to `join` (like a display
+        name change) will be filtered out since they result in no meaningful change.
+
+        Note: This function only works with "live" tokens with `stream_ordering` only.
+
+        We're looking for membership changes in the token range (> `from_key` and <=
+        `to_key`).
+
+        Args:
+            user_id: The user ID to fetch membership events for.
+            from_key: The point in the stream to sync from (fetching events > this point).
+            to_key: The token to fetch rooms up to (fetching events <= this point).
+            excluded_room_ids: Optional list of room IDs to exclude from the results.
+
+        Returns:
+            All meaningful membership changes to the current state in the token range.
+            Events are sorted by `stream_ordering` ascending.
+
+            `event_id`/`sender` can be `None` when the server leaves a room (meaning
+            everyone locally left) or a state reset which removed the person from the
+            room. We can't tell the difference between the two cases with what's
+            available in the `current_state_delta_stream` table. To actually check for a
+            state reset, you need to check if a membership still exists in the room.
+        """
+
+        assert from_key.topological is None
+        assert to_key.topological is None
+
+        # Start by ruling out cases where a DB query is not necessary.
+        if from_key == to_key:
+            return {}
+
+        if from_key:
+            has_changed = self._membership_stream_cache.has_entity_changed(
+                user_id, int(from_key.stream)
+            )
+            if not has_changed:
+                return {}
+
+        room_ids_to_exclude: AbstractSet[str] = set()
+        if excluded_room_ids is not None:
+            room_ids_to_exclude = excluded_room_ids
+
+        def f(txn: LoggingTransaction) -> Dict[str, RoomsForUserStateReset]:
+            # To handle tokens with a non-empty instance_map we fetch more
+            # results than necessary and then filter down
+            min_from_id = from_key.stream
+            max_to_id = to_key.get_max_stream_pos()
+
+            # This query looks at membership changes in
+            # `sliding_sync_membership_snapshots` which will not include users
+            # that were state reset out of rooms; so we need to look for that
+            # case in `current_state_delta_stream`.
+            sql = """
+                SELECT
+                    room_id,
+                    membership_event_id,
+                    event_instance_name,
+                    event_stream_ordering,
+                    membership,
+                    sender,
+                    prev_membership,
+                    room_version
+                FROM
+                (
+                    SELECT
+                        s.room_id,
+                        s.membership_event_id,
+                        s.event_instance_name,
+                        s.event_stream_ordering,
+                        s.membership,
+                        s.sender,
+                        m_prev.membership AS prev_membership
+                    FROM sliding_sync_membership_snapshots as s
+                        LEFT JOIN event_edges AS e ON e.event_id = s.membership_event_id
+                        LEFT JOIN room_memberships AS m_prev ON m_prev.event_id = e.prev_event_id
+                    WHERE s.user_id = ?
+
+                    UNION ALL
+
+                    SELECT
+                        s.room_id,
+                        e.event_id,
+                        s.instance_name,
+                        s.stream_id,
+                        m.membership,
+                        e.sender,
+                        m_prev.membership AS prev_membership
+                    FROM current_state_delta_stream AS s
+                        LEFT JOIN events AS e ON e.event_id = s.event_id
+                        LEFT JOIN room_memberships AS m ON m.event_id = s.event_id
+                        LEFT JOIN room_memberships AS m_prev ON m_prev.event_id = s.prev_event_id
+                    WHERE
+                        s.type = ?
+                        AND s.state_key = ?
+                ) AS c
+                INNER JOIN rooms USING (room_id)
+                WHERE event_stream_ordering > ? AND event_stream_ordering <= ?
+                ORDER BY event_stream_ordering ASC
+            """
+
+            txn.execute(
+                sql,
+                (user_id, EventTypes.Member, user_id, min_from_id, max_to_id),
+            )
+
+            membership_changes: Dict[str, RoomsForUserStateReset] = {}
+            for (
+                room_id,
+                membership_event_id,
+                event_instance_name,
+                event_stream_ordering,
+                membership,
+                sender,
+                prev_membership,
+                room_version_id,
+            ) in txn:
+                assert room_id is not None
+                assert event_stream_ordering is not None
+
+                if room_id in room_ids_to_exclude:
+                    continue
+
+                if _filter_results_by_stream(
+                    from_key,
+                    to_key,
+                    event_instance_name,
+                    event_stream_ordering,
+                ):
+                    # When the server leaves a room, it will insert new rows into the
+                    # `current_state_delta_stream` table with `event_id = null` for all
+                    # current state. This means we might already have a row for the
+                    # leave event and then another for the same leave where the
+                    # `event_id=null` but the `prev_event_id` is pointing back at the
+                    # earlier leave event. We don't want to report the leave, if we
+                    # already have a leave event.
+                    if (
+                        membership_event_id is None
+                        and prev_membership == Membership.LEAVE
+                    ):
+                        continue
+
+                    if membership_event_id is None and room_id in membership_changes:
+                        # SUSPICIOUS: if we join a room and get state reset out of it
+                        # in the same queried window,
+                        # won't this ignore the 'state reset out of it' part?
+                        continue
+
+                    # When `s.event_id = null`, we won't be able to get respective
+                    # `room_membership` but can assume the user has left the room
+                    # because this only happens when the server leaves a room
+                    # (meaning everyone locally left) or a state reset which removed
+                    # the person from the room.
+                    membership = (
+                        membership if membership is not None else Membership.LEAVE
+                    )
+
+                    if membership == prev_membership:
+                        # If `membership` and `prev_membership` are the same then this
+                        # is not a meaningful change so we can skip it.
+                        # An example of this happening is when the user changes their display name.
+                        continue
+
+                    membership_change = RoomsForUserStateReset(
+                        room_id=room_id,
+                        sender=sender,
+                        membership=membership,
+                        event_id=membership_event_id,
+                        event_pos=PersistedEventPosition(
+                            event_instance_name, event_stream_ordering
+                        ),
+                        room_version_id=room_version_id,
+                    )
+
+                    membership_changes[room_id] = membership_change
+
+            return membership_changes
+
+        membership_changes = await self.db_pool.runInteraction(
+            "get_sliding_sync_membership_changes", f
+        )
+
+        return membership_changes
+
    @cancellable
    async def get_membership_changes_for_user(
        self,
@@ -1037,11 +1037,11 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore):
                }
        """

+        join_args: Tuple[str, ...] = (user_id,)
+
        if self.hs.config.userdirectory.user_directory_search_all_users:
-            join_args = (user_id,)
            where_clause = "user_id != ?"
        else:
-            join_args = (user_id,)
            where_clause = """
                (
                    EXISTS (select 1 from users_in_public_rooms WHERE user_id = t.user_id)
@@ -1055,6 +1055,14 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore):
        if not show_locked_users:
            where_clause += " AND (u.locked IS NULL OR u.locked = FALSE)"

+        # Adjust the JOIN type based on the exclude_remote_users flag (the users
+        # table only contains local users so an inner join is a good way to
+        # to exclude remote users)
+        if self.hs.config.userdirectory.user_directory_exclude_remote_users:
+            join_type = "JOIN"
+        else:
+            join_type = "LEFT JOIN"
+
        # We allow manipulating the ranking algorithm by injecting statements
        # based on config options.
        additional_ordering_statements = []
@@ -1086,7 +1094,7 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore):
                SELECT d.user_id AS user_id, display_name, avatar_url
                FROM matching_users as t
                INNER JOIN user_directory AS d USING (user_id)
-                LEFT JOIN users AS u ON t.user_id = u.name
+                %(join_type)s users AS u ON t.user_id = u.name
                WHERE
                    %(where_clause)s
                ORDER BY
@@ -1115,6 +1123,7 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore):
            """ % {
                "where_clause": where_clause,
                "order_case_statements": " ".join(additional_ordering_statements),
+                "join_type": join_type,
            }
            args = (
                (full_query,)
@@ -1142,7 +1151,7 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore):
                SELECT d.user_id AS user_id, display_name, avatar_url
                FROM user_directory_search as t
                INNER JOIN user_directory AS d USING (user_id)
-                LEFT JOIN users AS u ON t.user_id = u.name
+                %(join_type)s users AS u ON t.user_id = u.name
                WHERE
                    %(where_clause)s
                    AND value MATCH ?
@@ -1155,6 +1164,7 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore):
            """ % {
                "where_clause": where_clause,
                "order_statements": " ".join(additional_ordering_statements),
+                "join_type": join_type,
            }
            args = join_args + (search_query,) + ordering_arguments + (limit + 1,)
        else:
@@ -20,7 +20,15 @@
 #

 import logging
-from typing import TYPE_CHECKING, Dict, List, Mapping, Optional, Tuple, Union
+from typing import (
+    TYPE_CHECKING,
+    Dict,
+    List,
+    Mapping,
+    Optional,
+    Tuple,
+    Union,
+)

 from synapse.logging.opentracing import tag_args, trace
 from synapse.storage._base import SQLBaseStore
@@ -321,18 +321,42 @@ class StateDeletionDataStore:
    async def mark_state_groups_as_pending_deletion(
        self, state_groups: Collection[int]
    ) -> None:
-        """Mark the given state groups as pending deletion"""
+        """Mark the given state groups as pending deletion.
+
+        If any of the state groups are already pending deletion, then those records are
+        left as is.
+        """
+
+        await self.db_pool.runInteraction(
+            "mark_state_groups_as_pending_deletion",
+            self._mark_state_groups_as_pending_deletion_txn,
+            state_groups,
+        )
+
+    def _mark_state_groups_as_pending_deletion_txn(
+        self,
+        txn: LoggingTransaction,
+        state_groups: Collection[int],
+    ) -> None:
+        sql = """
+        INSERT INTO state_groups_pending_deletion (state_group, insertion_ts)
+        VALUES %s
+        ON CONFLICT (state_group)
+        DO NOTHING
+        """

        now = self._clock.time_msec()
-
-        await self.db_pool.simple_upsert_many(
-            table="state_groups_pending_deletion",
-            key_names=("state_group",),
-            key_values=[(state_group,) for state_group in state_groups],
-            value_names=("insertion_ts",),
-            value_values=[(now,) for _ in state_groups],
-            desc="mark_state_groups_as_pending_deletion",
-        )
+        rows = [
+            (
+                state_group,
+                now,
+            )
+            for state_group in state_groups
+        ]
+        if isinstance(txn.database_engine, PostgresEngine):
+            txn.execute_values(sql % ("?",), rows, fetch=False)
+        else:
+            txn.execute_batch(sql % ("(?, ?)",), rows)

    async def mark_state_groups_as_used(self, state_groups: Collection[int]) -> None:
        """Mark the given state groups as now being referenced"""
@@ -48,6 +48,7 @@ from synapse.storage.database import (
    LoggingTransaction,
 )
 from synapse.storage.databases.state.bg_updates import StateBackgroundUpdateStore
+from synapse.storage.engines import PostgresEngine
 from synapse.storage.types import Cursor
 from synapse.storage.util.sequence import build_sequence_generator
 from synapse.types import MutableStateMap, StateKey, StateMap
@@ -914,6 +915,12 @@ class StateGroupDataStore(StateBackgroundUpdateStore, SQLBaseStore):
    ) -> None:
        # Delete all edges that reference a state group linked to room_id
        logger.info("[purge] removing %s from state_group_edges", room_id)
+
+        if isinstance(self.database_engine, PostgresEngine):
+            # Disable statement timeouts for this transaction; purging rooms can
+            # take a while!
+            txn.execute("SET LOCAL statement_timeout = 0")
+
        txn.execute(
            """
            DELETE FROM state_group_edges AS sge WHERE sge.state_group IN (
@@ -19,7 +19,7 @@
 #
 #

-SCHEMA_VERSION = 89  # remember to update the list below when updating
+SCHEMA_VERSION = 92  # remember to update the list below when updating
 """Represents the expectations made by the codebase about the database schema

 This should be incremented whenever the codebase changes its requirements on the
@@ -158,6 +158,16 @@ Changes in SCHEMA_VERSION = 88

 Changes in SCHEMA_VERSION = 89
    - Add `state_groups_pending_deletion` and `state_groups_persisting` tables.
+
+Changes in SCHEMA_VERSION = 90
+    - Add a column `participant` to `room_memberships` table
+    - Add background update to delete unreferenced state groups.
+
+Changes in SCHEMA_VERSION = 91
+    - Add a `sha256` column to the `local_media_repository` and `remote_media_cache` tables.
+
+Changes in SCHEMA_VERSION = 92
+    - Cleaned up a trigger that was added in #18260 and then reverted.
 """


@@ -0,0 +1,15 @@
+--
+-- This file is licensed under the Affero General Public License (AGPL) version 3.
+--
+-- Copyright (C) 2025 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>.
+
+INSERT INTO background_updates (ordering, update_name, progress_json) VALUES
+    (8901, 'sliding_sync_membership_snapshots_membership_event_id_idx', '{}');
--- a/Show More
+++ b/Show More
				`@@ -0,0 +1 @@`
				Explicitly enable PyPy builds in `cibuildwheel`s config to avoid it being disabled on a future upgrade to `cibuildwheel` v3.
				`@@ -0,0 +1 @@`
				`Update the PR review template to remove an erroneous line break from the final bullet point.`
				`@@ -0,0 +1 @@`
				Update `pyproject.toml` to remove deprecated keys.