Drop using _tlsbase and _tlstop in favor of access via NtCurrentTeb.

* cygtls.h (_tlsbase): Remove.  Replace throughout with
        NtCurrentTeb()->Tib.StackBase.
        (_tlstop): Remove. Replace throughout with
        NtCurrentTeb()->Tib.StackLimit.
        * dcrt0.cc (child_info_fork::alloc_stack): Move definition of local
        teb variable up to be used throughout.
        * include/cygwin/config.h (__getreent): Use inline function on both
        architectures.

Signed-off-by: Corinna Vinschen <corinna@vinschen.de>

winsup/cygwin/ChangeLog

@@ -1,3 +1,14 @@
+2015-12-02  Corinna Vinschen  <corinna@vinschen.de>
+
+	* cygtls.h (_tlsbase): Remove.  Replace throughout with
+	NtCurrentTeb()->Tib.StackBase.
+	(_tlstop): Remove. Replace throughout with
+	NtCurrentTeb()->Tib.StackLimit.
+	* dcrt0.cc (child_info_fork::alloc_stack): Move definition of local
+	teb variable up to be used throughout.
+	* include/cygwin/config.h (__getreent): Use inline function on both
+	architectures.
+
 2015-11-29  Corinna Vinschen  <corinna@vinschen.de>
 
 	* uinfo.cc (pwdgrp::fetch_account_from_windows): Only create 1-5-32-x

winsup/cygwin/cygtls.h

@@ -286,18 +286,7 @@ private:
 #include "cygerrno.h"
 #include "ntdll.h"
 
-#ifdef __x86_64__
-/* When just using a "gs:X" asm for the x86_64 code, gcc wrongly creates
-   pc-relative instructions.  However, NtCurrentTeb() is inline assembler
-   anyway, so using it here should be fast enough on x86_64. */
-#define _tlsbase (NtCurrentTeb()->Tib.StackBase)
-#define _tlstop (NtCurrentTeb()->Tib.StackLimit)
-#else
-extern PVOID _tlsbase __asm__ ("%fs:4");
-extern PVOID _tlstop __asm__ ("%fs:8");
-#endif
-
-#define _my_tls (*((_cygtls *) ((char *)_tlsbase - CYGTLS_PADSIZE)))
+#define _my_tls (*((_cygtls *) ((PBYTE) NtCurrentTeb()->Tib.StackBase - CYGTLS_PADSIZE)))
 extern _cygtls *_main_tls;
 extern _cygtls *_sig_tls;
 

winsup/cygwin/dcrt0.cc

@@ -411,15 +411,16 @@ child_info_fork::alloc_stack ()
 {
   /* Make sure not to try a hard allocation if we have been forked off from
      the main thread of a Cygwin process which has been started from a 64 bit
-     parent.  In that case the _tlsbase of the forked child is not the same
-     as the _tlsbase of the parent (== stackbottom), but only because the
+     parent.  In that case the StackBase of the forked child is not the same
+     as the StackBase of the parent (== stackbottom), but only because the
      stack of the parent has been slightly rearranged.  See comment in
      wow64_revert_to_original_stack for details. We check here if the
      parent stack fits into the child stack. */
-  if (_tlsbase != stackbottom
+  PTEB teb = NtCurrentTeb ();
+  if (teb->Tib.StackBase != stackbottom
       && (!wincap.is_wow64 ()
-	  || stacktop < NtCurrentTeb ()->DeallocationStack
-	  || stackbottom > _tlsbase))
+	  || stacktop < teb->DeallocationStack
+	  || stackbottom > teb->Tib.StackBase))
     {
       void *stack_ptr;
       size_t stacksize;
@@ -432,10 +433,10 @@ child_info_fork::alloc_stack ()
       stacksize = (PBYTE) stackbottom - (PBYTE) stackaddr;
       if (!VirtualAlloc (stackaddr, stacksize, MEM_RESERVE, PAGE_NOACCESS))
 	{
-	  PTEB teb = NtCurrentTeb ();
 	  api_fatal ("fork: can't reserve memory for parent stack "
 		     "%p - %p, (child has %p - %p), %E",
-		     stackaddr, stackbottom, teb->DeallocationStack, _tlsbase);
+		     stackaddr, stackbottom, teb->DeallocationStack,
+		     teb->Tib.StackBase);
 	}
       /* Commit the area commited in parent. */
       stacksize = (PBYTE) stackbottom - (PBYTE) stacktop;
@@ -471,9 +472,10 @@ child_info_fork::alloc_stack ()
       /* Fork has been called from main thread.  Simply commit the region
 	 of the stack commited in the parent but not yet commited in the
 	 child and create new guardpages. */
-      if (_tlstop > stacktop)
+      if (NtCurrentTeb()->Tib.StackLimit > stacktop)
 	{
-	  SIZE_T commitsize = (PBYTE) _tlstop - (PBYTE) stacktop;
+	  SIZE_T commitsize = (PBYTE) NtCurrentTeb()->Tib.StackLimit
+			      - (PBYTE) stacktop;
 	  if (!VirtualAlloc (stacktop, commitsize, MEM_COMMIT, PAGE_READWRITE))
 	    api_fatal ("can't commit child memory for stack %p(%ly), %E",
 		       stacktop, commitsize);
@@ -482,14 +484,14 @@ child_info_fork::alloc_stack ()
 			     MEM_COMMIT, PAGE_READWRITE | PAGE_GUARD))
 	    api_fatal ("fork: couldn't allocate new stack guard page %p, %E",
 		       guardpage);
-	  _tlstop = stacktop;
+	  NtCurrentTeb()->Tib.StackLimit = stacktop;
 	}
       stackaddr = 0;
       /* This only affects forked children of a process started from a native
 	 64 bit process, but it doesn't hurt to do it unconditionally.  Fix
 	 StackBase in the child to be the same as in the parent, so that the
 	 computation of _my_tls is correct. */
-      _tlsbase = (PVOID) stackbottom;
+      teb->Tib.StackBase = (PVOID) stackbottom;
     }
 }
 
@@ -918,8 +920,8 @@ dll_crt0_1 (void *)
 	 this step. */
       if (fork_info->stackaddr)
 	{
-	  _tlsbase = (PVOID) fork_info->stackbottom;
-	  _tlstop = (PVOID) fork_info->stacktop;
+	  NtCurrentTeb()->Tib.StackBase = (PVOID) fork_info->stackbottom;
+	  NtCurrentTeb()->Tib.StackLimit = (PVOID) fork_info->stacktop;
 	}
 
       /* Not resetting _my_tls.incyg here because presumably fork will overwrite

winsup/cygwin/fork.cc

@@ -307,7 +307,7 @@ frok::parent (volatile char * volatile stack_here)
 
   ch.forker_finished = forker_finished;
 
-  ch.stackbottom = _tlsbase;
+  ch.stackbottom = NtCurrentTeb()->Tib.StackBase;
   ch.stackaddr = NtCurrentTeb ()->DeallocationStack;
   if (!ch.stackaddr)
     {
@@ -315,7 +315,7 @@ frok::parent (volatile char * volatile stack_here)
 	 stack.  If so, the entire stack is committed anyway and StackLimit
 	 points to the allocation address of the stack.  Mark in guardsize that
 	 we must not set up guard pages. */
-      ch.stackaddr = ch.stacktop = _tlstop;
+      ch.stackaddr = ch.stacktop = NtCurrentTeb()->Tib.StackLimit;
       ch.guardsize = (size_t) -1;
     }
   else

winsup/cygwin/include/cygwin/config.h

@@ -40,20 +40,19 @@ extern "C" {
 #ifdef _COMPILING_NEWLIB
 #ifdef __x86_64__
 #include "../tlsoffsets64.h"
-/* We would like to use just "%gs:8", but on x86_64 gcc uses pc-relative
-   addressing and translates "gs:8" into the wrong addressing mode. */
-static inline char *___getreent (void)
-{
-  register char *ret;
-  __asm __volatile__ ("movq %%gs:8,%0" : "=r" (ret));
-  return ret + tls_local_clib;
-}
-#define __getreent() ((struct _reent *) ___getreent())
 #else
 #include "../tlsoffsets.h"
-extern char *_tlsbase __asm__ ("%fs:4");
-#define __getreent() (struct _reent *)(_tlsbase + tls_local_clib)
 #endif
+extern inline struct _reent *__getreent (void)
+{
+  register char *ret;
+#ifdef __x86_64__
+  __asm __volatile__ ("movq %%gs:8,%0" : "=r" (ret));
+#else
+  __asm __volatile__ ("movl %%fs:4,%0" : "=r" (ret));
+#endif
+  return (struct _reent *) (ret + tls_local_clib);
+}
 #endif  /* _COMPILING_NEWLIB */
 
 #ifdef __x86_64__

winsup/cygwin/init.cc

@@ -42,7 +42,7 @@ munge_threadfunc ()
   if (!threadfunc_ix[0])
     {
       char **peb;
-      char **top = (char **) _tlsbase;
+      char **top = (char **) NtCurrentTeb()->Tib.StackBase;
       for (peb = ebp, i = 0; peb < top && i < 7; peb++)
 	if (*peb == search_for)
 	  threadfunc_ix[i++] = peb - ebp;

winsup/cygwin/wow64.cc

@@ -171,10 +171,10 @@ wow64_revert_to_original_stack (PVOID &allocationbase)
      accordingly, and return the new, 16 byte aligned address for the
      stack pointer.  The second half of the stack move is done by the
      caller _dll_crt0. */
-  _tlsbase = (char *) newbase;
-  _tlstop = (char *) newtop;
+  NtCurrentTeb()->Tib.StackBase = (char *) newbase;
+  NtCurrentTeb()->Tib.StackLimit = (char *) newtop;
   _main_tls = &_my_tls;
-  return PTR_ADD (_tlsbase, -16);
+  return PTR_ADD (NtCurrentTeb()->Tib.StackBase, -16);
 }
 
 /* Respawn WOW64 process. This is only called if we can't reuse the original