runtime: return the error code as a return value in asmstdcall

This shaves off 8 bytes from the syscall_syscalln stack frame, which is
significant as that call path is almost over the nosplit limit.

Also, it follows the cgocall convention of returning the error code as
a return value, making it easier to reason about.

Cq-Include-Trybots: luci.golang.try:gotip-windows-amd64-longtest,gotip-windows-arm64,gotip-windows-386
Change-Id: I62acdb7c0d4cf9cb928bf3974d3300dd752f6c29
Reviewed-on: https://go-review.googlesource.com/c/go/+/751861
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Carlos Amedee <carlos@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

src/internal/runtime/syscall/windows/asm_windows_386.s

@@ -5,10 +5,14 @@
 #include "go_asm.h"
 #include "textflag.h"
 
-TEXT ·StdCall<ABIInternal>(SB),NOSPLIT,$0
-	JMP	·asmstdcall(SB)
+TEXT ·StdCall<ABIInternal>(SB),NOSPLIT,$4-8
+	MOVL	fn+0(FP), AX
+	MOVL	AX, 0(SP)
+	CALL	·asmstdcall(SB)
+	MOVL	AX, ret+4(FP)
+	RET
 
-TEXT ·asmstdcall(SB),NOSPLIT,$0
+TEXT ·asmstdcall(SB),NOSPLIT,$0-4
 	MOVL	fn+0(FP), BX
 	MOVL	SP, BP	// save stack pointer
 
@@ -43,6 +47,5 @@ docall:
 
 	// GetLastError().
 	MOVL	0x34(FS), AX
-	MOVL	AX, StdCallInfo_Err(BX)
 
 	RET

src/internal/runtime/syscall/windows/asm_windows_amd64.s

@@ -79,6 +79,5 @@ _0args:
 	// GetLastError().
 	MOVQ	0x30(GS), DI
 	MOVL	0x68(DI), AX
-	MOVQ	AX, StdCallInfo_Err(CX)
 
 	RET

src/internal/runtime/syscall/windows/asm_windows_arm64.s

@@ -83,7 +83,6 @@ _0args:
 
 	// GetLastError
 	MOVD	TEB_error(R18_PLATFORM), R0
-	MOVD	R0, StdCallInfo_Err(R19)
 
 	// Restore callee-saved registers.
 	LDP	16(RSP), (R19, R20)

src/internal/runtime/syscall/windows/syscall_windows.go

@@ -24,15 +24,20 @@ type StdCallInfo struct {
 	Args uintptr // parameters
 	R1   uintptr // return values
 	R2   uintptr
-	Err  uintptr // error number
 }
 
 // StdCall calls a function using Windows' stdcall convention.
+// The calling thread's last-error code value is cleared before calling the function,
+// and stored in the return value.
 //
 //go:noescape
-func StdCall(fn *StdCallInfo)
+func StdCall(fn *StdCallInfo) uint32
 
 // asmstdcall is the function pointer for [AsmStdCallAddr].
+// The calling thread's last-error code value is cleared before calling the function,
+// and returned in the C ABI return register (not via Go stack convention).
+// This function is not called directly from Go; it is either jumped to from
+// [StdCall] or called from C via [AsmStdCallAddr].
 func asmstdcall(fn *StdCallInfo)
 
 // AsmStdCallAddr is the address of a function that accepts a pointer

src/runtime/syscall_windows.go

@@ -428,10 +428,10 @@ func syscall_syscalln(fn, n uintptr, args ...uintptr) (r1, r2, err uintptr) {
 	if c.N != 0 {
 		c.Args = uintptr(noescape(unsafe.Pointer(&args[0])))
 	}
-	cgocall(asmstdcallAddr, unsafe.Pointer(c))
+	errno := cgocall(asmstdcallAddr, unsafe.Pointer(c))
 	// cgocall may reschedule us on to a different M,
 	// but it copies the return values into the new M's
 	// so we can read them from there.
 	c = &getg().m.winsyscall
-	return c.R1, c.R2, c.Err
+	return c.R1, c.R2, uintptr(errno)
 }