runtime: use newm for profileloop

This replaces the externalthreadhandler-based implementation of
profileloop with one that uses newm to start a new thread. This is a
step toward eliminating externalthreadhandler.

For #45530.

Change-Id: Id8e5540423fe2d2004024b649afec6998f77b092
Reviewed-on: https://go-review.googlesource.com/c/go/+/309633
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

5 files changed, 9 insertions, 32 deletions

diff --git a/src/runtime/os_windows.go b/src/runtime/os_windows.go
index 77ad075be1..5bff4b6646 100644
--- a/src/runtime/os_windows.go
+++ b/src/runtime/os_windows.go
@@ -1198,9 +1198,6 @@ func ctrlHandler(_type uint32) uintptr {
 	return 0
 }
 
-// in sys_windows_386.s and sys_windows_amd64.s
-func profileloop()
-
 // called from zcallback_windows_*.s to sys_windows_*.s
 func callbackasm1()
 
@@ -1233,13 +1230,18 @@ func gFromSP(mp *m, sp uintptr) *g {
 	return nil
 }
 
-func profileloop1(param uintptr) uint32 {
+func profileLoop() {
 	stdcall2(_SetThreadPriority, currentThread, _THREAD_PRIORITY_HIGHEST)
 
 	for {
 		stdcall2(_WaitForSingleObject, profiletimer, _INFINITE)
 		first := (*m)(atomic.Loadp(unsafe.Pointer(&allm)))
 		for mp := first; mp != nil; mp = mp.alllink {
+			if mp == getg().m {
+				// Don't profile ourselves.
+				continue
+			}
+
 			lock(&mp.threadLock)
 			// Do not profile threads blocked on Notes,
 			// this includes idle worker threads,
@@ -1251,8 +1253,8 @@ func profileloop1(param uintptr) uint32 {
 			// Acquire our own handle to the thread.
 			var thread uintptr
 			if stdcall7(_DuplicateHandle, currentProcess, mp.thread, currentProcess, uintptr(unsafe.Pointer(&thread)), 0, 0, _DUPLICATE_SAME_ACCESS) == 0 {
-				print("runtime.profileloop1: duplicatehandle failed; errno=", getlasterror(), "\n")
-				throw("runtime.profileloop1: duplicatehandle failed")
+				print("runtime: duplicatehandle failed; errno=", getlasterror(), "\n")
+				throw("duplicatehandle failed")
 			}
 			unlock(&mp.threadLock)
 
@@ -1280,9 +1282,7 @@ func setProcessCPUProfiler(hz int32) {
 	if profiletimer == 0 {
 		timer := stdcall3(_CreateWaitableTimerA, 0, 0, 0)
 		atomic.Storeuintptr(&profiletimer, timer)
-		thread := stdcall6(_CreateThread, 0, 0, funcPC(profileloop), 0, 0, 0)
-		stdcall2(_SetThreadPriority, thread, _THREAD_PRIORITY_HIGHEST)
-		stdcall1(_CloseHandle, thread)
+		newm(profileLoop, nil, -1)
 	}
 }
 
diff --git a/src/runtime/sys_windows_386.s b/src/runtime/sys_windows_386.s
index b1c27c40d9..c8839b9976 100644
--- a/src/runtime/sys_windows_386.s
+++ b/src/runtime/sys_windows_386.s
@@ -156,15 +156,6 @@ TEXT runtime·lastcontinuetramp<ABIInternal>(SB),NOSPLIT,$0-0
 	MOVL	$runtime·lastcontinuehandler(SB), AX
 	JMP	sigtramp<>(SB)
 
-// Called by OS using stdcall ABI: uint32 profileloop(void*).
-TEXT runtime·profileloop<ABIInternal>(SB),NOSPLIT,$0
-	PUSHL	$runtime·profileloop1(SB)
-	NOP	SP	// tell vet SP changed - stop checking offsets
-	CALL	runtime·externalthreadhandler(SB)
-	MOVL	4(SP), CX
-	ADDL	$12, SP
-	JMP	CX
-
 TEXT runtime·externalthreadhandler<ABIInternal>(SB),NOSPLIT|TOPFRAME,$0
 	PUSHL	BP
 	MOVL	SP, BP
diff --git a/src/runtime/sys_windows_amd64.s b/src/runtime/sys_windows_amd64.s
index e9428c36db..df1462b877 100644
--- a/src/runtime/sys_windows_amd64.s
+++ b/src/runtime/sys_windows_amd64.s
@@ -202,12 +202,6 @@ TEXT runtime·lastcontinuetramp<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-0
 	MOVQ	$runtime·lastcontinuehandler(SB), AX
 	JMP	sigtramp<>(SB)
 
-TEXT runtime·profileloop<ABIInternal>(SB),NOSPLIT|NOFRAME,$8
-	MOVQ	$runtime·profileloop1(SB), CX
-	MOVQ	CX, 0(SP)
-	CALL	runtime·externalthreadhandler<ABIInternal>(SB)
-	RET
-
 TEXT runtime·externalthreadhandler<ABIInternal>(SB),NOSPLIT|NOFRAME|TOPFRAME,$0
 	PUSHQ	BP
 	MOVQ	SP, BP
diff --git a/src/runtime/sys_windows_arm.s b/src/runtime/sys_windows_arm.s
index 304bba22e0..8914a2688f 100644
--- a/src/runtime/sys_windows_arm.s
+++ b/src/runtime/sys_windows_arm.s
@@ -233,10 +233,6 @@ TEXT runtime·lastcontinuetramp<ABIInternal>(SB),NOSPLIT|NOFRAME,$0
 	MOVW	$runtime·lastcontinuehandler(SB), R1
 	B	sigtramp<>(SB)
 
-TEXT runtime·profileloop<ABIInternal>(SB),NOSPLIT|NOFRAME,$0
-	MOVW	$runtime·profileloop1(SB), R1
-	B	runtime·externalthreadhandler<ABIInternal>(SB)
-
 // int32 externalthreadhandler(uint32 arg, int (*func)(uint32))
 // stack layout:
 //   +----------------+
diff --git a/src/runtime/sys_windows_arm64.s b/src/runtime/sys_windows_arm64.s
index 4512fbc8b0..70f628c27d 100644
--- a/src/runtime/sys_windows_arm64.s
+++ b/src/runtime/sys_windows_arm64.s
@@ -299,10 +299,6 @@ TEXT runtime·lastcontinuetramp(SB),NOSPLIT|NOFRAME,$0
 	MOVD	$runtime·lastcontinuehandler<ABIInternal>(SB), R1
 	B	sigtramp<>(SB)
 
-TEXT runtime·profileloop<ABIInternal>(SB),NOSPLIT|NOFRAME,$0
-	MOVD	$runtime·profileloop1(SB), R1
-	B	runtime·externalthreadhandler<ABIInternal>(SB)
-
 // externalthreadhander called with R0 = uint32 arg, R1 = Go function f.
 // Need to call f(arg), which returns a uint32, and return it in R0.
 TEXT runtime·externalthreadhandler<ABIInternal>(SB),NOSPLIT|TOPFRAME,$96-0