Merge branch 'net-dsa-link-aggregation-support'

Tobias Waldekranz says:

====================
net: dsa: Link aggregation support

Start of by adding an extra notification when adding a port to a bond,
this allows static LAGs to be offloaded using the bonding driver.

Then add the generic support required to offload link aggregates to
drivers built on top of the DSA subsystem.

Finally, implement offloading for the mv88e6xxx driver, i.e. Marvell's
LinkStreet family.

Supported LAG implementations:
- Bonding
- Team

Supported modes:
- Isolated. The LAG may be used as a regular interface outside of any
  bridge.
- Bridged. The LAG may be added to a bridge, in which case switching
  is offloaded between the LAG and any other switch ports. I.e. the
  LAG behaves just like a port from this perspective.

In bridged mode, the following is supported:
- STP filtering.
- VLAN filtering.
- Multicast filtering. The bridge correctly snoops IGMP and configures
  the proper groups if snooping is enabled. Static groups can also be
  configured. MLD seems to work, but has not been extensively tested.
- Unicast filtering. Automatic learning works. Static entries are
  _not_ supported. This will be added in a later series as it requires
  some more general refactoring in mv88e6xxx before I can test it.

v4 -> v5:
- Cleanup PVT configuration for LAGed ports in mv88e6xxx (Vladimir)
- Document dsa_lag_{map,unmap} (Vladimir)
====================

Link: https://lore.kernel.org/r/20210113084255.22675-1-tobias@waldekranz.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

drivers/net/bonding/bond_main.c

@@ -1922,6 +1922,8 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev,
 		goto err_unregister;
 	}
 
+	bond_lower_state_changed(new_slave);
+
 	res = bond_sysfs_slave_add(new_slave);
 	if (res) {
 		slave_dbg(bond_dev, slave_dev, "Error %d calling bond_sysfs_slave_add\n", res);

drivers/net/dsa/mv88e6xxx/chip.c

@@ -1396,15 +1396,32 @@ static int mv88e6xxx_mac_setup(struct mv88e6xxx_chip *chip)
 
 static int mv88e6xxx_pvt_map(struct mv88e6xxx_chip *chip, int dev, int port)
 {
+	struct dsa_switch_tree *dst = chip->ds->dst;
+	struct dsa_switch *ds;
+	struct dsa_port *dp;
 	u16 pvlan = 0;
 
 	if (!mv88e6xxx_has_pvt(chip))
 		return 0;
 
 	/* Skip the local source device, which uses in-chip port VLAN */
-	if (dev != chip->ds->index)
+	if (dev != chip->ds->index) {
 		pvlan = mv88e6xxx_port_vlan(chip, dev, port);
 
+		ds = dsa_switch_find(dst->index, dev);
+		dp = ds ? dsa_to_port(ds, port) : NULL;
+		if (dp && dp->lag_dev) {
+			/* As the PVT is used to limit flooding of
+			 * FORWARD frames, which use the LAG ID as the
+			 * source port, we must translate dev/port to
+			 * the special "LAG device" in the PVT, using
+			 * the LAG ID as the port number.
+			 */
+			dev = MV88E6XXX_G2_PVT_ADRR_DEV_TRUNK;
+			port = dsa_lag_id(dst, dp->lag_dev);
+		}
+	}
+
 	return mv88e6xxx_g2_pvt_write(chip, dev, port, pvlan);
 }
 
@@ -5364,6 +5381,271 @@ static int mv88e6xxx_port_egress_floods(struct dsa_switch *ds, int port,
 	return err;
 }
 
+static bool mv88e6xxx_lag_can_offload(struct dsa_switch *ds,
+				      struct net_device *lag,
+				      struct netdev_lag_upper_info *info)
+{
+	struct dsa_port *dp;
+	int id, members = 0;
+
+	id = dsa_lag_id(ds->dst, lag);
+	if (id < 0 || id >= ds->num_lag_ids)
+		return false;
+
+	dsa_lag_foreach_port(dp, ds->dst, lag)
+		/* Includes the port joining the LAG */
+		members++;
+
+	if (members > 8)
+		return false;
+
+	/* We could potentially relax this to include active
+	 * backup in the future.
+	 */
+	if (info->tx_type != NETDEV_LAG_TX_TYPE_HASH)
+		return false;
+
+	/* Ideally we would also validate that the hash type matches
+	 * the hardware. Alas, this is always set to unknown on team
+	 * interfaces.
+	 */
+	return true;
+}
+
+static int mv88e6xxx_lag_sync_map(struct dsa_switch *ds, struct net_device *lag)
+{
+	struct mv88e6xxx_chip *chip = ds->priv;
+	struct dsa_port *dp;
+	u16 map = 0;
+	int id;
+
+	id = dsa_lag_id(ds->dst, lag);
+
+	/* Build the map of all ports to distribute flows destined for
+	 * this LAG. This can be either a local user port, or a DSA
+	 * port if the LAG port is on a remote chip.
+	 */
+	dsa_lag_foreach_port(dp, ds->dst, lag)
+		map |= BIT(dsa_towards_port(ds, dp->ds->index, dp->index));
+
+	return mv88e6xxx_g2_trunk_mapping_write(chip, id, map);
+}
+
+static const u8 mv88e6xxx_lag_mask_table[8][8] = {
+	/* Row number corresponds to the number of active members in a
+	 * LAG. Each column states which of the eight hash buckets are
+	 * mapped to the column:th port in the LAG.
+	 *
+	 * Example: In a LAG with three active ports, the second port
+	 * ([2][1]) would be selected for traffic mapped to buckets
+	 * 3,4,5 (0x38).
+	 */
+	{ 0xff,    0,    0,    0,    0,    0,    0,    0 },
+	{ 0x0f, 0xf0,    0,    0,    0,    0,    0,    0 },
+	{ 0x07, 0x38, 0xc0,    0,    0,    0,    0,    0 },
+	{ 0x03, 0x0c, 0x30, 0xc0,    0,    0,    0,    0 },
+	{ 0x03, 0x0c, 0x30, 0x40, 0x80,    0,    0,    0 },
+	{ 0x03, 0x0c, 0x10, 0x20, 0x40, 0x80,    0,    0 },
+	{ 0x03, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,    0 },
+	{ 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 },
+};
+
+static void mv88e6xxx_lag_set_port_mask(u16 *mask, int port,
+					int num_tx, int nth)
+{
+	u8 active = 0;
+	int i;
+
+	num_tx = num_tx <= 8 ? num_tx : 8;
+	if (nth < num_tx)
+		active = mv88e6xxx_lag_mask_table[num_tx - 1][nth];
+
+	for (i = 0; i < 8; i++) {
+		if (BIT(i) & active)
+			mask[i] |= BIT(port);
+	}
+}
+
+static int mv88e6xxx_lag_sync_masks(struct dsa_switch *ds)
+{
+	struct mv88e6xxx_chip *chip = ds->priv;
+	unsigned int id, num_tx;
+	struct net_device *lag;
+	struct dsa_port *dp;
+	int i, err, nth;
+	u16 mask[8];
+	u16 ivec;
+
+	/* Assume no port is a member of any LAG. */
+	ivec = BIT(mv88e6xxx_num_ports(chip)) - 1;
+
+	/* Disable all masks for ports that _are_ members of a LAG. */
+	list_for_each_entry(dp, &ds->dst->ports, list) {
+		if (!dp->lag_dev || dp->ds != ds)
+			continue;
+
+		ivec &= ~BIT(dp->index);
+	}
+
+	for (i = 0; i < 8; i++)
+		mask[i] = ivec;
+
+	/* Enable the correct subset of masks for all LAG ports that
+	 * are in the Tx set.
+	 */
+	dsa_lags_foreach_id(id, ds->dst) {
+		lag = dsa_lag_dev(ds->dst, id);
+		if (!lag)
+			continue;
+
+		num_tx = 0;
+		dsa_lag_foreach_port(dp, ds->dst, lag) {
+			if (dp->lag_tx_enabled)
+				num_tx++;
+		}
+
+		if (!num_tx)
+			continue;
+
+		nth = 0;
+		dsa_lag_foreach_port(dp, ds->dst, lag) {
+			if (!dp->lag_tx_enabled)
+				continue;
+
+			if (dp->ds == ds)
+				mv88e6xxx_lag_set_port_mask(mask, dp->index,
+							    num_tx, nth);
+
+			nth++;
+		}
+	}
+
+	for (i = 0; i < 8; i++) {
+		err = mv88e6xxx_g2_trunk_mask_write(chip, i, true, mask[i]);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static int mv88e6xxx_lag_sync_masks_map(struct dsa_switch *ds,
+					struct net_device *lag)
+{
+	int err;
+
+	err = mv88e6xxx_lag_sync_masks(ds);
+
+	if (!err)
+		err = mv88e6xxx_lag_sync_map(ds, lag);
+
+	return err;
+}
+
+static int mv88e6xxx_port_lag_change(struct dsa_switch *ds, int port)
+{
+	struct mv88e6xxx_chip *chip = ds->priv;
+	int err;
+
+	mv88e6xxx_reg_lock(chip);
+	err = mv88e6xxx_lag_sync_masks(ds);
+	mv88e6xxx_reg_unlock(chip);
+	return err;
+}
+
+static int mv88e6xxx_port_lag_join(struct dsa_switch *ds, int port,
+				   struct net_device *lag,
+				   struct netdev_lag_upper_info *info)
+{
+	struct mv88e6xxx_chip *chip = ds->priv;
+	int err, id;
+
+	if (!mv88e6xxx_lag_can_offload(ds, lag, info))
+		return -EOPNOTSUPP;
+
+	id = dsa_lag_id(ds->dst, lag);
+
+	mv88e6xxx_reg_lock(chip);
+
+	err = mv88e6xxx_port_set_trunk(chip, port, true, id);
+	if (err)
+		goto err_unlock;
+
+	err = mv88e6xxx_lag_sync_masks_map(ds, lag);
+	if (err)
+		goto err_clear_trunk;
+
+	mv88e6xxx_reg_unlock(chip);
+	return 0;
+
+err_clear_trunk:
+	mv88e6xxx_port_set_trunk(chip, port, false, 0);
+err_unlock:
+	mv88e6xxx_reg_unlock(chip);
+	return err;
+}
+
+static int mv88e6xxx_port_lag_leave(struct dsa_switch *ds, int port,
+				    struct net_device *lag)
+{
+	struct mv88e6xxx_chip *chip = ds->priv;
+	int err_sync, err_trunk;
+
+	mv88e6xxx_reg_lock(chip);
+	err_sync = mv88e6xxx_lag_sync_masks_map(ds, lag);
+	err_trunk = mv88e6xxx_port_set_trunk(chip, port, false, 0);
+	mv88e6xxx_reg_unlock(chip);
+	return err_sync ? : err_trunk;
+}
+
+static int mv88e6xxx_crosschip_lag_change(struct dsa_switch *ds, int sw_index,
+					  int port)
+{
+	struct mv88e6xxx_chip *chip = ds->priv;
+	int err;
+
+	mv88e6xxx_reg_lock(chip);
+	err = mv88e6xxx_lag_sync_masks(ds);
+	mv88e6xxx_reg_unlock(chip);
+	return err;
+}
+
+static int mv88e6xxx_crosschip_lag_join(struct dsa_switch *ds, int sw_index,
+					int port, struct net_device *lag,
+					struct netdev_lag_upper_info *info)
+{
+	struct mv88e6xxx_chip *chip = ds->priv;
+	int err;
+
+	if (!mv88e6xxx_lag_can_offload(ds, lag, info))
+		return -EOPNOTSUPP;
+
+	mv88e6xxx_reg_lock(chip);
+
+	err = mv88e6xxx_lag_sync_masks_map(ds, lag);
+	if (err)
+		goto unlock;
+
+	err = mv88e6xxx_pvt_map(chip, sw_index, port);
+
+unlock:
+	mv88e6xxx_reg_unlock(chip);
+	return err;
+}
+
+static int mv88e6xxx_crosschip_lag_leave(struct dsa_switch *ds, int sw_index,
+					 int port, struct net_device *lag)
+{
+	struct mv88e6xxx_chip *chip = ds->priv;
+	int err_sync, err_pvt;
+
+	mv88e6xxx_reg_lock(chip);
+	err_sync = mv88e6xxx_lag_sync_masks_map(ds, lag);
+	err_pvt = mv88e6xxx_pvt_map(chip, sw_index, port);
+	mv88e6xxx_reg_unlock(chip);
+	return err_sync ? : err_pvt;
+}
+
 static const struct dsa_switch_ops mv88e6xxx_switch_ops = {
 	.get_tag_protocol	= mv88e6xxx_get_tag_protocol,
 	.setup			= mv88e6xxx_setup,
@@ -5416,6 +5698,12 @@ static const struct dsa_switch_ops mv88e6xxx_switch_ops = {
 	.devlink_param_get	= mv88e6xxx_devlink_param_get,
 	.devlink_param_set	= mv88e6xxx_devlink_param_set,
 	.devlink_info_get	= mv88e6xxx_devlink_info_get,
+	.port_lag_change	= mv88e6xxx_port_lag_change,
+	.port_lag_join		= mv88e6xxx_port_lag_join,
+	.port_lag_leave		= mv88e6xxx_port_lag_leave,
+	.crosschip_lag_change	= mv88e6xxx_crosschip_lag_change,
+	.crosschip_lag_join	= mv88e6xxx_crosschip_lag_join,
+	.crosschip_lag_leave	= mv88e6xxx_crosschip_lag_leave,
 };
 
 static int mv88e6xxx_register_switch(struct mv88e6xxx_chip *chip)
@@ -5435,6 +5723,12 @@ static int mv88e6xxx_register_switch(struct mv88e6xxx_chip *chip)
 	ds->ageing_time_min = chip->info->age_time_coeff;
 	ds->ageing_time_max = chip->info->age_time_coeff * U8_MAX;
 
+	/* Some chips support up to 32, but that requires enabling the
+	 * 5-bit port mode, which we do not support. 640k^W16 ought to
+	 * be enough for anyone.
+	 */
+	ds->num_lag_ids = 16;
+
 	dev_set_drvdata(dev, ds);
 
 	return dsa_register_switch(ds);

drivers/net/dsa/mv88e6xxx/global2.c

@@ -126,8 +126,8 @@ int mv88e6xxx_g2_device_mapping_write(struct mv88e6xxx_chip *chip, int target,
 
 /* Offset 0x07: Trunk Mask Table register */
 
-static int mv88e6xxx_g2_trunk_mask_write(struct mv88e6xxx_chip *chip, int num,
-					 bool hash, u16 mask)
+int mv88e6xxx_g2_trunk_mask_write(struct mv88e6xxx_chip *chip, int num,
+				  bool hash, u16 mask)
 {
 	u16 val = (num << 12) | (mask & mv88e6xxx_port_mask(chip));
 
@@ -140,8 +140,8 @@ static int mv88e6xxx_g2_trunk_mask_write(struct mv88e6xxx_chip *chip, int num,
 
 /* Offset 0x08: Trunk Mapping Table register */
 
-static int mv88e6xxx_g2_trunk_mapping_write(struct mv88e6xxx_chip *chip, int id,
-					    u16 map)
+int mv88e6xxx_g2_trunk_mapping_write(struct mv88e6xxx_chip *chip, int id,
+				     u16 map)
 {
 	const u16 port_mask = BIT(mv88e6xxx_num_ports(chip)) - 1;
 	u16 val = (id << 11) | (map & port_mask);

drivers/net/dsa/mv88e6xxx/global2.h

@@ -101,6 +101,7 @@
 #define MV88E6XXX_G2_PVT_ADDR_OP_WRITE_PVLAN	0x3000
 #define MV88E6XXX_G2_PVT_ADDR_OP_READ		0x4000
 #define MV88E6XXX_G2_PVT_ADDR_PTR_MASK		0x01ff
+#define MV88E6XXX_G2_PVT_ADRR_DEV_TRUNK		0x1f
 
 /* Offset 0x0C: Cross-chip Port VLAN Data Register */
 #define MV88E6XXX_G2_PVT_DATA		0x0c
@@ -345,6 +346,10 @@ int mv88e6352_g2_mgmt_rsvd2cpu(struct mv88e6xxx_chip *chip);
 
 int mv88e6xxx_g2_pot_clear(struct mv88e6xxx_chip *chip);
 
+int mv88e6xxx_g2_trunk_mask_write(struct mv88e6xxx_chip *chip, int num,
+				  bool hash, u16 mask);
+int mv88e6xxx_g2_trunk_mapping_write(struct mv88e6xxx_chip *chip, int id,
+				     u16 map);
 int mv88e6xxx_g2_trunk_clear(struct mv88e6xxx_chip *chip);
 
 int mv88e6xxx_g2_device_mapping_write(struct mv88e6xxx_chip *chip, int target,

drivers/net/dsa/mv88e6xxx/port.c

@@ -851,6 +851,27 @@ int mv88e6xxx_port_set_message_port(struct mv88e6xxx_chip *chip, int port,
 	return mv88e6xxx_port_write(chip, port, MV88E6XXX_PORT_CTL1, val);
 }
 
+int mv88e6xxx_port_set_trunk(struct mv88e6xxx_chip *chip, int port,
+			     bool trunk, u8 id)
+{
+	u16 val;
+	int err;
+
+	err = mv88e6xxx_port_read(chip, port, MV88E6XXX_PORT_CTL1, &val);
+	if (err)
+		return err;
+
+	val &= ~MV88E6XXX_PORT_CTL1_TRUNK_ID_MASK;
+
+	if (trunk)
+		val |= MV88E6XXX_PORT_CTL1_TRUNK_PORT |
+			(id << MV88E6XXX_PORT_CTL1_TRUNK_ID_SHIFT);
+	else
+		val &= ~MV88E6XXX_PORT_CTL1_TRUNK_PORT;
+
+	return mv88e6xxx_port_write(chip, port, MV88E6XXX_PORT_CTL1, val);
+}
+
 /* Offset 0x06: Port Based VLAN Map */
 
 int mv88e6xxx_port_set_vlan_map(struct mv88e6xxx_chip *chip, int port, u16 map)

drivers/net/dsa/mv88e6xxx/port.h

@@ -168,6 +168,9 @@
 /* Offset 0x05: Port Control 1 */
 #define MV88E6XXX_PORT_CTL1			0x05
 #define MV88E6XXX_PORT_CTL1_MESSAGE_PORT	0x8000
+#define MV88E6XXX_PORT_CTL1_TRUNK_PORT		0x4000
+#define MV88E6XXX_PORT_CTL1_TRUNK_ID_MASK	0x0f00
+#define MV88E6XXX_PORT_CTL1_TRUNK_ID_SHIFT	8
 #define MV88E6XXX_PORT_CTL1_FID_11_4_MASK	0x00ff
 
 /* Offset 0x06: Port Based VLAN Map */
@@ -351,6 +354,8 @@ int mv88e6351_port_set_ether_type(struct mv88e6xxx_chip *chip, int port,
 				  u16 etype);
 int mv88e6xxx_port_set_message_port(struct mv88e6xxx_chip *chip, int port,
 				    bool message_port);
+int mv88e6xxx_port_set_trunk(struct mv88e6xxx_chip *chip, int port,
+			     bool trunk, u8 id);
 int mv88e6165_port_set_jumbo_size(struct mv88e6xxx_chip *chip, int port,
 				  size_t size);
 int mv88e6095_port_egress_rate_limiting(struct mv88e6xxx_chip *chip, int port);

include/net/dsa.h

@@ -149,8 +149,41 @@ struct dsa_switch_tree {
 
 	/* List of DSA links composing the routing table */
 	struct list_head rtable;
+
+	/* Maps offloaded LAG netdevs to a zero-based linear ID for
+	 * drivers that need it.
+	 */
+	struct net_device **lags;
+	unsigned int lags_len;
 };
 
+#define dsa_lags_foreach_id(_id, _dst)				\
+	for ((_id) = 0; (_id) < (_dst)->lags_len; (_id)++)	\
+		if ((_dst)->lags[(_id)])
+
+#define dsa_lag_foreach_port(_dp, _dst, _lag)			\
+	list_for_each_entry((_dp), &(_dst)->ports, list)	\
+		if ((_dp)->lag_dev == (_lag))
+
+static inline struct net_device *dsa_lag_dev(struct dsa_switch_tree *dst,
+					     unsigned int id)
+{
+	return dst->lags[id];
+}
+
+static inline int dsa_lag_id(struct dsa_switch_tree *dst,
+			     struct net_device *lag)
+{
+	unsigned int id;
+
+	dsa_lags_foreach_id(id, dst) {
+		if (dsa_lag_dev(dst, id) == lag)
+			return id;
+	}
+
+	return -ENODEV;
+}
+
 /* TC matchall action types */
 enum dsa_port_mall_action_type {
 	DSA_PORT_MALL_MIRROR,
@@ -220,6 +253,8 @@ struct dsa_port {
 	bool			devlink_port_setup;
 	struct phylink		*pl;
 	struct phylink_config	pl_config;
+	struct net_device	*lag_dev;
+	bool			lag_tx_enabled;
 
 	struct list_head list;
 
@@ -340,6 +375,14 @@ struct dsa_switch {
 	 */
 	bool			mtu_enforcement_ingress;
 
+	/* Drivers that benefit from having an ID associated with each
+	 * offloaded LAG should set this to the maximum number of
+	 * supported IDs. DSA will then maintain a mapping of _at
+	 * least_ these many IDs, accessible to drivers via
+	 * dsa_lag_id().
+	 */
+	unsigned int		num_lag_ids;
+
 	size_t num_ports;
 };
 
@@ -626,6 +669,13 @@ struct dsa_switch_ops {
 	void	(*crosschip_bridge_leave)(struct dsa_switch *ds, int tree_index,
 					  int sw_index, int port,
 					  struct net_device *br);
+	int	(*crosschip_lag_change)(struct dsa_switch *ds, int sw_index,
+					int port);
+	int	(*crosschip_lag_join)(struct dsa_switch *ds, int sw_index,
+				      int port, struct net_device *lag,
+				      struct netdev_lag_upper_info *info);
+	int	(*crosschip_lag_leave)(struct dsa_switch *ds, int sw_index,
+				       int port, struct net_device *lag);
 
 	/*
 	 * PTP functionality
@@ -657,6 +707,16 @@ struct dsa_switch_ops {
 	int	(*port_change_mtu)(struct dsa_switch *ds, int port,
 				   int new_mtu);
 	int	(*port_max_mtu)(struct dsa_switch *ds, int port);
+
+	/*
+	 * LAG integration
+	 */
+	int	(*port_lag_change)(struct dsa_switch *ds, int port);
+	int	(*port_lag_join)(struct dsa_switch *ds, int port,
+				 struct net_device *lag,
+				 struct netdev_lag_upper_info *info);
+	int	(*port_lag_leave)(struct dsa_switch *ds, int port,
+				  struct net_device *lag);
 };
 
 #define DSA_DEVLINK_PARAM_DRIVER(_id, _name, _type, _cmodes)		\

net/dsa/dsa.c

@@ -219,11 +219,21 @@ static int dsa_switch_rcv(struct sk_buff *skb, struct net_device *dev,
 	}
 
 	skb = nskb;
-	p = netdev_priv(skb->dev);
 	skb_push(skb, ETH_HLEN);
 	skb->pkt_type = PACKET_HOST;
 	skb->protocol = eth_type_trans(skb, skb->dev);
 
+	if (unlikely(!dsa_slave_dev_check(skb->dev))) {
+		/* Packet is to be injected directly on an upper
+		 * device, e.g. a team/bond, so skip all DSA-port
+		 * specific actions.
+		 */
+		netif_rx(skb);
+		return 0;
+	}
+
+	p = netdev_priv(skb->dev);
+
 	if (unlikely(cpu_dp->ds->untag_bridge_pvid)) {
 		nskb = dsa_untag_bridge_pvid(skb);
 		if (!nskb) {

net/dsa/dsa2.c

@@ -21,6 +21,65 @@
 static DEFINE_MUTEX(dsa2_mutex);
 LIST_HEAD(dsa_tree_list);
 
+/**
+ * dsa_lag_map() - Map LAG netdev to a linear LAG ID
+ * @dst: Tree in which to record the mapping.
+ * @lag: Netdev that is to be mapped to an ID.
+ *
+ * dsa_lag_id/dsa_lag_dev can then be used to translate between the
+ * two spaces. The size of the mapping space is determined by the
+ * driver by setting ds->num_lag_ids. It is perfectly legal to leave
+ * it unset if it is not needed, in which case these functions become
+ * no-ops.
+ */
+void dsa_lag_map(struct dsa_switch_tree *dst, struct net_device *lag)
+{
+	unsigned int id;
+
+	if (dsa_lag_id(dst, lag) >= 0)
+		/* Already mapped */
+		return;
+
+	for (id = 0; id < dst->lags_len; id++) {
+		if (!dsa_lag_dev(dst, id)) {
+			dst->lags[id] = lag;
+			return;
+		}
+	}
+
+	/* No IDs left, which is OK. Some drivers do not need it. The
+	 * ones that do, e.g. mv88e6xxx, will discover that dsa_lag_id
+	 * returns an error for this device when joining the LAG. The
+	 * driver can then return -EOPNOTSUPP back to DSA, which will
+	 * fall back to a software LAG.
+	 */
+}
+
+/**
+ * dsa_lag_unmap() - Remove a LAG ID mapping
+ * @dst: Tree in which the mapping is recorded.
+ * @lag: Netdev that was mapped.
+ *
+ * As there may be multiple users of the mapping, it is only removed
+ * if there are no other references to it.
+ */
+void dsa_lag_unmap(struct dsa_switch_tree *dst, struct net_device *lag)
+{
+	struct dsa_port *dp;
+	unsigned int id;
+
+	dsa_lag_foreach_port(dp, dst, lag)
+		/* There are remaining users of this mapping */
+		return;
+
+	dsa_lags_foreach_id(id, dst) {
+		if (dsa_lag_dev(dst, id) == lag) {
+			dst->lags[id] = NULL;
+			break;
+		}
+	}
+}
+
 struct dsa_switch *dsa_switch_find(int tree_index, int sw_index)
 {
 	struct dsa_switch_tree *dst;
@@ -578,6 +637,32 @@ static void dsa_tree_teardown_master(struct dsa_switch_tree *dst)
 			dsa_master_teardown(dp->master);
 }
 
+static int dsa_tree_setup_lags(struct dsa_switch_tree *dst)
+{
+	unsigned int len = 0;
+	struct dsa_port *dp;
+
+	list_for_each_entry(dp, &dst->ports, list) {
+		if (dp->ds->num_lag_ids > len)
+			len = dp->ds->num_lag_ids;
+	}
+
+	if (!len)
+		return 0;
+
+	dst->lags = kcalloc(len, sizeof(*dst->lags), GFP_KERNEL);
+	if (!dst->lags)
+		return -ENOMEM;
+
+	dst->lags_len = len;
+	return 0;
+}
+
+static void dsa_tree_teardown_lags(struct dsa_switch_tree *dst)
+{
+	kfree(dst->lags);
+}
+
 static int dsa_tree_setup(struct dsa_switch_tree *dst)
 {
 	bool complete;
@@ -605,12 +690,18 @@ static int dsa_tree_setup(struct dsa_switch_tree *dst)
 	if (err)
 		goto teardown_switches;
 
+	err = dsa_tree_setup_lags(dst);
+	if (err)
+		goto teardown_master;
+
 	dst->setup = true;
 
 	pr_info("DSA: tree %d setup\n", dst->index);
 
 	return 0;
 
+teardown_master:
+	dsa_tree_teardown_master(dst);
 teardown_switches:
 	dsa_tree_teardown_switches(dst);
 teardown_default_cpu:
@@ -626,6 +717,8 @@ static void dsa_tree_teardown(struct dsa_switch_tree *dst)
 	if (!dst->setup)
 		return;
 
+	dsa_tree_teardown_lags(dst);
+
 	dsa_tree_teardown_master(dst);
 
 	dsa_tree_teardown_switches(dst);

net/dsa/dsa_priv.h

@@ -20,6 +20,9 @@ enum {
 	DSA_NOTIFIER_BRIDGE_LEAVE,
 	DSA_NOTIFIER_FDB_ADD,
 	DSA_NOTIFIER_FDB_DEL,
+	DSA_NOTIFIER_LAG_CHANGE,
+	DSA_NOTIFIER_LAG_JOIN,
+	DSA_NOTIFIER_LAG_LEAVE,
 	DSA_NOTIFIER_MDB_ADD,
 	DSA_NOTIFIER_MDB_DEL,
 	DSA_NOTIFIER_VLAN_ADD,
@@ -55,6 +58,15 @@ struct dsa_notifier_mdb_info {
 	int port;
 };
 
+/* DSA_NOTIFIER_LAG_* */
+struct dsa_notifier_lag_info {
+	struct net_device *lag;
+	int sw_index;
+	int port;
+
+	struct netdev_lag_upper_info *info;
+};
+
 /* DSA_NOTIFIER_VLAN_* */
 struct dsa_notifier_vlan_info {
 	const struct switchdev_obj_port_vlan *vlan;
@@ -134,6 +146,11 @@ void dsa_port_disable_rt(struct dsa_port *dp);
 void dsa_port_disable(struct dsa_port *dp);
 int dsa_port_bridge_join(struct dsa_port *dp, struct net_device *br);
 void dsa_port_bridge_leave(struct dsa_port *dp, struct net_device *br);
+int dsa_port_lag_change(struct dsa_port *dp,
+			struct netdev_lag_lower_state_info *linfo);
+int dsa_port_lag_join(struct dsa_port *dp, struct net_device *lag_dev,
+		      struct netdev_lag_upper_info *uinfo);
+void dsa_port_lag_leave(struct dsa_port *dp, struct net_device *lag_dev);
 int dsa_port_vlan_filtering(struct dsa_port *dp, bool vlan_filtering);
 bool dsa_port_skip_vlan_configuration(struct dsa_port *dp);
 int dsa_port_ageing_time(struct dsa_port *dp, clock_t ageing_clock);
@@ -159,6 +176,22 @@ int dsa_port_link_register_of(struct dsa_port *dp);
 void dsa_port_link_unregister_of(struct dsa_port *dp);
 extern const struct phylink_mac_ops dsa_port_phylink_mac_ops;
 
+static inline bool dsa_port_offloads_netdev(struct dsa_port *dp,
+					    struct net_device *dev)
+{
+	/* Switchdev offloading can be configured on: */
+
+	if (dev == dp->slave)
+		/* DSA ports directly connected to a bridge. */
+		return true;
+
+	if (dp->lag_dev == dev)
+		/* DSA ports connected to a bridge via a LAG */
+		return true;
+
+	return false;
+}
+
 /* slave.c */
 extern const struct dsa_device_ops notag_netdev_ops;
 void dsa_slave_mii_bus_init(struct dsa_switch *ds);
@@ -248,6 +281,9 @@ int dsa_switch_register_notifier(struct dsa_switch *ds);
 void dsa_switch_unregister_notifier(struct dsa_switch *ds);
 
 /* dsa2.c */
+void dsa_lag_map(struct dsa_switch_tree *dst, struct net_device *lag);
+void dsa_lag_unmap(struct dsa_switch_tree *dst, struct net_device *lag);
+
 extern struct list_head dsa_tree_list;
 
 #endif

net/dsa/port.c

@@ -191,6 +191,85 @@ void dsa_port_bridge_leave(struct dsa_port *dp, struct net_device *br)
 	dsa_port_set_state_now(dp, BR_STATE_FORWARDING);
 }
 
+int dsa_port_lag_change(struct dsa_port *dp,
+			struct netdev_lag_lower_state_info *linfo)
+{
+	struct dsa_notifier_lag_info info = {
+		.sw_index = dp->ds->index,
+		.port = dp->index,
+	};
+	bool tx_enabled;
+
+	if (!dp->lag_dev)
+		return 0;
+
+	/* On statically configured aggregates (e.g. loadbalance
+	 * without LACP) ports will always be tx_enabled, even if the
+	 * link is down. Thus we require both link_up and tx_enabled
+	 * in order to include it in the tx set.
+	 */
+	tx_enabled = linfo->link_up && linfo->tx_enabled;
+
+	if (tx_enabled == dp->lag_tx_enabled)
+		return 0;
+
+	dp->lag_tx_enabled = tx_enabled;
+
+	return dsa_port_notify(dp, DSA_NOTIFIER_LAG_CHANGE, &info);
+}
+
+int dsa_port_lag_join(struct dsa_port *dp, struct net_device *lag,
+		      struct netdev_lag_upper_info *uinfo)
+{
+	struct dsa_notifier_lag_info info = {
+		.sw_index = dp->ds->index,
+		.port = dp->index,
+		.lag = lag,
+		.info = uinfo,
+	};
+	int err;
+
+	dsa_lag_map(dp->ds->dst, lag);
+	dp->lag_dev = lag;
+
+	err = dsa_port_notify(dp, DSA_NOTIFIER_LAG_JOIN, &info);
+	if (err) {
+		dp->lag_dev = NULL;
+		dsa_lag_unmap(dp->ds->dst, lag);
+	}
+
+	return err;
+}
+
+void dsa_port_lag_leave(struct dsa_port *dp, struct net_device *lag)
+{
+	struct dsa_notifier_lag_info info = {
+		.sw_index = dp->ds->index,
+		.port = dp->index,
+		.lag = lag,
+	};
+	int err;
+
+	if (!dp->lag_dev)
+		return;
+
+	/* Port might have been part of a LAG that in turn was
+	 * attached to a bridge.
+	 */
+	if (dp->bridge_dev)
+		dsa_port_bridge_leave(dp, dp->bridge_dev);
+
+	dp->lag_tx_enabled = false;
+	dp->lag_dev = NULL;
+
+	err = dsa_port_notify(dp, DSA_NOTIFIER_LAG_LEAVE, &info);
+	if (err)
+		pr_err("DSA: failed to notify DSA_NOTIFIER_LAG_LEAVE: %d\n",
+		       err);
+
+	dsa_lag_unmap(dp->ds->dst, lag);
+}
+
 /* Must be called under rcu_read_lock() */
 static bool dsa_port_can_apply_vlan_filtering(struct dsa_port *dp,
 					      bool vlan_filtering)

net/dsa/slave.c

@@ -273,6 +273,9 @@ static int dsa_slave_port_attr_set(struct net_device *dev,
 	struct dsa_port *dp = dsa_slave_to_port(dev);
 	int ret;
 
+	if (!dsa_port_offloads_netdev(dp, attr->orig_dev))
+		return -EOPNOTSUPP;
+
 	switch (attr->id) {
 	case SWITCHDEV_ATTR_ID_PORT_STP_STATE:
 		ret = dsa_port_set_state(dp, attr->u.stp_state);
@@ -330,7 +333,7 @@ static int dsa_slave_vlan_add(struct net_device *dev,
 	struct switchdev_obj_port_vlan vlan;
 	int err;
 
-	if (obj->orig_dev != dev)
+	if (!dsa_port_offloads_netdev(dp, obj->orig_dev))
 		return -EOPNOTSUPP;
 
 	if (dsa_port_skip_vlan_configuration(dp))
@@ -375,7 +378,7 @@ static int dsa_slave_port_obj_add(struct net_device *dev,
 
 	switch (obj->id) {
 	case SWITCHDEV_OBJ_ID_PORT_MDB:
-		if (obj->orig_dev != dev)
+		if (!dsa_port_offloads_netdev(dp, obj->orig_dev))
 			return -EOPNOTSUPP;
 		err = dsa_port_mdb_add(dp, SWITCHDEV_OBJ_PORT_MDB(obj));
 		break;
@@ -404,7 +407,7 @@ static int dsa_slave_vlan_del(struct net_device *dev,
 	struct switchdev_obj_port_vlan *vlan;
 	int err;
 
-	if (obj->orig_dev != dev)
+	if (!dsa_port_offloads_netdev(dp, obj->orig_dev))
 		return -EOPNOTSUPP;
 
 	if (dsa_port_skip_vlan_configuration(dp))
@@ -432,7 +435,7 @@ static int dsa_slave_port_obj_del(struct net_device *dev,
 
 	switch (obj->id) {
 	case SWITCHDEV_OBJ_ID_PORT_MDB:
-		if (obj->orig_dev != dev)
+		if (!dsa_port_offloads_netdev(dp, obj->orig_dev))
 			return -EOPNOTSUPP;
 		err = dsa_port_mdb_del(dp, SWITCHDEV_OBJ_PORT_MDB(obj));
 		break;
@@ -1904,6 +1907,46 @@ static int dsa_slave_changeupper(struct net_device *dev,
 			dsa_port_bridge_leave(dp, info->upper_dev);
 			err = NOTIFY_OK;
 		}
+	} else if (netif_is_lag_master(info->upper_dev)) {
+		if (info->linking) {
+			err = dsa_port_lag_join(dp, info->upper_dev,
+						info->upper_info);
+			if (err == -EOPNOTSUPP) {
+				NL_SET_ERR_MSG_MOD(info->info.extack,
+						   "Offloading not supported");
+				err = 0;
+			}
+			err = notifier_from_errno(err);
+		} else {
+			dsa_port_lag_leave(dp, info->upper_dev);
+			err = NOTIFY_OK;
+		}
+	}
+
+	return err;
+}
+
+static int
+dsa_slave_lag_changeupper(struct net_device *dev,
+			  struct netdev_notifier_changeupper_info *info)
+{
+	struct net_device *lower;
+	struct list_head *iter;
+	int err = NOTIFY_DONE;
+	struct dsa_port *dp;
+
+	netdev_for_each_lower_dev(dev, lower, iter) {
+		if (!dsa_slave_dev_check(lower))
+			continue;
+
+		dp = dsa_slave_to_port(lower);
+		if (!dp->lag_dev)
+			/* Software LAG */
+			continue;
+
+		err = dsa_slave_changeupper(lower, info);
+		if (notifier_to_errno(err))
+			break;
 	}
 
 	return err;
@@ -2001,10 +2044,26 @@ static int dsa_slave_netdevice_event(struct notifier_block *nb,
 		break;
 	}
 	case NETDEV_CHANGEUPPER:
+		if (dsa_slave_dev_check(dev))
+			return dsa_slave_changeupper(dev, ptr);
+
+		if (netif_is_lag_master(dev))
+			return dsa_slave_lag_changeupper(dev, ptr);
+
+		break;
+	case NETDEV_CHANGELOWERSTATE: {
+		struct netdev_notifier_changelowerstate_info *info = ptr;
+		struct dsa_port *dp;
+		int err;
+
 		if (!dsa_slave_dev_check(dev))
-			return NOTIFY_DONE;
+			break;
 
-		return dsa_slave_changeupper(dev, ptr);
+		dp = dsa_slave_to_port(dev);
+
+		err = dsa_port_lag_change(dp, info->lower_state_info);
+		return notifier_from_errno(err);
+	}
 	}
 
 	return NOTIFY_DONE;

net/dsa/switch.c

@@ -166,6 +166,47 @@ static int dsa_switch_fdb_del(struct dsa_switch *ds,
 	return ds->ops->port_fdb_del(ds, port, info->addr, info->vid);
 }
 
+static int dsa_switch_lag_change(struct dsa_switch *ds,
+				 struct dsa_notifier_lag_info *info)
+{
+	if (ds->index == info->sw_index && ds->ops->port_lag_change)
+		return ds->ops->port_lag_change(ds, info->port);
+
+	if (ds->index != info->sw_index && ds->ops->crosschip_lag_change)
+		return ds->ops->crosschip_lag_change(ds, info->sw_index,
+						     info->port);
+
+	return 0;
+}
+
+static int dsa_switch_lag_join(struct dsa_switch *ds,
+			       struct dsa_notifier_lag_info *info)
+{
+	if (ds->index == info->sw_index && ds->ops->port_lag_join)
+		return ds->ops->port_lag_join(ds, info->port, info->lag,
+					      info->info);
+
+	if (ds->index != info->sw_index && ds->ops->crosschip_lag_join)
+		return ds->ops->crosschip_lag_join(ds, info->sw_index,
+						   info->port, info->lag,
+						   info->info);
+
+	return 0;
+}
+
+static int dsa_switch_lag_leave(struct dsa_switch *ds,
+				struct dsa_notifier_lag_info *info)
+{
+	if (ds->index == info->sw_index && ds->ops->port_lag_leave)
+		return ds->ops->port_lag_leave(ds, info->port, info->lag);
+
+	if (ds->index != info->sw_index && ds->ops->crosschip_lag_leave)
+		return ds->ops->crosschip_lag_leave(ds, info->sw_index,
+						    info->port, info->lag);
+
+	return 0;
+}
+
 static bool dsa_switch_mdb_match(struct dsa_switch *ds, int port,
 				 struct dsa_notifier_mdb_info *info)
 {
@@ -278,6 +319,15 @@ static int dsa_switch_event(struct notifier_block *nb,
 	case DSA_NOTIFIER_FDB_DEL:
 		err = dsa_switch_fdb_del(ds, info);
 		break;
+	case DSA_NOTIFIER_LAG_CHANGE:
+		err = dsa_switch_lag_change(ds, info);
+		break;
+	case DSA_NOTIFIER_LAG_JOIN:
+		err = dsa_switch_lag_join(ds, info);
+		break;
+	case DSA_NOTIFIER_LAG_LEAVE:
+		err = dsa_switch_lag_leave(ds, info);
+		break;
 	case DSA_NOTIFIER_MDB_ADD:
 		err = dsa_switch_mdb_add(ds, info);
 		break;

net/dsa/tag_dsa.c

@@ -163,6 +163,7 @@ static struct sk_buff *dsa_rcv_ll(struct sk_buff *skb, struct net_device *dev,
 				  u8 extra)
 {
 	int source_device, source_port;
+	bool trunk = false;
 	enum dsa_code code;
 	enum dsa_cmd cmd;
 	u8 *dsa_header;
@@ -174,6 +175,8 @@ static struct sk_buff *dsa_rcv_ll(struct sk_buff *skb, struct net_device *dev,
 	switch (cmd) {
 	case DSA_CMD_FORWARD:
 		skb->offload_fwd_mark = 1;
+
+		trunk = !!(dsa_header[1] & 7);
 		break;
 
 	case DSA_CMD_TO_CPU:
@@ -216,7 +219,19 @@ static struct sk_buff *dsa_rcv_ll(struct sk_buff *skb, struct net_device *dev,
 	source_device = dsa_header[0] & 0x1f;
 	source_port = (dsa_header[1] >> 3) & 0x1f;
 
-	skb->dev = dsa_master_find_slave(dev, source_device, source_port);
+	if (trunk) {
+		struct dsa_port *cpu_dp = dev->dsa_ptr;
+
+		/* The exact source port is not available in the tag,
+		 * so we inject the frame directly on the upper
+		 * team/bond.
+		 */
+		skb->dev = dsa_lag_dev(cpu_dp->dst, source_port);
+	} else {
+		skb->dev = dsa_master_find_slave(dev, source_device,
+						 source_port);
+	}
+
 	if (!skb->dev)
 		return NULL;