Merge branch 'add-notifications-when-route-hardware-flags-change'

Ido Schimmel says:

====================
Add notifications when route hardware flags change

Routes installed to the kernel can be programmed to capable devices, in
which case they are marked with one of two flags. RTM_F_OFFLOAD for
routes that offload traffic from the kernel and RTM_F_TRAP for routes
that trap packets to the kernel for processing (e.g., host routes).

These flags are of interest to routing daemons since they would like to
delay advertisement of routes until they are installed in hardware. This
allows them to avoid packet loss or misrouted packets. Currently,
routing daemons do not receive any notifications when these flags are
changed, requiring them to poll the kernel tables for changes which is
inefficient.

This series addresses the issue by having the kernel emit RTM_NEWROUTE
notifications whenever these flags change. The behavior is controlled by
two sysctls (net.ipv4.fib_notify_on_flag_change and
net.ipv6.fib_notify_on_flag_change) that default to 0 (no
notifications).

Note that even if route installation in hardware is improved to be more
synchronous, these notifications are still of interest. For example, a
multipath route can change from RTM_F_OFFLOAD to RTM_F_TRAP if its
neighbours become invalid. A routing daemon can choose to withdraw /
replace the route in that case. In addition, the deletion of a route
from the kernel can prompt the installation of an identical route
(already in kernel, with an higher metric) to hardware.

For testing purposes, netdevsim is aligned to simulate a "real" driver
that programs routes to hardware.

Series overview:

Patches #1-#2 align netdevsim to perform route programming in a
non-atomic context

Patches #3-#5 add sysctl to control IPv4 notifications

Patches #6-#8 add sysctl to control IPv6 notifications

Patch #9 extends existing fib tests to set sysctls before running tests

Patch #10 adds test for fib notifications over netdevsim
====================

Link: https://lore.kernel.org/r/20210201194757.3463461-1-idosch@idosch.orgSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Documentation/networking/ip-sysctl.rst

@@ -178,6 +178,26 @@ min_adv_mss - INTEGER
 	The advertised MSS depends on the first hop route MTU, but will
 	never be lower than this setting.
 
+fib_notify_on_flag_change - INTEGER
+        Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
+        RTM_F_TRAP flags are changed.
+
+        After installing a route to the kernel, user space receives an
+        acknowledgment, which means the route was installed in the kernel,
+        but not necessarily in hardware.
+        It is also possible for a route already installed in hardware to change
+        its action and therefore its flags. For example, a host route that is
+        trapping packets can be "promoted" to perform decapsulation following
+        the installation of an IPinIP/VXLAN tunnel.
+        The notifications will indicate to user-space the state of the route.
+
+        Default: 0 (Do not emit notifications.)
+
+        Possible values:
+
+        - 0 - Do not emit notifications.
+        - 1 - Emit notifications.
+
 IP Fragmentation:
 
 ipfrag_high_thresh - LONG INTEGER
@@ -1775,6 +1795,26 @@ nexthop_compat_mode - BOOLEAN
 	and extraneous notifications.
 	Default: true (backward compat mode)
 
+fib_notify_on_flag_change - INTEGER
+        Whether to emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/
+        RTM_F_TRAP flags are changed.
+
+        After installing a route to the kernel, user space receives an
+        acknowledgment, which means the route was installed in the kernel,
+        but not necessarily in hardware.
+        It is also possible for a route already installed in hardware to change
+        its action and therefore its flags. For example, a host route that is
+        trapping packets can be "promoted" to perform decapsulation following
+        the installation of an IPinIP/VXLAN tunnel.
+        The notifications will indicate to user-space the state of the route.
+
+        Default: 0 (Do not emit notifications.)
+
+        Possible values:
+
+        - 0 - Do not emit notifications.
+        - 1 - Emit notifications.
+
 IPv6 Fragmentation:
 
 ip6frag_high_thresh - INTEGER

drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c

@@ -4988,6 +4988,7 @@ mlxsw_sp_fib4_entry_hw_flags_clear(struct mlxsw_sp *mlxsw_sp,
 	fib_alias_hw_flags_set(mlxsw_sp_net(mlxsw_sp), &fri);
 }
 
+#if IS_ENABLED(CONFIG_IPV6)
 static void
 mlxsw_sp_fib6_entry_hw_flags_set(struct mlxsw_sp *mlxsw_sp,
 				 struct mlxsw_sp_fib_entry *fib_entry)
@@ -5004,10 +5005,18 @@ mlxsw_sp_fib6_entry_hw_flags_set(struct mlxsw_sp *mlxsw_sp,
 	fib6_entry = container_of(fib_entry, struct mlxsw_sp_fib6_entry,
 				  common);
 	list_for_each_entry(mlxsw_sp_rt6, &fib6_entry->rt6_list, list)
-		fib6_info_hw_flags_set(mlxsw_sp_rt6->rt, should_offload,
-				       !should_offload);
+		fib6_info_hw_flags_set(mlxsw_sp_net(mlxsw_sp), mlxsw_sp_rt6->rt,
+				       should_offload, !should_offload);
+}
+#else
+static void
+mlxsw_sp_fib6_entry_hw_flags_set(struct mlxsw_sp *mlxsw_sp,
+				 struct mlxsw_sp_fib_entry *fib_entry)
+{
 }
+#endif
 
+#if IS_ENABLED(CONFIG_IPV6)
 static void
 mlxsw_sp_fib6_entry_hw_flags_clear(struct mlxsw_sp *mlxsw_sp,
 				   struct mlxsw_sp_fib_entry *fib_entry)
@@ -5018,8 +5027,16 @@ mlxsw_sp_fib6_entry_hw_flags_clear(struct mlxsw_sp *mlxsw_sp,
 	fib6_entry = container_of(fib_entry, struct mlxsw_sp_fib6_entry,
 				  common);
 	list_for_each_entry(mlxsw_sp_rt6, &fib6_entry->rt6_list, list)
-		fib6_info_hw_flags_set(mlxsw_sp_rt6->rt, false, false);
+		fib6_info_hw_flags_set(mlxsw_sp_net(mlxsw_sp), mlxsw_sp_rt6->rt,
+				       false, false);
+}
+#else
+static void
+mlxsw_sp_fib6_entry_hw_flags_clear(struct mlxsw_sp *mlxsw_sp,
+				   struct mlxsw_sp_fib_entry *fib_entry)
+{
 }
+#endif
 
 static void
 mlxsw_sp_fib_entry_hw_flags_set(struct mlxsw_sp *mlxsw_sp,

drivers/net/netdevsim/fib.c

@@ -31,7 +31,7 @@
 
 struct nsim_fib_entry {
 	u64 max;
-	u64 num;
+	atomic64_t num;
 };
 
 struct nsim_per_fib_data {
@@ -46,10 +46,13 @@ struct nsim_fib_data {
 	struct nsim_fib_entry nexthops;
 	struct rhashtable fib_rt_ht;
 	struct list_head fib_rt_list;
-	spinlock_t fib_lock;	/* Protects hashtable, list and accounting */
+	struct mutex fib_lock; /* Protects hashtable and list */
 	struct notifier_block nexthop_nb;
 	struct rhashtable nexthop_ht;
 	struct devlink *devlink;
+	struct work_struct fib_event_work;
+	struct list_head fib_event_queue;
+	spinlock_t fib_event_queue_lock; /* Protects fib event queue list */
 };
 
 struct nsim_fib_rt_key {
@@ -83,6 +86,22 @@ struct nsim_fib6_rt_nh {
 	struct fib6_info *rt;
 };
 
+struct nsim_fib6_event {
+	struct fib6_info **rt_arr;
+	unsigned int nrt6;
+};
+
+struct nsim_fib_event {
+	struct list_head list; /* node in fib queue */
+	union {
+		struct fib_entry_notifier_info fen_info;
+		struct nsim_fib6_event fib6_event;
+	};
+	struct nsim_fib_data *data;
+	unsigned long event;
+	int family;
+};
+
 static const struct rhashtable_params nsim_fib_rt_ht_params = {
 	.key_offset = offsetof(struct nsim_fib_rt, key),
 	.head_offset = offsetof(struct nsim_fib_rt, ht_node),
@@ -128,7 +147,7 @@ u64 nsim_fib_get_val(struct nsim_fib_data *fib_data,
 		return 0;
 	}
 
-	return max ? entry->max : entry->num;
+	return max ? entry->max : atomic64_read(&entry->num);
 }
 
 static void nsim_fib_set_max(struct nsim_fib_data *fib_data,
@@ -165,14 +184,12 @@ static int nsim_fib_rule_account(struct nsim_fib_entry *entry, bool add,
 	int err = 0;
 
 	if (add) {
-		if (entry->num < entry->max) {
-			entry->num++;
-		} else {
+		if (!atomic64_add_unless(&entry->num, 1, entry->max)) {
 			err = -ENOSPC;
 			NL_SET_ERR_MSG_MOD(extack, "Exceeded number of supported fib rule entries");
 		}
 	} else {
-		entry->num--;
+		atomic64_dec_if_positive(&entry->num);
 	}
 
 	return err;
@@ -196,20 +213,15 @@ static int nsim_fib_rule_event(struct nsim_fib_data *data,
 	return err;
 }
 
-static int nsim_fib_account(struct nsim_fib_entry *entry, bool add,
-			    struct netlink_ext_ack *extack)
+static int nsim_fib_account(struct nsim_fib_entry *entry, bool add)
 {
 	int err = 0;
 
 	if (add) {
-		if (entry->num < entry->max) {
-			entry->num++;
-		} else {
+		if (!atomic64_add_unless(&entry->num, 1, entry->max))
 			err = -ENOSPC;
-			NL_SET_ERR_MSG_MOD(extack, "Exceeded number of supported fib entries");
-		}
 	} else {
-		entry->num--;
+		atomic64_dec_if_positive(&entry->num);
 	}
 
 	return err;
@@ -254,7 +266,7 @@ nsim_fib4_rt_create(struct nsim_fib_data *data,
 {
 	struct nsim_fib4_rt *fib4_rt;
 
-	fib4_rt = kzalloc(sizeof(*fib4_rt), GFP_ATOMIC);
+	fib4_rt = kzalloc(sizeof(*fib4_rt), GFP_KERNEL);
 	if (!fib4_rt)
 		return NULL;
 
@@ -311,51 +323,52 @@ static void nsim_fib4_rt_hw_flags_set(struct net *net,
 }
 
 static int nsim_fib4_rt_add(struct nsim_fib_data *data,
-			    struct nsim_fib4_rt *fib4_rt,
-			    struct netlink_ext_ack *extack)
+			    struct nsim_fib4_rt *fib4_rt)
 {
 	struct net *net = devlink_net(data->devlink);
 	int err;
 
-	err = nsim_fib_account(&data->ipv4.fib, true, extack);
-	if (err)
-		return err;
-
 	err = rhashtable_insert_fast(&data->fib_rt_ht,
 				     &fib4_rt->common.ht_node,
 				     nsim_fib_rt_ht_params);
-	if (err) {
-		NL_SET_ERR_MSG_MOD(extack, "Failed to insert IPv4 route");
+	if (err)
 		goto err_fib_dismiss;
-	}
 
+	/* Simulate hardware programming latency. */
+	msleep(1);
 	nsim_fib4_rt_hw_flags_set(net, fib4_rt, true);
 
 	return 0;
 
 err_fib_dismiss:
-	nsim_fib_account(&data->ipv4.fib, false, extack);
+	/* Drop the accounting that was increased from the notification
+	 * context when FIB_EVENT_ENTRY_REPLACE was triggered.
+	 */
+	nsim_fib_account(&data->ipv4.fib, false);
 	return err;
 }
 
 static int nsim_fib4_rt_replace(struct nsim_fib_data *data,
 				struct nsim_fib4_rt *fib4_rt,
-				struct nsim_fib4_rt *fib4_rt_old,
-				struct netlink_ext_ack *extack)
+				struct nsim_fib4_rt *fib4_rt_old)
 {
 	struct net *net = devlink_net(data->devlink);
 	int err;
 
-	/* We are replacing a route, so no need to change the accounting. */
+	/* We are replacing a route, so need to remove the accounting which
+	 * was increased when FIB_EVENT_ENTRY_REPLACE was triggered.
+	 */
+	err = nsim_fib_account(&data->ipv4.fib, false);
+	if (err)
+		return err;
 	err = rhashtable_replace_fast(&data->fib_rt_ht,
 				      &fib4_rt_old->common.ht_node,
 				      &fib4_rt->common.ht_node,
 				      nsim_fib_rt_ht_params);
-	if (err) {
-		NL_SET_ERR_MSG_MOD(extack, "Failed to replace IPv4 route");
+	if (err)
 		return err;
-	}
 
+	msleep(1);
 	nsim_fib4_rt_hw_flags_set(net, fib4_rt, true);
 
 	nsim_fib4_rt_hw_flags_set(net, fib4_rt_old, false);
@@ -367,7 +380,6 @@ static int nsim_fib4_rt_replace(struct nsim_fib_data *data,
 static int nsim_fib4_rt_insert(struct nsim_fib_data *data,
 			       struct fib_entry_notifier_info *fen_info)
 {
-	struct netlink_ext_ack *extack = fen_info->info.extack;
 	struct nsim_fib4_rt *fib4_rt, *fib4_rt_old;
 	int err;
 
@@ -377,9 +389,9 @@ static int nsim_fib4_rt_insert(struct nsim_fib_data *data,
 
 	fib4_rt_old = nsim_fib4_rt_lookup(&data->fib_rt_ht, fen_info);
 	if (!fib4_rt_old)
-		err = nsim_fib4_rt_add(data, fib4_rt, extack);
+		err = nsim_fib4_rt_add(data, fib4_rt);
 	else
-		err = nsim_fib4_rt_replace(data, fib4_rt, fib4_rt_old, extack);
+		err = nsim_fib4_rt_replace(data, fib4_rt, fib4_rt_old);
 
 	if (err)
 		nsim_fib4_rt_destroy(fib4_rt);
@@ -390,7 +402,6 @@ static int nsim_fib4_rt_insert(struct nsim_fib_data *data,
 static void nsim_fib4_rt_remove(struct nsim_fib_data *data,
 				const struct fib_entry_notifier_info *fen_info)
 {
-	struct netlink_ext_ack *extack = fen_info->info.extack;
 	struct nsim_fib4_rt *fib4_rt;
 
 	fib4_rt = nsim_fib4_rt_lookup(&data->fib_rt_ht, fen_info);
@@ -399,19 +410,15 @@ static void nsim_fib4_rt_remove(struct nsim_fib_data *data,
 
 	rhashtable_remove_fast(&data->fib_rt_ht, &fib4_rt->common.ht_node,
 			       nsim_fib_rt_ht_params);
-	nsim_fib_account(&data->ipv4.fib, false, extack);
 	nsim_fib4_rt_destroy(fib4_rt);
 }
 
 static int nsim_fib4_event(struct nsim_fib_data *data,
-			   struct fib_notifier_info *info,
+			   struct fib_entry_notifier_info *fen_info,
 			   unsigned long event)
 {
-	struct fib_entry_notifier_info *fen_info;
 	int err = 0;
 
-	fen_info = container_of(info, struct fib_entry_notifier_info, info);
-
 	switch (event) {
 	case FIB_EVENT_ENTRY_REPLACE:
 		err = nsim_fib4_rt_insert(data, fen_info);
@@ -445,7 +452,7 @@ static int nsim_fib6_rt_nh_add(struct nsim_fib6_rt *fib6_rt,
 {
 	struct nsim_fib6_rt_nh *fib6_rt_nh;
 
-	fib6_rt_nh = kzalloc(sizeof(*fib6_rt_nh), GFP_ATOMIC);
+	fib6_rt_nh = kzalloc(sizeof(*fib6_rt_nh), GFP_KERNEL);
 	if (!fib6_rt_nh)
 		return -ENOMEM;
 
@@ -457,6 +464,17 @@ static int nsim_fib6_rt_nh_add(struct nsim_fib6_rt *fib6_rt,
 	return 0;
 }
 
+#if IS_ENABLED(CONFIG_IPV6)
+static void nsim_rt6_release(struct fib6_info *rt)
+{
+	fib6_info_release(rt);
+}
+#else
+static void nsim_rt6_release(struct fib6_info *rt)
+{
+}
+#endif
+
 static void nsim_fib6_rt_nh_del(struct nsim_fib6_rt *fib6_rt,
 				const struct fib6_info *rt)
 {
@@ -468,22 +486,20 @@ static void nsim_fib6_rt_nh_del(struct nsim_fib6_rt *fib6_rt,
 
 	fib6_rt->nhs--;
 	list_del(&fib6_rt_nh->list);
-#if IS_ENABLED(CONFIG_IPV6)
-	fib6_info_release(fib6_rt_nh->rt);
-#endif
+	nsim_rt6_release(fib6_rt_nh->rt);
 	kfree(fib6_rt_nh);
 }
 
 static struct nsim_fib6_rt *
 nsim_fib6_rt_create(struct nsim_fib_data *data,
-		    struct fib6_entry_notifier_info *fen6_info)
+		    struct fib6_info **rt_arr, unsigned int nrt6)
 {
-	struct fib6_info *iter, *rt = fen6_info->rt;
+	struct fib6_info *rt = rt_arr[0];
 	struct nsim_fib6_rt *fib6_rt;
 	int i = 0;
 	int err;
 
-	fib6_rt = kzalloc(sizeof(*fib6_rt), GFP_ATOMIC);
+	fib6_rt = kzalloc(sizeof(*fib6_rt), GFP_KERNEL);
 	if (!fib6_rt)
 		return ERR_PTR(-ENOMEM);
 
@@ -497,32 +513,18 @@ nsim_fib6_rt_create(struct nsim_fib_data *data,
 	 */
 	INIT_LIST_HEAD(&fib6_rt->nh_list);
 
-	err = nsim_fib6_rt_nh_add(fib6_rt, rt);
-	if (err)
-		goto err_fib_rt_fini;
-
-	if (!fen6_info->nsiblings)
-		return fib6_rt;
-
-	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
-		if (i == fen6_info->nsiblings)
-			break;
-
-		err = nsim_fib6_rt_nh_add(fib6_rt, iter);
+	for (i = 0; i < nrt6; i++) {
+		err = nsim_fib6_rt_nh_add(fib6_rt, rt_arr[i]);
 		if (err)
 			goto err_fib6_rt_nh_del;
-		i++;
 	}
-	WARN_ON_ONCE(i != fen6_info->nsiblings);
 
 	return fib6_rt;
 
 err_fib6_rt_nh_del:
-	list_for_each_entry_continue_reverse(iter, &rt->fib6_siblings,
-					     fib6_siblings)
-		nsim_fib6_rt_nh_del(fib6_rt, iter);
-	nsim_fib6_rt_nh_del(fib6_rt, rt);
-err_fib_rt_fini:
+	for (i--; i >= 0; i--) {
+		nsim_fib6_rt_nh_del(fib6_rt, rt_arr[i]);
+	};
 	nsim_fib_rt_fini(&fib6_rt->common);
 	kfree(fib6_rt);
 	return ERR_PTR(err);
@@ -555,127 +557,125 @@ nsim_fib6_rt_lookup(struct rhashtable *fib_rt_ht, const struct fib6_info *rt)
 }
 
 static int nsim_fib6_rt_append(struct nsim_fib_data *data,
-			       struct fib6_entry_notifier_info *fen6_info)
+			       struct nsim_fib6_event *fib6_event)
 {
-	struct fib6_info *iter, *rt = fen6_info->rt;
+	struct fib6_info *rt = fib6_event->rt_arr[0];
 	struct nsim_fib6_rt *fib6_rt;
-	int i = 0;
-	int err;
+	int i, err;
 
 	fib6_rt = nsim_fib6_rt_lookup(&data->fib_rt_ht, rt);
 	if (WARN_ON_ONCE(!fib6_rt))
 		return -EINVAL;
 
-	err = nsim_fib6_rt_nh_add(fib6_rt, rt);
-	if (err)
-		return err;
-	rt->trap = true;
-
-	if (!fen6_info->nsiblings)
-		return 0;
-
-	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
-		if (i == fen6_info->nsiblings)
-			break;
-
-		err = nsim_fib6_rt_nh_add(fib6_rt, iter);
+	for (i = 0; i < fib6_event->nrt6; i++) {
+		err = nsim_fib6_rt_nh_add(fib6_rt, fib6_event->rt_arr[i]);
 		if (err)
 			goto err_fib6_rt_nh_del;
-		iter->trap = true;
-		i++;
+
+		fib6_event->rt_arr[i]->trap = true;
 	}
-	WARN_ON_ONCE(i != fen6_info->nsiblings);
 
 	return 0;
 
 err_fib6_rt_nh_del:
-	list_for_each_entry_continue_reverse(iter, &rt->fib6_siblings,
-					     fib6_siblings) {
-		iter->trap = false;
-		nsim_fib6_rt_nh_del(fib6_rt, iter);
+	for (i--; i >= 0; i--) {
+		fib6_event->rt_arr[i]->trap = false;
+		nsim_fib6_rt_nh_del(fib6_rt, fib6_event->rt_arr[i]);
 	}
-	rt->trap = false;
-	nsim_fib6_rt_nh_del(fib6_rt, rt);
 	return err;
 }
 
-static void nsim_fib6_rt_hw_flags_set(const struct nsim_fib6_rt *fib6_rt,
+#if IS_ENABLED(CONFIG_IPV6)
+static void nsim_fib6_rt_hw_flags_set(struct nsim_fib_data *data,
+				      const struct nsim_fib6_rt *fib6_rt,
 				      bool trap)
 {
+	struct net *net = devlink_net(data->devlink);
 	struct nsim_fib6_rt_nh *fib6_rt_nh;
 
 	list_for_each_entry(fib6_rt_nh, &fib6_rt->nh_list, list)
-		fib6_info_hw_flags_set(fib6_rt_nh->rt, false, trap);
+		fib6_info_hw_flags_set(net, fib6_rt_nh->rt, false, trap);
 }
+#else
+static void nsim_fib6_rt_hw_flags_set(struct nsim_fib_data *data,
+				      const struct nsim_fib6_rt *fib6_rt,
+				      bool trap)
+{
+}
+#endif
 
 static int nsim_fib6_rt_add(struct nsim_fib_data *data,
-			    struct nsim_fib6_rt *fib6_rt,
-			    struct netlink_ext_ack *extack)
+			    struct nsim_fib6_rt *fib6_rt)
 {
 	int err;
 
-	err = nsim_fib_account(&data->ipv6.fib, true, extack);
-	if (err)
-		return err;
-
 	err = rhashtable_insert_fast(&data->fib_rt_ht,
 				     &fib6_rt->common.ht_node,
 				     nsim_fib_rt_ht_params);
-	if (err) {
-		NL_SET_ERR_MSG_MOD(extack, "Failed to insert IPv6 route");
+
+	if (err)
 		goto err_fib_dismiss;
-	}
 
-	nsim_fib6_rt_hw_flags_set(fib6_rt, true);
+	msleep(1);
+	nsim_fib6_rt_hw_flags_set(data, fib6_rt, true);
 
 	return 0;
 
 err_fib_dismiss:
-	nsim_fib_account(&data->ipv6.fib, false, extack);
+	/* Drop the accounting that was increased from the notification
+	 * context when FIB_EVENT_ENTRY_REPLACE was triggered.
+	 */
+	nsim_fib_account(&data->ipv6.fib, false);
 	return err;
 }
 
 static int nsim_fib6_rt_replace(struct nsim_fib_data *data,
 				struct nsim_fib6_rt *fib6_rt,
-				struct nsim_fib6_rt *fib6_rt_old,
-				struct netlink_ext_ack *extack)
+				struct nsim_fib6_rt *fib6_rt_old)
 {
 	int err;
 
-	/* We are replacing a route, so no need to change the accounting. */
+	/* We are replacing a route, so need to remove the accounting which
+	 * was increased when FIB_EVENT_ENTRY_REPLACE was triggered.
+	 */
+	err = nsim_fib_account(&data->ipv6.fib, false);
+	if (err)
+		return err;
+
 	err = rhashtable_replace_fast(&data->fib_rt_ht,
 				      &fib6_rt_old->common.ht_node,
 				      &fib6_rt->common.ht_node,
 				      nsim_fib_rt_ht_params);
-	if (err) {
-		NL_SET_ERR_MSG_MOD(extack, "Failed to replace IPv6 route");
+
+	if (err)
 		return err;
-	}
 
-	nsim_fib6_rt_hw_flags_set(fib6_rt, true);
+	msleep(1);
+	nsim_fib6_rt_hw_flags_set(data, fib6_rt, true);
 
-	nsim_fib6_rt_hw_flags_set(fib6_rt_old, false);
+	nsim_fib6_rt_hw_flags_set(data, fib6_rt_old, false);
 	nsim_fib6_rt_destroy(fib6_rt_old);
 
 	return 0;
 }
 
 static int nsim_fib6_rt_insert(struct nsim_fib_data *data,
-			       struct fib6_entry_notifier_info *fen6_info)
+			       struct nsim_fib6_event *fib6_event)
 {
-	struct netlink_ext_ack *extack = fen6_info->info.extack;
+	struct fib6_info *rt = fib6_event->rt_arr[0];
 	struct nsim_fib6_rt *fib6_rt, *fib6_rt_old;
 	int err;
 
-	fib6_rt = nsim_fib6_rt_create(data, fen6_info);
+	fib6_rt = nsim_fib6_rt_create(data, fib6_event->rt_arr,
+				      fib6_event->nrt6);
 	if (IS_ERR(fib6_rt))
 		return PTR_ERR(fib6_rt);
 
-	fib6_rt_old = nsim_fib6_rt_lookup(&data->fib_rt_ht, fen6_info->rt);
+	fib6_rt_old = nsim_fib6_rt_lookup(&data->fib_rt_ht, rt);
 	if (!fib6_rt_old)
-		err = nsim_fib6_rt_add(data, fib6_rt, extack);
+		err = nsim_fib6_rt_add(data, fib6_rt);
 	else
-		err = nsim_fib6_rt_replace(data, fib6_rt, fib6_rt_old, extack);
+		err = nsim_fib6_rt_replace(data, fib6_rt, fib6_rt_old);
 
 	if (err)
 		nsim_fib6_rt_destroy(fib6_rt);
@@ -683,59 +683,100 @@ static int nsim_fib6_rt_insert(struct nsim_fib_data *data,
 	return err;
 }
 
-static void
-nsim_fib6_rt_remove(struct nsim_fib_data *data,
-		    const struct fib6_entry_notifier_info *fen6_info)
+static void nsim_fib6_rt_remove(struct nsim_fib_data *data,
+				struct nsim_fib6_event *fib6_event)
 {
-	struct netlink_ext_ack *extack = fen6_info->info.extack;
+	struct fib6_info *rt = fib6_event->rt_arr[0];
 	struct nsim_fib6_rt *fib6_rt;
+	int i;
 
 	/* Multipath routes are first added to the FIB trie and only then
 	 * notified. If we vetoed the addition, we will get a delete
 	 * notification for a route we do not have. Therefore, do not warn if
 	 * route was not found.
 	 */
-	fib6_rt = nsim_fib6_rt_lookup(&data->fib_rt_ht, fen6_info->rt);
+	fib6_rt = nsim_fib6_rt_lookup(&data->fib_rt_ht, rt);
 	if (!fib6_rt)
 		return;
 
 	/* If not all the nexthops are deleted, then only reduce the nexthop
 	 * group.
 	 */
-	if (fen6_info->nsiblings + 1 != fib6_rt->nhs) {
-		nsim_fib6_rt_nh_del(fib6_rt, fen6_info->rt);
+	if (fib6_event->nrt6 != fib6_rt->nhs) {
+		for (i = 0; i < fib6_event->nrt6; i++)
+			nsim_fib6_rt_nh_del(fib6_rt, fib6_event->rt_arr[i]);
 		return;
 	}
 
 	rhashtable_remove_fast(&data->fib_rt_ht, &fib6_rt->common.ht_node,
 			       nsim_fib_rt_ht_params);
-	nsim_fib_account(&data->ipv6.fib, false, extack);
 	nsim_fib6_rt_destroy(fib6_rt);
 }
 
+static int nsim_fib6_event_init(struct nsim_fib6_event *fib6_event,
+				struct fib6_entry_notifier_info *fen6_info)
+{
+	struct fib6_info *rt = fen6_info->rt;
+	struct fib6_info **rt_arr;
+	struct fib6_info *iter;
+	unsigned int nrt6;
+	int i = 0;
+
+	nrt6 = fen6_info->nsiblings + 1;
+
+	rt_arr = kcalloc(nrt6, sizeof(struct fib6_info *), GFP_ATOMIC);
+	if (!rt_arr)
+		return -ENOMEM;
+
+	fib6_event->rt_arr = rt_arr;
+	fib6_event->nrt6 = nrt6;
+
+	rt_arr[0] = rt;
+	fib6_info_hold(rt);
+
+	if (!fen6_info->nsiblings)
+		return 0;
+
+	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
+		if (i == fen6_info->nsiblings)
+			break;
+
+		rt_arr[i + 1] = iter;
+		fib6_info_hold(iter);
+		i++;
+	}
+	WARN_ON_ONCE(i != fen6_info->nsiblings);
+
+	return 0;
+}
+
+static void nsim_fib6_event_fini(struct nsim_fib6_event *fib6_event)
+{
+	int i;
+
+	for (i = 0; i < fib6_event->nrt6; i++)
+		nsim_rt6_release(fib6_event->rt_arr[i]);
+	kfree(fib6_event->rt_arr);
+}
+
 static int nsim_fib6_event(struct nsim_fib_data *data,
-			   struct fib_notifier_info *info,
+			   struct nsim_fib6_event *fib6_event,
 			   unsigned long event)
 {
-	struct fib6_entry_notifier_info *fen6_info;
 	int err = 0;
 
-	fen6_info = container_of(info, struct fib6_entry_notifier_info, info);
-
-	if (fen6_info->rt->fib6_src.plen) {
-		NL_SET_ERR_MSG_MOD(info->extack, "IPv6 source-specific route is not supported");
+	if (fib6_event->rt_arr[0]->fib6_src.plen)
 		return 0;
-	}
 
 	switch (event) {
 	case FIB_EVENT_ENTRY_REPLACE:
-		err = nsim_fib6_rt_insert(data, fen6_info);
+		err = nsim_fib6_rt_insert(data, fib6_event);
 		break;
 	case FIB_EVENT_ENTRY_APPEND:
-		err = nsim_fib6_rt_append(data, fen6_info);
+		err = nsim_fib6_rt_append(data, fib6_event);
 		break;
 	case FIB_EVENT_ENTRY_DEL:
-		nsim_fib6_rt_remove(data, fen6_info);
+		nsim_fib6_rt_remove(data, fib6_event);
 		break;
 	default:
 		break;
@@ -744,51 +785,165 @@ static int nsim_fib6_event(struct nsim_fib_data *data,
 	return err;
 }
 
-static int nsim_fib_event(struct nsim_fib_data *data,
-			  struct fib_notifier_info *info, unsigned long event)
+static int nsim_fib_event(struct nsim_fib_event *fib_event)
 {
 	int err = 0;
 
-	switch (info->family) {
+	switch (fib_event->family) {
 	case AF_INET:
-		err = nsim_fib4_event(data, info, event);
+		nsim_fib4_event(fib_event->data, &fib_event->fen_info,
+				fib_event->event);
+		fib_info_put(fib_event->fen_info.fi);
 		break;
 	case AF_INET6:
-		err = nsim_fib6_event(data, info, event);
+		nsim_fib6_event(fib_event->data, &fib_event->fib6_event,
+				fib_event->event);
+		nsim_fib6_event_fini(&fib_event->fib6_event);
 		break;
 	}
 
 	return err;
 }
 
+static int nsim_fib4_prepare_event(struct fib_notifier_info *info,
+				   struct nsim_fib_event *fib_event,
+				   unsigned long event)
+{
+	struct nsim_fib_data *data = fib_event->data;
+	struct fib_entry_notifier_info *fen_info;
+	struct netlink_ext_ack *extack;
+	int err = 0;
+
+	fen_info = container_of(info, struct fib_entry_notifier_info,
+				info);
+	fib_event->fen_info = *fen_info;
+	extack = info->extack;
+
+	switch (event) {
+	case FIB_EVENT_ENTRY_REPLACE:
+		err = nsim_fib_account(&data->ipv4.fib, true);
+		if (err) {
+			NL_SET_ERR_MSG_MOD(extack, "Exceeded number of supported fib entries");
+			return err;
+		}
+		break;
+	case FIB_EVENT_ENTRY_DEL:
+		nsim_fib_account(&data->ipv4.fib, false);
+		break;
+	}
+
+	/* Take reference on fib_info to prevent it from being
+	 * freed while event is queued. Release it afterwards.
+	 */
+	fib_info_hold(fib_event->fen_info.fi);
+
+	return 0;
+}
+
+static int nsim_fib6_prepare_event(struct fib_notifier_info *info,
+				   struct nsim_fib_event *fib_event,
+				   unsigned long event)
+{
+	struct nsim_fib_data *data = fib_event->data;
+	struct fib6_entry_notifier_info *fen6_info;
+	struct netlink_ext_ack *extack;
+	int err = 0;
+
+	fen6_info = container_of(info, struct fib6_entry_notifier_info,
+				 info);
+
+	err = nsim_fib6_event_init(&fib_event->fib6_event, fen6_info);
+	if (err)
+		return err;
+
+	extack = info->extack;
+	switch (event) {
+	case FIB_EVENT_ENTRY_REPLACE:
+		err = nsim_fib_account(&data->ipv6.fib, true);
+		if (err) {
+			NL_SET_ERR_MSG_MOD(extack, "Exceeded number of supported fib entries");
+			goto err_fib6_event_fini;
+		}
+		break;
+	case FIB_EVENT_ENTRY_DEL:
+		nsim_fib_account(&data->ipv6.fib, false);
+		break;
+	}
+
+	return 0;
+
+err_fib6_event_fini:
+	nsim_fib6_event_fini(&fib_event->fib6_event);
+	return err;
+}
+
+static int nsim_fib_event_schedule_work(struct nsim_fib_data *data,
+					struct fib_notifier_info *info,
+					unsigned long event)
+{
+	struct nsim_fib_event *fib_event;
+	int err;
+
+	if (info->family != AF_INET && info->family != AF_INET6)
+		/* netdevsim does not support 'RTNL_FAMILY_IP6MR' and
+		 * 'RTNL_FAMILY_IPMR' and should ignore them.
+		 */
+		return NOTIFY_DONE;
+
+	fib_event = kzalloc(sizeof(*fib_event), GFP_ATOMIC);
+	if (!fib_event)
+		return NOTIFY_BAD;
+
+	fib_event->data = data;
+	fib_event->event = event;
+	fib_event->family = info->family;
+
+	switch (info->family) {
+	case AF_INET:
+		err = nsim_fib4_prepare_event(info, fib_event, event);
+		break;
+	case AF_INET6:
+		err = nsim_fib6_prepare_event(info, fib_event, event);
+		break;
+	}
+
+	if (err)
+		goto err_fib_prepare_event;
+
+	/* Enqueue the event and trigger the work */
+	spin_lock_bh(&data->fib_event_queue_lock);
+	list_add_tail(&fib_event->list, &data->fib_event_queue);
+	spin_unlock_bh(&data->fib_event_queue_lock);
+	schedule_work(&data->fib_event_work);
+
+	return NOTIFY_DONE;
+
+err_fib_prepare_event:
+	kfree(fib_event);
+	return NOTIFY_BAD;
+}
+
 static int nsim_fib_event_nb(struct notifier_block *nb, unsigned long event,
 			     void *ptr)
 {
 	struct nsim_fib_data *data = container_of(nb, struct nsim_fib_data,
 						  fib_nb);
 	struct fib_notifier_info *info = ptr;
-	int err = 0;
-
-	/* IPv6 routes can be added via RAs from softIRQ. */
-	spin_lock_bh(&data->fib_lock);
+	int err;
 
 	switch (event) {
 	case FIB_EVENT_RULE_ADD:
 	case FIB_EVENT_RULE_DEL:
 		err = nsim_fib_rule_event(data, info,
 					  event == FIB_EVENT_RULE_ADD);
-		break;
-
+		return notifier_from_errno(err);
 	case FIB_EVENT_ENTRY_REPLACE:
 	case FIB_EVENT_ENTRY_APPEND:
 	case FIB_EVENT_ENTRY_DEL:
-		err = nsim_fib_event(data, info, event);
-		break;
+		return nsim_fib_event_schedule_work(data, info, event);
 	}
 
-	spin_unlock_bh(&data->fib_lock);
-
-	return notifier_from_errno(err);
+	return NOTIFY_DONE;
 }
 
 static void nsim_fib4_rt_free(struct nsim_fib_rt *fib_rt,
@@ -799,7 +954,7 @@ static void nsim_fib4_rt_free(struct nsim_fib_rt *fib_rt,
 
 	fib4_rt = container_of(fib_rt, struct nsim_fib4_rt, common);
 	nsim_fib4_rt_hw_flags_set(devlink_net(devlink), fib4_rt, false);
-	nsim_fib_account(&data->ipv4.fib, false, NULL);
+	nsim_fib_account(&data->ipv4.fib, false);
 	nsim_fib4_rt_destroy(fib4_rt);
 }
 
@@ -809,8 +964,8 @@ static void nsim_fib6_rt_free(struct nsim_fib_rt *fib_rt,
 	struct nsim_fib6_rt *fib6_rt;
 
 	fib6_rt = container_of(fib_rt, struct nsim_fib6_rt, common);
-	nsim_fib6_rt_hw_flags_set(fib6_rt, false);
-	nsim_fib_account(&data->ipv6.fib, false, NULL);
+	nsim_fib6_rt_hw_flags_set(data, fib6_rt, false);
+	nsim_fib_account(&data->ipv6.fib, false);
 	nsim_fib6_rt_destroy(fib6_rt);
 }
 
@@ -838,6 +993,9 @@ static void nsim_fib_dump_inconsistent(struct notifier_block *nb)
 						  fib_nb);
 	struct nsim_fib_rt *fib_rt, *fib_rt_tmp;
 
+	/* Flush the work to make sure there is no race with notifications. */
+	flush_work(&data->fib_event_work);
+
 	/* The notifier block is still not registered, so we do not need to
 	 * take any locks here.
 	 */
@@ -847,8 +1005,8 @@ static void nsim_fib_dump_inconsistent(struct notifier_block *nb)
 		nsim_fib_rt_free(fib_rt, data);
 	}
 
-	data->ipv4.rules.num = 0ULL;
-	data->ipv6.rules.num = 0ULL;
+	atomic64_set(&data->ipv4.rules.num, 0ULL);
+	atomic64_set(&data->ipv6.rules.num, 0ULL);
 }
 
 static struct nsim_nexthop *nsim_nexthop_create(struct nsim_fib_data *data,
@@ -894,22 +1052,28 @@ static void nsim_nexthop_destroy(struct nsim_nexthop *nexthop)
 static int nsim_nexthop_account(struct nsim_fib_data *data, u64 occ,
 				bool add, struct netlink_ext_ack *extack)
 {
-	int err = 0;
+	int i, err = 0;
 
 	if (add) {
-		if (data->nexthops.num + occ <= data->nexthops.max) {
-			data->nexthops.num += occ;
-		} else {
+		for (i = 0; i < occ; i++)
+			if (!atomic64_add_unless(&data->nexthops.num, 1,
+						 data->nexthops.max)) {
 				err = -ENOSPC;
 				NL_SET_ERR_MSG_MOD(extack, "Exceeded number of supported nexthops");
+				goto err_num_decrease;
 			}
 	} else {
-		if (WARN_ON(occ > data->nexthops.num))
+		if (WARN_ON(occ > atomic64_read(&data->nexthops.num)))
 			return -EINVAL;
-		data->nexthops.num -= occ;
+		atomic64_sub(occ, &data->nexthops.num);
 	}
 
 	return err;
+
+err_num_decrease:
+	atomic64_sub(i, &data->nexthops.num);
+	return err;
+
 }
 
 static int nsim_nexthop_add(struct nsim_fib_data *data,
@@ -1102,6 +1266,29 @@ static void nsim_fib_set_max_all(struct nsim_fib_data *data,
 	}
 }
 
+static void nsim_fib_event_work(struct work_struct *work)
+{
+	struct nsim_fib_data *data = container_of(work, struct nsim_fib_data,
+						  fib_event_work);
+	struct nsim_fib_event *fib_event, *next_fib_event;
+
+	LIST_HEAD(fib_event_queue);
+
+	spin_lock_bh(&data->fib_event_queue_lock);
+	list_splice_init(&data->fib_event_queue, &fib_event_queue);
+	spin_unlock_bh(&data->fib_event_queue_lock);
+
+	mutex_lock(&data->fib_lock);
+	list_for_each_entry_safe(fib_event, next_fib_event, &fib_event_queue,
+				 list) {
+		nsim_fib_event(fib_event);
+		list_del(&fib_event->list);
+		kfree(fib_event);
+		cond_resched();
+	}
+	mutex_unlock(&data->fib_lock);
+}
+
 struct nsim_fib_data *nsim_fib_create(struct devlink *devlink,
 				      struct netlink_ext_ack *extack)
 {
@@ -1117,12 +1304,16 @@ struct nsim_fib_data *nsim_fib_create(struct devlink *devlink,
 	if (err)
 		goto err_data_free;
 
-	spin_lock_init(&data->fib_lock);
+	mutex_init(&data->fib_lock);
 	INIT_LIST_HEAD(&data->fib_rt_list);
 	err = rhashtable_init(&data->fib_rt_ht, &nsim_fib_rt_ht_params);
 	if (err)
 		goto err_rhashtable_nexthop_destroy;
 
+	INIT_WORK(&data->fib_event_work, nsim_fib_event_work);
+	INIT_LIST_HEAD(&data->fib_event_queue);
+	spin_lock_init(&data->fib_event_queue_lock);
+
 	nsim_fib_set_max_all(data, devlink);
 
 	data->nexthop_nb.notifier_call = nsim_nexthop_event_nb;
@@ -1166,11 +1357,13 @@ struct nsim_fib_data *nsim_fib_create(struct devlink *devlink,
 err_nexthop_nb_unregister:
 	unregister_nexthop_notifier(devlink_net(devlink), &data->nexthop_nb);
 err_rhashtable_fib_destroy:
+	flush_work(&data->fib_event_work);
 	rhashtable_free_and_destroy(&data->fib_rt_ht, nsim_fib_rt_free,
 				    data);
 err_rhashtable_nexthop_destroy:
 	rhashtable_free_and_destroy(&data->nexthop_ht, nsim_nexthop_free,
 				    data);
+	mutex_destroy(&data->fib_lock);
 err_data_free:
 	kfree(data);
 	return ERR_PTR(err);
@@ -1190,10 +1383,13 @@ void nsim_fib_destroy(struct devlink *devlink, struct nsim_fib_data *data)
 					    NSIM_RESOURCE_IPV4_FIB);
 	unregister_fib_notifier(devlink_net(devlink), &data->fib_nb);
 	unregister_nexthop_notifier(devlink_net(devlink), &data->nexthop_nb);
+	flush_work(&data->fib_event_work);
 	rhashtable_free_and_destroy(&data->fib_rt_ht, nsim_fib_rt_free,
 				    data);
 	rhashtable_free_and_destroy(&data->nexthop_ht, nsim_nexthop_free,
 				    data);
+	WARN_ON_ONCE(!list_empty(&data->fib_event_queue));
 	WARN_ON_ONCE(!list_empty(&data->fib_rt_list));
+	mutex_destroy(&data->fib_lock);
 	kfree(data);
 }

include/net/ip6_fib.h

@@ -336,13 +336,6 @@ static inline void fib6_info_release(struct fib6_info *f6i)
 		call_rcu(&f6i->rcu, fib6_info_destroy_rcu);
 }
 
-static inline void fib6_info_hw_flags_set(struct fib6_info *f6i, bool offload,
-					  bool trap)
-{
-	f6i->offload = offload;
-	f6i->trap = trap;
-}
-
 enum fib6_walk_state {
 #ifdef CONFIG_IPV6_SUBTREES
 	FWS_S,
@@ -545,6 +538,8 @@ static inline bool fib6_metric_locked(struct fib6_info *f6i, int metric)
 {
 	return !!(f6i->fib6_metrics->metrics[RTAX_LOCK - 1] & (1 << metric));
 }
+void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
+			    bool offload, bool trap);
 
 #if IS_BUILTIN(CONFIG_IPV6) && defined(CONFIG_BPF_SYSCALL)
 struct bpf_iter__ipv6_route {

include/net/netns/ipv4.h

@@ -188,6 +188,8 @@ struct netns_ipv4 {
 	int sysctl_udp_wmem_min;
 	int sysctl_udp_rmem_min;
 
+	int sysctl_fib_notify_on_flag_change;
+
 #ifdef CONFIG_NET_L3_MASTER_DEV
 	int sysctl_udp_l3mdev_accept;
 #endif

include/net/netns/ipv6.h

@@ -51,6 +51,7 @@ struct netns_sysctl_ipv6 {
 	int max_hbh_opts_len;
 	int seg6_flowlabel;
 	bool skip_notify_on_dev_down;
+	int fib_notify_on_flag_change;
 };
 
 struct netns_ipv6 {

net/ipv4/af_inet.c

@@ -1871,6 +1871,8 @@ static __net_init int inet_init_net(struct net *net)
 	net->ipv4.sysctl_igmp_llm_reports = 1;
 	net->ipv4.sysctl_igmp_qrv = 2;
 
+	net->ipv4.sysctl_fib_notify_on_flag_change = 0;
+
 	return 0;
 }
 

net/ipv4/fib_lookup.h

@@ -39,9 +39,10 @@ int fib_nh_match(struct net *net, struct fib_config *cfg, struct fib_info *fi,
 		 struct netlink_ext_ack *extack);
 bool fib_metrics_match(struct fib_config *cfg, struct fib_info *fi);
 int fib_dump_info(struct sk_buff *skb, u32 pid, u32 seq, int event,
-		  struct fib_rt_info *fri, unsigned int flags);
+		  const struct fib_rt_info *fri, unsigned int flags);
 void rtmsg_fib(int event, __be32 key, struct fib_alias *fa, int dst_len,
 	       u32 tb_id, const struct nl_info *info, unsigned int nlm_flags);
+size_t fib_nlmsg_size(struct fib_info *fi);
 
 static inline void fib_result_assign(struct fib_result *res,
 				     struct fib_info *fi)

net/ipv4/fib_semantics.c

@@ -452,7 +452,7 @@ int ip_fib_check_default(__be32 gw, struct net_device *dev)
 	return -1;
 }
 
-static inline size_t fib_nlmsg_size(struct fib_info *fi)
+size_t fib_nlmsg_size(struct fib_info *fi)
 {
 	size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg))
 			 + nla_total_size(4) /* RTA_TABLE */
@@ -1733,7 +1733,7 @@ static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
 #endif
 
 int fib_dump_info(struct sk_buff *skb, u32 portid, u32 seq, int event,
-		  struct fib_rt_info *fri, unsigned int flags)
+		  const struct fib_rt_info *fri, unsigned int flags)
 {
 	unsigned int nhs = fib_info_num_path(fri->fi);
 	struct fib_info *fi = fri->fi;

net/ipv4/fib_trie.c

@@ -1038,6 +1038,8 @@ fib_find_matching_alias(struct net *net, const struct fib_rt_info *fri)
 void fib_alias_hw_flags_set(struct net *net, const struct fib_rt_info *fri)
 {
 	struct fib_alias *fa_match;
+	struct sk_buff *skb;
+	int err;
 
 	rcu_read_lock();
 
@@ -1045,9 +1047,34 @@ void fib_alias_hw_flags_set(struct net *net, const struct fib_rt_info *fri)
 	if (!fa_match)
 		goto out;
 
+	if (fa_match->offload == fri->offload && fa_match->trap == fri->trap)
+		goto out;
+
 	fa_match->offload = fri->offload;
 	fa_match->trap = fri->trap;
 
+	if (!net->ipv4.sysctl_fib_notify_on_flag_change)
+		goto out;
+
+	skb = nlmsg_new(fib_nlmsg_size(fa_match->fa_info), GFP_ATOMIC);
+	if (!skb) {
+		err = -ENOBUFS;
+		goto errout;
+	}
+
+	err = fib_dump_info(skb, 0, 0, RTM_NEWROUTE, fri, 0);
+	if (err < 0) {
+		/* -EMSGSIZE implies BUG in fib_nlmsg_size() */
+		WARN_ON(err == -EMSGSIZE);
+		kfree_skb(skb);
+		goto errout;
+	}
+
+	rtnl_notify(skb, net, 0, RTNLGRP_IPV4_ROUTE, NULL, GFP_ATOMIC);
+	goto out;
+
+errout:
+	rtnl_set_sk_err(net, RTNLGRP_IPV4_ROUTE, err);
 out:
 	rcu_read_unlock();
 }

net/ipv4/sysctl_net_ipv4.c

@@ -1354,6 +1354,15 @@ static struct ctl_table ipv4_net_table[] = {
 		.proc_handler	= proc_dointvec_minmax,
 		.extra1		= SYSCTL_ONE
 	},
+	{
+		.procname	= "fib_notify_on_flag_change",
+		.data		= &init_net.ipv4.sysctl_fib_notify_on_flag_change,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= SYSCTL_ZERO,
+		.extra2		= SYSCTL_ONE,
+	},
 	{ }
 };
 

net/ipv6/af_inet6.c

@@ -954,6 +954,7 @@ static int __net_init inet6_net_init(struct net *net)
 	net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
 	net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
 	net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
+	net->ipv6.sysctl.fib_notify_on_flag_change = 0;
 	atomic_set(&net->ipv6.fib6_sernum, 1);
 
 	err = ipv6_init_mibs(net);

net/ipv6/route.c

@@ -6064,6 +6064,50 @@ void fib6_rt_update(struct net *net, struct fib6_info *rt,
 		rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
 }
 
+void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
+			    bool offload, bool trap)
+{
+	struct sk_buff *skb;
+	int err;
+
+	if (f6i->offload == offload && f6i->trap == trap)
+		return;
+
+	f6i->offload = offload;
+	f6i->trap = trap;
+
+	if (!rcu_access_pointer(f6i->fib6_node))
+		/* The route was removed from the tree, do not send
+		 * notfication.
+		 */
+		return;
+
+	if (!net->ipv6.sysctl.fib_notify_on_flag_change)
+		return;
+
+	skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
+	if (!skb) {
+		err = -ENOBUFS;
+		goto errout;
+	}
+
+	err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
+			    0, 0);
+	if (err < 0) {
+		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
+		WARN_ON(err == -EMSGSIZE);
+		kfree_skb(skb);
+		goto errout;
+	}
+
+	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
+	return;
+
+errout:
+	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
+}
+EXPORT_SYMBOL(fib6_info_hw_flags_set);
+
 static int ip6_route_dev_notify(struct notifier_block *this,
 				unsigned long event, void *ptr)
 {

net/ipv6/sysctl_net_ipv6.c

@@ -160,6 +160,15 @@ static struct ctl_table ipv6_table_template[] = {
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec
 	},
+	{
+		.procname	= "fib_notify_on_flag_change",
+		.data		= &init_net.ipv6.sysctl.fib_notify_on_flag_change,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1         = SYSCTL_ZERO,
+		.extra2         = SYSCTL_ONE,
+	},
 	{ }
 };
 

tools/testing/selftests/drivers/net/mlxsw/fib.sh

@@ -225,6 +225,16 @@ ipv6_local_replace()
 	ip -n $ns link del dev dummy1
 }
 
+fib_notify_on_flag_change_set()
+{
+	local notify=$1; shift
+
+	ip netns exec testns1 sysctl -qw net.ipv4.fib_notify_on_flag_change=$notify
+	ip netns exec testns1 sysctl -qw net.ipv6.fib_notify_on_flag_change=$notify
+
+	log_info "Set fib_notify_on_flag_change to $notify"
+}
+
 setup_prepare()
 {
 	ip netns add testns1
@@ -251,6 +261,10 @@ trap cleanup EXIT
 
 setup_prepare
 
+fib_notify_on_flag_change_set 1
+tests_run
+
+fib_notify_on_flag_change_set 0
 tests_run
 
 exit $EXIT_STATUS

tools/testing/selftests/drivers/net/netdevsim/fib.sh

@@ -302,6 +302,16 @@ ipv6_error_path()
 	ipv6_error_path_replay
 }
 
+fib_notify_on_flag_change_set()
+{
+	local notify=$1; shift
+
+	ip netns exec testns1 sysctl -qw net.ipv4.fib_notify_on_flag_change=$notify
+	ip netns exec testns1 sysctl -qw net.ipv6.fib_notify_on_flag_change=$notify
+
+	log_info "Set fib_notify_on_flag_change to $notify"
+}
+
 setup_prepare()
 {
 	local netdev
@@ -336,6 +346,10 @@ trap cleanup EXIT
 
 setup_prepare
 
+fib_notify_on_flag_change_set 1
+tests_run
+
+fib_notify_on_flag_change_set 0
 tests_run
 
 exit $EXIT_STATUS

tools/testing/selftests/drivers/net/netdevsim/fib_notifications.sh

+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+lib_dir=$(dirname $0)/../../../net/forwarding
+
+ALL_TESTS="
+	ipv4_route_addition_test
+	ipv4_route_deletion_test
+	ipv4_route_replacement_test
+	ipv6_route_addition_test
+	ipv6_route_deletion_test
+	ipv6_route_replacement_test
+"
+
+NETDEVSIM_PATH=/sys/bus/netdevsim/
+DEV_ADDR=1337
+DEV=netdevsim${DEV_ADDR}
+DEVLINK_DEV=netdevsim/${DEV}
+SYSFS_NET_DIR=/sys/bus/netdevsim/devices/$DEV/net/
+NUM_NETIFS=0
+source $lib_dir/lib.sh
+
+check_rt_trap()
+{
+	local outfile=$1; shift
+	local line
+
+	# Make sure that the first notification was emitted without RTM_F_TRAP
+	# flag and the second with RTM_F_TRAP flag
+	head -n 1 $outfile | grep -q "rt_trap"
+	if [[ $? -eq 0 ]]; then
+		return 1
+	fi
+
+	head -n 2 $outfile | tail -n 1 | grep -q "rt_trap"
+}
+
+route_notify_check()
+{
+	local outfile=$1; shift
+	local expected_num_lines=$1; shift
+
+	# check the monitor results
+	lines=`wc -l $outfile | cut "-d " -f1`
+	test $lines -eq $expected_num_lines
+	check_err $? "$expected_num_lines notifications were expected but $lines were received"
+
+	if [[ $expected_num_lines -eq 2 ]]; then
+		check_rt_trap $outfile
+		check_err $? "Wrong RTM_F_TRAP flags in notifications"
+	fi
+}
+
+route_addition_check()
+{
+	local ip=$1; shift
+	local notify=$1; shift
+	local route=$1; shift
+	local expected_num_notifications=$1; shift
+
+	ip netns exec testns1 sysctl -qw net.$ip.fib_notify_on_flag_change=$notify
+
+	local outfile=$(mktemp)
+
+	$IP monitor route &> $outfile &
+	sleep 1
+	$IP route add $route dev dummy1
+	sleep 1
+	kill %% && wait %% &> /dev/null
+
+	route_notify_check $outfile $expected_num_notifications
+	rm -f $outfile
+
+	$IP route del $route dev dummy1
+}
+
+ipv4_route_addition_test()
+{
+	RET=0
+
+	local ip="ipv4"
+	local route=192.0.2.0/24
+
+	# Make sure a single notification will be emitted for the programmed
+	# route.
+	local notify=0
+	local expected_num_notifications=1
+	# route_addition_check will assign value to RET.
+	route_addition_check $ip $notify $route $expected_num_notifications
+
+	# Make sure two notifications will be emitted for the programmed route.
+	notify=1
+	expected_num_notifications=2
+	route_addition_check $ip $notify $route $expected_num_notifications
+
+	log_test "IPv4 route addition"
+}
+
+route_deletion_check()
+{
+	local ip=$1; shift
+	local notify=$1; shift
+	local route=$1; shift
+	local expected_num_notifications=$1; shift
+
+	ip netns exec testns1 sysctl -qw net.$ip.fib_notify_on_flag_change=$notify
+	$IP route add $route dev dummy1
+	sleep 1
+
+	local outfile=$(mktemp)
+
+	$IP monitor route &> $outfile &
+	sleep 1
+	$IP route del $route dev dummy1
+	sleep 1
+	kill %% && wait %% &> /dev/null
+
+	route_notify_check $outfile $expected_num_notifications
+	rm -f $outfile
+}
+
+ipv4_route_deletion_test()
+{
+	RET=0
+
+	local ip="ipv4"
+	local route=192.0.2.0/24
+	local expected_num_notifications=1
+
+	# Make sure a single notification will be emitted for the deleted route,
+	# regardless of fib_notify_on_flag_change value.
+	local notify=0
+	# route_deletion_check will assign value to RET.
+	route_deletion_check $ip $notify $route $expected_num_notifications
+
+	notify=1
+	route_deletion_check $ip $notify $route $expected_num_notifications
+
+	log_test "IPv4 route deletion"
+}
+
+route_replacement_check()
+{
+	local ip=$1; shift
+	local notify=$1; shift
+	local route=$1; shift
+	local expected_num_notifications=$1; shift
+
+	ip netns exec testns1 sysctl -qw net.$ip.fib_notify_on_flag_change=$notify
+	$IP route add $route dev dummy1
+	sleep 1
+
+	local outfile=$(mktemp)
+
+	$IP monitor route &> $outfile &
+	sleep 1
+	$IP route replace $route dev dummy2
+	sleep 1
+	kill %% && wait %% &> /dev/null
+
+	route_notify_check $outfile $expected_num_notifications
+	rm -f $outfile
+
+	$IP route del $route dev dummy2
+}
+
+ipv4_route_replacement_test()
+{
+	RET=0
+
+	local ip="ipv4"
+	local route=192.0.2.0/24
+
+	$IP link add name dummy2 type dummy
+	$IP link set dev dummy2 up
+
+	# Make sure a single notification will be emitted for the new route.
+	local notify=0
+	local expected_num_notifications=1
+	# route_replacement_check will assign value to RET.
+	route_replacement_check $ip $notify $route $expected_num_notifications
+
+	# Make sure two notifications will be emitted for the new route.
+	notify=1
+	expected_num_notifications=2
+	route_replacement_check $ip $notify $route $expected_num_notifications
+
+	$IP link del name dummy2
+
+	log_test "IPv4 route replacement"
+}
+
+ipv6_route_addition_test()
+{
+	RET=0
+
+	local ip="ipv6"
+	local route=2001:db8:1::/64
+
+	# Make sure a single notification will be emitted for the programmed
+	# route.
+	local notify=0
+	local expected_num_notifications=1
+	route_addition_check $ip $notify $route $expected_num_notifications
+
+	# Make sure two notifications will be emitted for the programmed route.
+	notify=1
+	expected_num_notifications=2
+	route_addition_check $ip $notify $route $expected_num_notifications
+
+	log_test "IPv6 route addition"
+}
+
+ipv6_route_deletion_test()
+{
+	RET=0
+
+	local ip="ipv6"
+	local route=2001:db8:1::/64
+	local expected_num_notifications=1
+
+	# Make sure a single notification will be emitted for the deleted route,
+	# regardless of fib_notify_on_flag_change value.
+	local notify=0
+	route_deletion_check $ip $notify $route $expected_num_notifications
+
+	notify=1
+	route_deletion_check $ip $notify $route $expected_num_notifications
+
+	log_test "IPv6 route deletion"
+}
+
+ipv6_route_replacement_test()
+{
+	RET=0
+
+	local ip="ipv6"
+	local route=2001:db8:1::/64
+
+	$IP link add name dummy2 type dummy
+	$IP link set dev dummy2 up
+
+	# Make sure a single notification will be emitted for the new route.
+	local notify=0
+	local expected_num_notifications=1
+	route_replacement_check $ip $notify $route $expected_num_notifications
+
+	# Make sure two notifications will be emitted for the new route.
+	notify=1
+	expected_num_notifications=2
+	route_replacement_check $ip $notify $route $expected_num_notifications
+
+	$IP link del name dummy2
+
+	log_test "IPv6 route replacement"
+}
+
+setup_prepare()
+{
+	modprobe netdevsim &> /dev/null
+	echo "$DEV_ADDR 1" > ${NETDEVSIM_PATH}/new_device
+	while [ ! -d $SYSFS_NET_DIR ] ; do :; done
+
+	ip netns add testns1
+
+	if [ $? -ne 0 ]; then
+		echo "Failed to add netns \"testns1\""
+		exit 1
+	fi
+
+	devlink dev reload $DEVLINK_DEV netns testns1
+
+	if [ $? -ne 0 ]; then
+		echo "Failed to reload into netns \"testns1\""
+		exit 1
+	fi
+
+	IP="ip -n testns1"
+
+	$IP link add name dummy1 type dummy
+	$IP link set dev dummy1 up
+}
+
+cleanup()
+{
+	pre_cleanup
+
+	$IP link del name dummy1
+	ip netns del testns1
+	echo "$DEV_ADDR" > ${NETDEVSIM_PATH}/del_device
+	modprobe -r netdevsim &> /dev/null
+}
+
+trap cleanup EXIT
+
+setup_prepare
+
+tests_run
+
+exit $EXIT_STATUS