Switch two calls from memset() to explicit_bzero()

This matches the documented behavior more obviously and ensures that
these aren't optimized away, although this is unlikely.

Discussed with deraadt and otto

diff --git a/lib/libc/stdlib/malloc.c b/lib/libc/stdlib/malloc.c
index 618554d..64273b4 100644 (file)
--- a/lib/libc/stdlib/malloc.c
+++ b/lib/libc/stdlib/malloc.c
@@ -1,4 +1,4 @@
-/*     $OpenBSD: malloc.c,v 1.271 2021/07/23 18:04:28 otto Exp $       */
+/*     $OpenBSD: malloc.c,v 1.272 2021/09/19 09:15:22 tb Exp $ */
 /*
  * Copyright (c) 2008, 2010, 2011, 2016 Otto Moerbeek <otto@drijf.net>
  * Copyright (c) 2012 Matthew Dempsky <matthew@openbsd.org>
@@ -741,7 +741,7 @@ unmap(struct dir_info *d, void *p, size_t sz, size_t clear)
 
        /* fill slot */
        if (clear > 0)
-               memset(p, 0, clear);
+               explicit_bzero(p, clear);
        if (mopts.malloc_freeunmap)
                mprotect(p, sz, PROT_NONE);
        else
@@ -1432,7 +1432,7 @@ ofree(struct dir_info **argpool, void *p, int clear, int check, size_t argsz)
                        }
                } else if (argsz > 0) {
                        r = find(pool, p);
-                       memset(p, 0, argsz);
+                       explicit_bzero(p, argsz);
                }
                if (p != NULL) {
                        if (r == NULL)