ci: harden fetch_ctk with manifest/archive verification and optional redist-base-url

.github/actions/fetch_ctk/action.yml

@@ -19,35 +19,76 @@ inputs:
     description: "where the CTK components will be installed to, relative to $PWD"
     required: false
     default: "./cuda_toolkit"
+  redist-base-url:
+    description: >
+      Base URL of a redistrib server (manifest and component archives).
+      Defaults to the public NVIDIA developer server.
+      Override to install from an alternate redistrib server.
+    required: false
+    default: "https://developer.download.nvidia.com/compute/cuda/redist"
 
 runs:
   using: composite
   steps:
+    - name: Install Unix dependencies
+      if: ${{ runner.os == 'Linux' }}
+      uses: ./.github/actions/install_unix_deps
+      with:
+        dependencies: "zstd curl xz-utils"
+        dependent_exes: "zstd curl xz"
+
+    - name: Install Windows dependencies
+      if: ${{ runner.os == 'Windows' }}
+      uses: ./.github/actions/install_windows_deps
+      with:
+        dependent_exes: "curl unzip tar sha256sum cygpath"
+
     - name: Set up CTK cache variable
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
-        # Pre-process the component list to ensure hash uniqueness
         # Use the runtime workspace mount so this also works inside container jobs.
         CTK_REDIST_TOOL="${GITHUB_WORKSPACE}/ci/tools/fetch_ctk_redistrib.py"
-        CTK_CACHE_COMPONENTS=${{ inputs.cuda-components }}
-        CTK_JSON_URL="https://developer.download.nvidia.com/compute/cuda/redist/redistrib_${{ inputs.cuda-version }}.json"
+        CTK_BASE_URL="${{ inputs.redist-base-url }}"
+        # A trailing slash would defeat the public-server comparison below and double up in URLs.
+        CTK_BASE_URL="${CTK_BASE_URL%/}"
+        # RUNNER_TEMP persists across steps, so "Get CUDA components" reuses this file.
+        CTK_JSON_FILE="$RUNNER_TEMP/redistrib-${{ inputs.cuda-version }}.json"
+        curl -LSs --fail --retry 3 --retry-delay 2 "${CTK_BASE_URL}/redistrib_${{ inputs.cuda-version }}.json" -o "$CTK_JSON_FILE"
+
+        # Pin the metadata JSON: the per-archive sha256 checks below are only
+        # trustworthy if the JSON that carries those hashes is itself trusted.
+        # The pins describe the public server; alternate servers only warn.
+        if [[ "$CTK_BASE_URL" == "https://developer.download.nvidia.com/compute/cuda/redist" ]]; then
+          METADATA_SOURCE=public
+        else
+          METADATA_SOURCE=custom
+        fi
+        python "$CTK_REDIST_TOOL" verify-metadata \
+          --cuda-version "${{ inputs.cuda-version }}" \
+          --metadata-path "$CTK_JSON_FILE" \
+          --metadata-source "$METADATA_SOURCE"
+
+        # Pre-process the component list to ensure hash uniqueness
         CTK_CACHE_COMPONENTS="$(python "$CTK_REDIST_TOOL" filter-components \
           --host-platform "${{ inputs.host-platform }}" \
           --cuda-version "${{ inputs.cuda-version }}" \
-          --components "$CTK_CACHE_COMPONENTS" \
-          --metadata-url "$CTK_JSON_URL")"
+          --components "${{ inputs.cuda-components }}" \
+          --metadata-path "$CTK_JSON_FILE")"
 
+        # Alternate redist servers (non-default) publish different manifests for the
+        # same CUDA version; keying on the full manifest hash keeps them from restoring
+        # each other's toolkit (a truncated hash could be collided on purpose).
+        # Hash via stdin: with a backslash in the path (Windows RUNNER_TEMP), sha256sum
+        # prefixes its output line with "\", which would end up in the cache key.
+        MANIFEST_SHA256="$(sha256sum < "$CTK_JSON_FILE" | awk '{print $1}')"
         HASH=$(echo -n "${CTK_CACHE_COMPONENTS}" | sha256sum | awk '{print $1}')
-        echo "CTK_CACHE_KEY=mini-ctk-${{ inputs.cuda-version }}-${{ inputs.host-platform }}-$HASH" >> $GITHUB_ENV
-        echo "CTK_CACHE_FILENAME=mini-ctk-${{ inputs.cuda-version }}-${{ inputs.host-platform }}-$HASH.tar.gz" >> $GITHUB_ENV
-        echo "CTK_CACHE_COMPONENTS=${CTK_CACHE_COMPONENTS}" >> $GITHUB_ENV
-
-    - name: Install dependencies
-      uses: ./.github/actions/install_unix_deps
-      continue-on-error: false
-      with:
-        dependencies: "zstd curl xz-utils"
-        dependent_exes: "zstd curl xz"
+        {
+          echo "CTK_CACHE_KEY=mini-ctk-${{ inputs.cuda-version }}-${{ inputs.host-platform }}-$HASH-m$MANIFEST_SHA256"
+          echo "CTK_CACHE_FILENAME=mini-ctk-${{ inputs.cuda-version }}-${{ inputs.host-platform }}-$HASH-m$MANIFEST_SHA256.tar.gz"
+          echo "CTK_CACHE_COMPONENTS=${CTK_CACHE_COMPONENTS}"
+          echo "CTK_BASE_URL=${CTK_BASE_URL}"
+          echo "CTK_JSON_FILE=${CTK_JSON_FILE}"
+        } >> "$GITHUB_ENV"
 
     - name: Download CTK cache
       id: ctk-get-cache
@@ -68,13 +109,10 @@ runs:
         rm -rf $CACHE_TMP_DIR
         mkdir $CACHE_TMP_DIR
 
-        # The binary archives (redist) are guaranteed to be updated as part of the release posting.
         # Use the runtime workspace mount so this also works inside container jobs.
         CTK_REDIST_TOOL="${GITHUB_WORKSPACE}/ci/tools/fetch_ctk_redistrib.py"
-        CTK_BASE_URL="https://developer.download.nvidia.com/compute/cuda/redist/"
-        CTK_JSON_URL="$CTK_BASE_URL/redistrib_${{ inputs.cuda-version }}.json"
-        CTK_JSON_FILE="$CACHE_TMP_DIR/redistrib.json"
-        curl -LSs "$CTK_JSON_URL" -o "$CTK_JSON_FILE"
+        # CTK_BASE_URL and the verified manifest at CTK_JSON_FILE come from
+        # "Set up CTK cache variable".
         if [[ "${{ inputs.host-platform }}" == linux* ]]; then
           function extract() {
             tar -xvf $1 -C $CACHE_TMP_DIR --strip-components=1
@@ -102,6 +140,13 @@ runs:
           CTK_COMPONENT_URL="${CTK_BASE_URL}/${CTK_COMPONENT_REL_PATH}"
           CTK_COMPONENT_COMPONENT_FILENAME="$(basename $CTK_COMPONENT_REL_PATH)"
           download $CTK_COMPONENT_URL $CTK_COMPONENT_COMPONENT_FILENAME
+          # Verify the downloaded archive against the sha256 in the (pinned)
+          # metadata before extracting; fail closed on any mismatch.
+          CTK_COMPONENT_SHA256="$(python "$CTK_REDIST_TOOL" component-sha256 \
+            --host-platform "${{ inputs.host-platform }}" \
+            --component "$CTK_COMPONENT" \
+            --metadata-path "$CTK_JSON_FILE")"
+          echo "${CTK_COMPONENT_SHA256}  ${CTK_COMPONENT_COMPONENT_FILENAME}" | sha256sum --check --strict -
           extract $CTK_COMPONENT_COMPONENT_FILENAME
           rm $CTK_COMPONENT_COMPONENT_FILENAME
         }

.github/actions/install_windows_deps/action.yml

@@ -0,0 +1,92 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+name: Install dependencies on Windows
+
+description: Install Git Bash tools from a verified portable Git bundle when required executables are missing.
+
+inputs:
+  dependent_exes:
+    description: Space-separated executable names provided by Git for Windows
+    required: true
+
+runs:
+  using: composite
+  steps:
+    - name: Install dependencies
+      shell: powershell
+      env:
+        DEPENDENT_EXES: ${{ inputs.dependent_exes }}
+      run: |
+        $ErrorActionPreference = 'Stop'
+        if ($env:DEPENDENT_EXES -notmatch '^\s*[a-zA-Z0-9_.-]+(\s+[a-zA-Z0-9_.-]+)*\s*$') {
+          throw 'Expected a space-separated list of executable names.'
+        }
+        # Fail the check if the list did not reach Bash (a WSL bash would not receive it).
+        # Keep double quotes out of $check: Windows PowerShell 5.1 strips them when
+        # passing arguments to native commands. The names are validated above.
+        $check = 'missing=0; set -- $DEPENDENT_EXES; [ $# -gt 0 ] || missing=1; for tool in $DEPENDENT_EXES; do command -v $tool || missing=1; done; exit $missing'
+        # Skip the WSL launcher in System32/WindowsApps: it is not a Git for Windows bash.
+        $bash = Get-Command bash -All -ErrorAction SilentlyContinue |
+          Where-Object { $_.Source -notmatch '\\(System32|SysWOW64|WindowsApps)\\' } |
+          Select-Object -First 1
+        if ($bash) {
+          & $bash.Source --noprofile --norc -c $check
+          if ($LASTEXITCODE -eq 0) {
+            # Later `shell: bash` steps resolve `bash` through PATH. If that is not the
+            # Git Bash validated above (e.g. the WSL launcher comes first), put it first.
+            $pathBash = Get-Command bash -ErrorAction SilentlyContinue
+            if ($pathBash -and ($pathBash.Source -ne $bash.Source)) {
+              $bashDir = Split-Path -Parent $bash.Source
+              $env:PATH = "$bashDir;$env:PATH"
+              $bashDir | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
+              Write-Host "Prepended $bashDir to PATH so steps use Git Bash instead of $($pathBash.Source)."
+            }
+            Write-Host 'All dependencies are found. Do nothing.'
+            exit 0
+          }
+        }
+
+        $architecture = [Runtime.InteropServices.RuntimeInformation]::OSArchitecture.ToString()
+        switch ($architecture) {
+          'Arm64' {
+            $package = 'arm64'
+            $expectedHash = '3bf26b94d9399b16a890776e468334f501742861576cbcdea2d9134643c374bd'
+          }
+          'X64' {
+            $package = '64-bit'
+            $expectedHash = 'ab00566336b5472120f9a52d34f2e79c5406535792acb0548001ffd0bd090e5d'
+          }
+          default { throw "Unsupported Windows architecture: $architecture" }
+        }
+        $root = Join-Path $env:RUNNER_TOOL_CACHE "WindowsDeps\Git-2.55.0.3-$package"
+        $complete = "$root.complete"
+        if (-not (Test-Path -LiteralPath $complete)) {
+          $archive = Join-Path $env:RUNNER_TEMP "PortableGit-2.55.0.3-$package.7z.exe"
+          try {
+            Invoke-WebRequest -UseBasicParsing -Uri "https://github.com/git-for-windows/git/releases/download/v2.55.0.windows.3/PortableGit-2.55.0.3-$package.7z.exe" -OutFile $archive
+            $actualHash = (Get-FileHash -LiteralPath $archive -Algorithm SHA256).Hash.ToLowerInvariant()
+            if ($actualHash -ne $expectedHash) {
+              throw "Portable Git checksum mismatch: expected $expectedHash, got $actualHash."
+            }
+            if (Test-Path -LiteralPath $root) {
+              Remove-Item -LiteralPath $root -Recurse -Force
+            }
+            $process = Start-Process -FilePath $archive -ArgumentList @('-y', ('-o"{0}"' -f $root)) -Wait -PassThru
+            if ($process.ExitCode -ne 0) {
+              throw "Portable Git extraction failed with exit code $($process.ExitCode)."
+            }
+          }
+          finally {
+            if (Test-Path -LiteralPath $archive) {
+              Remove-Item -LiteralPath $archive -Force
+            }
+          }
+        }
+        $env:PATH = "$root\bin;$root\usr\bin;$env:PATH"
+        & "$root\bin\bash.exe" --noprofile --norc -c $check
+        if ($LASTEXITCODE -ne 0) {
+          throw 'Required executables are still missing after installing portable Git.'
+        }
+        New-Item -ItemType File -Force -Path $complete | Out-Null
+        @("$root\bin", "$root\usr\bin") | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append