https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85230
--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
So (completely untested):
--- gcc/asan.c.jj 2018-01-09 21:53:38.821577722 +0100
+++ gcc/asan.c 2018-04-12 12:30:43.360840432 +0200
@@ -554,14 +554,14 @@ get_last_alloca_addr ()
return last_alloca_addr;
}
-/* Insert __asan_allocas_unpoison (top, bottom) call after
+/* Insert __asan_allocas_unpoison (top, bottom) call before
__builtin_stack_restore (new_sp) call.
The pseudocode of this routine should look like this:
- __builtin_stack_restore (new_sp);
top = last_alloca_addr;
bot = new_sp;
__asan_allocas_unpoison (top, bot);
last_alloca_addr = new_sp;
+ __builtin_stack_restore (new_sp);
In general, we can't use new_sp as bot parameter because on some
architectures SP has non zero offset from dynamic stack area. Moreover, on
some architectures this offset (STACK_DYNAMIC_OFFSET) becomes known for
each
@@ -584,9 +584,9 @@ handle_builtin_stack_restore (gcall *cal
tree restored_stack = gimple_call_arg (call, 0);
tree fn = builtin_decl_implicit (BUILT_IN_ASAN_ALLOCAS_UNPOISON);
gimple *g = gimple_build_call (fn, 2, last_alloca, restored_stack);
- gsi_insert_after (iter, g, GSI_NEW_STMT);
+ gsi_insert_before (iter, g, GSI_SAME_STMT);
g = gimple_build_assign (last_alloca, restored_stack);
- gsi_insert_after (iter, g, GSI_NEW_STMT);
+ gsi_insert_before (iter, g, GSI_SAME_STMT);
}
/* Deploy and poison redzones around __builtin_alloca call. To do this, we
? If the __builtin_stack_restore is at the end of function or isn't even
present at all (if __builtin_alloca is used), then it is already doing the
right thing. Small testcase:
void baz (char *, int);
void
foo (int n)
{
baz (__builtin_alloca (n), n);
}
void
bar (int n)
{
{
char a[n];
baz (a, n);
}
{
char b[n + 2];
baz (b, n + 2);
}
}
The patch only changes the order of __asan_allocas_unpoison and stack restore
after the first call to baz from bar above, the rest is unchanged.
