diff --git a/winsup/cygwin/ChangeLog b/winsup/cygwin/ChangeLog
index 3ead72e5d..7d9a24d31 100644
--- a/winsup/cygwin/ChangeLog
+++ b/winsup/cygwin/ChangeLog
@@ -1,3 +1,29 @@
+2006-03-12  Christopher Faylor  <cgf@timesys.com>
+
+	* cygtls.h (CYGTLS_INITIALIZED): Change to a little more unlikely value.
+	(CYGTLSMAGIC): Delete.
+	* dcrt0.cc (dll_crt0_0): Call sigproc_init during init startup.
+	(_dll_crt0): Don't worry about sync_startup.  Just wait for sigthread here.
+	* dll_init.cc (cygwin_detach_dll): Only pick up tls version of retaddr
+	if we have a valid tls.
+	* fork.cc (frok::child): Remove sigproc_init initialization since it
+	happens much earlier now.
+	* gendef: Recognize SIGFE_MAYBE.
+	(fefunc): Generate calls to _sigfe_maybe, if appropriate.
+	(_sigfe_maybe): New function.
+	* init.cc (search_for): Always initialize search_for, even on fork.
+	(calibration_thread): Delete.
+	(calibration_id): Delete.
+	(prime_threads): Delete.
+	(munge_threadfunc): Remove calibration_thread special case.  Avoid
+	calling thread function if we haven't yet hit the "search_for" thread.
+	(dll_entry): Remove prime_threads call.  Only call munge_threadfunc
+	when hwait_sig is active.  Ditto. for _my_tls.remove ();
+	* sigproc.cc (hwait_sig): Make global.
+	(sigproc_init): Don't bother with sync_startup.
+	(sig_send): Treat flush as a no-op when signals are held.
+	(wait_sig): Cause signals to be held after fork.
+
 2006-03-09  Corinna Vinschen  <corinna@vinschen.de>
 
 	* syscalls.cc (rename): Move existance check for oldpath further up
diff --git a/winsup/cygwin/cygtls.h b/winsup/cygwin/cygtls.h
index 5e3155a76..c21201434 100644
--- a/winsup/cygwin/cygtls.h
+++ b/winsup/cygwin/cygtls.h
@@ -23,8 +23,7 @@ details. */
 typedef unsigned int SOCKET;
 #endif
 
-#define CYGTLS_INITIALIZED 0x43227
-#define CYGTLSMAGIC "D0Ub313v31nm&G1c?";
+#define CYGTLS_INITIALIZED 0xc763173f
 
 #ifndef CYG_MAX_PATH
 # define CYG_MAX_PATH 260
diff --git a/winsup/cygwin/cygwin.din b/winsup/cygwin/cygwin.din
index ae91911cd..9fc71162a 100644
--- a/winsup/cygwin/cygwin.din
+++ b/winsup/cygwin/cygwin.din
@@ -301,8 +301,8 @@ cygwin_conv_to_posix_path SIGFE
 cygwin32_conv_to_posix_path = cygwin_conv_to_posix_path SIGFE
 cygwin_conv_to_win32_path SIGFE
 cygwin32_conv_to_win32_path = cygwin_conv_to_win32_path SIGFE
-cygwin_detach_dll SIGFE
-cygwin32_detach_dll = cygwin_detach_dll SIGFE
+cygwin_detach_dll SIGFE_MAYBE
+cygwin32_detach_dll = cygwin_detach_dll SIGFE_MAYBE
 cygwin_dll_init NOSIGFE
 endprotoent = cygwin_endprotoent SIGFE
 endservent = cygwin_endservent SIGFE
diff --git a/winsup/cygwin/dcrt0.cc b/winsup/cygwin/dcrt0.cc
index 8122939b6..9da090f30 100644
--- a/winsup/cygwin/dcrt0.cc
+++ b/winsup/cygwin/dcrt0.cc
@@ -750,6 +750,8 @@ dll_crt0_0 ()
     DuplicateTokenEx (hProcToken, MAXIMUM_ALLOWED, NULL,
 		      SecurityImpersonation, TokenImpersonation,
 		      &hProcImpToken);
+  /* Initialize signal/subprocess handling. */
+  sigproc_init ();
   debug_printf ("finished dll_crt0_0 initialization");
 }
 
@@ -835,9 +837,6 @@ dll_crt0_1 (char *)
   /* Initialize user info. */
   uinfo_init ();
 
-  /* Initialize signal/subprocess handling. */
-  sigproc_init ();
-
   /* Connect to tty. */
   tty_init ();
 
@@ -924,7 +923,6 @@ dll_crt0_1 (char *)
   /* Flush signals and ensure that signal thread is up and running. Can't
      do this for noncygwin case since the signal thread is blocked due to
      LoadLibrary serialization. */
-  wait_for_sigthread ();
   ld_preload ();
   if (user_data->main)
     cygwin_exit (user_data->main (__argc, __argv, *user_data->envptr));
@@ -950,14 +948,8 @@ initialize_main_tls (char *padding)
 extern "C" void __stdcall
 _dll_crt0 ()
 {
-  extern HANDLE sync_startup;
   extern DWORD threadfunc_ix;
-  if (sync_startup != INVALID_HANDLE_VALUE)
-    {
-      WaitForSingleObject (sync_startup, INFINITE);
-      CloseHandle (sync_startup);
-    }
-
+  wait_for_sigthread ();
   if (!threadfunc_ix)
     system_printf ("internal error: couldn't determine location of thread function on stack.  Expect signal problems.");
 
diff --git a/winsup/cygwin/dll_init.cc b/winsup/cygwin/dll_init.cc
index 7d35c1633..7c7fc6407 100644
--- a/winsup/cygwin/dll_init.cc
+++ b/winsup/cygwin/dll_init.cc
@@ -404,7 +404,12 @@ dll_noncygwin_dllcrt0 (HMODULE h, per_process *p)
 extern "C" void
 cygwin_detach_dll (dll *)
 {
-  dlls.detach ((HANDLE) _my_tls.retaddr ());
+  HANDLE retaddr;
+  if (_my_tls.isinitialized ())
+    retaddr = (HANDLE) _my_tls.retaddr ();
+  else
+    retaddr = __builtin_return_address (0);
+  dlls.detach (retaddr);
 }
 
 extern "C" void
diff --git a/winsup/cygwin/fork.cc b/winsup/cygwin/fork.cc
index e4c3c31bb..244bb71db 100644
--- a/winsup/cygwin/fork.cc
+++ b/winsup/cygwin/fork.cc
@@ -171,8 +171,6 @@ frok::child (void *)
 
   ForceCloseHandle1 (fork_info->forker_finished, forker_finished);
 
-  sigproc_init ();
-
   pthread::atforkchild ();
   fixup_timers_after_fork ();
   cygbench ("fork-child");
diff --git a/winsup/cygwin/gendef b/winsup/cygwin/gendef
index c6b6a7909..40796f094 100755
--- a/winsup/cygwin/gendef
+++ b/winsup/cygwin/gendef
@@ -1,5 +1,5 @@
 #!/usr/bin/perl
-# Copyright 2003, 2004, 2005 Red Hat, Inc.
+# Copyright 2003, 2004, 2005, 2006 Red Hat, Inc.
 #
 # This file is part of Cygwin.
 #
@@ -43,12 +43,14 @@ for (@in) {
     chomp;
     if (/=/o) {
 	if (s/\s+NOSIGFE\s*$//) {
-	} elsif (s/ SIGFE$//) {
-	  my $func = (split(' '))[2];
-	  $sigfe{$func} = '_sigfe_' . $func;
+	    # nothing
+	} elsif (s/ SIGFE(_MAYBE)?$//) {
+	    my $func = (split(' '))[2];
+	    my $maybe = lc $1 . '_';
+	    $sigfe{$func} = '_sigfe' . $maybe . $func;
 	}
     } else {
-	my ($func, $sigfe) = m%^\s*(\S+)(?:\s+((?:NO)?SIGR?FE))?$%o;
+	my ($func, $sigfe) = m%^\s*(\S+)(?:\s+((?:NO)?SIGFE(?:_MAYBE)?))?$%o;
 	if (defined($sigfe) && $sigfe =~ /^NO/o) {
 	    $_ = $func;
 	} else {
@@ -83,13 +85,14 @@ close SIGFE;
 sub fefunc {
     my $func = '_' . shift;
     my $fe = '_' . shift;
+    my $sigfe_func = ($fe =~ /^(.*)$func/)[0];
     my $extra;
     my $res = <<EOF;
 	.extern	$func
 	.global	$fe
 $fe:
 	pushl	\$$func
-	jmp	__sigfe
+	jmp	$sigfe_func
 
 EOF
     if (!$main::first++) {
@@ -97,6 +100,18 @@ EOF
 	.text
 
 	.stabs	"_sigfe:F(0,1)",36,0,0,__sigfe
+__sigfe_maybe:
+	pushl	%ebx
+	pushl	%edx
+	movl	%fs:4,%ebx				# location of bottom of stack
+	movl	$tls::initialized(%ebx),%eax
+	cmpl	\$0xc763173f,%eax			# initialized?
+	je	1f
+	popl	%edx
+	popl	%ebx
+	popl	%eax
+	jmp	*%eax
+
 __sigfe:
 	pushl	%ebx
 	pushl	%edx
diff --git a/winsup/cygwin/init.cc b/winsup/cygwin/init.cc
index 3fd91c1d6..342bb1cfe 100644
--- a/winsup/cygwin/init.cc
+++ b/winsup/cygwin/init.cc
@@ -19,11 +19,9 @@ details. */
 #include "ntdll.h"
 
 int NO_COPY dynamically_loaded;
-static char *search_for = (char *) cygthread::stub;
+static char NO_COPY *search_for = (char *) cygthread::stub;
 unsigned threadfunc_ix[8] __attribute__((section (".cygwin_dll_common"), shared));
-DWORD tls_func;
-
-HANDLE sync_startup;
+extern cygthread *hwait_sig;
 
 #define OLDFUNC_OFFSET -1
 
@@ -35,32 +33,6 @@ threadfunc_fe (VOID *arg)
   _cygtls::call ((DWORD (*)  (void *, void *)) (((char **) _tlsbase)[OLDFUNC_OFFSET]), arg);
 }
 
-static DWORD WINAPI
-calibration_thread (VOID *arg)
-{
-  ExitThread (0);
-}
-
-static DWORD calibration_id;
-
-/* We need to know where the OS stores the address of the thread function
-   on the stack so that we can intercept the call and insert some tls
-   stuff on the stack.  This function starts a known calibration thread.
-   When it starts, a call will be made to dll_entry which will call munge_threadfunc
-   looking for the calibration thread offset on the stack.  This offset will
-   be stored and used by all executing cygwin processes. */
-static void
-prime_threads ()
-{
-  if (threadfunc_ix[0])
-    sync_startup = INVALID_HANDLE_VALUE;
-  else
-    {
-      search_for = (char *) calibration_thread;
-      sync_startup = CreateThread (NULL, 0, calibration_thread, 0, 0, &calibration_id);
-    }
-}
-
 /* If possible, redirect the thread entry point to a cygwin routine which
    adds tls stuff to the stack. */
 static void
@@ -82,14 +54,16 @@ munge_threadfunc ()
 	}
     }
 
-  char *threadfunc = ebp[threadfunc_ix[0]];
-  if (threadfunc == (char *) calibration_thread)
-    /* no need for the overhead */;
-  else if (threadfunc_ix[0])
+  if (threadfunc_ix[0])
     {
-      for (i = 0; threadfunc_ix[i]; i++)
-	ebp[threadfunc_ix[i]] = (char *) threadfunc_fe;
-      ((char **) _tlsbase)[OLDFUNC_OFFSET] = threadfunc;
+      char *threadfunc = ebp[threadfunc_ix[0]];
+      if (!search_for || threadfunc == search_for)
+	{
+	  search_for = NULL;
+	  for (i = 0; threadfunc_ix[i]; i++)
+	    ebp[threadfunc_ix[i]] = (char *) threadfunc_fe;
+	  ((char **) _tlsbase)[OLDFUNC_OFFSET] = threadfunc;
+	}
     }
 }
 
@@ -170,16 +144,15 @@ dll_entry (HANDLE h, DWORD reason, void *static_load)
 	respawn_wow64_process ();
 
       dll_crt0_0 ();
-      prime_threads ();	// this should be the last thing to happen
       break;
     case DLL_PROCESS_DETACH:
       break;
     case DLL_THREAD_ATTACH:
-      if (!sync_startup || GetCurrentThreadId () == calibration_id)
+      if (hwait_sig)
 	munge_threadfunc ();
       break;
     case DLL_THREAD_DETACH:
-      if (!sync_startup)
+      if (hwait_sig)
 	_my_tls.remove (0);
       break;
     }
diff --git a/winsup/cygwin/sigproc.cc b/winsup/cygwin/sigproc.cc
index 5e9ad1915..845f7cd80 100644
--- a/winsup/cygwin/sigproc.cc
+++ b/winsup/cygwin/sigproc.cc
@@ -61,10 +61,10 @@ HANDLE NO_COPY signal_arrived;		// Event signaled when a signal has
 
 HANDLE NO_COPY sigCONT;			// Used to "STOP" a process
 
-Static cygthread *hwait_sig;
+cygthread *hwait_sig;
 Static HANDLE wait_sig_inited;		// Control synchronization of
 					//  message queue startup
-Static bool sigheld;			// True if holding signals
+static bool sigheld;			// True if holding signals
 
 Static int nprocs;			// Number of deceased children
 Static char cprocs[(NPROCS + 1) * sizeof (pinfo)];// All my children info
@@ -475,7 +475,6 @@ create_signal_arrived ()
 void __stdcall
 sigproc_init ()
 {
-  extern HANDLE sync_startup;
   wait_sig_inited = CreateEvent (&sec_none_nih, TRUE, FALSE, NULL);
   ProtectHandle (wait_sig_inited);
 
@@ -484,7 +483,6 @@ sigproc_init ()
    */
   sync_proc_subproc.init ("sync_proc_subproc");
 
-  sync_startup = NULL;
   hwait_sig = new cygthread (wait_sig, 0, cygself, "sig");
   hwait_sig->zap_h ();
 
@@ -523,6 +521,8 @@ sig_send (_pinfo *p, int sig)
 #endif
       return -1;
     }
+  else if (sig == __SIGFLUSH || sig == __SIGFLUSHFAST)
+    return 0;
   else
     {
       SetEvent (sigCONT);
@@ -1091,8 +1091,12 @@ wait_sig (VOID *)
 		readsig, myself->sendsig);
 
   sigpacket pack;
+  if (in_forkee)
+    pack.si.si_signo = __SIGHOLD;
   for (;;)
     {
+      if (pack.si.si_signo == __SIGHOLD)
+	WaitForSingleObject (sigCONT, INFINITE);
       DWORD nb;
       pack.tls = NULL;
       if (!ReadFile (readsig, &pack, sizeof (pack), &nb, NULL))
@@ -1194,8 +1198,6 @@ wait_sig (VOID *)
 	  sigproc_printf ("signalling pack.wakeup %p", pack.wakeup);
 	  SetEvent (pack.wakeup);
 	}
-      if (pack.si.si_signo == __SIGHOLD)
-	WaitForSingleObject (sigCONT, INFINITE);
       if (pack.si.si_signo == __SIGEXIT)
 	break;
     }