feat: Windows agent inventory

agent/go.mod

@@ -3,13 +3,13 @@ module gitea.hostxtra.co.uk/mrhid6/vantage/agent
 go 1.26
 
 require (
+	golang.org/x/sys v0.20.0
 	google.golang.org/grpc v1.64.0
 	gopkg.in/yaml.v3 v3.0.1
 )
 
 require (
 	golang.org/x/net v0.25.0 // indirect
-	golang.org/x/sys v0.20.0 // indirect
 	golang.org/x/text v0.15.0 // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20240521202816-d264139d666e // indirect
 	google.golang.org/protobuf v1.34.1 // indirect

agent/internal/inventory/collect_linux.go

@@ -144,11 +144,14 @@ func partitions() []pb.PartitionReport {
 		if syscall.Statfs(fields[1], &st) != nil {
 			continue
 		}
-		total := st.Blocks * uint64(st.Bsize)
-		free := st.Bavail * uint64(st.Bsize)
+		bsize := uint64(st.Bsize)
+		total := st.Blocks * bsize
+		// Bfree, not Bavail: the difference is the root-reserved 5% on ext4,
+		// which is not used space. df counts it the same way.
+		used := (st.Blocks - st.Bfree) * bsize
 		out = append(out, pb.PartitionReport{
 			Device: fields[0], Mountpoint: fields[1], Fstype: fields[2],
-			TotalBytes: total, UsedBytes: total - free,
+			TotalBytes: total, UsedBytes: used,
 		})
 	}
 	return out

agent/internal/inventory/collect_other.go

@@ -1,12 +1,13 @@
-//go:build !linux
+//go:build !linux && !windows
 
-// Inventory collection is Linux-only. This no-op stands in everywhere else.
+// Inventory collection has Linux and Windows implementations. This no-op stands
+// in everywhere else.
 //
 // The build constraint above is load-bearing: "_other" is not a GOOS suffix, so
 // without it this file compiles on Linux too and collides with collect_linux.go.
 package inventory
 
-import "github.com/mrhid6/vantage/agent/internal/grpc/pb"
+import "gitea.hostxtra.co.uk/mrhid6/vantage/agent/internal/grpc/pb"
 
 
 func collect(r *pb.InventoryReport, includeStatic bool) {}

agent/internal/inventory/collect_windows.go

@@ -0,0 +1,183 @@
+package inventory
+
+import (
+	"fmt"
+	"runtime"
+	"time"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+	"golang.org/x/sys/windows/registry"
+
+	"gitea.hostxtra.co.uk/mrhid6/vantage/agent/internal/grpc/pb"
+)
+
+var (
+	kernel32           = windows.NewLazySystemDLL("kernel32.dll")
+	procGetSystemTimes = kernel32.NewProc("GetSystemTimes")
+	// x/sys/windows exposes neither of these two, so they are bound by hand.
+	procGlobalMemoryStatusEx = kernel32.NewProc("GlobalMemoryStatusEx")
+)
+
+func collect(r *pb.InventoryReport, includeStatic bool) {
+	r.CPU.UsagePct = cpuUsage()
+	// Windows has no load average. Left at zero; the UI already treats it as
+	// optional because it is omitempty on the wire.
+
+	m := memoryStatus()
+	if m.TotalPhys > m.AvailPhys {
+		r.Memory.UsedBytes = m.TotalPhys - m.AvailPhys
+	}
+	// TotalPageFile is the commit limit — physical memory plus the pagefile —
+	// so the pagefile alone is the difference.
+	swapTotal := sub(m.TotalPageFile, m.TotalPhys)
+	swapUsed := sub(sub(m.TotalPageFile, m.AvailPageFile), sub(m.TotalPhys, m.AvailPhys))
+	if swapUsed > swapTotal {
+		swapUsed = swapTotal
+	}
+	r.SwapUsed = swapUsed
+
+	if includeStatic {
+		r.Memory.TotalBytes = m.TotalPhys
+		r.SwapTotal = swapTotal
+		r.CPU.Model, r.CPU.Cores = cpuStatic()
+		r.Kernel = kernel()
+		r.Partitions = partitions()
+	}
+}
+
+func sub(a, b uint64) uint64 {
+	if a > b {
+		return a - b
+	}
+	return 0
+}
+
+type memoryStatusEx struct {
+	Length               uint32
+	MemoryLoad           uint32
+	TotalPhys            uint64
+	AvailPhys            uint64
+	TotalPageFile        uint64
+	AvailPageFile        uint64
+	TotalVirtual         uint64
+	AvailVirtual         uint64
+	AvailExtendedVirtual uint64
+}
+
+func memoryStatus() memoryStatusEx {
+	var m memoryStatusEx
+	m.Length = uint32(unsafe.Sizeof(m))
+	r, _, _ := procGlobalMemoryStatusEx.Call(uintptr(unsafe.Pointer(&m)))
+	if r == 0 {
+		return memoryStatusEx{}
+	}
+	return m
+}
+
+func systemTimes() (idle, total uint64, ok bool) {
+	var idleFT, kernelFT, userFT windows.Filetime
+	r, _, _ := procGetSystemTimes.Call(
+		uintptr(unsafe.Pointer(&idleFT)),
+		uintptr(unsafe.Pointer(&kernelFT)),
+		uintptr(unsafe.Pointer(&userFT)),
+	)
+	if r == 0 {
+		return 0, 0, false
+	}
+	ft := func(f windows.Filetime) uint64 {
+		return uint64(f.HighDateTime)<<32 | uint64(f.LowDateTime)
+	}
+	// Kernel time already includes idle time, so kernel+user is the whole.
+	return ft(idleFT), ft(kernelFT) + ft(userFT), true
+}
+
+func cpuUsage() float64 {
+	i1, t1, ok := systemTimes()
+	if !ok {
+		return 0
+	}
+	time.Sleep(100 * time.Millisecond)
+	i2, t2, ok := systemTimes()
+	if !ok || t2 <= t1 {
+		return 0
+	}
+	return (1 - float64(i2-i1)/float64(t2-t1)) * 100
+}
+
+func cpuStatic() (model string, cores int) {
+	cores = runtime.NumCPU()
+	k, err := registry.OpenKey(registry.LOCAL_MACHINE,
+		`HARDWARE\DESCRIPTION\System\CentralProcessor\0`, registry.QUERY_VALUE)
+	if err != nil {
+		return
+	}
+	defer k.Close()
+	if s, _, err := k.GetStringValue("ProcessorNameString"); err == nil {
+		model = s
+	}
+	return
+}
+
+func kernel() string {
+	v := windows.RtlGetVersion()
+	return fmt.Sprintf("%d.%d.%d", v.MajorVersion, v.MinorVersion, v.BuildNumber)
+}
+
+func partitions() []pb.PartitionReport {
+	buf := make([]uint16, 256)
+	n, err := windows.GetLogicalDriveStrings(uint32(len(buf)), &buf[0])
+	if err != nil || n == 0 {
+		return nil
+	}
+
+	var out []pb.PartitionReport
+	for _, root := range splitNullStrings(buf[:n]) {
+		rootPtr, err := windows.UTF16PtrFromString(root)
+		if err != nil {
+			continue
+		}
+		// Fixed disks only: network shares can hang, and removable drives
+		// would appear and vanish between snapshots.
+		if windows.GetDriveType(rootPtr) != windows.DRIVE_FIXED {
+			continue
+		}
+
+		var free, total, totalFree uint64
+		if err := windows.GetDiskFreeSpaceEx(rootPtr, &free, &total, &totalFree); err != nil {
+			continue
+		}
+
+		fsBuf := make([]uint16, 32)
+		var fstype string
+		if err := windows.GetVolumeInformation(rootPtr, nil, 0, nil, nil, nil, &fsBuf[0], uint32(len(fsBuf))); err == nil {
+			fstype = windows.UTF16ToString(fsBuf)
+		}
+
+		out = append(out, pb.PartitionReport{
+			Device:     root,
+			Mountpoint: root,
+			Fstype:     fstype,
+			TotalBytes: total,
+			UsedBytes:  total - totalFree,
+		})
+	}
+	return out
+}
+
+// splitNullStrings splits the NUL-separated, double-NUL-terminated block that
+// GetLogicalDriveStrings writes.
+func splitNullStrings(b []uint16) []string {
+	var out []string
+	start := 0
+	for i, c := range b {
+		if c != 0 {
+			continue
+		}
+		if i > start {
+			out = append(out, windows.UTF16ToString(b[start:i]))
+		}
+		start = i + 1
+	}
+	return out
+}

agent/internal/sync/sync.go

@@ -412,13 +412,55 @@ func handleUpdateAgentWindows(cmd *pb.ServerCommand) {
 	}
 
 	logPath := filepath.Join(os.TempDir(), "vantage-agent-msi.log")
-	log.Printf("launching msiexec for upgrade to v%s (cmd=%s)", u.Version, cmd.CommandId)
 
-	up := exec.Command("cmd", "/c", "start", "", "/wait", "msiexec", "/i", msiPath, "/qn", "/norestart", "/l*v", logPath)
-	if err := up.Start(); err != nil {
+	// The MSI stops the vantage-agent service as part of the upgrade. Anything
+	// descended from this process is killed with it, so msiexec must not be a
+	// child: run it from a scheduled task, which is parented to the Task
+	// Scheduler service instead.
+	if err := launchDetachedUpdate(msiPath, logPath, cmd.CommandId); err != nil {
 		log.Printf("failed to launch msiexec (cmd=%s): %v", cmd.CommandId, err)
 		return
 	}
+	log.Printf("scheduled msiexec for upgrade to v%s (cmd=%s)", u.Version, cmd.CommandId)
+}
+
+const updateTaskName = "VantageAgentUpdate"
+
+func launchDetachedUpdate(msiPath, logPath, commandID string) error {
+	scriptPath := filepath.Join(os.TempDir(), "vantage-agent-update.cmd")
+	script := fmt.Sprintf("@echo off\r\n"+
+		"timeout /t 5 /nobreak >nul\r\n"+
+		"msiexec /i \"%s\" /qn /norestart /l*v \"%s\"\r\n"+
+		"schtasks /delete /tn %s /f >nul 2>&1\r\n"+
+		"del /f /q \"%s\" >nul 2>&1\r\n"+
+		"(goto) 2>nul & del /f /q \"%%~f0\"\r\n",
+		msiPath, logPath, updateTaskName, msiPath)
+	if err := os.WriteFile(scriptPath, []byte(script), 0o600); err != nil {
+		return fmt.Errorf("write update script: %w", err)
+	}
+
+	// Stale task from a previous attempt would make /create fail even with /f
+	// if it is still running, so tear it down first and ignore the result.
+	exec.Command("schtasks", "/end", "/tn", updateTaskName).Run()
+	exec.Command("schtasks", "/delete", "/tn", updateTaskName, "/f").Run()
+
+	create := exec.Command("schtasks", "/create",
+		"/tn", updateTaskName,
+		"/tr", `"`+scriptPath+`"`,
+		"/sc", "once",
+		// Already in the past: the task never fires on its own, only via /run.
+		"/st", "00:00",
+		"/ru", "SYSTEM",
+		"/rl", "HIGHEST",
+		"/f")
+	if out, err := create.CombinedOutput(); err != nil {
+		return fmt.Errorf("schtasks create: %v: %s", err, strings.TrimSpace(string(out)))
+	}
+
+	if out, err := exec.Command("schtasks", "/run", "/tn", updateTaskName).CombinedOutput(); err != nil {
+		return fmt.Errorf("schtasks run: %v: %s", err, strings.TrimSpace(string(out)))
+	}
+	return nil
 }
 
 func downloadFile(url, dest string) error {

go.work

@@ -2,6 +2,7 @@ go 1.26
 
 use (
 	./admin
+	./agent
 	./server
 	./shared
 	./sitesvc