iommu/omap: Fix -Woverflow warnings when compiling on 64-bit architectures

Although the OMAP IOMMU driver supports only ARMv7 (32-bit) platforms,
it can be compile tested for other architectures, including 64-bit ones.
In such case the warning appears:

       In file included from drivers/iommu/omap-iommu.c:33:0:
       drivers/iommu/omap-iommu.c: In function 'omap_iommu_iova_to_phys':
    >> drivers/iommu/omap-iopgtable.h:44:21: warning: large integer implicitly truncated to unsigned type [-Woverflow]
        #define IOPTE_MASK  (~(IOPTE_SIZE - 1))
                            ^
    >> drivers/iommu/omap-iommu.c:1641:41: note: in expansion of macro 'IOPTE_MASK'
           ret = omap_iommu_translate(*pte, da, IOPTE_MASK);
                                                ^~~~~~~~~~

Fix this by using architecture-depending types in omap_iommu_translate():
1. Pointer should be cast to unsigned long,
2. Virtual addresses should be cast to dma_addr_t.

On 32-bit this will be the same as original code (using u32).  On 64-bit
it should produce meaningful result, although it does not really matter.
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Joerg Roedel <jroedel@suse.de>

drivers/iommu/omap-iopgtable.h

@@ -63,7 +63,8 @@
  *
  * va to pa translation
  */
-static inline phys_addr_t omap_iommu_translate(u32 d, u32 va, u32 mask)
+static inline phys_addr_t omap_iommu_translate(unsigned long d, dma_addr_t va,
+					       dma_addr_t mask)
 {
 	return (d & mask) | (va & (~mask));
 }