iwlwifi: pcie: allocate and free rx cr's and tr's tails

The hw now refers to two new blocks:
* rx tr tail - The Tail index on the free buffers queue TR,
which is update by the device after reading the free buffer
from the tr.
* rx cr tail - Updated by the driver when completing
processing a new completion descriptor in the cr.

Add these two new struct to the rxq, allocate and free them
when needed.

In addition, the register for rx write pointer had been changed
to HBUS_TARG_WRPTR. The way to differentiate tx from rx is the
queue number. TX range is 0-511, and RX's is 512-527.

Signed-off-by: Golan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

1 files changed, 120 insertions, 81 deletions

diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/rx.c b/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
index d15f5ba2dc77..24ce2d6beb0f 100644
--- a/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
+++ b/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
@@ -209,7 +209,11 @@ static void iwl_pcie_rxq_inc_wr_ptr(struct iwl_trans *trans,
 	}
 
 	rxq->write_actual = round_down(rxq->write, 8);
-	if (trans->cfg->mq_rx_supported)
+	if (trans->cfg->device_family >= IWL_DEVICE_FAMILY_22560)
+		iwl_write32(trans, HBUS_TARG_WRPTR,
+			    (rxq->write_actual |
+			     ((FIRST_RX_QUEUE + rxq->id) << 16)));
+	else if (trans->cfg->mq_rx_supported)
 		iwl_write32(trans, RFH_Q_FRBDCB_WIDX_TRG(rxq->id),
 			    rxq->write_actual);
 	else
@@ -608,89 +612,146 @@ void iwl_pcie_rx_allocator_work(struct work_struct *data)
 	iwl_pcie_rx_allocator(trans_pcie->trans);
 }
 
-static int iwl_pcie_rx_alloc(struct iwl_trans *trans)
+static void iwl_pcie_free_rxq_dma(struct iwl_trans *trans,
+				  struct iwl_rxq *rxq)
+{
+	struct device *dev = trans->dev;
+	int free_size = trans->cfg->mq_rx_supported ? sizeof(__le64) :
+						      sizeof(__le32);
+
+	if (rxq->bd)
+		dma_free_coherent(dev, free_size * rxq->queue_size,
+				  rxq->bd, rxq->bd_dma);
+	rxq->bd_dma = 0;
+	rxq->bd = NULL;
+
+	if (rxq->rb_stts)
+		dma_free_coherent(trans->dev,
+				  sizeof(struct iwl_rb_status),
+				  rxq->rb_stts, rxq->rb_stts_dma);
+	rxq->rb_stts_dma = 0;
+	rxq->rb_stts = NULL;
+
+	if (rxq->used_bd)
+		dma_free_coherent(dev, sizeof(__le32) * rxq->queue_size,
+				  rxq->used_bd, rxq->used_bd_dma);
+	rxq->used_bd_dma = 0;
+	rxq->used_bd = NULL;
+
+	if (trans->cfg->device_family < IWL_DEVICE_FAMILY_22560)
+		return;
+
+	if (rxq->tr_tail)
+		dma_free_coherent(dev, sizeof(__le16),
+				  rxq->tr_tail, rxq->tr_tail_dma);
+	rxq->tr_tail_dma = 0;
+	rxq->tr_tail = NULL;
+
+	if (rxq->cr_tail)
+		dma_free_coherent(dev, sizeof(__le16),
+				  rxq->cr_tail, rxq->cr_tail_dma);
+	rxq->cr_tail_dma = 0;
+	rxq->cr_tail = NULL;
+}
+
+static int iwl_pcie_alloc_rxq_dma(struct iwl_trans *trans,
+				  struct iwl_rxq *rxq)
 {
 	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
-	struct iwl_rb_allocator *rba = &trans_pcie->rba;
 	struct device *dev = trans->dev;
 	int i;
 	int free_size = trans->cfg->mq_rx_supported ? sizeof(__le64) :
 						      sizeof(__le32);
 
-	if (WARN_ON(trans_pcie->rxq))
-		return -EINVAL;
+	spin_lock_init(&rxq->lock);
+	if (trans->cfg->mq_rx_supported)
+		rxq->queue_size = MQ_RX_TABLE_SIZE;
+	else
+		rxq->queue_size = RX_QUEUE_SIZE;
 
-	trans_pcie->rxq = kcalloc(trans->num_rx_queues, sizeof(struct iwl_rxq),
-				  GFP_KERNEL);
-	if (!trans_pcie->rxq)
-		return -EINVAL;
+	/*
+	 * Allocate the circular buffer of Read Buffer Descriptors
+	 * (RBDs)
+	 */
+	rxq->bd = dma_zalloc_coherent(dev,
+				      free_size * rxq->queue_size,
+				      &rxq->bd_dma, GFP_KERNEL);
+	if (!rxq->bd)
+		goto err;
 
-	spin_lock_init(&rba->lock);
+	if (trans->cfg->mq_rx_supported) {
+		rxq->used_bd = dma_zalloc_coherent(dev,
+						   sizeof(__le32) *
+						   rxq->queue_size,
+						   &rxq->used_bd_dma,
+						   GFP_KERNEL);
+		if (!rxq->used_bd)
+			goto err;
+	}
 
-	for (i = 0; i < trans->num_rx_queues; i++) {
-		struct iwl_rxq *rxq = &trans_pcie->rxq[i];
+	/* Allocate the driver's pointer to receive buffer status */
+	rxq->rb_stts = dma_zalloc_coherent(dev, sizeof(*rxq->rb_stts),
+					   &rxq->rb_stts_dma,
+					   GFP_KERNEL);
+	if (!rxq->rb_stts)
+		goto err;
 
-		spin_lock_init(&rxq->lock);
-		if (trans->cfg->mq_rx_supported)
-			rxq->queue_size = MQ_RX_TABLE_SIZE;
-		else
-			rxq->queue_size = RX_QUEUE_SIZE;
+	if (trans->cfg->device_family < IWL_DEVICE_FAMILY_22560)
+		return 0;
 
-		/*
-		 * Allocate the circular buffer of Read Buffer Descriptors
-		 * (RBDs)
-		 */
-		rxq->bd = dma_zalloc_coherent(dev,
-					     free_size * rxq->queue_size,
-					     &rxq->bd_dma, GFP_KERNEL);
-		if (!rxq->bd)
-			goto err;
+	/* Allocate the driver's pointer to TR tail */
+	rxq->tr_tail = dma_zalloc_coherent(dev, sizeof(__le16),
+					   &rxq->tr_tail_dma,
+					   GFP_KERNEL);
+	if (!rxq->tr_tail)
+		goto err;
 
-		if (trans->cfg->mq_rx_supported) {
-			rxq->used_bd = dma_zalloc_coherent(dev,
-							   sizeof(__le32) *
-							   rxq->queue_size,
-							   &rxq->used_bd_dma,
-							   GFP_KERNEL);
-			if (!rxq->used_bd)
-				goto err;
-		}
+	/* Allocate the driver's pointer to CR tail */
+	rxq->cr_tail = dma_zalloc_coherent(dev, sizeof(__le16),
+					   &rxq->cr_tail_dma,
+					   GFP_KERNEL);
+	if (!rxq->cr_tail)
+		goto err;
 
-		/*Allocate the driver's pointer to receive buffer status */
-		rxq->rb_stts = dma_zalloc_coherent(dev, sizeof(*rxq->rb_stts),
-						   &rxq->rb_stts_dma,
-						   GFP_KERNEL);
-		if (!rxq->rb_stts)
-			goto err;
-	}
 	return 0;
 
 err:
 	for (i = 0; i < trans->num_rx_queues; i++) {
 		struct iwl_rxq *rxq = &trans_pcie->rxq[i];
 
-		if (rxq->bd)
-			dma_free_coherent(dev, free_size * rxq->queue_size,
-					  rxq->bd, rxq->bd_dma);
-		rxq->bd_dma = 0;
-		rxq->bd = NULL;
-
-		if (rxq->rb_stts)
-			dma_free_coherent(trans->dev,
-					  sizeof(struct iwl_rb_status),
-					  rxq->rb_stts, rxq->rb_stts_dma);
-
-		if (rxq->used_bd)
-			dma_free_coherent(dev, sizeof(__le32) * rxq->queue_size,
-					  rxq->used_bd, rxq->used_bd_dma);
-		rxq->used_bd_dma = 0;
-		rxq->used_bd = NULL;
+		iwl_pcie_free_rxq_dma(trans, rxq);
 	}
 	kfree(trans_pcie->rxq);
 
 	return -ENOMEM;
 }
 
+static int iwl_pcie_rx_alloc(struct iwl_trans *trans)
+{
+	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
+	struct iwl_rb_allocator *rba = &trans_pcie->rba;
+	int i, ret;
+
+	if (WARN_ON(trans_pcie->rxq))
+		return -EINVAL;
+
+	trans_pcie->rxq = kcalloc(trans->num_rx_queues, sizeof(struct iwl_rxq),
+				  GFP_KERNEL);
+	if (!trans_pcie->rxq)
+		return -EINVAL;
+
+	spin_lock_init(&rba->lock);
+
+	for (i = 0; i < trans->num_rx_queues; i++) {
+		struct iwl_rxq *rxq = &trans_pcie->rxq[i];
+
+		ret = iwl_pcie_alloc_rxq_dma(trans, rxq);
+		if (ret)
+			return ret;
+	}
+	return 0;
+}
+
 static void iwl_pcie_rx_hw_init(struct iwl_trans *trans, struct iwl_rxq *rxq)
 {
 	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
@@ -1002,8 +1063,6 @@ void iwl_pcie_rx_free(struct iwl_trans *trans)
 {
 	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
 	struct iwl_rb_allocator *rba = &trans_pcie->rba;
-	int free_size = trans->cfg->mq_rx_supported ? sizeof(__le64) :
-					      sizeof(__le32);
 	int i;
 
 	/*
@@ -1022,27 +1081,7 @@ void iwl_pcie_rx_free(struct iwl_trans *trans)
 	for (i = 0; i < trans->num_rx_queues; i++) {
 		struct iwl_rxq *rxq = &trans_pcie->rxq[i];
 
-		if (rxq->bd)
-			dma_free_coherent(trans->dev,
-					  free_size * rxq->queue_size,
-					  rxq->bd, rxq->bd_dma);
-		rxq->bd_dma = 0;
-		rxq->bd = NULL;
-
-		if (rxq->rb_stts)
-			dma_free_coherent(trans->dev,
-					  sizeof(struct iwl_rb_status),
-					  rxq->rb_stts, rxq->rb_stts_dma);
-		else
-			IWL_DEBUG_INFO(trans,
-				       "Free rxq->rb_stts which is NULL\n");
-
-		if (rxq->used_bd)
-			dma_free_coherent(trans->dev,
-					  sizeof(__le32) * rxq->queue_size,
-					  rxq->used_bd, rxq->used_bd_dma);
-		rxq->used_bd_dma = 0;
-		rxq->used_bd = NULL;
+		iwl_pcie_free_rxq_dma(trans, rxq);
 
 		if (rxq->napi.poll)
 			netif_napi_del(&rxq->napi);