[PATCH v3 3/6] Initialize l_mach field from rtld link_map

[Jovan Dmitrovic]

Hello everyone,
sorry for the long delay.

> That's what my original version of the patch did....  🙂 

Almost the same, I only changed it to zero out only the l_mach field. Of 
course, the l_info
zeroing remained. AFAIK, only the l_mach and l_info fields are 
problematic, and there is no need
to zero out bootstrap_map in its entirety.

> Even if compiler inlines the zero clearing, I think it would be better to
> follow the same strategy done for l_info and either user __builtin_memset
> or a open-coded memset. 

You would be correct, this part is not portable. What we can do is what 
Sandra did, but only on the l_mach field. Something like this:

--- a/elf/rtld.c
+++ b/elf/rtld.c
@@ -534,11 +534,17 @@ _dl_start (void *arg)
  #ifndef DONT_USE_BOOTSTRAP_MAP
  # ifdef HAVE_BUILTIN_MEMSET
    __builtin_memset (bootstrap_map.l_info, '\0', sizeof 
(bootstrap_map.l_info));
+  __builtin_memset (bootstrap_map.l_mach, '\0', sizeof 
(bootstrap_map.l_mach));
  # else
    for (size_t cnt = 0;
         cnt < sizeof (bootstrap_map.l_info) / sizeof 
(bootstrap_map.l_info[0]);
         ++cnt)
      bootstrap_map.l_info[cnt] = 0;
+
+  char *p = (char*) &bootstrap_map.l_mach;
+  char *pend = p + sizeof (bootstrap_map.l_mach);
+  while (p != pend)
+    *(p++) = '\0';
  # endif
  #endif

Regards,
Jovan