Merge branch 'mlx5-fixes'

Saeed Mahameed says:

====================
Mellanox 100G mlx5 resiliency and xmit path fixes

This series provides two set of fixes to the mlx5 driver:
	- Resiliency fixes for reset flow and internal pci errors
	- xmit path fixes

Please consider queuing those patches for -stable (4.6).

Reset flow fixes for core driver:
	- Add more commands to the list of error simulated commands
	  when pci errors occur
	- Avoid calling sleeping function by the health poll thread
	- Fix incorrect page count when in internal error
	- Fix timeout in wait vital for VFs
	- Deadlock fix and Timeout handling in commands interface

Reset flow and resiliency fixes for mlx5e netdev driver:
	- Handle RQ flush in error cases
	- Implement ndo_tx_timeout callback
	- Timeout if SQ doesn't flush during close
	- Log link state changes
	- Validate BW weight values of ETS

xmit path fixes:
	- Fix wrong fallback assumption in select queue callback
	- Account for all L2 headers when copying headers into inline segment
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/mellanox/mlx5/core/cmd.c

@@ -295,6 +295,12 @@ static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
 	case MLX5_CMD_OP_DESTROY_FLOW_GROUP:
 	case MLX5_CMD_OP_DELETE_FLOW_TABLE_ENTRY:
 	case MLX5_CMD_OP_DEALLOC_FLOW_COUNTER:
+	case MLX5_CMD_OP_2ERR_QP:
+	case MLX5_CMD_OP_2RST_QP:
+	case MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT:
+	case MLX5_CMD_OP_MODIFY_FLOW_TABLE:
+	case MLX5_CMD_OP_SET_FLOW_TABLE_ENTRY:
+	case MLX5_CMD_OP_SET_FLOW_TABLE_ROOT:
 		return MLX5_CMD_STAT_OK;
 
 	case MLX5_CMD_OP_QUERY_HCA_CAP:
@@ -321,8 +327,6 @@ static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
 	case MLX5_CMD_OP_RTR2RTS_QP:
 	case MLX5_CMD_OP_RTS2RTS_QP:
 	case MLX5_CMD_OP_SQERR2RTS_QP:
-	case MLX5_CMD_OP_2ERR_QP:
-	case MLX5_CMD_OP_2RST_QP:
 	case MLX5_CMD_OP_QUERY_QP:
 	case MLX5_CMD_OP_SQD_RTS_QP:
 	case MLX5_CMD_OP_INIT2INIT_QP:
@@ -342,7 +346,6 @@ static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
 	case MLX5_CMD_OP_QUERY_ESW_VPORT_CONTEXT:
 	case MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT:
 	case MLX5_CMD_OP_QUERY_NIC_VPORT_CONTEXT:
-	case MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT:
 	case MLX5_CMD_OP_QUERY_ROCE_ADDRESS:
 	case MLX5_CMD_OP_SET_ROCE_ADDRESS:
 	case MLX5_CMD_OP_QUERY_HCA_VPORT_CONTEXT:
@@ -390,11 +393,12 @@ static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
 	case MLX5_CMD_OP_CREATE_RQT:
 	case MLX5_CMD_OP_MODIFY_RQT:
 	case MLX5_CMD_OP_QUERY_RQT:
+
 	case MLX5_CMD_OP_CREATE_FLOW_TABLE:
 	case MLX5_CMD_OP_QUERY_FLOW_TABLE:
 	case MLX5_CMD_OP_CREATE_FLOW_GROUP:
 	case MLX5_CMD_OP_QUERY_FLOW_GROUP:
-	case MLX5_CMD_OP_SET_FLOW_TABLE_ENTRY:
+
 	case MLX5_CMD_OP_QUERY_FLOW_TABLE_ENTRY:
 	case MLX5_CMD_OP_ALLOC_FLOW_COUNTER:
 	case MLX5_CMD_OP_QUERY_FLOW_COUNTER:
@@ -602,11 +606,36 @@ static void dump_command(struct mlx5_core_dev *dev,
 		pr_debug("\n");
 }
 
+static u16 msg_to_opcode(struct mlx5_cmd_msg *in)
+{
+	struct mlx5_inbox_hdr *hdr = (struct mlx5_inbox_hdr *)(in->first.data);
+
+	return be16_to_cpu(hdr->opcode);
+}
+
+static void cb_timeout_handler(struct work_struct *work)
+{
+	struct delayed_work *dwork = container_of(work, struct delayed_work,
+						  work);
+	struct mlx5_cmd_work_ent *ent = container_of(dwork,
+						     struct mlx5_cmd_work_ent,
+						     cb_timeout_work);
+	struct mlx5_core_dev *dev = container_of(ent->cmd, struct mlx5_core_dev,
+						 cmd);
+
+	ent->ret = -ETIMEDOUT;
+	mlx5_core_warn(dev, "%s(0x%x) timeout. Will cause a leak of a command resource\n",
+		       mlx5_command_str(msg_to_opcode(ent->in)),
+		       msg_to_opcode(ent->in));
+	mlx5_cmd_comp_handler(dev, 1UL << ent->idx);
+}
+
 static void cmd_work_handler(struct work_struct *work)
 {
 	struct mlx5_cmd_work_ent *ent = container_of(work, struct mlx5_cmd_work_ent, work);
 	struct mlx5_cmd *cmd = ent->cmd;
 	struct mlx5_core_dev *dev = container_of(cmd, struct mlx5_core_dev, cmd);
+	unsigned long cb_timeout = msecs_to_jiffies(MLX5_CMD_TIMEOUT_MSEC);
 	struct mlx5_cmd_layout *lay;
 	struct semaphore *sem;
 	unsigned long flags;
@@ -647,6 +676,9 @@ static void cmd_work_handler(struct work_struct *work)
 	dump_command(dev, ent, 1);
 	ent->ts1 = ktime_get_ns();
 
+	if (ent->callback)
+		schedule_delayed_work(&ent->cb_timeout_work, cb_timeout);
+
 	/* ring doorbell after the descriptor is valid */
 	mlx5_core_dbg(dev, "writing 0x%x to command doorbell\n", 1 << ent->idx);
 	wmb();
@@ -691,13 +723,6 @@ static const char *deliv_status_to_str(u8 status)
 	}
 }
 
-static u16 msg_to_opcode(struct mlx5_cmd_msg *in)
-{
-	struct mlx5_inbox_hdr *hdr = (struct mlx5_inbox_hdr *)(in->first.data);
-
-	return be16_to_cpu(hdr->opcode);
-}
-
 static int wait_func(struct mlx5_core_dev *dev, struct mlx5_cmd_work_ent *ent)
 {
 	unsigned long timeout = msecs_to_jiffies(MLX5_CMD_TIMEOUT_MSEC);
@@ -706,13 +731,13 @@ static int wait_func(struct mlx5_core_dev *dev, struct mlx5_cmd_work_ent *ent)
 
 	if (cmd->mode == CMD_MODE_POLLING) {
 		wait_for_completion(&ent->done);
-		err = ent->ret;
-	} else {
-		if (!wait_for_completion_timeout(&ent->done, timeout))
-			err = -ETIMEDOUT;
-		else
-			err = 0;
+	} else if (!wait_for_completion_timeout(&ent->done, timeout)) {
+		ent->ret = -ETIMEDOUT;
+		mlx5_cmd_comp_handler(dev, 1UL << ent->idx);
 	}
+
+	err = ent->ret;
+
 	if (err == -ETIMEDOUT) {
 		mlx5_core_warn(dev, "%s(0x%x) timeout. Will cause a leak of a command resource\n",
 			       mlx5_command_str(msg_to_opcode(ent->in)),
@@ -761,6 +786,7 @@ static int mlx5_cmd_invoke(struct mlx5_core_dev *dev, struct mlx5_cmd_msg *in,
 	if (!callback)
 		init_completion(&ent->done);
 
+	INIT_DELAYED_WORK(&ent->cb_timeout_work, cb_timeout_handler);
 	INIT_WORK(&ent->work, cmd_work_handler);
 	if (page_queue) {
 		cmd_work_handler(&ent->work);
@@ -770,10 +796,12 @@ static int mlx5_cmd_invoke(struct mlx5_core_dev *dev, struct mlx5_cmd_msg *in,
 		goto out_free;
 	}
 
-	if (!callback) {
+	if (callback)
+		goto out;
+
 	err = wait_func(dev, ent);
 	if (err == -ETIMEDOUT)
-			goto out;
+		goto out_free;
 
 	ds = ent->ts2 - ent->ts1;
 	op = be16_to_cpu(((struct mlx5_inbox_hdr *)in->first.data)->opcode);
@@ -788,10 +816,6 @@ static int mlx5_cmd_invoke(struct mlx5_core_dev *dev, struct mlx5_cmd_msg *in,
 			   "fw exec time for %s is %lld nsec\n",
 			   mlx5_command_str(op), ds);
 	*status = ent->status;
-		free_cmd(ent);
-	}
-
-	return err;
 
 out_free:
 	free_cmd(ent);
@@ -1181,41 +1205,30 @@ static int create_debugfs_files(struct mlx5_core_dev *dev)
 	return err;
 }
 
-void mlx5_cmd_use_events(struct mlx5_core_dev *dev)
+static void mlx5_cmd_change_mod(struct mlx5_core_dev *dev, int mode)
 {
 	struct mlx5_cmd *cmd = &dev->cmd;
 	int i;
 
 	for (i = 0; i < cmd->max_reg_cmds; i++)
 		down(&cmd->sem);
-
 	down(&cmd->pages_sem);
 
-	flush_workqueue(cmd->wq);
-
-	cmd->mode = CMD_MODE_EVENTS;
+	cmd->mode = mode;
 
 	up(&cmd->pages_sem);
 	for (i = 0; i < cmd->max_reg_cmds; i++)
 		up(&cmd->sem);
 }
 
-void mlx5_cmd_use_polling(struct mlx5_core_dev *dev)
+void mlx5_cmd_use_events(struct mlx5_core_dev *dev)
 {
-	struct mlx5_cmd *cmd = &dev->cmd;
-	int i;
-
-	for (i = 0; i < cmd->max_reg_cmds; i++)
-		down(&cmd->sem);
-
-	down(&cmd->pages_sem);
-
-	flush_workqueue(cmd->wq);
-	cmd->mode = CMD_MODE_POLLING;
+	mlx5_cmd_change_mod(dev, CMD_MODE_EVENTS);
+}
 
-	up(&cmd->pages_sem);
-	for (i = 0; i < cmd->max_reg_cmds; i++)
-		up(&cmd->sem);
+void mlx5_cmd_use_polling(struct mlx5_core_dev *dev)
+{
+	mlx5_cmd_change_mod(dev, CMD_MODE_POLLING);
 }
 
 static void free_msg(struct mlx5_core_dev *dev, struct mlx5_cmd_msg *msg)
@@ -1251,6 +1264,8 @@ void mlx5_cmd_comp_handler(struct mlx5_core_dev *dev, u64 vec)
 			struct semaphore *sem;
 
 			ent = cmd->ent_arr[i];
+			if (ent->callback)
+				cancel_delayed_work(&ent->cb_timeout_work);
 			if (ent->page_queue)
 				sem = &cmd->pages_sem;
 			else

drivers/net/ethernet/mellanox/mlx5/core/en.h

@@ -145,7 +145,6 @@ struct mlx5e_umr_wqe {
 
 #ifdef CONFIG_MLX5_CORE_EN_DCB
 #define MLX5E_MAX_BW_ALLOC 100 /* Max percentage of BW allocation */
-#define MLX5E_MIN_BW_ALLOC 1   /* Min percentage of BW allocation */
 #endif
 
 struct mlx5e_params {
@@ -191,6 +190,7 @@ struct mlx5e_tstamp {
 enum {
 	MLX5E_RQ_STATE_POST_WQES_ENABLE,
 	MLX5E_RQ_STATE_UMR_WQE_IN_PROGRESS,
+	MLX5E_RQ_STATE_FLUSH_TIMEOUT,
 };
 
 struct mlx5e_cq {
@@ -220,6 +220,8 @@ typedef void (*mlx5e_fp_handle_rx_cqe)(struct mlx5e_rq *rq,
 typedef int (*mlx5e_fp_alloc_wqe)(struct mlx5e_rq *rq, struct mlx5e_rx_wqe *wqe,
 				  u16 ix);
 
+typedef void (*mlx5e_fp_dealloc_wqe)(struct mlx5e_rq *rq, u16 ix);
+
 struct mlx5e_dma_info {
 	struct page	*page;
 	dma_addr_t	addr;
@@ -241,6 +243,7 @@ struct mlx5e_rq {
 	struct mlx5e_cq        cq;
 	mlx5e_fp_handle_rx_cqe handle_rx_cqe;
 	mlx5e_fp_alloc_wqe     alloc_wqe;
+	mlx5e_fp_dealloc_wqe   dealloc_wqe;
 
 	unsigned long          state;
 	int                    ix;
@@ -305,6 +308,7 @@ struct mlx5e_sq_dma {
 enum {
 	MLX5E_SQ_STATE_WAKE_TXQ_ENABLE,
 	MLX5E_SQ_STATE_BF_ENABLE,
+	MLX5E_SQ_STATE_TX_TIMEOUT,
 };
 
 struct mlx5e_ico_wqe_info {
@@ -538,6 +542,7 @@ struct mlx5e_priv {
 	struct workqueue_struct    *wq;
 	struct work_struct         update_carrier_work;
 	struct work_struct         set_rx_mode_work;
+	struct work_struct         tx_timeout_work;
 	struct delayed_work        update_stats_work;
 
 	struct mlx5_core_dev      *mdev;
@@ -589,12 +594,16 @@ void mlx5e_cq_error_event(struct mlx5_core_cq *mcq, enum mlx5_event event);
 int mlx5e_napi_poll(struct napi_struct *napi, int budget);
 bool mlx5e_poll_tx_cq(struct mlx5e_cq *cq, int napi_budget);
 int mlx5e_poll_rx_cq(struct mlx5e_cq *cq, int budget);
+void mlx5e_free_tx_descs(struct mlx5e_sq *sq);
+void mlx5e_free_rx_descs(struct mlx5e_rq *rq);
 
 void mlx5e_handle_rx_cqe(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe);
 void mlx5e_handle_rx_cqe_mpwrq(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe);
 bool mlx5e_post_rx_wqes(struct mlx5e_rq *rq);
 int mlx5e_alloc_rx_wqe(struct mlx5e_rq *rq, struct mlx5e_rx_wqe *wqe, u16 ix);
 int mlx5e_alloc_rx_mpwqe(struct mlx5e_rq *rq, struct mlx5e_rx_wqe *wqe, u16 ix);
+void mlx5e_dealloc_rx_wqe(struct mlx5e_rq *rq, u16 ix);
+void mlx5e_dealloc_rx_mpwqe(struct mlx5e_rq *rq, u16 ix);
 void mlx5e_post_rx_fragmented_mpwqe(struct mlx5e_rq *rq);
 void mlx5e_complete_rx_linear_mpwqe(struct mlx5e_rq *rq,
 				    struct mlx5_cqe64 *cqe,

drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c

@@ -96,7 +96,7 @@ 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] ?: MLX5E_MIN_BW_ALLOC;
+			tc_tx_bw[i] = ets->tc_tx_bw[i];
 			break;
 		}
 	}
@@ -140,9 +140,13 @@ static int mlx5e_dbcnl_validate_ets(struct ieee_ets *ets)
 
 	/* Validate Bandwidth Sum */
 	for (i = 0; i < IEEE_8021QAZ_MAX_TCS; i++) {
-		if (ets->tc_tsa[i] == IEEE_8021QAZ_TSA_ETS)
+		if (ets->tc_tsa[i] == IEEE_8021QAZ_TSA_ETS) {
+			if (!ets->tc_tx_bw[i])
+				return -EINVAL;
+
 			bw_sum += ets->tc_tx_bw[i];
 		}
+	}
 
 	if (bw_sum != 0 && bw_sum != 100)
 		return -EINVAL;

drivers/net/ethernet/mellanox/mlx5/core/en_main.c

@@ -39,6 +39,13 @@
 #include "eswitch.h"
 #include "vxlan.h"
 
+enum {
+	MLX5_EN_QP_FLUSH_TIMEOUT_MS	= 5000,
+	MLX5_EN_QP_FLUSH_MSLEEP_QUANT	= 20,
+	MLX5_EN_QP_FLUSH_MAX_ITER	= MLX5_EN_QP_FLUSH_TIMEOUT_MS /
+					  MLX5_EN_QP_FLUSH_MSLEEP_QUANT,
+};
+
 struct mlx5e_rq_param {
 	u32                        rqc[MLX5_ST_SZ_DW(rqc)];
 	struct mlx5_wq_param       wq;
@@ -74,10 +81,13 @@ static void mlx5e_update_carrier(struct mlx5e_priv *priv)
 	port_state = mlx5_query_vport_state(mdev,
 		MLX5_QUERY_VPORT_STATE_IN_OP_MOD_VNIC_VPORT, 0);
 
-	if (port_state == VPORT_STATE_UP)
+	if (port_state == VPORT_STATE_UP) {
+		netdev_info(priv->netdev, "Link up\n");
 		netif_carrier_on(priv->netdev);
-	else
+	} else {
+		netdev_info(priv->netdev, "Link down\n");
 		netif_carrier_off(priv->netdev);
+	}
 }
 
 static void mlx5e_update_carrier_work(struct work_struct *work)
@@ -91,6 +101,26 @@ static void mlx5e_update_carrier_work(struct work_struct *work)
 	mutex_unlock(&priv->state_lock);
 }
 
+static void mlx5e_tx_timeout_work(struct work_struct *work)
+{
+	struct mlx5e_priv *priv = container_of(work, struct mlx5e_priv,
+					       tx_timeout_work);
+	int err;
+
+	rtnl_lock();
+	mutex_lock(&priv->state_lock);
+	if (!test_bit(MLX5E_STATE_OPENED, &priv->state))
+		goto unlock;
+	mlx5e_close_locked(priv->netdev);
+	err = mlx5e_open_locked(priv->netdev);
+	if (err)
+		netdev_err(priv->netdev, "mlx5e_open_locked failed recovering from a tx_timeout, err(%d).\n",
+			   err);
+unlock:
+	mutex_unlock(&priv->state_lock);
+	rtnl_unlock();
+}
+
 static void mlx5e_update_sw_counters(struct mlx5e_priv *priv)
 {
 	struct mlx5e_sw_stats *s = &priv->stats.sw;
@@ -305,6 +335,7 @@ static int mlx5e_create_rq(struct mlx5e_channel *c,
 		}
 		rq->handle_rx_cqe = mlx5e_handle_rx_cqe_mpwrq;
 		rq->alloc_wqe = mlx5e_alloc_rx_mpwqe;
+		rq->dealloc_wqe = mlx5e_dealloc_rx_mpwqe;
 
 		rq->mpwqe_stride_sz = BIT(priv->params.mpwqe_log_stride_sz);
 		rq->mpwqe_num_strides = BIT(priv->params.mpwqe_log_num_strides);
@@ -320,6 +351,7 @@ static int mlx5e_create_rq(struct mlx5e_channel *c,
 		}
 		rq->handle_rx_cqe = mlx5e_handle_rx_cqe;
 		rq->alloc_wqe = mlx5e_alloc_rx_wqe;
+		rq->dealloc_wqe = mlx5e_dealloc_rx_wqe;
 
 		rq->wqe_sz = (priv->params.lro_en) ?
 				priv->params.lro_wqe_sz :
@@ -525,17 +557,25 @@ static int mlx5e_open_rq(struct mlx5e_channel *c,
 
 static void mlx5e_close_rq(struct mlx5e_rq *rq)
 {
+	int tout = 0;
+	int err;
+
 	clear_bit(MLX5E_RQ_STATE_POST_WQES_ENABLE, &rq->state);
 	napi_synchronize(&rq->channel->napi); /* prevent mlx5e_post_rx_wqes */
 
-	mlx5e_modify_rq_state(rq, MLX5_RQC_STATE_RDY, MLX5_RQC_STATE_ERR);
-	while (!mlx5_wq_ll_is_empty(&rq->wq))
-		msleep(20);
+	err = mlx5e_modify_rq_state(rq, MLX5_RQC_STATE_RDY, MLX5_RQC_STATE_ERR);
+	while (!mlx5_wq_ll_is_empty(&rq->wq) && !err &&
+	       tout++ < MLX5_EN_QP_FLUSH_MAX_ITER)
+		msleep(MLX5_EN_QP_FLUSH_MSLEEP_QUANT);
+
+	if (err || tout == MLX5_EN_QP_FLUSH_MAX_ITER)
+		set_bit(MLX5E_RQ_STATE_FLUSH_TIMEOUT, &rq->state);
 
 	/* avoid destroying rq before mlx5e_poll_rx_cq() is done with it */
 	napi_synchronize(&rq->channel->napi);
 
 	mlx5e_disable_rq(rq);
+	mlx5e_free_rx_descs(rq);
 	mlx5e_destroy_rq(rq);
 }
 
@@ -782,6 +822,9 @@ static inline void netif_tx_disable_queue(struct netdev_queue *txq)
 
 static void mlx5e_close_sq(struct mlx5e_sq *sq)
 {
+	int tout = 0;
+	int err;
+
 	if (sq->txq) {
 		clear_bit(MLX5E_SQ_STATE_WAKE_TXQ_ENABLE, &sq->state);
 		/* prevent netif_tx_wake_queue */
@@ -792,15 +835,24 @@ static void mlx5e_close_sq(struct mlx5e_sq *sq)
 		if (mlx5e_sq_has_room_for(sq, 1))
 			mlx5e_send_nop(sq, true);
 
-		mlx5e_modify_sq(sq, MLX5_SQC_STATE_RDY, MLX5_SQC_STATE_ERR);
+		err = mlx5e_modify_sq(sq, MLX5_SQC_STATE_RDY,
+				      MLX5_SQC_STATE_ERR);
+		if (err)
+			set_bit(MLX5E_SQ_STATE_TX_TIMEOUT, &sq->state);
 	}
 
-	while (sq->cc != sq->pc) /* wait till sq is empty */
-		msleep(20);
+	/* wait till sq is empty, unless a TX timeout occurred on this SQ */
+	while (sq->cc != sq->pc &&
+	       !test_bit(MLX5E_SQ_STATE_TX_TIMEOUT, &sq->state)) {
+		msleep(MLX5_EN_QP_FLUSH_MSLEEP_QUANT);
+		if (tout++ > MLX5_EN_QP_FLUSH_MAX_ITER)
+			set_bit(MLX5E_SQ_STATE_TX_TIMEOUT, &sq->state);
+	}
 
 	/* avoid destroying sq before mlx5e_poll_tx_cq() is done with it */
 	napi_synchronize(&sq->channel->napi);
 
+	mlx5e_free_tx_descs(sq);
 	mlx5e_disable_sq(sq);
 	mlx5e_destroy_sq(sq);
 }
@@ -1658,8 +1710,11 @@ static void mlx5e_netdev_set_tcs(struct net_device *netdev)
 
 	netdev_set_num_tc(netdev, ntc);
 
+	/* Map netdev TCs to offset 0
+	 * We have our own UP to TXQ mapping for QoS
+	 */
 	for (tc = 0; tc < ntc; tc++)
-		netdev_set_tc_queue(netdev, tc, nch, tc * nch);
+		netdev_set_tc_queue(netdev, tc, nch, 0);
 }
 
 int mlx5e_open_locked(struct net_device *netdev)
@@ -2590,6 +2645,29 @@ static netdev_features_t mlx5e_features_check(struct sk_buff *skb,
 	return features;
 }
 
+static void mlx5e_tx_timeout(struct net_device *dev)
+{
+	struct mlx5e_priv *priv = netdev_priv(dev);
+	bool sched_work = false;
+	int i;
+
+	netdev_err(dev, "TX timeout detected\n");
+
+	for (i = 0; i < priv->params.num_channels * priv->params.num_tc; i++) {
+		struct mlx5e_sq *sq = priv->txq_to_sq_map[i];
+
+		if (!netif_tx_queue_stopped(netdev_get_tx_queue(dev, i)))
+			continue;
+		sched_work = true;
+		set_bit(MLX5E_SQ_STATE_TX_TIMEOUT, &sq->state);
+		netdev_err(dev, "TX timeout on queue: %d, SQ: 0x%x, CQ: 0x%x, SQ Cons: 0x%x SQ Prod: 0x%x\n",
+			   i, sq->sqn, sq->cq.mcq.cqn, sq->cc, sq->pc);
+	}
+
+	if (sched_work && test_bit(MLX5E_STATE_OPENED, &priv->state))
+		schedule_work(&priv->tx_timeout_work);
+}
+
 static const struct net_device_ops mlx5e_netdev_ops_basic = {
 	.ndo_open                = mlx5e_open,
 	.ndo_stop                = mlx5e_close,
@@ -2607,6 +2685,7 @@ static const struct net_device_ops mlx5e_netdev_ops_basic = {
 #ifdef CONFIG_RFS_ACCEL
 	.ndo_rx_flow_steer	 = mlx5e_rx_flow_steer,
 #endif
+	.ndo_tx_timeout          = mlx5e_tx_timeout,
 };
 
 static const struct net_device_ops mlx5e_netdev_ops_sriov = {
@@ -2636,6 +2715,7 @@ static const struct net_device_ops mlx5e_netdev_ops_sriov = {
 	.ndo_get_vf_config       = mlx5e_get_vf_config,
 	.ndo_set_vf_link_state   = mlx5e_set_vf_link_state,
 	.ndo_get_vf_stats        = mlx5e_get_vf_stats,
+	.ndo_tx_timeout          = mlx5e_tx_timeout,
 };
 
 static int mlx5e_check_required_hca_cap(struct mlx5_core_dev *mdev)
@@ -2838,6 +2918,7 @@ static void mlx5e_build_netdev_priv(struct mlx5_core_dev *mdev,
 
 	INIT_WORK(&priv->update_carrier_work, mlx5e_update_carrier_work);
 	INIT_WORK(&priv->set_rx_mode_work, mlx5e_set_rx_mode_work);
+	INIT_WORK(&priv->tx_timeout_work, mlx5e_tx_timeout_work);
 	INIT_DELAYED_WORK(&priv->update_stats_work, mlx5e_update_stats_work);
 }
 

drivers/net/ethernet/mellanox/mlx5/core/en_rx.c

@@ -212,6 +212,20 @@ int mlx5e_alloc_rx_wqe(struct mlx5e_rq *rq, struct mlx5e_rx_wqe *wqe, u16 ix)
 	return -ENOMEM;
 }
 
+void mlx5e_dealloc_rx_wqe(struct mlx5e_rq *rq, u16 ix)
+{
+	struct sk_buff *skb = rq->skb[ix];
+
+	if (skb) {
+		rq->skb[ix] = NULL;
+		dma_unmap_single(rq->pdev,
+				 *((dma_addr_t *)skb->cb),
+				 rq->wqe_sz,
+				 DMA_FROM_DEVICE);
+		dev_kfree_skb(skb);
+	}
+}
+
 static inline int mlx5e_mpwqe_strides_per_page(struct mlx5e_rq *rq)
 {
 	return rq->mpwqe_num_strides >> MLX5_MPWRQ_WQE_PAGE_ORDER;
@@ -574,6 +588,30 @@ int mlx5e_alloc_rx_mpwqe(struct mlx5e_rq *rq, struct mlx5e_rx_wqe *wqe, u16 ix)
 	return 0;
 }
 
+void mlx5e_dealloc_rx_mpwqe(struct mlx5e_rq *rq, u16 ix)
+{
+	struct mlx5e_mpw_info *wi = &rq->wqe_info[ix];
+
+	wi->free_wqe(rq, wi);
+}
+
+void mlx5e_free_rx_descs(struct mlx5e_rq *rq)
+{
+	struct mlx5_wq_ll *wq = &rq->wq;
+	struct mlx5e_rx_wqe *wqe;
+	__be16 wqe_ix_be;
+	u16 wqe_ix;
+
+	while (!mlx5_wq_ll_is_empty(wq)) {
+		wqe_ix_be = *wq->tail_next;
+		wqe_ix    = be16_to_cpu(wqe_ix_be);
+		wqe       = mlx5_wq_ll_get_wqe(&rq->wq, wqe_ix);
+		rq->dealloc_wqe(rq, wqe_ix);
+		mlx5_wq_ll_pop(&rq->wq, wqe_ix_be,
+			       &wqe->next.next_wqe_index);
+	}
+}
+
 #define RQ_CANNOT_POST(rq) \
 		(!test_bit(MLX5E_RQ_STATE_POST_WQES_ENABLE, &rq->state) || \
 		 test_bit(MLX5E_RQ_STATE_UMR_WQE_IN_PROGRESS, &rq->state))
@@ -878,6 +916,9 @@ int mlx5e_poll_rx_cq(struct mlx5e_cq *cq, int budget)
 	struct mlx5e_rq *rq = container_of(cq, struct mlx5e_rq, cq);
 	int work_done = 0;
 
+	if (unlikely(test_bit(MLX5E_RQ_STATE_FLUSH_TIMEOUT, &rq->state)))
+		return 0;
+
 	if (cq->decmprs_left)
 		work_done += mlx5e_decompress_cqes_cont(rq, cq, 0, budget);
 

drivers/net/ethernet/mellanox/mlx5/core/en_tx.c

@@ -110,8 +110,20 @@ u16 mlx5e_select_queue(struct net_device *dev, struct sk_buff *skb,
 {
 	struct mlx5e_priv *priv = netdev_priv(dev);
 	int channel_ix = fallback(dev, skb);
-	int up = (netdev_get_num_tc(dev) && skb_vlan_tag_present(skb)) ?
-		 skb->vlan_tci >> VLAN_PRIO_SHIFT : 0;
+	int up = 0;
+
+	if (!netdev_get_num_tc(dev))
+		return channel_ix;
+
+	if (skb_vlan_tag_present(skb))
+		up = skb->vlan_tci >> VLAN_PRIO_SHIFT;
+
+	/* channel_ix can be larger than num_channels since
+	 * dev->num_real_tx_queues = num_channels * num_tc
+	 */
+	if (channel_ix >= priv->params.num_channels)
+		channel_ix = reciprocal_scale(channel_ix,
+					      priv->params.num_channels);
 
 	return priv->channeltc_to_txq_map[channel_ix][up];
 }
@@ -123,7 +135,7 @@ static inline u16 mlx5e_get_inline_hdr_size(struct mlx5e_sq *sq,
 	 * headers and occur before the data gather.
 	 * Therefore these headers must be copied into the WQE
 	 */
-#define MLX5E_MIN_INLINE ETH_HLEN
+#define MLX5E_MIN_INLINE (ETH_HLEN + VLAN_HLEN)
 
 	if (bf) {
 		u16 ihs = skb_headlen(skb);
@@ -135,7 +147,7 @@ static inline u16 mlx5e_get_inline_hdr_size(struct mlx5e_sq *sq,
 			return skb_headlen(skb);
 	}
 
-	return MLX5E_MIN_INLINE;
+	return max(skb_network_offset(skb), MLX5E_MIN_INLINE);
 }
 
 static inline void mlx5e_tx_skb_pull_inline(unsigned char **skb_data,
@@ -341,6 +353,35 @@ netdev_tx_t mlx5e_xmit(struct sk_buff *skb, struct net_device *dev)
 	return mlx5e_sq_xmit(sq, skb);
 }
 
+void mlx5e_free_tx_descs(struct mlx5e_sq *sq)
+{
+	struct mlx5e_tx_wqe_info *wi;
+	struct sk_buff *skb;
+	u16 ci;
+	int i;
+
+	while (sq->cc != sq->pc) {
+		ci = sq->cc & sq->wq.sz_m1;
+		skb = sq->skb[ci];
+		wi = &sq->wqe_info[ci];
+
+		if (!skb) { /* nop */
+			sq->cc++;
+			continue;
+		}
+
+		for (i = 0; i < wi->num_dma; i++) {
+			struct mlx5e_sq_dma *dma =
+				mlx5e_dma_get(sq, sq->dma_fifo_cc++);
+
+			mlx5e_tx_dma_unmap(sq->pdev, dma);
+		}
+
+		dev_kfree_skb_any(skb);
+		sq->cc += wi->num_wqebbs;
+	}
+}
+
 bool mlx5e_poll_tx_cq(struct mlx5e_cq *cq, int napi_budget)
 {
 	struct mlx5e_sq *sq;
@@ -352,6 +393,9 @@ bool mlx5e_poll_tx_cq(struct mlx5e_cq *cq, int napi_budget)
 
 	sq = container_of(cq, struct mlx5e_sq, cq);
 
+	if (unlikely(test_bit(MLX5E_SQ_STATE_TX_TIMEOUT, &sq->state)))
+		return false;
+
 	npkts = 0;
 	nbytes = 0;
 

drivers/net/ethernet/mellanox/mlx5/core/health.c

@@ -108,15 +108,21 @@ static int in_fatal(struct mlx5_core_dev *dev)
 
 void mlx5_enter_error_state(struct mlx5_core_dev *dev)
 {
+	mutex_lock(&dev->intf_state_mutex);
 	if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR)
-		return;
+		goto unlock;
 
 	mlx5_core_err(dev, "start\n");
-	if (pci_channel_offline(dev->pdev) || in_fatal(dev))
+	if (pci_channel_offline(dev->pdev) || in_fatal(dev)) {
 		dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR;
+		trigger_cmd_completions(dev);
+	}
 
 	mlx5_core_event(dev, MLX5_DEV_EVENT_SYS_ERROR, 0);
 	mlx5_core_err(dev, "end\n");
+
+unlock:
+	mutex_unlock(&dev->intf_state_mutex);
 }
 
 static void mlx5_handle_bad_state(struct mlx5_core_dev *dev)
@@ -245,7 +251,6 @@ static void poll_health(unsigned long data)
 	u32 count;
 
 	if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) {
-		trigger_cmd_completions(dev);
 		mod_timer(&health->timer, get_next_poll_jiffies());
 		return;
 	}

drivers/net/ethernet/mellanox/mlx5/core/main.c

@@ -1422,46 +1422,31 @@ void mlx5_disable_device(struct mlx5_core_dev *dev)
 	mlx5_pci_err_detected(dev->pdev, 0);
 }
 
-/* wait for the device to show vital signs. For now we check
- * that we can read the device ID and that the health buffer
- * shows a non zero value which is different than 0xffffffff
+/* wait for the device to show vital signs by waiting
+ * for the health counter to start counting.
  */
-static void wait_vital(struct pci_dev *pdev)
+static int wait_vital(struct pci_dev *pdev)
 {
 	struct mlx5_core_dev *dev = pci_get_drvdata(pdev);
 	struct mlx5_core_health *health = &dev->priv.health;
 	const int niter = 100;
+	u32 last_count = 0;
 	u32 count;
-	u16 did;
 	int i;
 
-	/* Wait for firmware to be ready after reset */
-	msleep(1000);
-	for (i = 0; i < niter; i++) {
-		if (pci_read_config_word(pdev, 2, &did)) {
-			dev_warn(&pdev->dev, "failed reading config word\n");
-			break;
-		}
-		if (did == pdev->device) {
-			dev_info(&pdev->dev, "device ID correctly read after %d iterations\n", i);
-			break;
-		}
-		msleep(50);
-	}
-	if (i == niter)
-		dev_warn(&pdev->dev, "%s-%d: could not read device ID\n", __func__, __LINE__);
-
 	for (i = 0; i < niter; i++) {
 		count = ioread32be(health->health_counter);
 		if (count && count != 0xffffffff) {
+			if (last_count && last_count != count) {
 				dev_info(&pdev->dev, "Counter value 0x%x after %d iterations\n", count, i);
-			break;
+				return 0;
+			}
+			last_count = count;
 		}
 		msleep(50);
 	}
 
-	if (i == niter)
-		dev_warn(&pdev->dev, "%s-%d: could not read device ID\n", __func__, __LINE__);
+	return -ETIMEDOUT;
 }
 
 static void mlx5_pci_resume(struct pci_dev *pdev)
@@ -1473,7 +1458,11 @@ static void mlx5_pci_resume(struct pci_dev *pdev)
 	dev_info(&pdev->dev, "%s was called\n", __func__);
 
 	pci_save_state(pdev);
-	wait_vital(pdev);
+	err = wait_vital(pdev);
+	if (err) {
+		dev_err(&pdev->dev, "%s: wait_vital timed out\n", __func__);
+		return;
+	}
 
 	err = mlx5_load_one(dev, priv);
 	if (err)

drivers/net/ethernet/mellanox/mlx5/core/pagealloc.c

@@ -345,7 +345,6 @@ static int give_pages(struct mlx5_core_dev *dev, u16 func_id, int npages,
 			       func_id, npages, err);
 		goto out_4k;
 	}
-	dev->priv.fw_pages += npages;
 
 	err = mlx5_cmd_status_to_err(&out.hdr);
 	if (err) {
@@ -373,6 +372,33 @@ static int give_pages(struct mlx5_core_dev *dev, u16 func_id, int npages,
 	return err;
 }
 
+static int reclaim_pages_cmd(struct mlx5_core_dev *dev,
+			     struct mlx5_manage_pages_inbox *in, int in_size,
+			     struct mlx5_manage_pages_outbox *out, int out_size)
+{
+	struct fw_page *fwp;
+	struct rb_node *p;
+	u32 npages;
+	u32 i = 0;
+
+	if (dev->state != MLX5_DEVICE_STATE_INTERNAL_ERROR)
+		return mlx5_cmd_exec_check_status(dev, (u32 *)in, in_size,
+						  (u32 *)out, out_size);
+
+	npages = be32_to_cpu(in->num_entries);
+
+	p = rb_first(&dev->priv.page_root);
+	while (p && i < npages) {
+		fwp = rb_entry(p, struct fw_page, rb_node);
+		out->pas[i] = cpu_to_be64(fwp->addr);
+		p = rb_next(p);
+		i++;
+	}
+
+	out->num_entries = cpu_to_be32(i);
+	return 0;
+}
+
 static int reclaim_pages(struct mlx5_core_dev *dev, u32 func_id, int npages,
 			 int *nclaimed)
 {
@@ -398,15 +424,9 @@ static int reclaim_pages(struct mlx5_core_dev *dev, u32 func_id, int npages,
 	in.func_id = cpu_to_be16(func_id);
 	in.num_entries = cpu_to_be32(npages);
 	mlx5_core_dbg(dev, "npages %d, outlen %d\n", npages, outlen);
-	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, outlen);
+	err = reclaim_pages_cmd(dev, &in, sizeof(in), out, outlen);
 	if (err) {
-		mlx5_core_err(dev, "failed reclaiming pages\n");
-		goto out_free;
-	}
-	dev->priv.fw_pages -= npages;
-
-	if (out->hdr.status) {
-		err = mlx5_cmd_status_to_err(&out->hdr);
+		mlx5_core_err(dev, "failed reclaiming pages: err %d\n", err);
 		goto out_free;
 	}
 
@@ -417,13 +437,15 @@ static int reclaim_pages(struct mlx5_core_dev *dev, u32 func_id, int npages,
 		err = -EINVAL;
 		goto out_free;
 	}
-	if (nclaimed)
-		*nclaimed = num_claimed;
 
 	for (i = 0; i < num_claimed; i++) {
 		addr = be64_to_cpu(out->pas[i]);
 		free_4k(dev, addr);
 	}
+
+	if (nclaimed)
+		*nclaimed = num_claimed;
+
 	dev->priv.fw_pages -= num_claimed;
 	if (func_id)
 		dev->priv.vfs_pages -= num_claimed;
@@ -514,14 +536,10 @@ int mlx5_reclaim_startup_pages(struct mlx5_core_dev *dev)
 		p = rb_first(&dev->priv.page_root);
 		if (p) {
 			fwp = rb_entry(p, struct fw_page, rb_node);
-			if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) {
-				free_4k(dev, fwp->addr);
-				nclaimed = 1;
-			} else {
 			err = reclaim_pages(dev, fwp->func_id,
 					    optimal_reclaimed_pages(),
 					    &nclaimed);
-			}
+
 			if (err) {
 				mlx5_core_warn(dev, "failed reclaiming pages (%d)\n",
 					       err);
@@ -536,6 +554,13 @@ int mlx5_reclaim_startup_pages(struct mlx5_core_dev *dev)
 		}
 	} while (p);
 
+	WARN(dev->priv.fw_pages,
+	     "FW pages counter is %d after reclaiming all pages\n",
+	     dev->priv.fw_pages);
+	WARN(dev->priv.vfs_pages,
+	     "VFs FW pages counter is %d after reclaiming all pages\n",
+	     dev->priv.vfs_pages);
+
 	return 0;
 }
 

include/linux/mlx5/driver.h

@@ -629,6 +629,7 @@ struct mlx5_cmd_work_ent {
 	void		       *uout;
 	int			uout_size;
 	mlx5_cmd_cbk_t		callback;
+	struct delayed_work	cb_timeout_work;
 	void		       *context;
 	int			idx;
 	struct completion	done;