Merge branch 'dev_base_lock-remove'

Eric Dumazet says:

====================
net: complete dev_base_lock removal

Back in 2009 we started an effort to get rid of dev_base_lock
in favor of RCU.

It is time to finish this work.

Say goodbye to dev_base_lock !

v4: rebase, and move dev_addr_sem to net/core/dev.h in patch 06/13 (Jakub)

v3: I misread kbot reports, the issue was with dev->operstate (patch 10/13)
    So dev->reg_state is back to u8, and dev->operstate becomes an u32.
    Sorry for the noise.

v2: dev->reg_state must be a standard enum, some arches
    do not support cmpxchg() on u8.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/networking/netdevices.rst

@@ -252,8 +252,8 @@ ndo_eth_ioctl:
 	Context: process
 
 ndo_get_stats:
-	Synchronization: rtnl_lock() semaphore, dev_base_lock rwlock, or RCU.
+	Synchronization: rtnl_lock() semaphore, or RCU.
-	Context: atomic (can't sleep under rwlock or RCU)
+	Context: atomic (can't sleep under RCU)
 
 ndo_start_xmit:
 	Synchronization: __netif_tx_lock spinlock.

drivers/net/ethernet/cisco/enic/enic_main.c

@@ -872,7 +872,7 @@ error:
 	return NETDEV_TX_OK;
 }
 
-/* dev_base_lock rwlock held, nominally process context */
+/* rcu_read_lock potentially held, nominally process context */
 static void enic_get_stats(struct net_device *netdev,
 			   struct rtnl_link_stats64 *net_stats)
 {

drivers/net/ethernet/nvidia/forcedeth.c

@@ -1761,7 +1761,7 @@ static void nv_get_stats(int cpu, struct fe_priv *np,
 /*
  * nv_get_stats64: dev->ndo_get_stats64 function
  * Get latest stats value from the nic.
- * Called with read_lock(&dev_base_lock) held for read -
+ * Called with rcu_read_lock() held -
  * only synchronized against unregister_netdevice.
  */
 static void
@@ -3090,7 +3090,7 @@ static void set_bufsize(struct net_device *dev)
 
 /*
  * nv_change_mtu: dev->change_mtu function
- * Called with dev_base_lock held for read.
+ * Called with RTNL held for read.
  */
 static int nv_change_mtu(struct net_device *dev, int new_mtu)
 {

drivers/net/ethernet/sfc/efx_common.c

@@ -595,7 +595,7 @@ void efx_stop_all(struct efx_nic *efx)
 	efx_stop_datapath(efx);
 }
 
-/* Context: process, dev_base_lock or RTNL held, non-blocking. */
+/* Context: process, rcu_read_lock or RTNL held, non-blocking. */
 void efx_net_stats(struct net_device *net_dev, struct rtnl_link_stats64 *stats)
 {
 	struct efx_nic *efx = efx_netdev_priv(net_dev);

drivers/net/ethernet/sfc/falcon/efx.c

@@ -2085,7 +2085,7 @@ int ef4_net_stop(struct net_device *net_dev)
 	return 0;
 }
 
-/* Context: process, dev_base_lock or RTNL held, non-blocking. */
+/* Context: process, rcu_read_lock or RTNL held, non-blocking. */
 static void ef4_net_stats(struct net_device *net_dev,
 			  struct rtnl_link_stats64 *stats)
 {

drivers/net/ethernet/sfc/siena/efx_common.c

@@ -605,7 +605,7 @@ static size_t efx_siena_update_stats_atomic(struct efx_nic *efx, u64 *full_stats
 	return efx->type->update_stats(efx, full_stats, core_stats);
 }
 
-/* Context: process, dev_base_lock or RTNL held, non-blocking. */
+/* Context: process, rcu_read_lock or RTNL held, non-blocking. */
 void efx_siena_net_stats(struct net_device *net_dev,
 			 struct rtnl_link_stats64 *stats)
 {

include/linux/netdevice.h

@@ -1815,6 +1815,15 @@ enum netdev_stat_type {
 	NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
 };
 
+enum netdev_reg_state {
+	NETREG_UNINITIALIZED = 0,
+	NETREG_REGISTERED,	/* completed register_netdevice */
+	NETREG_UNREGISTERING,	/* called unregister_netdevice */
+	NETREG_UNREGISTERED,	/* completed unregister todo */
+	NETREG_RELEASED,	/* called free_netdev */
+	NETREG_DUMMY,		/* dummy device for NAPI poll */
+};
+
 /**
  *	struct net_device - The DEVICE structure.
  *
@@ -2249,7 +2258,7 @@ struct net_device {
 	const struct tlsdev_ops *tlsdev_ops;
 #endif
 
-	unsigned char		operstate;
+	unsigned int		operstate;
 	unsigned char		link_mode;
 
 	unsigned char		if_port;
@@ -2372,13 +2381,7 @@ struct net_device {
 
 	struct list_head	link_watch_list;
 
-	enum { NETREG_UNINITIALIZED=0,
+	u8 reg_state;
-	       NETREG_REGISTERED,	/* completed register_netdevice */
-	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
-	       NETREG_UNREGISTERED,	/* completed unregister todo */
-	       NETREG_RELEASED,		/* called free_netdev */
-	       NETREG_DUMMY,		/* dummy device for NAPI poll */
-	} reg_state:8;
 
 	bool dismantle;
 
@@ -3074,8 +3077,6 @@ int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
 int call_netdevice_notifiers_info(unsigned long val,
 				  struct netdev_notifier_info *info);
 
-extern rwlock_t				dev_base_lock;		/* Device list lock */
-
 #define for_each_netdev(net, d)		\
 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
 #define for_each_netdev_reverse(net, d)	\
@@ -5254,7 +5255,9 @@ static inline const char *netdev_name(const struct net_device *dev)
 
 static inline const char *netdev_reg_state(const struct net_device *dev)
 {
-	switch (dev->reg_state) {
+	u8 reg_state = READ_ONCE(dev->reg_state);
+
+	switch (reg_state) {
 	case NETREG_UNINITIALIZED: return " (uninitialized)";
 	case NETREG_REGISTERED: return "";
 	case NETREG_UNREGISTERING: return " (unregistering)";
@@ -5263,7 +5266,7 @@ static inline const char *netdev_reg_state(const struct net_device *dev)
 	case NETREG_DUMMY: return " (dummy)";
 	}
 
-	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
+	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
 	return " (unknown)";
 }
 

include/linux/rtnetlink.h

@@ -172,4 +172,6 @@ rtnl_notify_needed(const struct net *net, u16 nlflags, u32 group)
 	return (nlflags & NLM_F_ECHO) || rtnl_has_listeners(net, group);
 }
 
+void netdev_set_operstate(struct net_device *dev, int newstate);
+
 #endif	/* __LINUX_RTNETLINK_H */

net/bridge/br_netlink.c

@@ -455,7 +455,8 @@ static int br_fill_ifinfo(struct sk_buff *skb,
 			  u32 filter_mask, const struct net_device *dev,
 			  bool getlink)
 {
-	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
+	u8 operstate = netif_running(dev) ? READ_ONCE(dev->operstate) :
+					    IF_OPER_DOWN;
 	struct nlattr *af = NULL;
 	struct net_bridge *br;
 	struct ifinfomsg *hdr;

net/core/dev.c

@@ -168,28 +168,6 @@ static int call_netdevice_notifiers_extack(unsigned long val,
 					   struct net_device *dev,
 					   struct netlink_ext_ack *extack);
 
-/*
- * The @dev_base_head list is protected by @dev_base_lock and the rtnl
- * semaphore.
- *
- * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
- *
- * Writers must hold the rtnl semaphore while they loop through the
- * dev_base_head list, and hold dev_base_lock for writing when they do the
- * actual updates.  This allows pure readers to access the list even
- * while a writer is preparing to update it.
- *
- * To put it another way, dev_base_lock is held for writing only to
- * protect against pure readers; the rtnl semaphore provides the
- * protection against other writers.
- *
- * See, for example usages, register_netdevice() and
- * unregister_netdevice(), which must be called with the rtnl
- * semaphore held.
- */
-DEFINE_RWLOCK(dev_base_lock);
-EXPORT_SYMBOL(dev_base_lock);
-
 static DEFINE_MUTEX(ifalias_mutex);
 
 /* protects napi_hash addition/deletion and napi_gen_id */
@@ -395,12 +373,10 @@ static void list_netdevice(struct net_device *dev)
 
 	ASSERT_RTNL();
 
-	write_lock(&dev_base_lock);
 	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
 	netdev_name_node_add(net, dev->name_node);
 	hlist_add_head_rcu(&dev->index_hlist,
 			   dev_index_hash(net, dev->ifindex));
-	write_unlock(&dev_base_lock);
 
 	netdev_for_each_altname(dev, name_node)
 		netdev_name_node_add(net, name_node);
@@ -414,7 +390,7 @@ static void list_netdevice(struct net_device *dev)
 /* Device list removal
  * caller must respect a RCU grace period before freeing/reusing dev
  */
-static void unlist_netdevice(struct net_device *dev, bool lock)
+static void unlist_netdevice(struct net_device *dev)
 {
 	struct netdev_name_node *name_node;
 	struct net *net = dev_net(dev);
@@ -427,13 +403,9 @@ static void unlist_netdevice(struct net_device *dev, bool lock)
 		netdev_name_node_del(name_node);
 
 	/* Unlink dev from the device chain */
-	if (lock)
-		write_lock(&dev_base_lock);
 	list_del_rcu(&dev->dev_list);
 	netdev_name_node_del(dev->name_node);
 	hlist_del_rcu(&dev->index_hlist);
-	if (lock)
-		write_unlock(&dev_base_lock);
 
 	dev_base_seq_inc(dev_net(dev));
 }
@@ -754,9 +726,9 @@ EXPORT_SYMBOL_GPL(dev_fill_forward_path);
  *	@net: the applicable net namespace
  *	@name: name to find
  *
- *	Find an interface by name. Must be called under RTNL semaphore
+ *	Find an interface by name. Must be called under RTNL semaphore.
- *	or @dev_base_lock. If the name is found a pointer to the device
+ *	If the name is found a pointer to the device is returned.
- *	is returned. If the name is not found then %NULL is returned. The
+ *	If the name is not found then %NULL is returned. The
  *	reference counters are not incremented so the caller must be
  *	careful with locks.
  */
@@ -837,8 +809,7 @@ EXPORT_SYMBOL(netdev_get_by_name);
  *	Search for an interface by index. Returns %NULL if the device
  *	is not found or a pointer to the device. The device has not
  *	had its reference counter increased so the caller must be careful
- *	about locking. The caller must hold either the RTNL semaphore
+ *	about locking. The caller must hold the RTNL semaphore.
- *	or @dev_base_lock.
  */
 
 struct net_device *__dev_get_by_index(struct net *net, int ifindex)
@@ -1228,13 +1199,13 @@ int dev_change_name(struct net_device *dev, const char *newname)
 			    dev->flags & IFF_UP ? " (while UP)" : "");
 
 	old_assign_type = dev->name_assign_type;
-	dev->name_assign_type = NET_NAME_RENAMED;
+	WRITE_ONCE(dev->name_assign_type, NET_NAME_RENAMED);
 
 rollback:
 	ret = device_rename(&dev->dev, dev->name);
 	if (ret) {
 		memcpy(dev->name, oldname, IFNAMSIZ);
-		dev->name_assign_type = old_assign_type;
+		WRITE_ONCE(dev->name_assign_type, old_assign_type);
 		up_write(&devnet_rename_sem);
 		return ret;
 	}
@@ -1243,15 +1214,11 @@ rollback:
 
 	netdev_adjacent_rename_links(dev, oldname);
 
-	write_lock(&dev_base_lock);
 	netdev_name_node_del(dev->name_node);
-	write_unlock(&dev_base_lock);
 
 	synchronize_net();
 
-	write_lock(&dev_base_lock);
 	netdev_name_node_add(net, dev->name_node);
-	write_unlock(&dev_base_lock);
 
 	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
 	ret = notifier_to_errno(ret);
@@ -1263,7 +1230,7 @@ rollback:
 			down_write(&devnet_rename_sem);
 			memcpy(dev->name, oldname, IFNAMSIZ);
 			memcpy(oldname, newname, IFNAMSIZ);
-			dev->name_assign_type = old_assign_type;
+			WRITE_ONCE(dev->name_assign_type, old_assign_type);
 			old_assign_type = NET_NAME_RENAMED;
 			goto rollback;
 		} else {
@@ -8993,7 +8960,7 @@ int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
 }
 EXPORT_SYMBOL(dev_set_mac_address);
 
-static DECLARE_RWSEM(dev_addr_sem);
+DECLARE_RWSEM(dev_addr_sem);
 
 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
 			     struct netlink_ext_ack *extack)
@@ -10338,9 +10305,9 @@ int register_netdevice(struct net_device *dev)
 		goto err_ifindex_release;
 
 	ret = netdev_register_kobject(dev);
-	write_lock(&dev_base_lock);
+
-	dev->reg_state = ret ? NETREG_UNREGISTERED : NETREG_REGISTERED;
+	WRITE_ONCE(dev->reg_state, ret ? NETREG_UNREGISTERED : NETREG_REGISTERED);
-	write_unlock(&dev_base_lock);
+
 	if (ret)
 		goto err_uninit_notify;
 
@@ -10629,9 +10596,7 @@ void netdev_run_todo(void)
 			continue;
 		}
 
-		write_lock(&dev_base_lock);
+		WRITE_ONCE(dev->reg_state, NETREG_UNREGISTERED);
-		dev->reg_state = NETREG_UNREGISTERED;
-		write_unlock(&dev_base_lock);
 		linkwatch_sync_dev(dev);
 	}
 
@@ -11050,7 +11015,7 @@ void free_netdev(struct net_device *dev)
 	}
 
 	BUG_ON(dev->reg_state != NETREG_UNREGISTERED);
-	dev->reg_state = NETREG_RELEASED;
+	WRITE_ONCE(dev->reg_state, NETREG_RELEASED);
 
 	/* will free via device release */
 	put_device(&dev->dev);
@@ -11138,10 +11103,8 @@ void unregister_netdevice_many_notify(struct list_head *head,
 
 	list_for_each_entry(dev, head, unreg_list) {
 		/* And unlink it from device chain. */
-		write_lock(&dev_base_lock);
+		unlist_netdevice(dev);
-		unlist_netdevice(dev, false);
+		WRITE_ONCE(dev->reg_state, NETREG_UNREGISTERING);
-		dev->reg_state = NETREG_UNREGISTERING;
-		write_unlock(&dev_base_lock);
 	}
 	flush_all_backlogs();
 
@@ -11323,7 +11286,7 @@ int __dev_change_net_namespace(struct net_device *dev, struct net *net,
 	dev_close(dev);
 
 	/* And unlink it from device chain */
-	unlist_netdevice(dev, true);
+	unlist_netdevice(dev);
 
 	synchronize_net();
 

net/core/dev.h

@@ -3,6 +3,7 @@
 #define _NET_CORE_DEV_H
 
 #include <linux/types.h>
+#include <linux/rwsem.h>
 
 struct net;
 struct net_device;
@@ -46,6 +47,8 @@ extern int		weight_p;
 extern int		dev_weight_rx_bias;
 extern int		dev_weight_tx_bias;
 
+extern struct rw_semaphore dev_addr_sem;
+
 /* rtnl helpers */
 extern struct list_head net_todo_list;
 void netdev_run_todo(void);

net/core/link_watch.c

@@ -33,7 +33,7 @@ static DECLARE_DELAYED_WORK(linkwatch_work, linkwatch_event);
 static LIST_HEAD(lweventlist);
 static DEFINE_SPINLOCK(lweventlist_lock);
 
-static unsigned char default_operstate(const struct net_device *dev)
+static unsigned int default_operstate(const struct net_device *dev)
 {
 	if (netif_testing(dev))
 		return IF_OPER_TESTING;
@@ -62,16 +62,13 @@ static unsigned char default_operstate(const struct net_device *dev)
 	return IF_OPER_UP;
 }
 
-
 static void rfc2863_policy(struct net_device *dev)
 {
-	unsigned char operstate = default_operstate(dev);
+	unsigned int operstate = default_operstate(dev);
 
-	if (operstate == dev->operstate)
+	if (operstate == READ_ONCE(dev->operstate))
 		return;
 
-	write_lock(&dev_base_lock);
-
 	switch(dev->link_mode) {
 	case IF_LINK_MODE_TESTING:
 		if (operstate == IF_OPER_UP)
@@ -87,9 +84,7 @@ static void rfc2863_policy(struct net_device *dev)
 		break;
 	}
 
-	dev->operstate = operstate;
+	WRITE_ONCE(dev->operstate, operstate);
-
-	write_unlock(&dev_base_lock);
 }
 
 

net/core/net-sysfs.c

@@ -34,10 +34,10 @@ static const char fmt_dec[] = "%d\n";
 static const char fmt_ulong[] = "%lu\n";
 static const char fmt_u64[] = "%llu\n";
 
-/* Caller holds RTNL or dev_base_lock */
+/* Caller holds RTNL or RCU */
 static inline int dev_isalive(const struct net_device *dev)
 {
-	return dev->reg_state <= NETREG_REGISTERED;
+	return READ_ONCE(dev->reg_state) <= NETREG_REGISTERED;
 }
 
 /* use same locking rules as GIF* ioctl's */
@@ -48,10 +48,10 @@ static ssize_t netdev_show(const struct device *dev,
 	struct net_device *ndev = to_net_dev(dev);
 	ssize_t ret = -EINVAL;
 
-	read_lock(&dev_base_lock);
+	rcu_read_lock();
 	if (dev_isalive(ndev))
 		ret = (*format)(ndev, buf);
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 
 	return ret;
 }
@@ -60,7 +60,7 @@ static ssize_t netdev_show(const struct device *dev,
 #define NETDEVICE_SHOW(field, format_string)				\
 static ssize_t format_##field(const struct net_device *dev, char *buf)	\
 {									\
-	return sysfs_emit(buf, format_string, dev->field);		\
+	return sysfs_emit(buf, format_string, READ_ONCE(dev->field));		\
 }									\
 static ssize_t field##_show(struct device *dev,				\
 			    struct device_attribute *attr, char *buf)	\
@@ -125,7 +125,7 @@ static DEVICE_ATTR_RO(iflink);
 
 static ssize_t format_name_assign_type(const struct net_device *dev, char *buf)
 {
-	return sysfs_emit(buf, fmt_dec, dev->name_assign_type);
+	return sysfs_emit(buf, fmt_dec, READ_ONCE(dev->name_assign_type));
 }
 
 static ssize_t name_assign_type_show(struct device *dev,
@@ -135,24 +135,28 @@ static ssize_t name_assign_type_show(struct device *dev,
 	struct net_device *ndev = to_net_dev(dev);
 	ssize_t ret = -EINVAL;
 
-	if (ndev->name_assign_type != NET_NAME_UNKNOWN)
+	if (READ_ONCE(ndev->name_assign_type) != NET_NAME_UNKNOWN)
 		ret = netdev_show(dev, attr, buf, format_name_assign_type);
 
 	return ret;
 }
 static DEVICE_ATTR_RO(name_assign_type);
 
-/* use same locking rules as GIFHWADDR ioctl's */
+/* use same locking rules as GIFHWADDR ioctl's (dev_get_mac_address()) */
 static ssize_t address_show(struct device *dev, struct device_attribute *attr,
 			    char *buf)
 {
 	struct net_device *ndev = to_net_dev(dev);
 	ssize_t ret = -EINVAL;
 
-	read_lock(&dev_base_lock);
+	down_read(&dev_addr_sem);
+
+	rcu_read_lock();
 	if (dev_isalive(ndev))
 		ret = sysfs_format_mac(buf, ndev->dev_addr, ndev->addr_len);
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
+
+	up_read(&dev_addr_sem);
 	return ret;
 }
 static DEVICE_ATTR_RO(address);
@@ -161,10 +165,13 @@ static ssize_t broadcast_show(struct device *dev,
 			      struct device_attribute *attr, char *buf)
 {
 	struct net_device *ndev = to_net_dev(dev);
+	int ret = -EINVAL;
 
+	rcu_read_lock();
 	if (dev_isalive(ndev))
-		return sysfs_format_mac(buf, ndev->broadcast, ndev->addr_len);
+		ret = sysfs_format_mac(buf, ndev->broadcast, ndev->addr_len);
-	return -EINVAL;
+	rcu_read_unlock();
+	return ret;
 }
 static DEVICE_ATTR_RO(broadcast);
 
@@ -318,11 +325,9 @@ static ssize_t operstate_show(struct device *dev,
 	const struct net_device *netdev = to_net_dev(dev);
 	unsigned char operstate;
 
-	read_lock(&dev_base_lock);
+	operstate = READ_ONCE(netdev->operstate);
-	operstate = netdev->operstate;
 	if (!netif_running(netdev))
 		operstate = IF_OPER_DOWN;
-	read_unlock(&dev_base_lock);
 
 	if (operstate >= ARRAY_SIZE(operstates))
 		return -EINVAL; /* should not happen */
@@ -680,14 +685,14 @@ static ssize_t netstat_show(const struct device *d,
 	WARN_ON(offset > sizeof(struct rtnl_link_stats64) ||
 		offset % sizeof(u64) != 0);
 
-	read_lock(&dev_base_lock);
+	rcu_read_lock();
 	if (dev_isalive(dev)) {
 		struct rtnl_link_stats64 temp;
 		const struct rtnl_link_stats64 *stats = dev_get_stats(dev, &temp);
 
 		ret = sysfs_emit(buf, fmt_u64, *(u64 *)(((u8 *)stats) + offset));
 	}
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 	return ret;
 }
 

net/core/rtnetlink.c

@@ -842,9 +842,22 @@ int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 }
 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 
+void netdev_set_operstate(struct net_device *dev, int newstate)
+{
+	unsigned int old = READ_ONCE(dev->operstate);
+
+	do {
+		if (old == newstate)
+			return;
+	} while (!try_cmpxchg(&dev->operstate, &old, newstate));
+
+	netdev_state_change(dev);
+}
+EXPORT_SYMBOL(netdev_set_operstate);
+
 static void set_operstate(struct net_device *dev, unsigned char transition)
 {
-	unsigned char operstate = dev->operstate;
+	unsigned char operstate = READ_ONCE(dev->operstate);
 
 	switch (transition) {
 	case IF_OPER_UP:
@@ -866,12 +879,7 @@ static void set_operstate(struct net_device *dev, unsigned char transition)
 		break;
 	}
 
-	if (dev->operstate != operstate) {
+	netdev_set_operstate(dev, operstate);
-		write_lock(&dev_base_lock);
-		dev->operstate = operstate;
-		write_unlock(&dev_base_lock);
-		netdev_state_change(dev);
-	}
 }
 
 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
@@ -2953,11 +2961,9 @@ static int do_setlink(const struct sk_buff *skb,
 	if (tb[IFLA_LINKMODE]) {
 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
 
-		write_lock(&dev_base_lock);
 		if (dev->link_mode ^ value)
 			status |= DO_SETLINK_NOTIFY;
-		dev->link_mode = value;
+		WRITE_ONCE(dev->link_mode, value);
-		write_unlock(&dev_base_lock);
 	}
 
 	if (tb[IFLA_VFINFO_LIST]) {

net/hsr/hsr_device.c

@@ -28,29 +28,19 @@ static bool is_slave_up(struct net_device *dev)
 	return dev && is_admin_up(dev) && netif_oper_up(dev);
 }
 
-static void __hsr_set_operstate(struct net_device *dev, int transition)
-{
-	write_lock(&dev_base_lock);
-	if (dev->operstate != transition) {
-		dev->operstate = transition;
-		write_unlock(&dev_base_lock);
-		netdev_state_change(dev);
-	} else {
-		write_unlock(&dev_base_lock);
-	}
-}
-
 static void hsr_set_operstate(struct hsr_port *master, bool has_carrier)
 {
-	if (!is_admin_up(master->dev)) {
+	struct net_device *dev = master->dev;
-		__hsr_set_operstate(master->dev, IF_OPER_DOWN);
+
+	if (!is_admin_up(dev)) {
+		netdev_set_operstate(dev, IF_OPER_DOWN);
 		return;
 	}
 
 	if (has_carrier)
-		__hsr_set_operstate(master->dev, IF_OPER_UP);
+		netdev_set_operstate(dev, IF_OPER_UP);
 	else
-		__hsr_set_operstate(master->dev, IF_OPER_LOWERLAYERDOWN);
+		netdev_set_operstate(dev, IF_OPER_LOWERLAYERDOWN);
 }
 
 static bool hsr_check_carrier(struct hsr_port *master)
@@ -78,14 +68,14 @@ static void hsr_check_announce(struct net_device *hsr_dev,
 
 	hsr = netdev_priv(hsr_dev);
 
-	if (hsr_dev->operstate == IF_OPER_UP && old_operstate != IF_OPER_UP) {
+	if (READ_ONCE(hsr_dev->operstate) == IF_OPER_UP && old_operstate != IF_OPER_UP) {
 		/* Went up */
 		hsr->announce_count = 0;
 		mod_timer(&hsr->announce_timer,
 			  jiffies + msecs_to_jiffies(HSR_ANNOUNCE_INTERVAL));
 	}
 
-	if (hsr_dev->operstate != IF_OPER_UP && old_operstate == IF_OPER_UP)
+	if (READ_ONCE(hsr_dev->operstate) != IF_OPER_UP && old_operstate == IF_OPER_UP)
 		/* Went down */
 		del_timer(&hsr->announce_timer);
 }
@@ -100,7 +90,7 @@ void hsr_check_carrier_and_operstate(struct hsr_priv *hsr)
 	/* netif_stacked_transfer_operstate() cannot be used here since
 	 * it doesn't set IF_OPER_LOWERLAYERDOWN (?)
 	 */
-	old_operstate = master->dev->operstate;
+	old_operstate = READ_ONCE(master->dev->operstate);
 	has_carrier = hsr_check_carrier(master);
 	hsr_set_operstate(master, has_carrier);
 	hsr_check_announce(master->dev, old_operstate);

net/ipv4/ip_tunnel.c

@@ -102,10 +102,9 @@ struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
 		if (!ip_tunnel_key_match(&t->parms, flags, key))
 			continue;
 
-		if (t->parms.link == link)
+		if (READ_ONCE(t->parms.link) == link)
 			return t;
-		else
+		cand = t;
-			cand = t;
 	}
 
 	hlist_for_each_entry_rcu(t, head, hash_node) {
@@ -117,9 +116,9 @@ struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
 		if (!ip_tunnel_key_match(&t->parms, flags, key))
 			continue;
 
-		if (t->parms.link == link)
+		if (READ_ONCE(t->parms.link) == link)
 			return t;
-		else if (!cand)
+		if (!cand)
 			cand = t;
 	}
 
@@ -137,9 +136,9 @@ struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
 		if (!ip_tunnel_key_match(&t->parms, flags, key))
 			continue;
 
-		if (t->parms.link == link)
+		if (READ_ONCE(t->parms.link) == link)
 			return t;
-		else if (!cand)
+		if (!cand)
 			cand = t;
 	}
 
@@ -150,9 +149,9 @@ struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
 		    !(t->dev->flags & IFF_UP))
 			continue;
 
-		if (t->parms.link == link)
+		if (READ_ONCE(t->parms.link) == link)
 			return t;
-		else if (!cand)
+		if (!cand)
 			cand = t;
 	}
 
@@ -221,7 +220,7 @@ static struct ip_tunnel *ip_tunnel_find(struct ip_tunnel_net *itn,
 	hlist_for_each_entry_rcu(t, head, hash_node) {
 		if (local == t->parms.iph.saddr &&
 		    remote == t->parms.iph.daddr &&
-		    link == t->parms.link &&
+		    link == READ_ONCE(t->parms.link) &&
 		    type == t->dev->type &&
 		    ip_tunnel_key_match(&t->parms, flags, key))
 			break;
@@ -747,7 +746,7 @@ void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 
 	ip_tunnel_init_flow(&fl4, protocol, dst, tnl_params->saddr,
 			    tunnel->parms.o_key, RT_TOS(tos),
-			    dev_net(dev), tunnel->parms.link,
+			    dev_net(dev), READ_ONCE(tunnel->parms.link),
 			    tunnel->fwmark, skb_get_hash(skb), 0);
 
 	if (ip_tunnel_encap(skb, &tunnel->encap, &protocol, &fl4) < 0)
@@ -867,7 +866,7 @@ static void ip_tunnel_update(struct ip_tunnel_net *itn,
 	if (t->parms.link != p->link || t->fwmark != fwmark) {
 		int mtu;
 
-		t->parms.link = p->link;
+		WRITE_ONCE(t->parms.link, p->link);
 		t->fwmark = fwmark;
 		mtu = ip_tunnel_bind_dev(dev);
 		if (set_mtu)
@@ -1057,9 +1056,9 @@ EXPORT_SYMBOL(ip_tunnel_get_link_net);
 
 int ip_tunnel_get_iflink(const struct net_device *dev)
 {
-	struct ip_tunnel *tunnel = netdev_priv(dev);
+	const struct ip_tunnel *tunnel = netdev_priv(dev);
 
-	return tunnel->parms.link;
+	return READ_ONCE(tunnel->parms.link);
 }
 EXPORT_SYMBOL(ip_tunnel_get_iflink);
 

net/ipv6/addrconf.c

@@ -6025,7 +6025,7 @@ static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
 	    (dev->ifindex != dev_get_iflink(dev) &&
 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
 	    nla_put_u8(skb, IFLA_OPERSTATE,
-		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
+		       netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
 		goto nla_put_failure;
 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
 	if (!protoinfo)