Integration tests for windows service support (#3733)

* Create integration tests for windows service support

Signed-off-by: Guilherme Carvalho <[email protected]>

Author: guilhermocc

pkg/common/entrypoint/entrypoint_windows.go

@@ -5,8 +5,13 @@ package entrypoint
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"os"
+	"strings"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
 
 	"golang.org/x/sys/windows/svc"
 )
@@ -20,7 +25,10 @@ type systemCall struct {
 }
 
 func (s *systemCall) IsWindowsService() (bool, error) {
-	return svc.IsWindowsService()
+	// We are using a custom function because the svc.IsWindowsService() one still has an open issue in which it states
+	// that it is not working properly in Windows containers: https://github.com/golang/go/issues/56335. Soon as we have
+	// a fix for that, we can use the original function.
+	return isWindowsService()
 }
 
 func (s *systemCall) Run(name string, handler svc.Handler) error {
@@ -51,11 +59,11 @@ func (e *EntryPoint) Main() int {
 	// Determining if SPIRE is running as a Windows service is done
 	// with a best-effort approach. If there is an error, just fallback
 	// to the behavior of not running as a Windows service.
-	isWindowsService, err := e.sc.IsWindowsService()
+	isWindowsSvc, err := e.sc.IsWindowsService()
 	if err != nil {
 		fmt.Fprintf(os.Stderr, "Could not determine if running as a Windows service: %v", err)
 	}
-	if isWindowsService {
+	if isWindowsSvc {
 		errChan := make(chan error)
 		go func() {
 			// Since the service runs in its own process, the service name is ignored.
@@ -71,3 +79,38 @@ func (e *EntryPoint) Main() int {
 
 	return e.runCmdFn(context.Background(), os.Args[1:])
 }
+
+// isWindowsService is a copy of the svc.IsWindowsService() function, but without the parentProcess.SessionID == 0 check
+// that is causing the issue in Windows containers, this logic is exactly the same from .NET runtime (>= 6.0.10).
+func isWindowsService() (bool, error) {
+	// The below technique looks a bit hairy, but it's actually
+	// exactly what the .NET runtime (>= 6.0.10) does for the similarly named function:
+	// https://github.com/dotnet/runtime/blob/36bf84fc4a89209f4fdbc1fc201e81afd8be49b0/src/libraries/Microsoft.Extensions.Hosting.WindowsServices/src/WindowsServiceHelpers.cs#L20-L33
+	// Specifically, it looks up whether the parent process is called "services".
+
+	var currentProcess windows.PROCESS_BASIC_INFORMATION
+	infoSize := uint32(unsafe.Sizeof(currentProcess))
+	err := windows.NtQueryInformationProcess(windows.CurrentProcess(), windows.ProcessBasicInformation, unsafe.Pointer(&currentProcess), infoSize, &infoSize)
+	if err != nil {
+		return false, err
+	}
+	var parentProcess *windows.SYSTEM_PROCESS_INFORMATION
+	for infoSize = uint32((unsafe.Sizeof(*parentProcess) + unsafe.Sizeof(uintptr(0))) * 1024); ; {
+		parentProcess = (*windows.SYSTEM_PROCESS_INFORMATION)(unsafe.Pointer(&make([]byte, infoSize)[0]))
+		err = windows.NtQuerySystemInformation(windows.SystemProcessInformation, unsafe.Pointer(parentProcess), infoSize, &infoSize)
+		if err == nil {
+			break
+		} else if !errors.Is(err, windows.STATUS_INFO_LENGTH_MISMATCH) {
+			return false, err
+		}
+	}
+	for ; ; parentProcess = (*windows.SYSTEM_PROCESS_INFORMATION)(unsafe.Pointer(uintptr(unsafe.Pointer(parentProcess)) + uintptr(parentProcess.NextEntryOffset))) {
+		if parentProcess.UniqueProcessID == currentProcess.InheritedFromUniqueProcessId {
+			return strings.EqualFold("services.exe", parentProcess.ImageName.String()), nil
+		}
+		if parentProcess.NextEntryOffset == 0 {
+			break
+		}
+	}
+	return false, nil
+}

test/integration/README.md

@@ -89,3 +89,4 @@ The following environment variables are available to the teardown script:
 * [Self Test](suites/self-test/README.md)
 * [SPIRE Server CLI](suites/spire-server-cli/README.md)
 * [Upgrade](suites/upgrade/README.md)
+* [Windows Service](suites-windows/windows-service/README.md)

test/integration/suites-windows/windows-service/00-setup.sh

@@ -0,0 +1,3 @@
+#!/bin/bash
+
+"${ROOTDIR}/setup/x509pop/setup.sh" conf/server conf/agent

test/integration/suites-windows/windows-service/01-start-server-service

@@ -0,0 +1,8 @@
+#!/bin/bash
+source ./common
+
+docker-up spire-server
+
+create-service spire-server C:/spire/bin/spire-server.exe
+start-service spire-server run -config C:/spire/conf/server/server.conf
+assert-service-status spire-server RUNNING

test/integration/suites-windows/windows-service/02-bootstrap-agent

@@ -0,0 +1,5 @@
+#!/bin/bash
+
+log-debug "bootstrapping agent..."
+docker-compose exec -T spire-server \
+    c:/spire/bin/spire-server bundle show > conf/agent/bootstrap.crt || fail-now "failed to bootstrap agent"

test/integration/suites-windows/windows-service/03-start-agent-service

@@ -0,0 +1,8 @@
+#!/bin/bash
+source ./common
+
+docker-up spire-agent
+
+create-service spire-agent C:/spire/bin/spire-agent.exe
+start-service spire-agent run -config C:/spire/conf/agent/agent.conf
+assert-service-status spire-agent RUNNING

test/integration/suites-windows/windows-service/04-create-registration-entries

@@ -0,0 +1,12 @@
+#!/bin/bash
+source ./common
+
+log-debug "creating regular registration entry..."
+docker-compose exec -T spire-server \
+    c:/spire/bin/spire-server entry create \
+    -parentID "spiffe://domain.test/spire/agent/x509pop/$(fingerprint conf/agent/agent.crt.pem)" \
+    -spiffeID "spiffe://domain.test/workload" \
+    -selector "windows:user_name:User Manager\ContainerUser" \
+    -ttl 0
+
+assert-synced-entry "spiffe://domain.test/workload"

test/integration/suites-windows/windows-service/05-test-fetch-svid

@@ -0,0 +1,9 @@
+#!/bin/bash
+
+log-debug "test fetch x509 SVID..."
+docker-compose exec -T -u ContainerUser spire-agent  \
+    c:/spire/bin/spire-agent api fetch x509 || fail-now "failed to fetch x509"
+
+log-debug "test fetch JWT SVID..."
+docker-compose exec -T -u ContainerUser spire-agent \
+    c:/spire/bin/spire-agent api fetch jwt -audience mydb || fail-now "failed to fetch JWT"

test/integration/suites-windows/windows-service/06-test-graceful-shutdown

@@ -0,0 +1,10 @@
+#!/bin/bash
+source ./common
+
+stop-service spire-agent
+assert-service-status spire-agent STOPPED
+assert-graceful-shutdown agent
+
+stop-service spire-server
+assert-service-status spire-server STOPPED
+assert-graceful-shutdown server

test/integration/suites-windows/windows-service/07-test-service-failing-to-start

@@ -0,0 +1,8 @@
+#!/bin/bash
+source ./common
+
+start-service spire-server run -config invalid-config-path
+assert-service-status spire-server STOPPED
+
+start-service spire-agent run -config invalid-config-path
+assert-service-status spire-agent STOPPED

test/integration/suites-windows/windows-service/README.md

@@ -0,0 +1,19 @@
+# SPIRE Server CLI Suite
+
+## Description
+
+This suite validates that we can run both spire agent and spire server natively on Windows OS, asserting that spire components
+can run as a [windows service application](https://learn.microsoft.com/en-us/dotnet/framework/windows-services/introduction-to-windows-service-applications#service-applications-vs-other-visual-studio-applications),
+and perform [service state transitions](https://learn.microsoft.com/en-us/windows/win32/services/service-status-transitions).
+
+The suite steps are structured as follows:
+
+1. Spire server and agent are installed as Windows services.
+2. Spire server and agent services starts, their respective status is asserted as **_RUNNING_**, and the node attestation
+is performed with x509pop.
+3. Workload registration entries are created.
+4. The feature of fetching SVIDs (x509 and JWT) is asserted with the running spire agent service.
+5. Spire server and agent services are stopped, their respective status is asserted as **_STOPPED_**, and graceful
+shutdown is verified via application logs.
+6. Spire server and agent services are started again, but this time with an invalid config; their respective status is
+asserted as **_STOPPED_**.

test/integration/suites-windows/windows-service/common

@@ -0,0 +1,66 @@
+#!/bin/bash
+
+assert-synced-entry() {
+  # Check at most 30 times (with one second in between) that the agent has
+  # successfully synced down the workload entry.
+  MAXCHECKS=30
+  CHECKINTERVAL=1
+  for ((i=1;i<=MAXCHECKS;i++)); do
+      log-info "checking for synced entry ($i of $MAXCHECKS max)..."
+      if grep -wq "$1" conf/agent/logs.txt; then
+        return 0
+      fi
+      sleep "${CHECKINTERVAL}"
+  done
+
+  fail-now "timed out waiting for agent to sync down entry"
+}
+
+assert-service-status() {
+  MAXCHECKS=10
+  CHECKINTERVAL=1
+  for ((i=1;i<=MAXCHECKS;i++)); do
+    log-info "checking for $1 service $2 ($i of $MAXCHECKS max)..."
+    scCommand=$([ "$2" == "STOPPED" ] && echo "query" || echo "interrogate")
+    if docker-compose exec -T -u ContainerAdministrator "$1" sc "$scCommand" "$1" | grep -wq "$2"; then
+      log-info "$1 is in $2 state"
+      return 0
+    fi
+    sleep "${CHECKINTERVAL}"
+  done
+
+  fail-now "$1 service failed to reach $2 state"
+}
+
+assert-graceful-shutdown() {
+  MAXCHECKS=10
+  CHECKINTERVAL=1
+  for ((i=1;i<=MAXCHECKS;i++)); do
+      log-info "checking for graceful shutdown ($i of $MAXCHECKS max)..."
+      if grep -wq "stopped gracefully" conf/"$1"/logs.txt; then
+        log-info "$1 stopped gracefully"
+        return 0
+      fi
+      sleep "${CHECKINTERVAL}"
+  done
+
+  fail-now "timed out waiting for $1 graceful shutdown"
+}
+
+create-service() {
+  log-info "creating $1 service..."
+  docker-compose exec -T -u ContainerAdministrator "$1" \
+      sc create "$1" binPath="$2" ||  grep "STOPPED" fail-now "failed to create $1 service"
+}
+
+stop-service() {
+  log-info "stopping $1 service..."
+  docker-compose exec -T -u ContainerAdministrator "$1" \
+      sc stop "$1" || fail-now "failed to stop $1 service"
+}
+
+start-service(){
+  log-info "starting $1 service..."
+  docker-compose exec -T -u ContainerAdministrator "$1" \
+      sc start "$@" | grep -wq  "START_PENDING" || fail-now "failed to start $2 service"
+}

test/integration/suites-windows/windows-service/conf/agent/agent.conf

@@ -0,0 +1,25 @@
+agent {
+    data_dir = "c:/spire/data/agent"
+    log_level = "DEBUG"
+    server_address = "spire-server"
+    log_file ="c:/spire/conf/agent/logs.txt"
+    server_port = "8081"
+    trust_bundle_path = "c:/spire/conf/agent/bootstrap.crt"
+    trust_domain = "domain.test"
+}
+
+plugins {
+    NodeAttestor "x509pop" {
+        plugin_data {
+            private_key_path = "c:/spire/conf/agent/agent.key.pem"
+            certificate_path = "c:/spire/conf/agent/agent.crt.pem"
+        }
+    }
+    KeyManager "disk" {
+        plugin_data {
+            directory = "c:/spire/data/agent"
+        }
+    }
+    WorkloadAttestor "windows" {
+    }
+}

test/integration/suites-windows/windows-service/conf/server/server.conf

@@ -0,0 +1,25 @@
+server {
+	bind_address = "0.0.0.0"
+	bind_port = "8081"
+	trust_domain = "domain.test"
+	log_file ="c:/spire/conf/server/logs.txt"
+	data_dir = "c:/spire/data/server"
+	log_level = "DEBUG"
+}
+
+plugins {
+	DataStore "sql" {
+		plugin_data {
+			database_type = "sqlite3"
+			connection_string = "c:/spire/data/server/datastore.sqlite3"
+		}
+	}
+	NodeAttestor "x509pop" {
+		plugin_data {
+			ca_bundle_path = "c:/spire/conf/server/agent-cacert.pem"
+		}
+	}
+	KeyManager "memory" {
+		plugin_data = {}
+	}
+}

test/integration/suites-windows/windows-service/docker-compose.yaml

@@ -0,0 +1,24 @@
+version: '3'
+
+services:
+  spire-server:
+    image: spire-server-windows:latest-local
+    hostname: spire-server
+    volumes:
+      - ./conf/server:c:/spire/conf/server
+    user: ContainerAdministrator
+    entrypoint:
+      - cmd
+    command:
+      - cmd /c ping -t localhost > NUL
+  spire-agent:
+    image: spire-agent-windows:latest-local
+    hostname: spire-agent
+    depends_on: ["spire-server"]
+    volumes:
+      - ./conf/agent:c:/spire/conf/agent
+    user: ContainerAdministrator
+    entrypoint:
+      - cmd
+    command:
+      - cmd /c ping -t localhost > NUL

test/integration/suites-windows/windows-service/teardown

@@ -0,0 +1,6 @@
+#!/bin/bash
+
+if [ -z "$SUCCESS" ]; then
+    docker-compose logs
+fi
+docker-down