mm: kmemleak: treat vm_struct as alternative reference to vmalloc'ed objects

Kmemleak requires that vmalloc'ed objects have a minimum reference count
of 2: one in the corresponding vm_struct object and the other owned by
the vmalloc() caller.  There are cases, however, where the original
vmalloc() returned pointer is lost and, instead, a pointer to vm_struct
is stored (see free_thread_stack()).  Kmemleak currently reports such
objects as leaks.

This patch adds support for treating any surplus references to an object
as additional references to a specified object.  It introduces the
kmemleak_vmalloc() API function which takes a vm_struct pointer and sets
its surplus reference passing to the actual vmalloc() returned pointer.
The __vmalloc_node_range() calling site has been modified accordingly.

Link: http://lkml.kernel.org/r/1495726937-23557-4-git-send-email-catalin.marinas@arm.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Reported-by: "Luis R. Rodriguez" <mcgrof@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: "Luis R. Rodriguez" <mcgrof@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
diff --git a/mm/kmemleak.c b/mm/kmemleak.c
index 266482f..7780cd8 100644
--- a/mm/kmemleak.c
+++ b/mm/kmemleak.c
@@ -159,6 +159,8 @@
 	atomic_t use_count;
 	unsigned long pointer;
 	size_t size;
+	/* pass surplus references to this pointer */
+	unsigned long excess_ref;
 	/* minimum number of a pointers found before it is considered leak */
 	int min_count;
 	/* the total number of pointers found pointing to this object */
@@ -253,7 +255,8 @@
 	KMEMLEAK_NOT_LEAK,
 	KMEMLEAK_IGNORE,
 	KMEMLEAK_SCAN_AREA,
-	KMEMLEAK_NO_SCAN
+	KMEMLEAK_NO_SCAN,
+	KMEMLEAK_SET_EXCESS_REF
 };
 
 /*
@@ -264,7 +267,10 @@
 	int op_type;			/* kmemleak operation type */
 	int min_count;			/* minimum reference count */
 	const void *ptr;		/* allocated/freed memory block */
-	size_t size;			/* memory block size */
+	union {
+		size_t size;		/* memory block size */
+		unsigned long excess_ref; /* surplus reference passing */
+	};
 	unsigned long trace[MAX_TRACE];	/* stack trace */
 	unsigned int trace_len;		/* stack trace length */
 };
@@ -562,6 +568,7 @@
 	object->flags = OBJECT_ALLOCATED;
 	object->pointer = ptr;
 	object->size = size;
+	object->excess_ref = 0;
 	object->min_count = min_count;
 	object->count = 0;			/* white color initially */
 	object->jiffies = jiffies;
@@ -795,6 +802,30 @@
 }
 
 /*
+ * Any surplus references (object already gray) to 'ptr' are passed to
+ * 'excess_ref'. This is used in the vmalloc() case where a pointer to
+ * vm_struct may be used as an alternative reference to the vmalloc'ed object
+ * (see free_thread_stack()).
+ */
+static void object_set_excess_ref(unsigned long ptr, unsigned long excess_ref)
+{
+	unsigned long flags;
+	struct kmemleak_object *object;
+
+	object = find_and_get_object(ptr, 0);
+	if (!object) {
+		kmemleak_warn("Setting excess_ref on unknown object at 0x%08lx\n",
+			      ptr);
+		return;
+	}
+
+	spin_lock_irqsave(&object->lock, flags);
+	object->excess_ref = excess_ref;
+	spin_unlock_irqrestore(&object->lock, flags);
+	put_object(object);
+}
+
+/*
  * Set the OBJECT_NO_SCAN flag for the object corresponding to the give
  * pointer. Such object will not be scanned by kmemleak but references to it
  * are searched.
@@ -908,7 +939,7 @@
  * @gfp:	kmalloc() flags used for kmemleak internal memory allocations
  *
  * This function is called from the kernel allocators when a new object
- * (memory block) is allocated (kmem_cache_alloc, kmalloc, vmalloc etc.).
+ * (memory block) is allocated (kmem_cache_alloc, kmalloc etc.).
  */
 void __ref kmemleak_alloc(const void *ptr, size_t size, int min_count,
 			  gfp_t gfp)
@@ -952,6 +983,36 @@
 EXPORT_SYMBOL_GPL(kmemleak_alloc_percpu);
 
 /**
+ * kmemleak_vmalloc - register a newly vmalloc'ed object
+ * @area:	pointer to vm_struct
+ * @size:	size of the object
+ * @gfp:	__vmalloc() flags used for kmemleak internal memory allocations
+ *
+ * This function is called from the vmalloc() kernel allocator when a new
+ * object (memory block) is allocated.
+ */
+void __ref kmemleak_vmalloc(const struct vm_struct *area, size_t size, gfp_t gfp)
+{
+	pr_debug("%s(0x%p, %zu)\n", __func__, area, size);
+
+	/*
+	 * A min_count = 2 is needed because vm_struct contains a reference to
+	 * the virtual address of the vmalloc'ed block.
+	 */
+	if (kmemleak_enabled) {
+		create_object((unsigned long)area->addr, size, 2, gfp);
+		object_set_excess_ref((unsigned long)area,
+				      (unsigned long)area->addr);
+	} else if (kmemleak_early_log) {
+		log_early(KMEMLEAK_ALLOC, area->addr, size, 2);
+		/* reusing early_log.size for storing area->addr */
+		log_early(KMEMLEAK_SET_EXCESS_REF,
+			  area, (unsigned long)area->addr, 0);
+	}
+}
+EXPORT_SYMBOL_GPL(kmemleak_vmalloc);
+
+/**
  * kmemleak_free - unregister a previously registered object
  * @ptr:	pointer to beginning of the object
  *
@@ -1248,6 +1309,7 @@
 	for (ptr = start; ptr < end; ptr++) {
 		struct kmemleak_object *object;
 		unsigned long pointer;
+		unsigned long excess_ref;
 
 		if (scan_should_stop())
 			break;
@@ -1283,8 +1345,27 @@
 		 * enclosed by scan_mutex.
 		 */
 		spin_lock_nested(&object->lock, SINGLE_DEPTH_NESTING);
-		update_refs(object);
+		/* only pass surplus references (object already gray) */
+		if (color_gray(object)) {
+			excess_ref = object->excess_ref;
+			/* no need for update_refs() if object already gray */
+		} else {
+			excess_ref = 0;
+			update_refs(object);
+		}
 		spin_unlock(&object->lock);
+
+		if (excess_ref) {
+			object = lookup_object(excess_ref, 0);
+			if (!object)
+				continue;
+			if (object == scanned)
+				/* circular reference, ignore */
+				continue;
+			spin_lock_nested(&object->lock, SINGLE_DEPTH_NESTING);
+			update_refs(object);
+			spin_unlock(&object->lock);
+		}
 	}
 	read_unlock_irqrestore(&kmemleak_lock, flags);
 }
@@ -1987,6 +2068,10 @@
 		case KMEMLEAK_NO_SCAN:
 			kmemleak_no_scan(log->ptr);
 			break;
+		case KMEMLEAK_SET_EXCESS_REF:
+			object_set_excess_ref((unsigned long)log->ptr,
+					      log->excess_ref);
+			break;
 		default:
 			kmemleak_warn("Unknown early log operation: %d\n",
 				      log->op_type);