kernel: Add assertions to ensure vm_page_reservation is destroyed with a count of 0.

Inspired by some of jpelczar's changes.

headers/private/kernel/vm/vm_page.h

@@ -20,6 +20,14 @@ extern int32 gMappedPagesCount;
 
 struct vm_page_reservation {
 	uint32	count;
+
+#if KDEBUG && defined(__cplusplus)
+	vm_page_reservation() : count(0) {}
+	~vm_page_reservation()
+	{
+		ASSERT(count == 0);
+	}
+#endif
 };
 
 

src/system/kernel/vm/vm_page.cpp

@@ -3990,8 +3990,10 @@ vm_page_allocate_page_run(uint32 flags, page_num_t length,
 
 		if (foundRun) {
 			i = allocate_page_run(start, length, flags, freeClearQueueLocker);
-			if (i == length)
+			if (i == length) {
+				reservation.count = 0;
 				return &sPages[start];
+			}
 
 			// apparently a cached page couldn't be allocated -- skip it and
 			// continue