@@ -398,25 +398,64 @@ static void mlx5e_sqs2vport_stop(struct mlx5_eswitch *esw,
}
}
+static int mlx5e_sqs2vport_add_peers_rules(struct mlx5_eswitch *esw, struct mlx5_eswitch_rep *rep,
+ struct mlx5_devcom *devcom,
+ struct mlx5e_rep_sq *rep_sq, int i)
+{
+ struct mlx5_eswitch *peer_esw = NULL;
+ struct mlx5_flow_handle *flow_rule;
+ int tmp;
+
+ mlx5_devcom_for_each_peer_entry(devcom, MLX5_DEVCOM_ESW_OFFLOADS,
+ peer_esw, tmp) {
+ int peer_rule_idx = mlx5_get_dev_index(peer_esw->dev);
+ struct mlx5e_rep_sq_peer *sq_peer;
+ int err;
+
+ sq_peer = kzalloc(sizeof(*sq_peer), GFP_KERNEL);
+ if (!sq_peer)
+ return -ENOMEM;
+
+ flow_rule = mlx5_eswitch_add_send_to_vport_rule(peer_esw, esw,
+ rep, rep_sq->sqn);
+ if (IS_ERR(flow_rule)) {
+ kfree(sq_peer);
+ return PTR_ERR(flow_rule);
+ }
+
+ sq_peer->rule = flow_rule;
+ sq_peer->peer = peer_esw;
+ err = xa_insert(&rep_sq->sq_peer, peer_rule_idx, sq_peer, GFP_KERNEL);
+ if (err) {
+ kfree(sq_peer);
+ mlx5_eswitch_del_send_to_vport_rule(flow_rule);
+ return err;
+ }
+ }
+
+ return 0;
+}
+
static int mlx5e_sqs2vport_start(struct mlx5_eswitch *esw,
struct mlx5_eswitch_rep *rep,
u32 *sqns_array, int sqns_num)
{
- struct mlx5_eswitch *peer_esw = NULL;
struct mlx5_flow_handle *flow_rule;
- struct mlx5e_rep_sq_peer *sq_peer;
struct mlx5e_rep_priv *rpriv;
struct mlx5e_rep_sq *rep_sq;
+ struct mlx5_devcom *devcom;
+ bool devcom_locked = false;
int err;
int i;
if (esw->mode != MLX5_ESWITCH_OFFLOADS)
return 0;
+ devcom = esw->dev->priv.devcom;
rpriv = mlx5e_rep_to_rep_priv(rep);
- if (mlx5_devcom_comp_is_ready(esw->dev->priv.devcom, MLX5_DEVCOM_ESW_OFFLOADS))
- peer_esw = mlx5_devcom_get_peer_data(esw->dev->priv.devcom,
- MLX5_DEVCOM_ESW_OFFLOADS);
+ if (mlx5_devcom_comp_is_ready(devcom, MLX5_DEVCOM_ESW_OFFLOADS) &&
+ mlx5_devcom_for_each_peer_begin(devcom, MLX5_DEVCOM_ESW_OFFLOADS))
+ devcom_locked = true;
for (i = 0; i < sqns_num; i++) {
rep_sq = kzalloc(sizeof(*rep_sq), GFP_KERNEL);
@@ -424,7 +463,6 @@ static int mlx5e_sqs2vport_start(struct mlx5_eswitch *esw,
err = -ENOMEM;
goto out_err;
}
- xa_init(&rep_sq->sq_peer);
/* Add re-inject rule to the PF/representor sqs */
flow_rule = mlx5_eswitch_add_send_to_vport_rule(esw, esw, rep,
@@ -437,50 +475,30 @@ static int mlx5e_sqs2vport_start(struct mlx5_eswitch *esw,
rep_sq->send_to_vport_rule = flow_rule;
rep_sq->sqn = sqns_array[i];
- if (peer_esw) {
- int peer_rule_idx = mlx5_get_dev_index(peer_esw->dev);
-
- sq_peer = kzalloc(sizeof(*sq_peer), GFP_KERNEL);
- if (!sq_peer) {
- err = -ENOMEM;
- goto out_sq_peer_err;
- }
-
- flow_rule = mlx5_eswitch_add_send_to_vport_rule(peer_esw, esw,
- rep, sqns_array[i]);
- if (IS_ERR(flow_rule)) {
- err = PTR_ERR(flow_rule);
- goto out_flow_rule_err;
+ xa_init(&rep_sq->sq_peer);
+ if (devcom_locked) {
+ err = mlx5e_sqs2vport_add_peers_rules(esw, rep, devcom, rep_sq, i);
+ if (err) {
+ mlx5_eswitch_del_send_to_vport_rule(rep_sq->send_to_vport_rule);
+ xa_destroy(&rep_sq->sq_peer);
+ kfree(rep_sq);
+ goto out_err;
}
-
- sq_peer->rule = flow_rule;
- sq_peer->peer = peer_esw;
- err = xa_insert(&rep_sq->sq_peer, peer_rule_idx, sq_peer, GFP_KERNEL);
- if (err)
- goto out_xa_err;
}
list_add(&rep_sq->list, &rpriv->vport_sqs_list);
}
- if (peer_esw)
- mlx5_devcom_release_peer_data(esw->dev->priv.devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ if (devcom_locked)
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
return 0;
-out_xa_err:
- mlx5_eswitch_del_send_to_vport_rule(flow_rule);
-out_flow_rule_err:
- kfree(sq_peer);
-out_sq_peer_err:
- mlx5_eswitch_del_send_to_vport_rule(rep_sq->send_to_vport_rule);
- xa_destroy(&rep_sq->sq_peer);
- kfree(rep_sq);
out_err:
mlx5e_sqs2vport_stop(esw, rep);
- if (peer_esw)
- mlx5_devcom_release_peer_data(esw->dev->priv.devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ if (devcom_locked)
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
return err;
}
@@ -1670,6 +1670,7 @@ int mlx5e_tc_query_route_vport(struct net_device *out_dev, struct net_device *ro
struct mlx5_eswitch *esw;
u16 vhca_id;
int err;
+ int i;
out_priv = netdev_priv(out_dev);
esw = out_priv->mdev->priv.eswitch;
@@ -1686,8 +1687,13 @@ int mlx5e_tc_query_route_vport(struct net_device *out_dev, struct net_device *ro
rcu_read_lock();
devcom = out_priv->mdev->priv.devcom;
- esw = mlx5_devcom_get_peer_data_rcu(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
- err = esw ? mlx5_eswitch_vhca_id_to_vport(esw, vhca_id, vport) : -ENODEV;
+ err = -ENODEV;
+ mlx5_devcom_for_each_peer_entry_rcu(devcom, MLX5_DEVCOM_ESW_OFFLOADS,
+ esw, i) {
+ err = mlx5_eswitch_vhca_id_to_vport(esw, vhca_id, vport);
+ if (!err)
+ break;
+ }
rcu_read_unlock();
return err;
@@ -2025,15 +2031,14 @@ static void mlx5e_tc_del_flow(struct mlx5e_priv *priv,
{
if (mlx5e_is_eswitch_flow(flow)) {
struct mlx5_devcom *devcom = flow->priv->mdev->priv.devcom;
- struct mlx5_eswitch *peer_esw;
- peer_esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
- if (!peer_esw) {
+ if (!mlx5_devcom_for_each_peer_begin(devcom, MLX5_DEVCOM_ESW_OFFLOADS)) {
mlx5e_tc_del_fdb_flow(priv, flow);
return;
}
+
mlx5e_tc_del_fdb_peers_flow(flow);
- mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
mlx5e_tc_del_fdb_flow(priv, flow);
} else {
mlx5e_tc_del_nic_flow(priv, flow);
@@ -4472,6 +4477,7 @@ mlx5e_add_fdb_flow(struct mlx5e_priv *priv,
struct mlx5_eswitch *peer_esw;
struct mlx5e_tc_flow *flow;
int err;
+ int i;
flow = __mlx5e_add_fdb_flow(priv, f, flow_flags, filter_dev, in_rep,
in_mdev);
@@ -4483,23 +4489,27 @@ mlx5e_add_fdb_flow(struct mlx5e_priv *priv,
return 0;
}
- peer_esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
- if (!peer_esw) {
+ if (!mlx5_devcom_for_each_peer_begin(devcom, MLX5_DEVCOM_ESW_OFFLOADS)) {
err = -ENODEV;
goto clean_flow;
}
- err = mlx5e_tc_add_fdb_peer_flow(f, flow, flow_flags, peer_esw);
- if (err)
- goto peer_clean;
- mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ mlx5_devcom_for_each_peer_entry(devcom,
+ MLX5_DEVCOM_ESW_OFFLOADS,
+ peer_esw, i) {
+ err = mlx5e_tc_add_fdb_peer_flow(f, flow, flow_flags, peer_esw);
+ if (err)
+ goto peer_clean;
+ }
- *__flow = flow;
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ *__flow = flow;
return 0;
peer_clean:
- mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ mlx5e_tc_del_fdb_peers_flow(flow);
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
clean_flow:
mlx5e_tc_del_fdb_flow(priv, flow);
return err;
@@ -4719,7 +4729,6 @@ int mlx5e_stats_flower(struct net_device *dev, struct mlx5e_priv *priv,
{
struct mlx5_devcom *devcom = priv->mdev->priv.devcom;
struct rhashtable *tc_ht = get_tc_ht(priv, flags);
- struct mlx5_eswitch *peer_esw;
struct mlx5e_tc_flow *flow;
struct mlx5_fc *counter;
u64 lastuse = 0;
@@ -4754,8 +4763,7 @@ int mlx5e_stats_flower(struct net_device *dev, struct mlx5e_priv *priv,
/* Under multipath it's possible for one rule to be currently
* un-offloaded while the other rule is offloaded.
*/
- peer_esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
- if (!peer_esw)
+ if (!mlx5_devcom_for_each_peer_begin(devcom, MLX5_DEVCOM_ESW_OFFLOADS))
goto out;
if (flow_flag_test(flow, DUP)) {
@@ -4786,7 +4794,7 @@ int mlx5e_stats_flower(struct net_device *dev, struct mlx5e_priv *priv,
}
no_peer_counter:
- mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
out:
flow_stats_update(&f->stats, bytes, packets, 0, lastuse,
FLOW_ACTION_HW_STATS_DELAYED);
@@ -647,22 +647,35 @@ mlx5_esw_bridge_ingress_flow_create(u16 vport_num, const unsigned char *addr,
}
static struct mlx5_flow_handle *
-mlx5_esw_bridge_ingress_flow_peer_create(u16 vport_num, const unsigned char *addr,
+mlx5_esw_bridge_ingress_flow_peer_create(u16 vport_num, u16 esw_owner_vhca_id,
+ const unsigned char *addr,
struct mlx5_esw_bridge_vlan *vlan, u32 counter_id,
struct mlx5_esw_bridge *bridge)
{
struct mlx5_devcom *devcom = bridge->br_offloads->esw->dev->priv.devcom;
+ struct mlx5_eswitch *tmp, *peer_esw = NULL;
static struct mlx5_flow_handle *handle;
- struct mlx5_eswitch *peer_esw;
+ int i;
- peer_esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
- if (!peer_esw)
+ if (!mlx5_devcom_for_each_peer_begin(devcom, MLX5_DEVCOM_ESW_OFFLOADS))
return ERR_PTR(-ENODEV);
+ mlx5_devcom_for_each_peer_entry(devcom,
+ MLX5_DEVCOM_ESW_OFFLOADS,
+ tmp, i) {
+ if (mlx5_esw_is_owner(tmp, vport_num, esw_owner_vhca_id)) {
+ peer_esw = tmp;
+ break;
+ }
+ }
+ if (!peer_esw) {
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ return ERR_PTR(-ENODEV);
+ }
+
handle = mlx5_esw_bridge_ingress_flow_with_esw_create(vport_num, addr, vlan, counter_id,
bridge, peer_esw);
-
- mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
return handle;
}
@@ -1369,8 +1382,9 @@ mlx5_esw_bridge_fdb_entry_init(struct net_device *dev, u16 vport_num, u16 esw_ow
entry->ingress_counter = counter;
handle = peer ?
- mlx5_esw_bridge_ingress_flow_peer_create(vport_num, addr, vlan,
- mlx5_fc_id(counter), bridge) :
+ mlx5_esw_bridge_ingress_flow_peer_create(vport_num, esw_owner_vhca_id,
+ addr, vlan, mlx5_fc_id(counter),
+ bridge) :
mlx5_esw_bridge_ingress_flow_create(vport_num, addr, vlan,
mlx5_fc_id(counter), bridge);
if (IS_ERR(handle)) {
@@ -540,16 +540,29 @@ static struct mlx5_flow_handle *
mlx5_esw_bridge_mcast_filter_flow_peer_create(struct mlx5_esw_bridge_port *port)
{
struct mlx5_devcom *devcom = port->bridge->br_offloads->esw->dev->priv.devcom;
+ struct mlx5_eswitch *tmp, *peer_esw = NULL;
static struct mlx5_flow_handle *handle;
- struct mlx5_eswitch *peer_esw;
+ int i;
- peer_esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
- if (!peer_esw)
+ if (!mlx5_devcom_for_each_peer_begin(devcom, MLX5_DEVCOM_ESW_OFFLOADS))
return ERR_PTR(-ENODEV);
+ mlx5_devcom_for_each_peer_entry(devcom,
+ MLX5_DEVCOM_ESW_OFFLOADS,
+ tmp, i) {
+ if (mlx5_esw_is_owner(tmp, port->vport_num, port->esw_owner_vhca_id)) {
+ peer_esw = tmp;
+ break;
+ }
+ }
+ if (!peer_esw) {
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ return ERR_PTR(-ENODEV);
+ }
+
handle = mlx5_esw_bridge_mcast_flow_with_esw_create(port, peer_esw);
- mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+ mlx5_devcom_for_each_peer_end(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
return handle;
}
@@ -585,6 +585,13 @@ mlx5_esw_is_manager_vport(const struct mlx5_eswitch *esw, u16 vport_num)
return esw->manager_vport == vport_num;
}
+static inline bool mlx5_esw_is_owner(struct mlx5_eswitch *esw, u16 vport_num,
+ u16 esw_owner_vhca_id)
+{
+ return esw_owner_vhca_id == MLX5_CAP_GEN(esw->dev, vhca_id) ||
+ (vport_num == MLX5_VPORT_UPLINK && mlx5_lag_is_master(esw->dev));
+}
+
static inline u16 mlx5_eswitch_first_host_vport_num(struct mlx5_core_dev *dev)
{
return mlx5_core_is_ecpf_esw_manager(dev) ?
@@ -239,55 +239,92 @@ bool mlx5_devcom_comp_is_ready(struct mlx5_devcom *devcom,
return READ_ONCE(devcom->priv->components[id].ready);
}
-void *mlx5_devcom_get_peer_data(struct mlx5_devcom *devcom,
- enum mlx5_devcom_components id)
+bool mlx5_devcom_for_each_peer_begin(struct mlx5_devcom *devcom,
+ enum mlx5_devcom_components id)
{
struct mlx5_devcom_component *comp;
- int i;
if (IS_ERR_OR_NULL(devcom))
- return NULL;
+ return false;
comp = &devcom->priv->components[id];
down_read(&comp->sem);
if (!READ_ONCE(comp->ready)) {
up_read(&comp->sem);
- return NULL;
+ return false;
}
- for (i = 0; i < MLX5_DEVCOM_PORTS_SUPPORTED; i++)
- if (i != devcom->idx)
- break;
+ return true;
+}
+
+void mlx5_devcom_for_each_peer_end(struct mlx5_devcom *devcom,
+ enum mlx5_devcom_components id)
+{
+ struct mlx5_devcom_component *comp = &devcom->priv->components[id];
- return rcu_dereference_protected(comp->device[i].data, lockdep_is_held(&comp->sem));
+ up_read(&comp->sem);
}
-void *mlx5_devcom_get_peer_data_rcu(struct mlx5_devcom *devcom, enum mlx5_devcom_components id)
+void *mlx5_devcom_get_next_peer_data(struct mlx5_devcom *devcom,
+ enum mlx5_devcom_components id,
+ int *i)
{
struct mlx5_devcom_component *comp;
- int i;
+ void *ret;
+ int idx;
- if (IS_ERR_OR_NULL(devcom))
- return NULL;
+ comp = &devcom->priv->components[id];
- for (i = 0; i < MLX5_DEVCOM_PORTS_SUPPORTED; i++)
- if (i != devcom->idx)
- break;
+ if (*i == MLX5_DEVCOM_PORTS_SUPPORTED)
+ return NULL;
+ for (idx = *i; idx < MLX5_DEVCOM_PORTS_SUPPORTED; idx++) {
+ if (idx != devcom->idx) {
+ ret = rcu_dereference_protected(comp->device[idx].data,
+ lockdep_is_held(&comp->sem));
+ if (ret)
+ break;
+ }
+ }
- comp = &devcom->priv->components[id];
- /* This can change concurrently, however 'data' pointer will remain
- * valid for the duration of RCU read section.
- */
- if (!READ_ONCE(comp->ready))
+ if (idx == MLX5_DEVCOM_PORTS_SUPPORTED) {
+ *i = idx;
return NULL;
+ }
+ *i = idx + 1;
- return rcu_dereference(comp->device[i].data);
+ return ret;
}
-void mlx5_devcom_release_peer_data(struct mlx5_devcom *devcom,
- enum mlx5_devcom_components id)
+void *mlx5_devcom_get_next_peer_data_rcu(struct mlx5_devcom *devcom,
+ enum mlx5_devcom_components id,
+ int *i)
{
- struct mlx5_devcom_component *comp = &devcom->priv->components[id];
+ struct mlx5_devcom_component *comp;
+ void *ret;
+ int idx;
- up_read(&comp->sem);
+ comp = &devcom->priv->components[id];
+
+ if (*i == MLX5_DEVCOM_PORTS_SUPPORTED)
+ return NULL;
+ for (idx = *i; idx < MLX5_DEVCOM_PORTS_SUPPORTED; idx++) {
+ if (idx != devcom->idx) {
+ /* This can change concurrently, however 'data' pointer will remain
+ * valid for the duration of RCU read section.
+ */
+ if (!READ_ONCE(comp->ready))
+ return NULL;
+ ret = rcu_dereference(comp->device[idx].data);
+ if (ret)
+ break;
+ }
+ }
+
+ if (idx == MLX5_DEVCOM_PORTS_SUPPORTED) {
+ *i = idx;
+ return NULL;
+ }
+ *i = idx + 1;
+
+ return ret;
}
@@ -39,11 +39,24 @@ void mlx5_devcom_comp_set_ready(struct mlx5_devcom *devcom,
bool mlx5_devcom_comp_is_ready(struct mlx5_devcom *devcom,
enum mlx5_devcom_components id);
-void *mlx5_devcom_get_peer_data(struct mlx5_devcom *devcom,
- enum mlx5_devcom_components id);
-void *mlx5_devcom_get_peer_data_rcu(struct mlx5_devcom *devcom, enum mlx5_devcom_components id);
-void mlx5_devcom_release_peer_data(struct mlx5_devcom *devcom,
+bool mlx5_devcom_for_each_peer_begin(struct mlx5_devcom *devcom,
+ enum mlx5_devcom_components id);
+void mlx5_devcom_for_each_peer_end(struct mlx5_devcom *devcom,
enum mlx5_devcom_components id);
+void *mlx5_devcom_get_next_peer_data(struct mlx5_devcom *devcom,
+ enum mlx5_devcom_components id, int *i);
-#endif
+#define mlx5_devcom_for_each_peer_entry(devcom, id, data, i) \
+ for (i = 0, data = mlx5_devcom_get_next_peer_data(devcom, id, &i); \
+ data; \
+ data = mlx5_devcom_get_next_peer_data(devcom, id, &i))
+
+void *mlx5_devcom_get_next_peer_data_rcu(struct mlx5_devcom *devcom,
+ enum mlx5_devcom_components id, int *i);
+#define mlx5_devcom_for_each_peer_entry_rcu(devcom, id, data, i) \
+ for (i = 0, data = mlx5_devcom_get_next_peer_data_rcu(devcom, id, &i); \
+ data; \
+ data = mlx5_devcom_get_next_peer_data_rcu(devcom, id, &i))
+
+#endif