plat/common/arm: Do not use a barrier on every invalidation

The current implementation of clean and invalidate by region uses
a barrier after every cache line. This is unnecessary and expensive.
Use a barrier once, at the end of the operation.

Signed-off-by: Michalis Pappas <michalis@unikraft.io>

plat/common/arm/cache64.S

@@ -60,17 +60,14 @@ ENTRY(clean_and_invalidate_dcache_range)
 1:
 	/* clean and invalidate D cache by D cache line size */
 	dc	civac, x0
-	dsb	nsh
 
 	/* Move to next line and reduce the size */
 	add	x0, x0, x3
 	subs	x1, x1, x3
 
 	/* Check if all range has been invalidated */
 	b.hi	1b
-
-	isb
-
+	dsb	sy
 	ret
 END(clean_and_invalidate_dcache_range)
 
@@ -99,8 +96,6 @@ ENTRY(invalidate_dcache_range)
 
 	/* Check if all range has been invalidated */
 	b.hi	1b
-
-	isb
-
+	dsb	sy
 	ret
 END(invalidate_dcache_range)