Merge branch 'net-fix-issues-around-register_netdevice-failures'

Jakub Kicinski says:

====================
net: fix issues around register_netdevice() failures

This series attempts to clean up the life cycle of struct
net_device. Dave has added dev->needs_free_netdev in the
past to fix double frees, we can lean on that mechanism
a little more to fix remaining issues with register_netdevice().

This is the next chapter of the saga which already includes:
commit 0e0eee2465df ("net: correct error path in rtnl_newlink()")
commit e51fb152318e ("rtnetlink: fix a memory leak when ->newlink fails")
commit cf124db566e6 ("net: Fix inconsistent teardown and release of private netdev state.")
commit 93ee31f14f6f ("[NET]: Fix free_netdev on register_netdev failure.")
commit 814152a89ed5 ("net: fix memleak in register_netdevice()")
commit 10cc514f451a ("net: Fix null de-reference of device refcount")

The immediate problem which gets fixed here is that calling
free_netdev() right after unregister_netdevice() is illegal
because we need to release rtnl_lock first, to let the
unregistration finish. Note that unregister_netdevice() is
just a wrapper of unregister_netdevice_queue(), it only
does half of the job.

Where this limitation becomes most problematic is in failure
modes of register_netdevice(). There is a notifier call right
at the end of it, which lets other subsystems veto the entire
thing. At which point we should really go through a full
unregister_netdevice(), but we can't because callers may
go straight to free_netdev() after the failure, and that's
no bueno (see the previous paragraph).

This set makes free_netdev() more lenient, when device
is still being unregistered free_netdev() will simply set
dev->needs_free_netdev and let the unregister process do
the freeing.

With the free_netdev() problem out of the way failures in
register_netdevice() can make use of net_todo, again.
Users are still expected to call free_netdev() right after
failure but that will only set dev->needs_free_netdev.

To prevent the pathological case of:

 dev->needs_free_netdev = true;
 if (register_netdevice(dev)) {
   rtnl_unlock();
   free_netdev(dev);
 }

make register_netdevice()'s failure clear dev->needs_free_netdev.

Problems described above are only present with register_netdevice() /
unregister_netdevice(). We have two parallel APIs for registration
of devices:
 - those called outside rtnl_lock (register_netdev(), and
   unregister_netdev());
 - and those to be used under rtnl_lock - register_netdevice()
   and unregister_netdevice().
The former is trivial and has no problems. The alternative
approach to fix the latter would be to also separate the
freeing functions - i.e. add free_netdevice(). This has been
implemented (incl. converting all relevant calls in the tree)
but it feels a little unnecessary to put the burden of choosing
the right free_netdev{,ice}() call on the programmer when we
can "just do the right thing" by default.
====================

Link: https://lore.kernel.org/r/20210106184007.1821480-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

4 files changed, 187 insertions, 36 deletions

diff --git a/Documentation/networking/netdevices.rst b/Documentation/networking/netdevices.rst
index e65665c5ab50..17bdcb746dcf 100644
--- a/Documentation/networking/netdevices.rst
+++ b/Documentation/networking/netdevices.rst
@@ -10,18 +10,177 @@ Introduction
 The following is a random collection of documentation regarding
 network devices.
 
-struct net_device allocation rules
-==================================
+struct net_device lifetime rules
+================================
 Network device structures need to persist even after module is unloaded and
 must be allocated with alloc_netdev_mqs() and friends.
 If device has registered successfully, it will be freed on last use
-by free_netdev(). This is required to handle the pathologic case cleanly
-(example: rmmod mydriver </sys/class/net/myeth/mtu )
+by free_netdev(). This is required to handle the pathological case cleanly
+(example: ``rmmod mydriver </sys/class/net/myeth/mtu``)
 
-alloc_netdev_mqs()/alloc_netdev() reserve extra space for driver
+alloc_netdev_mqs() / alloc_netdev() reserve extra space for driver
 private data which gets freed when the network device is freed. If
 separately allocated data is attached to the network device
-(netdev_priv(dev)) then it is up to the module exit handler to free that.
+(netdev_priv()) then it is up to the module exit handler to free that.
+
+There are two groups of APIs for registering struct net_device.
+First group can be used in normal contexts where ``rtnl_lock`` is not already
+held: register_netdev(), unregister_netdev().
+Second group can be used when ``rtnl_lock`` is already held:
+register_netdevice(), unregister_netdevice(), free_netdevice().
+
+Simple drivers
+--------------
+
+Most drivers (especially device drivers) handle lifetime of struct net_device
+in context where ``rtnl_lock`` is not held (e.g. driver probe and remove paths).
+
+In that case the struct net_device registration is done using
+the register_netdev(), and unregister_netdev() functions:
+
+.. code-block:: c
+
+  int probe()
+  {
+    struct my_device_priv *priv;
+    int err;
+
+    dev = alloc_netdev_mqs(...);
+    if (!dev)
+      return -ENOMEM;
+    priv = netdev_priv(dev);
+
+    /* ... do all device setup before calling register_netdev() ...
+     */
+
+    err = register_netdev(dev);
+    if (err)
+      goto err_undo;
+
+    /* net_device is visible to the user! */
+
+  err_undo:
+    /* ... undo the device setup ... */
+    free_netdev(dev);
+    return err;
+  }
+
+  void remove()
+  {
+    unregister_netdev(dev);
+    free_netdev(dev);
+  }
+
+Note that after calling register_netdev() the device is visible in the system.
+Users can open it and start sending / receiving traffic immediately,
+or run any other callback, so all initialization must be done prior to
+registration.
+
+unregister_netdev() closes the device and waits for all users to be done
+with it. The memory of struct net_device itself may still be referenced
+by sysfs but all operations on that device will fail.
+
+free_netdev() can be called after unregister_netdev() returns on when
+register_netdev() failed.
+
+Device management under RTNL
+----------------------------
+
+Registering struct net_device while in context which already holds
+the ``rtnl_lock`` requires extra care. In those scenarios most drivers
+will want to make use of struct net_device's ``needs_free_netdev``
+and ``priv_destructor`` members for freeing of state.
+
+Example flow of netdev handling under ``rtnl_lock``:
+
+.. code-block:: c
+
+  static void my_setup(struct net_device *dev)
+  {
+    dev->needs_free_netdev = true;
+  }
+
+  static void my_destructor(struct net_device *dev)
+  {
+    some_obj_destroy(priv->obj);
+    some_uninit(priv);
+  }
+
+  int create_link()
+  {
+    struct my_device_priv *priv;
+    int err;
+
+    ASSERT_RTNL();
+
+    dev = alloc_netdev(sizeof(*priv), "net%d", NET_NAME_UNKNOWN, my_setup);
+    if (!dev)
+      return -ENOMEM;
+    priv = netdev_priv(dev);
+
+    /* Implicit constructor */
+    err = some_init(priv);
+    if (err)
+      goto err_free_dev;
+
+    priv->obj = some_obj_create();
+    if (!priv->obj) {
+      err = -ENOMEM;
+      goto err_some_uninit;
+    }
+    /* End of constructor, set the destructor: */
+    dev->priv_destructor = my_destructor;
+
+    err = register_netdevice(dev);
+    if (err)
+      /* register_netdevice() calls destructor on failure */
+      goto err_free_dev;
+
+    /* If anything fails now unregister_netdevice() (or unregister_netdev())
+     * will take care of calling my_destructor and free_netdev().
+     */
+
+    return 0;
+
+  err_some_uninit:
+    some_uninit(priv);
+  err_free_dev:
+    free_netdev(dev);
+    return err;
+  }
+
+If struct net_device.priv_destructor is set it will be called by the core
+some time after unregister_netdevice(), it will also be called if
+register_netdevice() fails. The callback may be invoked with or without
+``rtnl_lock`` held.
+
+There is no explicit constructor callback, driver "constructs" the private
+netdev state after allocating it and before registration.
+
+Setting struct net_device.needs_free_netdev makes core call free_netdevice()
+automatically after unregister_netdevice() when all references to the device
+are gone. It only takes effect after a successful call to register_netdevice()
+so if register_netdevice() fails driver is responsible for calling
+free_netdev().
+
+free_netdev() is safe to call on error paths right after unregister_netdevice()
+or when register_netdevice() fails. Parts of netdev (de)registration process
+happen after ``rtnl_lock`` is released, therefore in those cases free_netdev()
+will defer some of the processing until ``rtnl_lock`` is released.
+
+Devices spawned from struct rtnl_link_ops should never free the
+struct net_device directly.
+
+.ndo_init and .ndo_uninit
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
+``.ndo_init`` and ``.ndo_uninit`` callbacks are called during net_device
+registration and de-registration, under ``rtnl_lock``. Drivers can use
+those e.g. when parts of their init process need to run under ``rtnl_lock``.
+
+``.ndo_init`` runs before device is visible in the system, ``.ndo_uninit``
+runs during de-registering after device is closed but other subsystems
+may still have outstanding references to the netdevice.
 
 MTU
 ===
diff --git a/net/8021q/vlan.c b/net/8021q/vlan.c
index 15bbfaf943fd..8b644113715e 100644
--- a/net/8021q/vlan.c
+++ b/net/8021q/vlan.c
@@ -284,9 +284,7 @@ static int register_vlan_device(struct net_device *real_dev, u16 vlan_id)
 	return 0;
 
 out_free_newdev:
-	if (new_dev->reg_state == NETREG_UNINITIALIZED ||
-	    new_dev->reg_state == NETREG_UNREGISTERED)
-		free_netdev(new_dev);
+	free_netdev(new_dev);
 	return err;
 }
 
diff --git a/net/core/dev.c b/net/core/dev.c
index 8fa739259041..0071a11a6dc3 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -10077,17 +10077,11 @@ int register_netdevice(struct net_device *dev)
 	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
 	ret = notifier_to_errno(ret);
 	if (ret) {
+		/* Expect explicit free_netdev() on failure */
+		dev->needs_free_netdev = false;
 		rollback_registered(dev);
-		rcu_barrier();
-
-		dev->reg_state = NETREG_UNREGISTERED;
-		/* We should put the kobject that hold in
-		 * netdev_unregister_kobject(), otherwise
-		 * the net device cannot be freed when
-		 * driver calls free_netdev(), because the
-		 * kobject is being hold.
-		 */
-		kobject_put(&dev->dev.kobj);
+		net_set_todo(dev);
+		goto out;
 	}
 	/*
 	 *	Prevent userspace races by waiting until the network
@@ -10631,6 +10625,17 @@ void free_netdev(struct net_device *dev)
 	struct napi_struct *p, *n;
 
 	might_sleep();
+
+	/* When called immediately after register_netdevice() failed the unwind
+	 * handling may still be dismantling the device. Handle that case by
+	 * deferring the free.
+	 */
+	if (dev->reg_state == NETREG_UNREGISTERING) {
+		ASSERT_RTNL();
+		dev->needs_free_netdev = true;
+		return;
+	}
+
 	netif_free_tx_queues(dev);
 	netif_free_rx_queues(dev);
 
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index bb0596c41b3e..3d6ab194d0f5 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -3439,26 +3439,15 @@ replay:
 
 	dev->ifindex = ifm->ifi_index;
 
-	if (ops->newlink) {
+	if (ops->newlink)
 		err = ops->newlink(link_net ? : net, dev, tb, data, extack);
-		/* Drivers should call free_netdev() in ->destructor
-		 * and unregister it on failure after registration
-		 * so that device could be finally freed in rtnl_unlock.
-		 */
-		if (err < 0) {
-			/* If device is not registered at all, free it now */
-			if (dev->reg_state == NETREG_UNINITIALIZED ||
-			    dev->reg_state == NETREG_UNREGISTERED)
-				free_netdev(dev);
-			goto out;
-		}
-	} else {
+	else
 		err = register_netdevice(dev);
-		if (err < 0) {
-			free_netdev(dev);
-			goto out;
-		}
+	if (err < 0) {
+		free_netdev(dev);
+		goto out;
 	}
+
 	err = rtnl_configure_link(dev, ifm);
 	if (err < 0)
 		goto out_unregister;