diff --git a/winsup/cygwin/ChangeLog b/winsup/cygwin/ChangeLog
index 91575625d..de66583c7 100644
--- a/winsup/cygwin/ChangeLog
+++ b/winsup/cygwin/ChangeLog
@@ -1,3 +1,8 @@
+2005-04-01  Corinna Vinschen  <corinna@vinschen.de>
+
+	* fork.cc (fork_child): Fixup SYSV IPC shared memory before fixing
+	up fdtab.
+
 2005-03-31  Corinna Vinschen  <corinna@vinschen.de>
 
 	* gentls_offsets: Add '-m32' flag to native compiler call to satisfy
diff --git a/winsup/cygwin/fork.cc b/winsup/cygwin/fork.cc
index 4ae2d0a21..4a842131a 100644
--- a/winsup/cygwin/fork.cc
+++ b/winsup/cygwin/fork.cc
@@ -199,6 +199,17 @@ fork_child (HANDLE& hParent, dll *&first_dll, bool& load_dlls)
   if (fixup_mmaps_after_fork (hParent))
     api_fatal ("recreate_mmaps_after_fork_failed");
 
+#ifdef USE_SERVER
+  /* Incredible but true:  If we use sockets and SYSV IPC shared memory,
+     there's a good chance that a duplicated socket in the child occupies
+     memory which is needed to duplicate shared memory from the parent
+     process, if the shared memory hasn't been duplicated already.
+     The same goes very likely for "normal" mmap shared memory, too, but
+     with SYSV IPC it was the first time observed.  So, *never* fixup
+     fdtab before fixing up shared memory. */
+  if (fixup_shms_after_fork ())
+    api_fatal ("recreate_shm areas after fork failed");
+#endif
 
   MALLOC_CHECK;
 
@@ -226,11 +237,6 @@ fork_child (HANDLE& hParent, dll *&first_dll, bool& load_dlls)
   _my_tls.fixup_after_fork ();
   sigproc_init ();
 
-#ifdef USE_SERVER
-  if (fixup_shms_after_fork ())
-    api_fatal ("recreate_shm areas after fork failed");
-#endif
-
   pthread::atforkchild ();
   fixup_timers_after_fork ();
   fixup_hooks_after_fork ();