iommu: Disable caching of IOVA rb-tree node

Cached rb-tree node is maintained in IOVA allocation/free code.
This node is updated whenever new IOVA allocation is requested or
and IOVA is freed. There is a bug is caching logic which is causing
some freed IOVAs not to be used as during allocation request, free
IOVAs are checked in rb-tree beginning from cached node.
Fix this by traversing whole rb-tree for free node from start during
each IOVA allocation request.

Bug 3697066

Change-Id: I9a3248fd2e39988aefdb54521641f37b22db4b78
Signed-off-by: Ashish Mhetre <[email protected]>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-4.9/+/2874969
Reviewed-by: Krishna Reddy <[email protected]>
Reviewed-by: Sachin Nikam <[email protected]>
GVS: Gerrit_Virtual_Submit <[email protected]>

drivers/iommu/iova.c

@@ -57,16 +57,7 @@ EXPORT_SYMBOL_GPL(init_iova_domain);
 static struct rb_node *
 __get_cached_rbnode(struct iova_domain *iovad, unsigned long *limit_pfn)
 {
-	if ((*limit_pfn != iovad->dma_32bit_pfn) ||
-		(iovad->cached32_node == NULL))
-		return rb_last(&iovad->rbroot);
-	else {
-		struct rb_node *prev_node = rb_prev(iovad->cached32_node);
-		struct iova *curr_iova =
-			container_of(iovad->cached32_node, struct iova, node);
-		*limit_pfn = curr_iova->pfn_lo - 1;
-		return prev_node;
-	}
+	return rb_last(&iovad->rbroot);
 }
 
 static void