[release-branch.go1.12] runtime: monitor for suspend/resume to kick timeouts

Starting in Windows 8, the wait functions don't take into account
suspend time, even though the monotonic counters do. This results in
timer buckets stalling on resume. Therefore, this commit makes it so
that on resume, we return from the wait functions and recalculate the
amount of time left to wait.

This is a cherry pick of CL 191957 and its cleanup, CL 198417.

Updates #31528
Fixes #36376

Change-Id: I0db02cc72188cb620954e87a0180e0a3c83f4a56
Reviewed-on: https://go-review.googlesource.com/c/go/+/193607
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-on: https://go-review.googlesource.com/c/go/+/213197

2 files changed, 80 insertions, 13 deletions

diff --git a/src/runtime/os_windows.go b/src/runtime/os_windows.go
index d3e84fe3dc..a3292c427b 100644
--- a/src/runtime/os_windows.go
+++ b/src/runtime/os_windows.go
@@ -49,6 +49,7 @@ const (
 //go:cgo_import_dynamic runtime._VirtualFree VirtualFree%3 "kernel32.dll"
 //go:cgo_import_dynamic runtime._VirtualQuery VirtualQuery%3 "kernel32.dll"
 //go:cgo_import_dynamic runtime._WaitForSingleObject WaitForSingleObject%2 "kernel32.dll"
+//go:cgo_import_dynamic runtime._WaitForMultipleObjects WaitForMultipleObjects%4 "kernel32.dll"
 //go:cgo_import_dynamic runtime._WriteConsoleW WriteConsoleW%5 "kernel32.dll"
 //go:cgo_import_dynamic runtime._WriteFile WriteFile%5 "kernel32.dll"
 
@@ -96,6 +97,7 @@ var (
 	_VirtualFree,
 	_VirtualQuery,
 	_WaitForSingleObject,
+	_WaitForMultipleObjects,
 	_WriteConsoleW,
 	_WriteFile,
 	_ stdFunction
@@ -138,7 +140,8 @@ func tstart_stdcall(newm *m) uint32
 func ctrlhandler(_type uint32) uint32
 
 type mOS struct {
-	waitsema uintptr // semaphore for parking on locks
+	waitsema   uintptr // semaphore for parking on locks
+	resumesema uintptr // semaphore to indicate suspend/resume
 }
 
 //go:linkname os_sigpipe os.sigpipe
@@ -257,6 +260,40 @@ func loadOptionalSyscalls() {
 	}
 }
 
+func monitorSuspendResume() {
+	const _DEVICE_NOTIFY_CALLBACK = 2
+	type _DEVICE_NOTIFY_SUBSCRIBE_PARAMETERS struct {
+		callback uintptr
+		context  uintptr
+	}
+
+	powrprof := windowsLoadSystemLib([]byte("powrprof.dll\000"))
+	if powrprof == 0 {
+		return // Running on Windows 7, where we don't need it anyway.
+	}
+	powerRegisterSuspendResumeNotification := windowsFindfunc(powrprof, []byte("PowerRegisterSuspendResumeNotification\000"))
+	if powerRegisterSuspendResumeNotification == nil {
+		return // Running on Windows 7, where we don't need it anyway.
+	}
+	var fn interface{} = func(context uintptr, changeType uint32, setting uintptr) uintptr {
+		for mp := (*m)(atomic.Loadp(unsafe.Pointer(&allm))); mp != nil; mp = mp.alllink {
+			if mp.resumesema != 0 {
+				stdcall1(_SetEvent, mp.resumesema)
+			}
+		}
+		return 0
+	}
+	params := _DEVICE_NOTIFY_SUBSCRIBE_PARAMETERS{
+		callback: compileCallback(*efaceOf(&fn), true),
+	}
+	handle := uintptr(0)
+	if stdcall3(powerRegisterSuspendResumeNotification, _DEVICE_NOTIFY_CALLBACK,
+		uintptr(unsafe.Pointer(&params)),
+		uintptr(unsafe.Pointer(&handle))) != 0 {
+		throw("PowerRegisterSuspendResumeNotification failure")
+	}
+}
+
 //go:nosplit
 func getLoadLibrary() uintptr {
 	return uintptr(unsafe.Pointer(_LoadLibraryW))
@@ -487,6 +524,10 @@ func goenvs() {
 	}
 
 	stdcall1(_FreeEnvironmentStringsW, uintptr(strings))
+
+	// We call this all the way here, late in init, so that malloc works
+	// for the callback function this generates.
+	monitorSuspendResume()
 }
 
 // exiting is set to non-zero when the process is exiting.
@@ -605,19 +646,32 @@ func semasleep(ns int64) int32 {
 		_WAIT_FAILED    = 0xFFFFFFFF
 	)
 
-	// store ms in ns to save stack space
+	var result uintptr
 	if ns < 0 {
-		ns = _INFINITE
+		result = stdcall2(_WaitForSingleObject, getg().m.waitsema, uintptr(_INFINITE))
 	} else {
-		ns = int64(timediv(ns, 1000000, nil))
-		if ns == 0 {
-			ns = 1
+		start := nanotime()
+		elapsed := int64(0)
+		for {
+			ms := int64(timediv(ns-elapsed, 1000000, nil))
+			if ms == 0 {
+				ms = 1
+			}
+			result = stdcall4(_WaitForMultipleObjects, 2,
+				uintptr(unsafe.Pointer(&[2]uintptr{getg().m.waitsema, getg().m.resumesema})),
+				0, uintptr(ms))
+			if result != _WAIT_OBJECT_0+1 {
+				// Not a suspend/resume event
+				break
+			}
+			elapsed = nanotime() - start
+			if elapsed >= ns {
+				return -1
+			}
 		}
 	}
-
-	result := stdcall2(_WaitForSingleObject, getg().m.waitsema, uintptr(ns))
 	switch result {
-	case _WAIT_OBJECT_0: //signaled
+	case _WAIT_OBJECT_0: // Signaled
 		return 0
 
 	case _WAIT_TIMEOUT:
@@ -666,6 +720,15 @@ func semacreate(mp *m) {
 			throw("runtime.semacreate")
 		})
 	}
+	mp.resumesema = stdcall4(_CreateEventA, 0, 0, 0, 0)
+	if mp.resumesema == 0 {
+		systemstack(func() {
+			print("runtime: createevent failed; errno=", getlasterror(), "\n")
+			throw("runtime.semacreate")
+		})
+		stdcall1(_CloseHandle, mp.waitsema)
+		mp.waitsema = 0
+	}
 }
 
 // May run with m.p==nil, so write barriers are not allowed. This
diff --git a/src/runtime/syscall_windows.go b/src/runtime/syscall_windows.go
index 36ad7511af..920468286b 100644
--- a/src/runtime/syscall_windows.go
+++ b/src/runtime/syscall_windows.go
@@ -74,16 +74,18 @@ func compileCallback(fn eface, cleanstack bool) (code uintptr) {
 		argsize += uintptrSize
 	}
 
-	lock(&cbs.lock)
-	defer unlock(&cbs.lock)
+	lock(&cbs.lock) // We don't unlock this in a defer because this is used from the system stack.
 
 	n := cbs.n
 	for i := 0; i < n; i++ {
 		if cbs.ctxt[i].gobody == fn.data && cbs.ctxt[i].isCleanstack() == cleanstack {
-			return callbackasmAddr(i)
+			r := callbackasmAddr(i)
+			unlock(&cbs.lock)
+			return r
 		}
 	}
 	if n >= cb_max {
+		unlock(&cbs.lock)
 		throw("too many callback functions")
 	}
 
@@ -99,7 +101,9 @@ func compileCallback(fn eface, cleanstack bool) (code uintptr) {
 	cbs.ctxt[n] = c
 	cbs.n++
 
-	return callbackasmAddr(n)
+	r := callbackasmAddr(n)
+	unlock(&cbs.lock)
+	return r
 }
 
 const _LOAD_LIBRARY_SEARCH_SYSTEM32 = 0x00000800