Avoid potential race between realloc and GC_block_was_dirty

GC_realloc might be changing the block size while GC_block_was_dirty
is examining it.  The change to the size field is benign, in that
GC_block_was_dirty would work correctly with either value, since we
are not changing the number of objects in the block.  But seeing
a half-updated value (though unlikely to occur in practice) could be
probably bad.  Using unordered atomic accesses on the size field
should solve the issue.

* mark.c [!GC_DISABLE_INCREMENTAL && AO_HAVE_load]
(GC_block_was_dirty): Use AO_load() to get hhdr->hb_sz value; add
comment.

1 files changed, 7 insertions, 1 deletions

diff --git a/mark.c b/mark.c
index 9f39af5c..f33ec90c 100644
--- a/mark.c
+++ b/mark.c
@@ -1962,9 +1962,15 @@ STATIC void GC_push_marked(struct hblk *h, hdr *hhdr)
   /* Test whether any page in the given block is dirty.   */
   STATIC GC_bool GC_block_was_dirty(struct hblk *h, hdr *hhdr)
   {
-    word sz = hhdr -> hb_sz;
+    word sz;
     ptr_t p;
 
+#   ifdef AO_HAVE_load
+      /* Atomic access is used to avoid racing with GC_realloc. */
+      sz = (word)AO_load((volatile AO_t *)&(hhdr -> hb_sz));
+#   else
+      sz = hhdr -> hb_sz;
+#   endif
     if (sz <= MAXOBJBYTES) {
       return GC_page_was_dirty(h);
     }