omap iommu: fix incorrect address for largepage 1st entry

There was a potential risk that the following "memset" could override
its range if a given address was not the 1st entry of a
largepage. This is not the case for "iovmm".
Reported-by: NHari Nagalla <hnagalla@ti.com>
Signed-off-by: NHiroshi DOYU <Hiroshi.DOYU@nokia.com>
Signed-off-by: NHari Nagalla <hnagalla@ti.com>
Signed-off-by: NTony Lindgren <tony@atomide.com>

arch/arm/plat-omap/iommu.c

 /*
  * omap iommu: tlb and pagetable primitives
  *
- * Copyright (C) 2008-2009 Nokia Corporation
+ * Copyright (C) 2008-2010 Nokia Corporation
  *
  * Written by Hiroshi DOYU <Hiroshi.DOYU@nokia.com>,
  *		Paul Mundt and Toshihiro Kobayashi
@@ -646,7 +646,7 @@ static size_t iopgtable_clear_entry_core(struct iommu *obj, u32 da)
 		if (*iopte & IOPTE_LARGE) {
 			nent *= 16;
 			/* rewind to the 1st entry */
-			iopte = (u32 *)((u32)iopte & IOLARGE_MASK);
+			iopte = iopte_offset(iopgd, (da & IOLARGE_MASK));
 		}
 		bytes *= nent;
 		memset(iopte, 0, nent * sizeof(*iopte));