net: refine dev_put()/dev_hold() debugging

We are still chasing some syzbot reports where we think a rogue dev_put()
is called with no corresponding prior dev_hold().
Unfortunately it eats a reference on dev->dev_refcnt taken by innocent
dev_hold_track(), meaning that the refcount saturation splat comes
too late to be useful.

Make sure that 'not tracked' dev_put() and dev_hold() better use
CONFIG_NET_DEV_REFCNT_TRACKER=y debug infrastructure:

Prior patch in the series allowed ref_tracker_alloc() and ref_tracker_free()
to be called with a NULL @trackerp parameter, and to use a separate refcount
only to detect too many put() even in the following case:

dev_hold_track(dev, tracker_1, GFP_ATOMIC);
 dev_hold(dev);
 dev_put(dev);
 dev_put(dev); // Should complain loudly here.
dev_put_track(dev, tracker_1); // instead of here

Add clarification about netdev_tracker_alloc() role.

v2: I replaced the dev_put() in linkwatch_do_dev()
    with __dev_put() because callers called netdev_tracker_free().

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

3 files changed, 50 insertions, 27 deletions

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index e490b84732d1..3fb6fb67ed77 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3817,14 +3817,7 @@ extern unsigned int	netdev_budget_usecs;
 /* Called by rtnetlink.c:rtnl_unlock() */
 void netdev_run_todo(void);
 
-/**
- *	dev_put - release reference to device
- *	@dev: network device
- *
- * Release reference to device to allow it to be freed.
- * Try using dev_put_track() instead.
- */
-static inline void dev_put(struct net_device *dev)
+static inline void __dev_put(struct net_device *dev)
 {
 	if (dev) {
 #ifdef CONFIG_PCPU_DEV_REFCNT
@@ -3835,14 +3828,7 @@ static inline void dev_put(struct net_device *dev)
 	}
 }
 
-/**
- *	dev_hold - get reference to device
- *	@dev: network device
- *
- * Hold reference to device to keep it from being freed.
- * Try using dev_hold_track() instead.
- */
-static inline void dev_hold(struct net_device *dev)
+static inline void __dev_hold(struct net_device *dev)
 {
 	if (dev) {
 #ifdef CONFIG_PCPU_DEV_REFCNT
@@ -3853,11 +3839,24 @@ static inline void dev_hold(struct net_device *dev)
 	}
 }
 
+static inline void __netdev_tracker_alloc(struct net_device *dev,
+					  netdevice_tracker *tracker,
+					  gfp_t gfp)
+{
+#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
+	ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
+#endif
+}
+
+/* netdev_tracker_alloc() can upgrade a prior untracked reference
+ * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
+ */
 static inline void netdev_tracker_alloc(struct net_device *dev,
 					netdevice_tracker *tracker, gfp_t gfp)
 {
 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
-	ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
+	refcount_dec(&dev->refcnt_tracker.no_tracker);
+	__netdev_tracker_alloc(dev, tracker, gfp);
 #endif
 }
 
@@ -3873,8 +3872,8 @@ static inline void dev_hold_track(struct net_device *dev,
 				  netdevice_tracker *tracker, gfp_t gfp)
 {
 	if (dev) {
-		dev_hold(dev);
-		netdev_tracker_alloc(dev, tracker, gfp);
+		__dev_hold(dev);
+		__netdev_tracker_alloc(dev, tracker, gfp);
 	}
 }
 
@@ -3883,10 +3882,34 @@ static inline void dev_put_track(struct net_device *dev,
 {
 	if (dev) {
 		netdev_tracker_free(dev, tracker);
-		dev_put(dev);
+		__dev_put(dev);
 	}
 }
 
+/**
+ *	dev_hold - get reference to device
+ *	@dev: network device
+ *
+ * Hold reference to device to keep it from being freed.
+ * Try using dev_hold_track() instead.
+ */
+static inline void dev_hold(struct net_device *dev)
+{
+	dev_hold_track(dev, NULL, GFP_ATOMIC);
+}
+
+/**
+ *	dev_put - release reference to device
+ *	@dev: network device
+ *
+ * Release reference to device to allow it to be freed.
+ * Try using dev_put_track() instead.
+ */
+static inline void dev_put(struct net_device *dev)
+{
+	dev_put_track(dev, NULL);
+}
+
 static inline void dev_replace_track(struct net_device *odev,
 				     struct net_device *ndev,
 				     netdevice_tracker *tracker,
@@ -3895,11 +3918,11 @@ static inline void dev_replace_track(struct net_device *odev,
 	if (odev)
 		netdev_tracker_free(odev, tracker);
 
-	dev_hold(ndev);
-	dev_put(odev);
+	__dev_hold(ndev);
+	__dev_put(odev);
 
 	if (ndev)
-		netdev_tracker_alloc(ndev, tracker, gfp);
+		__netdev_tracker_alloc(ndev, tracker, gfp);
 }
 
 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
diff --git a/net/core/dev.c b/net/core/dev.c
index f79744d99413..1eaa0b88e3ba 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -10172,7 +10172,7 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 	dev->pcpu_refcnt = alloc_percpu(int);
 	if (!dev->pcpu_refcnt)
 		goto free_dev;
-	dev_hold(dev);
+	__dev_hold(dev);
 #else
 	refcount_set(&dev->dev_refcnt, 1);
 #endif
diff --git a/net/core/link_watch.c b/net/core/link_watch.c
index b0f5344d1185..95098d1a49bd 100644
--- a/net/core/link_watch.c
+++ b/net/core/link_watch.c
@@ -166,10 +166,10 @@ static void linkwatch_do_dev(struct net_device *dev)
 
 		netdev_state_change(dev);
 	}
-	/* Note: our callers are responsible for
-	 * calling netdev_tracker_free().
+	/* Note: our callers are responsible for calling netdev_tracker_free().
+	 * This is the reason we use __dev_put() instead of dev_put().
 	 */
-	dev_put(dev);
+	__dev_put(dev);
 }
 
 static void __linkwatch_run_queue(int urgent_only)