Fix stack alignment of child threads on ARM.

src/nonsfi/linux/linux_pthread_private.c

 /*
  * Copyright 2014 The Native Client Authors. All rights reserved.
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #include <errno.h>
 
 #include "native_client/src/nonsfi/linux/linux_syscall_defines.h"
 #include "native_client/src/nonsfi/linux/linux_syscall_structs.h"
@@ skipping 77 matching lines @@
   __asm__ __volatile__("svc #0\n"
                        /*
                         * If the return value of clone is non-zero, we are
                         * in the parent thread of clone.
                         */
                        "cmp r0, #0\n"
                        "bne 0f\n"
                        /*
                         * In child thread. Clear the frame pointer to
                         * prevent debuggers from unwinding beyond this,
-                        * pop the stack to get start_func and call it.
+                        * load start_func from the stack and call it.
                         */
                        "mov fp, #0\n"
-                       "pop {r0}\n"
+                       "ldr r0, [sp]\n"
                        "blx r0\n"
                        /* start_func never finishes. */
                        "bkpt #0\n"
                        "0:\n"
                        : "=r"(result)
                        : "r"(sysno),
                          "r"(a1), "r"(a2), "r"(a3), "r"(a4), "r"(a5)
                        : "memory");
 #else
 # error Unsupported architecture
@@ skipping 38 matching lines @@
 }
 
 void __nc_initialize_interfaces(void) {
   const struct nacl_irt_thread init = {
     nacl_irt_thread_create,
     nacl_irt_thread_exit,
     nacl_irt_thread_nice,
   };
   __libnacl_irt_thread = init;
 }