[Bug sanitizer/88291] New: [9 Regression] asan ICE in asan_clear_shadow

["jakub at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88291

            Bug ID: 88291
           Summary: [9 Regression] asan ICE in asan_clear_shadow
           Product: gcc
           Version: 9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: sanitizer
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jakub at gcc dot gnu.org
                CC: dodji at gcc dot gnu.org, dvyukov at gcc dot gnu.org,
                    jakub at gcc dot gnu.org, kcc at gcc dot gnu.org, marxin at gcc dot gnu.org
  Target Milestone: ---

As mentioned in PR88289, e.g. on clone-test-1.c with -Os gcc ICEs on
powerpc64-linux.

I get the same ICE on x86_64-linux with -fsanitize=address -Os
-mstringop-strategy=libcall on:

void bar (void *, void *);
void
foo (void)
{
  int b;
  char __attribute__((aligned(16))) a[(1 << 20) + 1];
  bar (&a, &b);
}

The assert that the length is a multiple of 4 is done only if clear_storage
decided to use a libcall, which is why it doesn't trigger all the time
everywhere.
Like for the initialization of the shadow memory at the start of the function,
the clearing at the end should also do the rounding to multiples of 32 of the
offset and similarly pad the size to multiples of 32 (before the >> 3).