The patch in this Bugzilla entry was requested by a customer:
  https://sourceware.org/bugzilla/show_bug.cgi?id=4578

If a thread happens to hold dl_load_lock and have r_state set to RT_ADD or
RT_DELETE at the time another thread calls fork(), then the child exit code
from fork (in nptl/sysdeps/unix/sysv/linux/fork.c in our case) re-initializes
dl_load_lock but does not restore r_state to RT_CONSISTENT. If the child
subsequently requires ld.so functionality before calling exec(), then the
assertion will fire.

The patch acquires dl_load_lock on entry to fork() and releases it on exit
from the parent path.  The child path is initialized as currently done.
This is essentially pthreads_atfork, but forced to be first because the
acquisition of dl_load_lock must happen before malloc_atfork is active
to avoid a deadlock.
The patch has not yet been integrated upstream.

Upstream-Status: Pending [ Not Author See bugzilla]

Signed-off-by: Raghunath Lolur <Raghunath.Lolur@kpit.com>
Signed-off-by: Yuanjie Huang <yuanjie.huang@windriver.com>
Signed-off-by: Zhixiong Chi <zhixiong.chi@windriver.com>

Index: git/sysdeps/nptl/fork.c
===================================================================
--- git.orig/sysdeps/nptl/fork.c       2017-08-03 16:02:15.674704080 +0800
+++ git/sysdeps/nptl/fork.c    2017-08-04 18:15:02.463362015 +0800
@@ -25,6 +25,7 @@
 #include <tls.h>
 #include <hp-timing.h>
 #include <ldsodefs.h>
+#include <libc-lock.h>
 #include <stdio-lock.h>
 #include <atomic.h>
 #include <nptl/pthreadP.h>
@@ -60,6 +61,10 @@
      but our current fork implementation is not.  */
   bool multiple_threads = THREAD_GETMEM (THREAD_SELF, header.multiple_threads);
 
+  /* grab ld.so lock BEFORE switching to malloc_atfork */
+  __rtld_lock_lock_recursive (GL(dl_load_lock));
+  __rtld_lock_lock_recursive (GL(dl_load_write_lock));
+
   /* Run all the registered preparation handlers.  In reverse order.
      While doing this we build up a list of all the entries.  */
   struct fork_handler *runp;
@@ -247,6 +252,10 @@
 
 	  allp = allp->next;
 	}
+
+      /* unlock ld.so last, because we locked it first */
+      __rtld_lock_unlock_recursive (GL(dl_load_write_lock));
+      __rtld_lock_unlock_recursive (GL(dl_load_lock));
     }
 
   return pid;