src/hotspot/os/linux/os_linux.cpp

@@ -798,10 +798,66 @@
     os::current_thread_id(), (uintx) pthread_self());
 
   return 0;
 }
 
+// On Linux, glibc places static TLS blocks (for __thread variables) on
+// the thread stack. This decreases the stack size actually available
+// to threads.
+//
+// For large static TLS sizes, this may cause threads to malfunction due
+// to insufficient stack space. This is a well-known issue in glibc:
+// http://sourceware.org/bugzilla/show_bug.cgi?id=11787.
+//
+// As a workaround, we call a private but assumed-stable glibc function,
+// __pthread_get_minstack() to obtain the minstack size and derive the
+// static TLS size from it. We then increase the user requested stack
+// size by this TLS size.
+//
+// Due to compatibility concerns, this size adjustment is opt-in and
+// controlled via AdjustStackSizeForTLS.
+typedef size_t (*GetMinStack)(const pthread_attr_t *attr);
+
+GetMinStack _get_minstack_func = NULL;
+
+static void get_minstack_init() {
+  _get_minstack_func =
+        (GetMinStack)dlsym(RTLD_DEFAULT, "__pthread_get_minstack");
+}
+
+// Returns the size of the static TLS area glibc puts on thread stacks.
+static size_t get_static_tls_area_size(const pthread_attr_t *attr) {
+  size_t tls_size = 0;
+  if (_get_minstack_func != NULL) {
+    // Obtain the pthread minstack size by calling __pthread_get_minstack.
+    size_t minstack_size = _get_minstack_func(attr);
+
+    // Remove non-TLS area size included in minstack size returned
+    // by __pthread_get_minstack() to get the static TLS size.
+    // In glibc before 2.27, minstack size includes guard_size.
+    // In glibc 2.27 and later, guard_size is automatically added
+    // to the stack size by pthread_create and is no longer included
+    // in minstack size. In both cases, the guard_size is taken into
+    // account, so there is no need to adjust the result for that.
+    //
+    // Although __pthread_get_minstack() is a private glibc function,
+    // it is expected to have a stable behavior across future glibc
+    // versions while glibc still allocats the static TLS blocks off
+    // the stack. Following is glibc 2.28 __pthread_get_minstack(): 
+    //
+    // size_t
+    // __pthread_get_minstack (const pthread_attr_t *attr)
+    // {
+    //   return GLRO(dl_pagesize) + __static_tls_size + PTHREAD_STACK_MIN;
+    // }
+    //
+    tls_size = minstack_size - os::vm_page_size() - PTHREAD_STACK_MIN;
+    assert(tls_size > 0, "unexpected size");
+  }
+  return tls_size;
+}
+
 bool os::create_thread(Thread* thread, ThreadType thr_type,
                        size_t req_stack_size) {
   assert(thread->osthread() == NULL, "caller responsible");
 
   // Allocate the OSThread object

@@ -831,21 +887,33 @@
   // takes the space out of 'stacksize'. Thus we adapt the requested
   // stack_size by the size of the guard pages to mimick proper
   // behaviour. However, be careful not to end up with a size
   // of zero due to overflow. Don't add the guard page in that case.
   size_t guard_size = os::Linux::default_guard_size(thr_type);
-  if (stack_size <= SIZE_MAX - guard_size) {
-    stack_size += guard_size;
+  // Configure glibc guard page.
+  pthread_attr_setguardsize(&attr, guard_size);
+ 
+  size_t stack_adjust_size = 0;
+  if (AdjustStackSizeForTLS) {
+    // Adjust the stack_size by adding the on-stack TLS size if
+    // AdjustStackSizeForTLS is true. The guard size is already
+    // accounted in this case, please see comments in
+    // get_static_tls_area_size().
+    stack_adjust_size += get_static_tls_area_size(&attr);
+  } else {
+    stack_adjust_size += guard_size;
+  }
+
+  stack_adjust_size = align_up(stack_adjust_size, os::vm_page_size()); 
+  if (stack_size <= SIZE_MAX - stack_adjust_size) {
+    stack_size += stack_adjust_size;
   }
   assert(is_aligned(stack_size, os::vm_page_size()), "stack_size not aligned");
 
   int status = pthread_attr_setstacksize(&attr, stack_size);
   assert_status(status == 0, status, "pthread_attr_setstacksize");
 
-  // Configure glibc guard page.
-  pthread_attr_setguardsize(&attr, os::Linux::default_guard_size(thr_type));
-
   ThreadState state;
 
   {
     pthread_t tid;
     int ret = pthread_create(&tid, &attr, (void* (*)(void*)) thread_native_entry, thread);

@@ -5120,10 +5188,12 @@
   // Initialize data for jdk.internal.misc.Signal
   if (!ReduceSignalUsage) {
     jdk_misc_signal_init();
   }
 
+  get_minstack_init();
+
   // Check and sets minimum stack sizes against command line options
   if (Posix::set_minimum_stack_sizes() == JNI_ERR) {
     return JNI_ERR;
   }