[Bug libc/4737] fork is not async-signal-safe

["tom dot honermann at oracle dot com" <sourceware-bugzilla at sourceware dot org>]

------- Additional Comments From tom dot honermann at oracle dot com  2008-11-11 22:03 -------
Another solution for this issue would be to use error checked mutexes (ie,
PTHREAD_MUTEX_ERRORCHECK).  However, glibc 2.3.4 doesn't seem to have a generic
infrastructure for supporting error checked mutexes as it does for recursive
mutexes (ie, __libc_lock_init vs __libc_lock_init_recursive - there is no
__libc_lock_init_error_checked or similar that I can find), so this approach
might require a more invasive change.

If error checked mutexes were used, then ptmalloc_lock_all could be modified
similarly to the code below.  (This assumes mutex_lock would return the same
values as pthread_mutex_lock)

static void
ptmalloc_lock_all (void)
{
  mstate ar_ptr;
  int rc;

  if(__malloc_initialized < 1)
    return;
  rc = mutex_lock(&list_lock);
  if(rc != 0 && rc != EDEADLK) {
    /* Error handling */
  }
  for(ar_ptr = &main_arena;;) {
    rc = mutex_lock(&ar_ptr->mutex);
    if(rc != 0 && rc != EDEADLK) {
      /* Error handling */
    }
    ar_ptr = ar_ptr->next;
    if(ar_ptr == &main_arena) break;
  }
  save_malloc_hook = __malloc_hook;
  save_free_hook = __free_hook;
  __malloc_hook = malloc_atfork;
  __free_hook = free_atfork;
  /* Only the current thread may perform malloc/free calls now. */
  tsd_getspecific(arena_key, save_arena);
  tsd_setspecific(arena_key, ATFORK_ARENA_PTR);
}

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=4737

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.