From 6377ed0bbae6fa28853e1679d068a9106c8a8908 Mon Sep 17 00:00:00 2001 From: Moshe Shemesh Date: Thu, 19 Oct 2017 14:14:29 +0300 Subject: net/mlx5: Fix health work queue spin lock to IRQ safe spin_lock/unlock of health->wq_lock should be IRQ safe. It was changed to spin_lock_irqsave since adding commit 0179720d6be2 ("net/mlx5: Introduce trigger_health_work function") which uses spin_lock from asynchronous event (IRQ) context. Thus, all spin_lock/unlock of health->wq_lock should have been moved to IRQ safe mode. However, one occurrence on new code using this lock missed that change, resulting in possible deadlock: kernel: Possible unsafe locking scenario: kernel: CPU0 kernel: ---- kernel: lock(&(&health->wq_lock)->rlock); kernel: kernel: lock(&(&health->wq_lock)->rlock); kernel: #012 *** DEADLOCK *** Fixes: 2a0165a034ac ("net/mlx5: Cancel delayed recovery work when unloading the driver") Signed-off-by: Moshe Shemesh Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/health.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/health.c b/drivers/net/ethernet/mellanox/mlx5/core/health.c index 8aea0a065e56..db86e1506c8b 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/health.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/health.c @@ -356,10 +356,11 @@ void mlx5_drain_health_wq(struct mlx5_core_dev *dev) void mlx5_drain_health_recovery(struct mlx5_core_dev *dev) { struct mlx5_core_health *health = &dev->priv.health; + unsigned long flags; - spin_lock(&health->wq_lock); + spin_lock_irqsave(&health->wq_lock, flags); set_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags); - spin_unlock(&health->wq_lock); + spin_unlock_irqrestore(&health->wq_lock, flags); cancel_delayed_work_sync(&dev->priv.health.recover_work); } -- cgit v1.2.3 From 4ca637a20a524cd8ddbca696f12bfa92111c96e3 Mon Sep 17 00:00:00 2001 From: Huy Nguyen Date: Wed, 4 Oct 2017 17:58:21 -0500 Subject: net/mlx5: Delay events till mlx5 interface's add complete for pci resume mlx5_ib_add is called during mlx5_pci_resume after a pci error. Before mlx5_ib_add completes, there are multiple events which trigger function mlx5_ib_event. This cause kernel panic because mlx5_ib_event accesses unitialized resources. The fix is to extend Erez Shitrit's patch <97834eba7c19> ("net/mlx5: Delay events till ib registration ends") to cover the pci resume code path. Trace: mlx5_core 0001:01:00.6: mlx5_pci_resume was called mlx5_core 0001:01:00.6: firmware version: 16.20.1011 mlx5_core 0001:01:00.6: mlx5_attach_interface:164:(pid 779): mlx5_ib_event:2996:(pid 34777): warning: event on port 1 mlx5_ib_event:2996:(pid 34782): warning: event on port 1 Unable to handle kernel paging request for data at address 0x0001c104 Faulting instruction address: 0xd000000008f411fc Oops: Kernel access of bad area, sig: 11 [#1] ... ... Call Trace: [c000000fff77bb70] [d000000008f4119c] mlx5_ib_event+0x64/0x470 [mlx5_ib] (unreliable) [c000000fff77bc60] [d000000008e67130] mlx5_core_event+0xb8/0x210 [mlx5_core] [c000000fff77bd10] [d000000008e4bd00] mlx5_eq_int+0x528/0x860[mlx5_core] Fixes: 97834eba7c19 ("net/mlx5: Delay events till ib registration ends") Signed-off-by: Huy Nguyen Reviewed-by: Saeed Mahameed Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/dev.c | 70 ++++++++++++++++----------- 1 file changed, 41 insertions(+), 29 deletions(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/dev.c b/drivers/net/ethernet/mellanox/mlx5/core/dev.c index ff60cf7342ca..fc281712869b 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/dev.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/dev.c @@ -77,35 +77,41 @@ static void add_delayed_event(struct mlx5_priv *priv, list_add_tail(&delayed_event->list, &priv->waiting_events_list); } -static void fire_delayed_event_locked(struct mlx5_device_context *dev_ctx, - struct mlx5_core_dev *dev, - struct mlx5_priv *priv) +static void delayed_event_release(struct mlx5_device_context *dev_ctx, + struct mlx5_priv *priv) { + struct mlx5_core_dev *dev = container_of(priv, struct mlx5_core_dev, priv); struct mlx5_delayed_event *de; struct mlx5_delayed_event *n; + struct list_head temp; - /* stop delaying events */ - priv->is_accum_events = false; + INIT_LIST_HEAD(&temp); + + spin_lock_irq(&priv->ctx_lock); - /* fire all accumulated events before new event comes */ - list_for_each_entry_safe(de, n, &priv->waiting_events_list, list) { + priv->is_accum_events = false; + list_splice_init(&priv->waiting_events_list, &temp); + if (!dev_ctx->context) + goto out; + list_for_each_entry_safe(de, n, &priv->waiting_events_list, list) dev_ctx->intf->event(dev, dev_ctx->context, de->event, de->param); + +out: + spin_unlock_irq(&priv->ctx_lock); + + list_for_each_entry_safe(de, n, &temp, list) { list_del(&de->list); kfree(de); } } -static void cleanup_delayed_evets(struct mlx5_priv *priv) +/* accumulating events that can come after mlx5_ib calls to + * ib_register_device, till adding that interface to the events list. + */ +static void delayed_event_start(struct mlx5_priv *priv) { - struct mlx5_delayed_event *de; - struct mlx5_delayed_event *n; - spin_lock_irq(&priv->ctx_lock); - priv->is_accum_events = false; - list_for_each_entry_safe(de, n, &priv->waiting_events_list, list) { - list_del(&de->list); - kfree(de); - } + priv->is_accum_events = true; spin_unlock_irq(&priv->ctx_lock); } @@ -122,11 +128,8 @@ void mlx5_add_device(struct mlx5_interface *intf, struct mlx5_priv *priv) return; dev_ctx->intf = intf; - /* accumulating events that can come after mlx5_ib calls to - * ib_register_device, till adding that interface to the events list. - */ - priv->is_accum_events = true; + delayed_event_start(priv); dev_ctx->context = intf->add(dev); set_bit(MLX5_INTERFACE_ADDED, &dev_ctx->state); @@ -137,8 +140,6 @@ void mlx5_add_device(struct mlx5_interface *intf, struct mlx5_priv *priv) spin_lock_irq(&priv->ctx_lock); list_add_tail(&dev_ctx->list, &priv->ctx_list); - fire_delayed_event_locked(dev_ctx, dev, priv); - #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING if (dev_ctx->intf->pfault) { if (priv->pfault) { @@ -150,11 +151,12 @@ void mlx5_add_device(struct mlx5_interface *intf, struct mlx5_priv *priv) } #endif spin_unlock_irq(&priv->ctx_lock); - } else { - kfree(dev_ctx); - /* delete all accumulated events */ - cleanup_delayed_evets(priv); } + + delayed_event_release(dev_ctx, priv); + + if (!dev_ctx->context) + kfree(dev_ctx); } static struct mlx5_device_context *mlx5_get_device(struct mlx5_interface *intf, @@ -205,17 +207,21 @@ static void mlx5_attach_interface(struct mlx5_interface *intf, struct mlx5_priv if (!dev_ctx) return; + delayed_event_start(priv); if (intf->attach) { if (test_bit(MLX5_INTERFACE_ATTACHED, &dev_ctx->state)) - return; + goto out; intf->attach(dev, dev_ctx->context); set_bit(MLX5_INTERFACE_ATTACHED, &dev_ctx->state); } else { if (test_bit(MLX5_INTERFACE_ADDED, &dev_ctx->state)) - return; + goto out; dev_ctx->context = intf->add(dev); set_bit(MLX5_INTERFACE_ADDED, &dev_ctx->state); } + +out: + delayed_event_release(dev_ctx, priv); } void mlx5_attach_device(struct mlx5_core_dev *dev) @@ -414,8 +420,14 @@ void mlx5_core_event(struct mlx5_core_dev *dev, enum mlx5_dev_event event, if (priv->is_accum_events) add_delayed_event(priv, dev, event, param); + /* After mlx5_detach_device, the dev_ctx->intf is still set and dev_ctx is + * still in priv->ctx_list. In this case, only notify the dev_ctx if its + * ADDED or ATTACHED bit are set. + */ list_for_each_entry(dev_ctx, &priv->ctx_list, list) - if (dev_ctx->intf->event) + if (dev_ctx->intf->event && + (test_bit(MLX5_INTERFACE_ADDED, &dev_ctx->state) || + test_bit(MLX5_INTERFACE_ATTACHED, &dev_ctx->state))) dev_ctx->intf->event(dev, dev_ctx->context, event, param); spin_unlock_irqrestore(&priv->ctx_lock, flags); -- cgit v1.2.3 From 3c37745ec614ff048d5dce38f976804b05d307ee Mon Sep 17 00:00:00 2001 From: Or Gerlitz Date: Tue, 17 Oct 2017 12:33:43 +0200 Subject: net/mlx5e: Properly deal with encap flows add/del under neigh update Currently, the encap action offload is handled in the actions parse function and not in mlx5e_tc_add_fdb_flow() where we deal with all the other aspects of offloading actions (vlan, modify header) and the rule itself. When the neigh update code (mlx5e_tc_encap_flows_add()) recreates the encap entry and offloads the related flows, we wrongly call again into mlx5e_tc_add_fdb_flow(), this for itself would cause us to handle again the offloading of vlans and header re-write which puts things in non consistent state and step on freed memory (e.g the modify header parse buffer which is already freed). Since on error, mlx5e_tc_add_fdb_flow() detaches and may release the encap entry, it causes a corruption at the neigh update code which goes over the list of flows associated with this encap entry, or double free when the tc flow is later deleted by user-space. When neigh update (mlx5e_tc_encap_flows_del()) unoffloads the flows related to an encap entry which is now invalid, we do a partial repeat of the eswitch flow removal code which is wrong too. To fix things up we do the following: (1) handle the encap action offload in the eswitch flow add function mlx5e_tc_add_fdb_flow() as done for the other actions and the rule itself. (2) modify the neigh update code (mlx5e_tc_encap_flows_add/del) to only deal with the encap entry and rules delete/add and not with any of the other offloaded actions. Fixes: 232c001398ae ('net/mlx5e: Add support to neighbour update flow') Signed-off-by: Or Gerlitz Reviewed-by: Paul Blakey Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 89 +++++++++++++++---------- 1 file changed, 54 insertions(+), 35 deletions(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c index 1aa2028ed995..9ba1f72060aa 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c @@ -78,9 +78,11 @@ struct mlx5e_tc_flow { }; struct mlx5e_tc_flow_parse_attr { + struct ip_tunnel_info tun_info; struct mlx5_flow_spec spec; int num_mod_hdr_actions; void *mod_hdr_actions; + int mirred_ifindex; }; enum { @@ -322,6 +324,12 @@ static void mlx5e_tc_del_nic_flow(struct mlx5e_priv *priv, static void mlx5e_detach_encap(struct mlx5e_priv *priv, struct mlx5e_tc_flow *flow); +static int mlx5e_attach_encap(struct mlx5e_priv *priv, + struct ip_tunnel_info *tun_info, + struct net_device *mirred_dev, + struct net_device **encap_dev, + struct mlx5e_tc_flow *flow); + static struct mlx5_flow_handle * mlx5e_tc_add_fdb_flow(struct mlx5e_priv *priv, struct mlx5e_tc_flow_parse_attr *parse_attr, @@ -329,9 +337,27 @@ mlx5e_tc_add_fdb_flow(struct mlx5e_priv *priv, { struct mlx5_eswitch *esw = priv->mdev->priv.eswitch; struct mlx5_esw_flow_attr *attr = flow->esw_attr; - struct mlx5_flow_handle *rule; + struct net_device *out_dev, *encap_dev = NULL; + struct mlx5_flow_handle *rule = NULL; + struct mlx5e_rep_priv *rpriv; + struct mlx5e_priv *out_priv; int err; + if (attr->action & MLX5_FLOW_CONTEXT_ACTION_ENCAP) { + out_dev = __dev_get_by_index(dev_net(priv->netdev), + attr->parse_attr->mirred_ifindex); + err = mlx5e_attach_encap(priv, &parse_attr->tun_info, + out_dev, &encap_dev, flow); + if (err) { + rule = ERR_PTR(err); + if (err != -EAGAIN) + goto err_attach_encap; + } + out_priv = netdev_priv(encap_dev); + rpriv = out_priv->ppriv; + attr->out_rep = rpriv->rep; + } + err = mlx5_eswitch_add_vlan_action(esw, attr); if (err) { rule = ERR_PTR(err); @@ -347,10 +373,14 @@ mlx5e_tc_add_fdb_flow(struct mlx5e_priv *priv, } } - rule = mlx5_eswitch_add_offloaded_rule(esw, &parse_attr->spec, attr); - if (IS_ERR(rule)) - goto err_add_rule; - + /* we get here if (1) there's no error (rule being null) or when + * (2) there's an encap action and we're on -EAGAIN (no valid neigh) + */ + if (rule != ERR_PTR(-EAGAIN)) { + rule = mlx5_eswitch_add_offloaded_rule(esw, &parse_attr->spec, attr); + if (IS_ERR(rule)) + goto err_add_rule; + } return rule; err_add_rule: @@ -361,6 +391,7 @@ err_mod_hdr: err_add_vlan: if (attr->action & MLX5_FLOW_CONTEXT_ACTION_ENCAP) mlx5e_detach_encap(priv, flow); +err_attach_encap: return rule; } @@ -389,6 +420,8 @@ static void mlx5e_tc_del_fdb_flow(struct mlx5e_priv *priv, void mlx5e_tc_encap_flows_add(struct mlx5e_priv *priv, struct mlx5e_encap_entry *e) { + struct mlx5_eswitch *esw = priv->mdev->priv.eswitch; + struct mlx5_esw_flow_attr *esw_attr; struct mlx5e_tc_flow *flow; int err; @@ -404,10 +437,9 @@ void mlx5e_tc_encap_flows_add(struct mlx5e_priv *priv, mlx5e_rep_queue_neigh_stats_work(priv); list_for_each_entry(flow, &e->flows, encap) { - flow->esw_attr->encap_id = e->encap_id; - flow->rule = mlx5e_tc_add_fdb_flow(priv, - flow->esw_attr->parse_attr, - flow); + esw_attr = flow->esw_attr; + esw_attr->encap_id = e->encap_id; + flow->rule = mlx5_eswitch_add_offloaded_rule(esw, &esw_attr->parse_attr->spec, esw_attr); if (IS_ERR(flow->rule)) { err = PTR_ERR(flow->rule); mlx5_core_warn(priv->mdev, "Failed to update cached encapsulation flow, %d\n", @@ -421,15 +453,13 @@ void mlx5e_tc_encap_flows_add(struct mlx5e_priv *priv, void mlx5e_tc_encap_flows_del(struct mlx5e_priv *priv, struct mlx5e_encap_entry *e) { + struct mlx5_eswitch *esw = priv->mdev->priv.eswitch; struct mlx5e_tc_flow *flow; - struct mlx5_fc *counter; list_for_each_entry(flow, &e->flows, encap) { if (flow->flags & MLX5E_TC_FLOW_OFFLOADED) { flow->flags &= ~MLX5E_TC_FLOW_OFFLOADED; - counter = mlx5_flow_rule_counter(flow->rule); - mlx5_del_flow_rules(flow->rule); - mlx5_fc_destroy(priv->mdev, counter); + mlx5_eswitch_del_offloaded_rule(esw, flow->rule, flow->esw_attr); } } @@ -1942,7 +1972,7 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts, if (is_tcf_mirred_egress_redirect(a)) { int ifindex = tcf_mirred_ifindex(a); - struct net_device *out_dev, *encap_dev = NULL; + struct net_device *out_dev; struct mlx5e_priv *out_priv; out_dev = __dev_get_by_index(dev_net(priv->netdev), ifindex); @@ -1955,17 +1985,13 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts, rpriv = out_priv->ppriv; attr->out_rep = rpriv->rep; } else if (encap) { - err = mlx5e_attach_encap(priv, info, - out_dev, &encap_dev, flow); - if (err && err != -EAGAIN) - return err; + parse_attr->mirred_ifindex = ifindex; + parse_attr->tun_info = *info; + attr->parse_attr = parse_attr; attr->action |= MLX5_FLOW_CONTEXT_ACTION_ENCAP | MLX5_FLOW_CONTEXT_ACTION_FWD_DEST | MLX5_FLOW_CONTEXT_ACTION_COUNT; - out_priv = netdev_priv(encap_dev); - rpriv = out_priv->ppriv; - attr->out_rep = rpriv->rep; - attr->parse_attr = parse_attr; + /* attr->out_rep is resolved when we handle encap */ } else { pr_err("devices %s %s not on same switch HW, can't offload forwarding\n", priv->netdev->name, out_dev->name); @@ -2047,7 +2073,7 @@ int mlx5e_configure_flower(struct mlx5e_priv *priv, if (flow->flags & MLX5E_TC_FLOW_ESWITCH) { err = parse_tc_fdb_actions(priv, f->exts, parse_attr, flow); if (err < 0) - goto err_handle_encap_flow; + goto err_free; flow->rule = mlx5e_tc_add_fdb_flow(priv, parse_attr, flow); } else { err = parse_tc_nic_actions(priv, f->exts, parse_attr, flow); @@ -2058,10 +2084,13 @@ int mlx5e_configure_flower(struct mlx5e_priv *priv, if (IS_ERR(flow->rule)) { err = PTR_ERR(flow->rule); - goto err_free; + if (err != -EAGAIN) + goto err_free; } - flow->flags |= MLX5E_TC_FLOW_OFFLOADED; + if (err != -EAGAIN) + flow->flags |= MLX5E_TC_FLOW_OFFLOADED; + err = rhashtable_insert_fast(&tc->ht, &flow->node, tc->ht_params); if (err) @@ -2075,16 +2104,6 @@ int mlx5e_configure_flower(struct mlx5e_priv *priv, err_del_rule: mlx5e_tc_del_flow(priv, flow); -err_handle_encap_flow: - if (err == -EAGAIN) { - err = rhashtable_insert_fast(&tc->ht, &flow->node, - tc->ht_params); - if (err) - mlx5e_tc_del_flow(priv, flow); - else - return 0; - } - err_free: kvfree(parse_attr); kfree(flow); -- cgit v1.2.3 From be0f161ef141e4df368aa3f417a1c2ab9c362e75 Mon Sep 17 00:00:00 2001 From: Huy Nguyen Date: Thu, 28 Sep 2017 15:33:50 -0500 Subject: net/mlx5e: DCBNL, Implement tc with ets type and zero bandwidth Previously, tc with ets type and zero bandwidth is not accepted by driver. This behavior does not follow the IEEE802.1qaz spec. If there are tcs with ets type and zero bandwidth, these tcs are assigned to the lowest priority tc_group #0. We equally distribute 100% bw of the tc_group #0 to these zero bandwidth ets tcs. Also, the non zero bandwidth ets tcs are assigned to tc_group #1. If there is no zero bandwidth ets tc, the non zero bandwidth ets tcs are assigned to tc_group #0. Fixes: cdcf11212b22 ("net/mlx5e: Validate BW weight values of ETS") Signed-off-by: Huy Nguyen Reviewed-by: Parav Pandit Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c | 113 +++++++++++++++------ drivers/net/ethernet/mellanox/mlx5/core/port.c | 21 ++++ 2 files changed, 104 insertions(+), 30 deletions(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c b/drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c index c1d384fca4dc..51c4cc00a186 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c @@ -41,6 +41,11 @@ #define MLX5E_CEE_STATE_UP 1 #define MLX5E_CEE_STATE_DOWN 0 +enum { + MLX5E_VENDOR_TC_GROUP_NUM = 7, + MLX5E_LOWEST_PRIO_GROUP = 0, +}; + /* If dcbx mode is non-host set the dcbx mode to host. */ static int mlx5e_dcbnl_set_dcbx_mode(struct mlx5e_priv *priv, @@ -85,6 +90,9 @@ static int mlx5e_dcbnl_ieee_getets(struct net_device *netdev, { struct mlx5e_priv *priv = netdev_priv(netdev); struct mlx5_core_dev *mdev = priv->mdev; + u8 tc_group[IEEE_8021QAZ_MAX_TCS]; + bool is_tc_group_6_exist = false; + bool is_zero_bw_ets_tc = false; int err = 0; int i; @@ -96,37 +104,64 @@ static int mlx5e_dcbnl_ieee_getets(struct net_device *netdev, err = mlx5_query_port_prio_tc(mdev, i, &ets->prio_tc[i]); if (err) return err; - } - for (i = 0; i < ets->ets_cap; i++) { + err = mlx5_query_port_tc_group(mdev, i, &tc_group[i]); + if (err) + return err; + err = mlx5_query_port_tc_bw_alloc(mdev, i, &ets->tc_tx_bw[i]); if (err) return err; + + if (ets->tc_tx_bw[i] < MLX5E_MAX_BW_ALLOC && + tc_group[i] == (MLX5E_LOWEST_PRIO_GROUP + 1)) + is_zero_bw_ets_tc = true; + + if (tc_group[i] == (MLX5E_VENDOR_TC_GROUP_NUM - 1)) + is_tc_group_6_exist = true; + } + + /* Report 0% ets tc if exits*/ + if (is_zero_bw_ets_tc) { + for (i = 0; i < ets->ets_cap; i++) + if (tc_group[i] == MLX5E_LOWEST_PRIO_GROUP) + ets->tc_tx_bw[i] = 0; + } + + /* Update tc_tsa based on fw setting*/ + for (i = 0; i < ets->ets_cap; i++) { if (ets->tc_tx_bw[i] < MLX5E_MAX_BW_ALLOC) priv->dcbx.tc_tsa[i] = IEEE_8021QAZ_TSA_ETS; + else if (tc_group[i] == MLX5E_VENDOR_TC_GROUP_NUM && + !is_tc_group_6_exist) + priv->dcbx.tc_tsa[i] = IEEE_8021QAZ_TSA_VENDOR; } - memcpy(ets->tc_tsa, priv->dcbx.tc_tsa, sizeof(ets->tc_tsa)); return err; } -enum { - MLX5E_VENDOR_TC_GROUP_NUM = 7, - MLX5E_ETS_TC_GROUP_NUM = 0, -}; - static void mlx5e_build_tc_group(struct ieee_ets *ets, u8 *tc_group, int max_tc) { bool any_tc_mapped_to_ets = false; + bool ets_zero_bw = false; int strict_group; int i; - for (i = 0; i <= max_tc; i++) - if (ets->tc_tsa[i] == IEEE_8021QAZ_TSA_ETS) + for (i = 0; i <= max_tc; i++) { + if (ets->tc_tsa[i] == IEEE_8021QAZ_TSA_ETS) { any_tc_mapped_to_ets = true; + if (!ets->tc_tx_bw[i]) + ets_zero_bw = true; + } + } - strict_group = any_tc_mapped_to_ets ? 1 : 0; + /* strict group has higher priority than ets group */ + strict_group = MLX5E_LOWEST_PRIO_GROUP; + if (any_tc_mapped_to_ets) + strict_group++; + if (ets_zero_bw) + strict_group++; for (i = 0; i <= max_tc; i++) { switch (ets->tc_tsa[i]) { @@ -137,7 +172,9 @@ static void mlx5e_build_tc_group(struct ieee_ets *ets, u8 *tc_group, int max_tc) tc_group[i] = strict_group++; break; case IEEE_8021QAZ_TSA_ETS: - tc_group[i] = MLX5E_ETS_TC_GROUP_NUM; + tc_group[i] = MLX5E_LOWEST_PRIO_GROUP; + if (ets->tc_tx_bw[i] && ets_zero_bw) + tc_group[i] = MLX5E_LOWEST_PRIO_GROUP + 1; break; } } @@ -146,8 +183,22 @@ static void mlx5e_build_tc_group(struct ieee_ets *ets, u8 *tc_group, int max_tc) static void mlx5e_build_tc_tx_bw(struct ieee_ets *ets, u8 *tc_tx_bw, u8 *tc_group, int max_tc) { + int bw_for_ets_zero_bw_tc = 0; + int last_ets_zero_bw_tc = -1; + int num_ets_zero_bw = 0; int i; + for (i = 0; i <= max_tc; i++) { + if (ets->tc_tsa[i] == IEEE_8021QAZ_TSA_ETS && + !ets->tc_tx_bw[i]) { + num_ets_zero_bw++; + last_ets_zero_bw_tc = i; + } + } + + if (num_ets_zero_bw) + bw_for_ets_zero_bw_tc = MLX5E_MAX_BW_ALLOC / num_ets_zero_bw; + for (i = 0; i <= max_tc; i++) { switch (ets->tc_tsa[i]) { case IEEE_8021QAZ_TSA_VENDOR: @@ -157,12 +208,26 @@ static void mlx5e_build_tc_tx_bw(struct ieee_ets *ets, u8 *tc_tx_bw, tc_tx_bw[i] = MLX5E_MAX_BW_ALLOC; break; case IEEE_8021QAZ_TSA_ETS: - tc_tx_bw[i] = ets->tc_tx_bw[i]; + tc_tx_bw[i] = ets->tc_tx_bw[i] ? + ets->tc_tx_bw[i] : + bw_for_ets_zero_bw_tc; break; } } + + /* Make sure the total bw for ets zero bw group is 100% */ + if (last_ets_zero_bw_tc != -1) + tc_tx_bw[last_ets_zero_bw_tc] += + MLX5E_MAX_BW_ALLOC % num_ets_zero_bw; } +/* If there are ETS BW 0, + * Set ETS group # to 1 for all ETS non zero BW tcs. Their sum must be 100%. + * Set group #0 to all the ETS BW 0 tcs and + * equally splits the 100% BW between them + * Report both group #0 and #1 as ETS type. + * All the tcs in group #0 will be reported with 0% BW. + */ int mlx5e_dcbnl_ieee_setets_core(struct mlx5e_priv *priv, struct ieee_ets *ets) { struct mlx5_core_dev *mdev = priv->mdev; @@ -188,7 +253,6 @@ int mlx5e_dcbnl_ieee_setets_core(struct mlx5e_priv *priv, struct ieee_ets *ets) return err; memcpy(priv->dcbx.tc_tsa, ets->tc_tsa, sizeof(ets->tc_tsa)); - return err; } @@ -209,17 +273,9 @@ static int mlx5e_dbcnl_validate_ets(struct net_device *netdev, } /* Validate Bandwidth Sum */ - for (i = 0; i < IEEE_8021QAZ_MAX_TCS; i++) { - if (ets->tc_tsa[i] == IEEE_8021QAZ_TSA_ETS) { - if (!ets->tc_tx_bw[i]) { - netdev_err(netdev, - "Failed to validate ETS: BW 0 is illegal\n"); - return -EINVAL; - } - + for (i = 0; i < IEEE_8021QAZ_MAX_TCS; i++) + if (ets->tc_tsa[i] == IEEE_8021QAZ_TSA_ETS) bw_sum += ets->tc_tx_bw[i]; - } - } if (bw_sum != 0 && bw_sum != 100) { netdev_err(netdev, @@ -533,8 +589,7 @@ static void mlx5e_dcbnl_getpgtccfgtx(struct net_device *netdev, static void mlx5e_dcbnl_getpgbwgcfgtx(struct net_device *netdev, int pgid, u8 *bw_pct) { - struct mlx5e_priv *priv = netdev_priv(netdev); - struct mlx5_core_dev *mdev = priv->mdev; + struct ieee_ets ets; if (pgid >= CEE_DCBX_MAX_PGS) { netdev_err(netdev, @@ -542,8 +597,8 @@ static void mlx5e_dcbnl_getpgbwgcfgtx(struct net_device *netdev, return; } - if (mlx5_query_port_tc_bw_alloc(mdev, pgid, bw_pct)) - *bw_pct = 0; + mlx5e_dcbnl_ieee_getets(netdev, &ets); + *bw_pct = ets.tc_tx_bw[pgid]; } static void mlx5e_dcbnl_setpfccfg(struct net_device *netdev, @@ -739,8 +794,6 @@ static void mlx5e_ets_init(struct mlx5e_priv *priv) ets.prio_tc[i] = i; } - memcpy(priv->dcbx.tc_tsa, ets.tc_tsa, sizeof(ets.tc_tsa)); - /* tclass[prio=0]=1, tclass[prio=1]=0, tclass[prio=i]=i (for i>1) */ ets.prio_tc[0] = 1; ets.prio_tc[1] = 0; diff --git a/drivers/net/ethernet/mellanox/mlx5/core/port.c b/drivers/net/ethernet/mellanox/mlx5/core/port.c index 1975d4388d4f..e07061f565d6 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/port.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/port.c @@ -677,6 +677,27 @@ int mlx5_set_port_tc_group(struct mlx5_core_dev *mdev, u8 *tc_group) } EXPORT_SYMBOL_GPL(mlx5_set_port_tc_group); +int mlx5_query_port_tc_group(struct mlx5_core_dev *mdev, + u8 tc, u8 *tc_group) +{ + u32 out[MLX5_ST_SZ_DW(qetc_reg)]; + void *ets_tcn_conf; + int err; + + err = mlx5_query_port_qetcr_reg(mdev, out, sizeof(out)); + if (err) + return err; + + ets_tcn_conf = MLX5_ADDR_OF(qetc_reg, out, + tc_configuration[tc]); + + *tc_group = MLX5_GET(ets_tcn_config_reg, ets_tcn_conf, + group); + + return 0; +} +EXPORT_SYMBOL_GPL(mlx5_query_port_tc_group); + int mlx5_set_port_tc_bw_alloc(struct mlx5_core_dev *mdev, u8 *tc_bw) { u32 in[MLX5_ST_SZ_DW(qetc_reg)] = {0}; -- cgit v1.2.3