ath10k: disable TX complete indication of htt for sdio

For sdio chip, it is high latency bus, all the TX packet's content will
be tranferred from HOST memory to firmware memory via sdio bus, then it
need much more memory in firmware than low latency bus chip, for low
latency chip, such as PCI-E, it only need to transfer the TX descriptor
via PCI-E bus to firmware memory. For sdio chip, reduce the complexity of
TX logic will help TX efficiency since its memory is limited, and it will
reduce the TX circle's time of each packet and then firmware will have more
memory for TX since TX complete also need memeory.

This patch disable TX complete indication from firmware for htt data
packet, it will not have TX complete indication from firmware to ath10k.
It will cut the cost of bus bandwidth of TX complete and make the TX
logic of firmware simpler, it results in significant performance
improvement on TX path.

Udp TX throughout is 130Mbps without this patch, and it arrives
400Mbps with this patch.

The downside of this patch is the command "iw wlan0 station dump" will
show 0 for "tx retries" and "tx failed" since all tx packet's status
is success.

This patch only effect sdio chip, it will not effect PCI, SNOC etc.

Tested with QCA6174 SDIO with firmware
WLAN.RMH.4.4.1-00017-QCARMSWPZ-1

Signed-off-by: Wen Gong <wgong@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200212080415.31265-2-wgong@codeaurora.org

1 files changed, 33 insertions, 1 deletions

diff --git a/drivers/net/wireless/ath/ath10k/htt_rx.c b/drivers/net/wireless/ath/ath10k/htt_rx.c
index f883f2a724dd..64e45bfa5d05 100644
--- a/drivers/net/wireless/ath/ath10k/htt_rx.c
+++ b/drivers/net/wireless/ath/ath10k/htt_rx.c
@@ -3789,6 +3789,9 @@ bool ath10k_htt_t2h_msg_handler(struct ath10k *ar, struct sk_buff *skb)
 	}
 	case HTT_T2H_MSG_TYPE_MGMT_TX_COMPLETION: {
 		struct htt_tx_done tx_done = {};
+		struct ath10k_htt *htt = &ar->htt;
+		struct ath10k_htc *htc = &ar->htc;
+		struct ath10k_htc_ep *ep = &ar->htc.endpoint[htt->eid];
 		int status = __le32_to_cpu(resp->mgmt_tx_completion.status);
 		int info = __le32_to_cpu(resp->mgmt_tx_completion.info);
 
@@ -3814,6 +3817,12 @@ bool ath10k_htt_t2h_msg_handler(struct ath10k *ar, struct sk_buff *skb)
 			break;
 		}
 
+		if (htt->disable_tx_comp) {
+			spin_lock_bh(&htc->tx_lock);
+			ep->tx_credits++;
+			spin_unlock_bh(&htc->tx_lock);
+		}
+
 		status = ath10k_txrx_tx_unref(htt, &tx_done);
 		if (!status) {
 			spin_lock_bh(&htt->tx_lock);
@@ -3888,8 +3897,31 @@ bool ath10k_htt_t2h_msg_handler(struct ath10k *ar, struct sk_buff *skb)
 		skb_queue_tail(&htt->rx_in_ord_compl_q, skb);
 		return false;
 	}
-	case HTT_T2H_MSG_TYPE_TX_CREDIT_UPDATE_IND:
+	case HTT_T2H_MSG_TYPE_TX_CREDIT_UPDATE_IND: {
+		struct ath10k_htt *htt = &ar->htt;
+		struct ath10k_htc *htc = &ar->htc;
+		struct ath10k_htc_ep *ep = &ar->htc.endpoint[htt->eid];
+		u32 msg_word = __le32_to_cpu(*(__le32 *)resp);
+		int htt_credit_delta;
+
+		htt_credit_delta = HTT_TX_CREDIT_DELTA_ABS_GET(msg_word);
+		if (HTT_TX_CREDIT_SIGN_BIT_GET(msg_word))
+			htt_credit_delta = -htt_credit_delta;
+
+		ath10k_dbg(ar, ATH10K_DBG_HTT,
+			   "htt credit update delta %d\n",
+			   htt_credit_delta);
+
+		if (htt->disable_tx_comp) {
+			spin_lock_bh(&htc->tx_lock);
+			ep->tx_credits += htt_credit_delta;
+			spin_unlock_bh(&htc->tx_lock);
+			ath10k_dbg(ar, ATH10K_DBG_HTT,
+				   "htt credit total %d\n",
+				   ep->tx_credits);
+		}
 		break;
+	}
 	case HTT_T2H_MSG_TYPE_CHAN_CHANGE: {
 		u32 phymode = __le32_to_cpu(resp->chan_change.phymode);
 		u32 freq = __le32_to_cpu(resp->chan_change.freq);