qed: FW 8.42.2.0 Additional ll2 type

LL2 queues were a limited resource due to FW constraints.
This FW introduced a new resource which is a context based ll2 queue
(memory on host). The additional ll2 queues are required for RDMA SRIOV.
The code refers to the previous ll2 queues as ram-based or legacy, and the
new queues as ctx-based.
This change decreased the "legacy" ram-based queues therefore the first ll2
queue used for iWARP was converted to the ctx-based ll2 queue.
This feature also exposed a bug in the DIRECT_REG_WR64 macro implementation
which didn't have an effect in other use cases.

Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 132 insertions, 17 deletions

diff --git a/drivers/net/ethernet/qlogic/qed/qed_ll2.c b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
index 19a1a58d60f8..037e5978787e 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_ll2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
@@ -962,7 +962,7 @@ static int qed_sp_ll2_rx_queue_start(struct qed_hwfn *p_hwfn,
 		return rc;
 
 	p_ramrod = &p_ent->ramrod.core_rx_queue_start;
-
+	memset(p_ramrod, 0, sizeof(*p_ramrod));
 	p_ramrod->sb_id = cpu_to_le16(qed_int_get_sp_sb_id(p_hwfn));
 	p_ramrod->sb_index = p_rx->rx_sb_index;
 	p_ramrod->complete_event_flg = 1;
@@ -996,6 +996,8 @@ static int qed_sp_ll2_rx_queue_start(struct qed_hwfn *p_hwfn,
 
 	p_ramrod->action_on_error.error_type = action_on_error;
 	p_ramrod->gsi_offload_flag = p_ll2_conn->input.gsi_enable;
+	p_ramrod->zero_prod_flg = 1;
+
 	return qed_spq_post(p_hwfn, p_ent, NULL);
 }
 
@@ -1317,6 +1319,25 @@ qed_ll2_set_cbs(struct qed_ll2_info *p_ll2_info, const struct qed_ll2_cbs *cbs)
 	return 0;
 }
 
+static void _qed_ll2_calc_allowed_conns(struct qed_hwfn *p_hwfn,
+					struct qed_ll2_acquire_data *data,
+					u8 *start_idx, u8 *last_idx)
+{
+	/* LL2 queues handles will be split as follows:
+	 * First will be the legacy queues, and then the ctx based.
+	 */
+	if (data->input.rx_conn_type == QED_LL2_RX_TYPE_LEGACY) {
+		*start_idx = QED_LL2_LEGACY_CONN_BASE_PF;
+		*last_idx = *start_idx +
+			QED_MAX_NUM_OF_LEGACY_LL2_CONNS_PF;
+	} else {
+		/* QED_LL2_RX_TYPE_CTX */
+		*start_idx = QED_LL2_CTX_CONN_BASE_PF;
+		*last_idx = *start_idx +
+			QED_MAX_NUM_OF_CTX_LL2_CONNS_PF;
+	}
+}
+
 static enum core_error_handle
 qed_ll2_get_error_choice(enum qed_ll2_error_handle err)
 {
@@ -1337,14 +1358,16 @@ int qed_ll2_acquire_connection(void *cxt, struct qed_ll2_acquire_data *data)
 	struct qed_hwfn *p_hwfn = cxt;
 	qed_int_comp_cb_t comp_rx_cb, comp_tx_cb;
 	struct qed_ll2_info *p_ll2_info = NULL;
-	u8 i, *p_tx_max;
+	u8 i, first_idx, last_idx, *p_tx_max;
 	int rc;
 
 	if (!data->p_connection_handle || !p_hwfn->p_ll2_info)
 		return -EINVAL;
 
+	_qed_ll2_calc_allowed_conns(p_hwfn, data, &first_idx, &last_idx);
+
 	/* Find a free connection to be used */
-	for (i = 0; (i < QED_MAX_NUM_OF_LL2_CONNECTIONS); i++) {
+	for (i = first_idx; i < last_idx; i++) {
 		mutex_lock(&p_hwfn->p_ll2_info[i].mutex);
 		if (p_hwfn->p_ll2_info[i].b_active) {
 			mutex_unlock(&p_hwfn->p_ll2_info[i].mutex);
@@ -1448,6 +1471,7 @@ static int qed_ll2_establish_connection_rx(struct qed_hwfn *p_hwfn,
 	enum qed_ll2_error_handle error_input;
 	enum core_error_handle error_mode;
 	u8 action_on_error = 0;
+	int rc;
 
 	if (!QED_LL2_RX_REGISTERED(p_ll2_conn))
 		return 0;
@@ -1461,7 +1485,18 @@ static int qed_ll2_establish_connection_rx(struct qed_hwfn *p_hwfn,
 	error_mode = qed_ll2_get_error_choice(error_input);
 	SET_FIELD(action_on_error, CORE_RX_ACTION_ON_ERROR_NO_BUFF, error_mode);
 
-	return qed_sp_ll2_rx_queue_start(p_hwfn, p_ll2_conn, action_on_error);
+	rc = qed_sp_ll2_rx_queue_start(p_hwfn, p_ll2_conn, action_on_error);
+	if (rc)
+		return rc;
+
+	if (p_ll2_conn->rx_queue.ctx_based) {
+		rc = qed_db_recovery_add(p_hwfn->cdev,
+					 p_ll2_conn->rx_queue.set_prod_addr,
+					 &p_ll2_conn->rx_queue.db_data,
+					 DB_REC_WIDTH_64B, DB_REC_KERNEL);
+	}
+
+	return rc;
 }
 
 static void
@@ -1475,13 +1510,41 @@ qed_ll2_establish_connection_ooo(struct qed_hwfn *p_hwfn,
 	qed_ooo_submit_rx_buffers(p_hwfn, p_ll2_conn);
 }
 
+static inline u8 qed_ll2_handle_to_queue_id(struct qed_hwfn *p_hwfn,
+					    u8 handle,
+					    u8 ll2_queue_type)
+{
+	u8 qid;
+
+	if (ll2_queue_type == QED_LL2_RX_TYPE_LEGACY)
+		return p_hwfn->hw_info.resc_start[QED_LL2_RAM_QUEUE] + handle;
+
+	/* QED_LL2_RX_TYPE_CTX
+	 * FW distinguishes between the legacy queues (ram based) and the
+	 * ctx based queues by the queue_id.
+	 * The first MAX_NUM_LL2_RX_RAM_QUEUES queues are legacy
+	 * and the queue ids above that are ctx base.
+	 */
+	qid = p_hwfn->hw_info.resc_start[QED_LL2_CTX_QUEUE] +
+	      MAX_NUM_LL2_RX_RAM_QUEUES;
+
+	/* See comment on the acquire connection for how the ll2
+	 * queues handles are divided.
+	 */
+	qid += (handle - QED_MAX_NUM_OF_LEGACY_LL2_CONNS_PF);
+
+	return qid;
+}
+
 int qed_ll2_establish_connection(void *cxt, u8 connection_handle)
 {
-	struct qed_hwfn *p_hwfn = cxt;
-	struct qed_ll2_info *p_ll2_conn;
+	struct e4_core_conn_context *p_cxt;
 	struct qed_ll2_tx_packet *p_pkt;
+	struct qed_ll2_info *p_ll2_conn;
+	struct qed_hwfn *p_hwfn = cxt;
 	struct qed_ll2_rx_queue *p_rx;
 	struct qed_ll2_tx_queue *p_tx;
+	struct qed_cxt_info cxt_info;
 	struct qed_ptt *p_ptt;
 	int rc = -EINVAL;
 	u32 i, capacity;
@@ -1539,13 +1602,46 @@ int qed_ll2_establish_connection(void *cxt, u8 connection_handle)
 	rc = qed_cxt_acquire_cid(p_hwfn, PROTOCOLID_CORE, &p_ll2_conn->cid);
 	if (rc)
 		goto out;
+	cxt_info.iid = p_ll2_conn->cid;
+	rc = qed_cxt_get_cid_info(p_hwfn, &cxt_info);
+	if (rc) {
+		DP_NOTICE(p_hwfn, "Cannot find context info for cid=%d\n",
+			  p_ll2_conn->cid);
+		goto out;
+	}
+
+	p_cxt = cxt_info.p_cxt;
+
+	memset(p_cxt, 0, sizeof(*p_cxt));
 
-	qid = p_hwfn->hw_info.resc_start[QED_LL2_QUEUE] + connection_handle;
+	qid = qed_ll2_handle_to_queue_id(p_hwfn, connection_handle,
+					 p_ll2_conn->input.rx_conn_type);
 	p_ll2_conn->queue_id = qid;
 	p_ll2_conn->tx_stats_id = qid;
-	p_rx->set_prod_addr = (u8 __iomem *)p_hwfn->regview +
-					    GTT_BAR0_MAP_REG_TSDM_RAM +
-					    TSTORM_LL2_RX_PRODS_OFFSET(qid);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_LL2,
+		   "Establishing ll2 queue. PF %d ctx_based=%d abs qid=%d\n",
+		   p_hwfn->rel_pf_id, p_ll2_conn->input.rx_conn_type, qid);
+
+	if (p_ll2_conn->input.rx_conn_type == QED_LL2_RX_TYPE_LEGACY) {
+		p_rx->set_prod_addr = p_hwfn->regview +
+		    GTT_BAR0_MAP_REG_TSDM_RAM + TSTORM_LL2_RX_PRODS_OFFSET(qid);
+	} else {
+		/* QED_LL2_RX_TYPE_CTX - using doorbell */
+		p_rx->ctx_based = 1;
+
+		p_rx->set_prod_addr = p_hwfn->doorbells +
+			p_hwfn->dpi_start_offset +
+			DB_ADDR_SHIFT(DQ_PWM_OFFSET_TCM_LL2_PROD_UPDATE);
+
+		/* prepare db data */
+		p_rx->db_data.icid = cpu_to_le16((u16)p_ll2_conn->cid);
+		SET_FIELD(p_rx->db_data.params,
+			  CORE_PWM_PROD_UPDATE_DATA_AGG_CMD, DB_AGG_CMD_SET);
+		SET_FIELD(p_rx->db_data.params,
+			  CORE_PWM_PROD_UPDATE_DATA_RESERVED1, 0);
+	}
+
 	p_tx->doorbell_addr = (u8 __iomem *)p_hwfn->doorbells +
 					    qed_db_addr(p_ll2_conn->cid,
 							DQ_DEMS_LEGACY);
@@ -1556,7 +1652,6 @@ int qed_ll2_establish_connection(void *cxt, u8 connection_handle)
 		  DQ_XCM_CORE_TX_BD_PROD_CMD);
 	p_tx->db_msg.agg_flags = DQ_XCM_CORE_DQ_CF_CMD;
 
-
 	rc = qed_ll2_establish_connection_rx(p_hwfn, p_ll2_conn);
 	if (rc)
 		goto out;
@@ -1590,7 +1685,7 @@ static void qed_ll2_post_rx_buffer_notify_fw(struct qed_hwfn *p_hwfn,
 					     struct qed_ll2_rx_packet *p_curp)
 {
 	struct qed_ll2_rx_packet *p_posting_packet = NULL;
-	struct core_ll2_rx_prod rx_prod = { 0, 0, 0 };
+	struct core_ll2_rx_prod rx_prod = { 0, 0 };
 	bool b_notify_fw = false;
 	u16 bd_prod, cq_prod;
 
@@ -1615,13 +1710,27 @@ static void qed_ll2_post_rx_buffer_notify_fw(struct qed_hwfn *p_hwfn,
 
 	bd_prod = qed_chain_get_prod_idx(&p_rx->rxq_chain);
 	cq_prod = qed_chain_get_prod_idx(&p_rx->rcq_chain);
-	rx_prod.bd_prod = cpu_to_le16(bd_prod);
-	rx_prod.cqe_prod = cpu_to_le16(cq_prod);
+	if (p_rx->ctx_based) {
+		/* update producer by giving a doorbell */
+		p_rx->db_data.prod.bd_prod = cpu_to_le16(bd_prod);
+		p_rx->db_data.prod.cqe_prod = cpu_to_le16(cq_prod);
+		/* Make sure chain element is updated before ringing the
+		 * doorbell
+		 */
+		dma_wmb();
+		DIRECT_REG_WR64(p_rx->set_prod_addr,
+				*((u64 *)&p_rx->db_data));
+	} else {
+		rx_prod.bd_prod = cpu_to_le16(bd_prod);
+		rx_prod.cqe_prod = cpu_to_le16(cq_prod);
 
-	/* Make sure chain element is updated before ringing the doorbell */
-	dma_wmb();
+		/* Make sure chain element is updated before ringing the
+		 * doorbell
+		 */
+		dma_wmb();
 
-	DIRECT_REG_WR(p_rx->set_prod_addr, *((u32 *)&rx_prod));
+		DIRECT_REG_WR(p_rx->set_prod_addr, *((u32 *)&rx_prod));
+	}
 }
 
 int qed_ll2_post_rx_buffer(void *cxt,
@@ -1965,6 +2074,12 @@ int qed_ll2_terminate_connection(void *cxt, u8 connection_handle)
 	if (QED_LL2_RX_REGISTERED(p_ll2_conn)) {
 		p_ll2_conn->rx_queue.b_cb_registered = false;
 		smp_wmb(); /* Make sure this is seen by ll2_lb_rxq_completion */
+
+		if (p_ll2_conn->rx_queue.ctx_based)
+			qed_db_recovery_del(p_hwfn->cdev,
+					    p_ll2_conn->rx_queue.set_prod_addr,
+					    &p_ll2_conn->rx_queue.db_data);
+
 		rc = qed_sp_ll2_rx_queue_stop(p_hwfn, p_ll2_conn);
 		if (rc)
 			goto out;