runtime: fallback to TEB arbitrary pointer when TLS slots are full

The Go runtime allocates the TLS slot in the TEB TLS slots instead of
using the TEB arbitrary pointer. See CL 431775 for more context.

The problem is that the TEB TLS slots array only has capacity for 64
indices, allocating more requires some complex logic that we don't
support yet.

Although the Go runtime only allocates one index, a Go DLL can be
loaded in a process with more than 64 TLS slots allocated,
in which case it abort.

This CL avoids aborting by falling back to the older behavior, that
is to use the TEB arbitrary pointer.

Fixes #59213

Change-Id: I39c73286fe2da95aa9c5ec5657ee0979ecbec533
Reviewed-on: https://go-review.googlesource.com/c/go/+/486816
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Quim Muntal <quimmuntal@gmail.com>
Reviewed-by: Bryan Mills <bcmills@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Gopher Robot <gobot@golang.org>

src/runtime/signal_windows_test.go

@@ -257,3 +257,59 @@ func TestLibraryCtrlHandler(t *testing.T) {
 		t.Fatalf("Program exited with error: %v\n%s", err, &stderr)
 	}
 }
+
+func TestIssue59213(t *testing.T) {
+	if runtime.GOOS != "windows" {
+		t.Skip("skipping windows only test")
+	}
+	if *flagQuick {
+		t.Skip("-quick")
+	}
+	testenv.MustHaveGoBuild(t)
+	testenv.MustHaveCGO(t)
+
+	goEnv := func(arg string) string {
+		cmd := testenv.Command(t, testenv.GoToolPath(t), "env", arg)
+		cmd.Stderr = new(bytes.Buffer)
+
+		line, err := cmd.Output()
+		if err != nil {
+			t.Fatalf("%v: %v\n%s", cmd, err, cmd.Stderr)
+		}
+		out := string(bytes.TrimSpace(line))
+		t.Logf("%v: %q", cmd, out)
+		return out
+	}
+
+	cc := goEnv("CC")
+	cgoCflags := goEnv("CGO_CFLAGS")
+
+	t.Parallel()
+
+	tmpdir := t.TempDir()
+	dllfile := filepath.Join(tmpdir, "test.dll")
+	exefile := filepath.Join(tmpdir, "gotest.exe")
+
+	// build go dll
+	cmd := testenv.Command(t, testenv.GoToolPath(t), "build", "-o", dllfile, "-buildmode", "c-shared", "testdata/testwintls/main.go")
+	out, err := testenv.CleanCmdEnv(cmd).CombinedOutput()
+	if err != nil {
+		t.Fatalf("failed to build go library: %s\n%s", err, out)
+	}
+
+	// build c program
+	cmd = testenv.Command(t, cc, "-o", exefile, "testdata/testwintls/main.c")
+	testenv.CleanCmdEnv(cmd)
+	cmd.Env = append(cmd.Env, "CGO_CFLAGS="+cgoCflags)
+	out, err = cmd.CombinedOutput()
+	if err != nil {
+		t.Fatalf("failed to build c exe: %s\n%s", err, out)
+	}
+
+	// run test program
+	cmd = testenv.Command(t, exefile, dllfile, "GoFunc")
+	out, err = testenv.CleanCmdEnv(cmd).CombinedOutput()
+	if err != nil {
+		t.Fatalf("failed: %s\n%s", err, out)
+	}
+}

src/runtime/sys_windows_386.s

@@ -9,6 +9,7 @@
 
 // Offsets into Thread Environment Block (pointer in FS)
 #define TEB_TlsSlots 0xE10
+#define TEB_ArbitraryPtr 0x14
 
 // void runtime·asmstdcall(void *c);
 TEXT runtime·asmstdcall(SB),NOSPLIT,$0
@@ -286,7 +287,10 @@ TEXT runtime·wintls(SB),NOSPLIT,$0
 	// Assert that slot is less than 64 so we can use _TEB->TlsSlots
 	CMPL	CX, $64
 	JB	ok
-	CALL	runtime·abort(SB)
+	// Fallback to the TEB arbitrary pointer.
+	// TODO: don't use the arbitrary pointer (see go.dev/issue/59824)
+	MOVL	$TEB_ArbitraryPtr, CX
+	JMP	settls
 ok:
 	// Convert the TLS index at CX into
 	// an offset from TEB_TlsSlots.
@@ -294,5 +298,6 @@ ok:
 
 	// Save offset from TLS into tls_g.
 	ADDL	$TEB_TlsSlots, CX
+settls:
 	MOVL	CX, runtime·tls_g(SB)
 	RET

src/runtime/sys_windows_amd64.s

@@ -10,6 +10,7 @@
 
 // Offsets into Thread Environment Block (pointer in GS)
 #define TEB_TlsSlots 0x1480
+#define TEB_ArbitraryPtr 0x28
 
 // void runtime·asmstdcall(void *c);
 TEXT runtime·asmstdcall(SB),NOSPLIT,$16
@@ -301,7 +302,11 @@ TEXT runtime·wintls(SB),NOSPLIT,$0
 	// Assert that slot is less than 64 so we can use _TEB->TlsSlots
 	CMPQ	CX, $64
 	JB	ok
-	CALL	runtime·abort(SB)
+
+	// Fallback to the TEB arbitrary pointer.
+	// TODO: don't use the arbitrary pointer (see go.dev/issue/59824)
+	MOVQ	$TEB_ArbitraryPtr, CX
+	JMP	settls
 ok:
 	// Convert the TLS index at CX into
 	// an offset from TEB_TlsSlots.
@@ -309,5 +314,6 @@ ok:
 
 	// Save offset from TLS into tls_g.
 	ADDQ	$TEB_TlsSlots, CX
+settls:
 	MOVQ	CX, runtime·tls_g(SB)
 	RET

src/runtime/sys_windows_arm64.s

@@ -12,6 +12,7 @@
 // Offsets into Thread Environment Block (pointer in R18)
 #define TEB_error 0x68
 #define TEB_TlsSlots 0x1480
+#define TEB_ArbitraryPtr 0x28
 
 // Note: R0-R7 are args, R8 is indirect return value address,
 // R9-R15 are caller-save, R19-R29 are callee-save.
@@ -273,12 +274,15 @@ TEXT runtime·wintls(SB),NOSPLIT,$0
 	// Assert that slot is less than 64 so we can use _TEB->TlsSlots
 	CMP	$64, R0
 	BLT	ok
-	MOVD	$runtime·abort(SB), R1
+	// Fallback to the TEB arbitrary pointer.
-	BL	(R1)
+	// TODO: don't use the arbitrary pointer (see go.dev/issue/59824)
+	MOVD	$TEB_ArbitraryPtr, R0
+	B	settls
 ok:
 
 	// Save offset from R18 into tls_g.
 	LSL	$3, R0
 	ADD	$TEB_TlsSlots, R0
+settls:
 	MOVD	R0, runtime·tls_g(SB)
 	RET

src/runtime/testdata/testwintls/main.c

@@ -0,0 +1,29 @@
+// Copyright 2023 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+#include <windows.h>
+
+int main(int argc, char **argv) {
+    if (argc < 3) {
+        return 1;
+    }
+    // Allocate more than 64 TLS indices
+    // so the Go runtime doesn't find
+    // enough space in the TEB TLS slots.
+    for (int i = 0; i < 65; i++) {
+        TlsAlloc();
+    }
+    HMODULE hlib = LoadLibrary(argv[1]);
+    if (hlib == NULL) {
+        return 2;
+    }
+    FARPROC proc = GetProcAddress(hlib, argv[2]);
+    if (proc == NULL) {
+        return 3;
+    }
+    if (proc() != 42) {
+        return 4;
+    }
+    return 0;
+}

src/runtime/testdata/testwintls/main.go

@@ -0,0 +1,12 @@
+// Copyright 2023 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package main
+
+import "C"
+
+//export GoFunc
+func GoFunc() int { return 42 }
+
+func main() {}