diff --git a/lib/bcheck.c b/lib/bcheck.c
index 993e687a..d64159b3 100644
--- a/lib/bcheck.c
+++ b/lib/bcheck.c
@@ -257,6 +257,8 @@ void splay_printtree(Tree * t, int d);
 
 /* external interface */
 void __bounds_checking (int no_check);
+void __bound_checking_lock (void);
+void __bound_checking_unlock (void);
 void __bound_never_fatal (int no_check);
 DLL_EXPORT void * __bound_ptr_add(void *p, size_t offset);
 DLL_EXPORT void * __bound_ptr_indir1(void *p, size_t offset);
@@ -445,7 +447,11 @@ int tcc_backtrace(const char *fmt, ...);
 
 /* print a bound error message */
 #define bound_warning(...) \
-    tcc_backtrace("^bcheck.c^BCHECK: " __VA_ARGS__)
+    do {                                                 \
+        WAIT_SEM ();                                     \
+        tcc_backtrace("^bcheck.c^BCHECK: " __VA_ARGS__); \
+        POST_SEM ();                                     \
+    } while (0)
 
 #define bound_error(...)            \
     do {                            \
@@ -498,6 +504,16 @@ void __bounds_checking (int no_check)
 #endif
 }
 
+void __bound_checking_lock(void)
+{
+    WAIT_SEM ();
+}
+
+void __bound_checking_unlock(void)
+{
+    POST_SEM ();
+}
+
 /* enable/disable checking. This can be used in signal handlers. */
 void __bound_never_fatal (int neverfatal)
 {
@@ -1264,6 +1280,7 @@ void __bound_exit_dll(size_t *p)
     dprintf(stderr, "%s, %s()\n", __FILE__, __FUNCTION__);
 
     if (p) {
+        WAIT_SEM ();
 	while (p[0] != 0) {
 	    tree = splay_delete(p[0], tree);
 #if BOUND_DEBUG
@@ -1275,6 +1292,7 @@ void __bound_exit_dll(size_t *p)
 #endif
 	    p += 2;
 	}
+        POST_SEM ();
     }
 }
 
diff --git a/lib/bt-exe.c b/lib/bt-exe.c
index b3bf2e67..f7584765 100644
--- a/lib/bt-exe.c
+++ b/lib/bt-exe.c
@@ -7,6 +7,8 @@
 #include "../tccrun.c"
 
 int (*__rt_error)(void*, void*, const char *, va_list);
+__attribute__((weak)) void __bound_checking_lock(void);
+__attribute__((weak)) void __bound_checking_unlock(void);
 
 #ifndef _WIN32
 # define __declspec(n)
@@ -21,8 +23,10 @@ void __bt_init(rt_context *p, int num_callers)
     //fprintf(stderr, "__bt_init %d %p %p\n", num_callers, p->stab_sym, p->bounds_start), fflush(stderr);
     /* call __bound_init here due to redirection of sigaction */
     /* needed to add global symbols */
-    if (__bound_init && p->bounds_start)
+    if (p->bounds_start) {
 	__bound_init(p->bounds_start, -1);
+        __bound_checking_lock();
+    }
     if (num_callers) {
         memcpy(rc, p, offsetof(rt_context, next));
         rc->num_callers = num_callers - 1;
@@ -32,6 +36,8 @@ void __bt_init(rt_context *p, int num_callers)
     } else {
         p->next = rc->next, rc->next = p;
     }
+    if (p->bounds_start)
+        __bound_checking_unlock();
 }
 
 __declspec(dllexport)
@@ -40,8 +46,10 @@ void __bt_exit(rt_context *p)
     __attribute__((weak)) void __bound_exit_dll(void*);
     struct rt_context *rc = &g_rtctxt;
 
-    if (__bound_exit_dll && p->bounds_start)
+    if (p->bounds_start) {
 	__bound_exit_dll(p->bounds_start);
+        __bound_checking_lock();
+    }
     while (rc) {
         if (rc->next == p) {
 	    rc->next = rc->next->next;
@@ -49,6 +57,8 @@ void __bt_exit(rt_context *p)
 	}
 	rc = rc->next;
     }
+    if (p->bounds_start)
+        __bound_checking_unlock();
 }
 
 /* copy a string and truncate it. */