[PATCH 0/7] ath10k: add copy engine fast path support

[PATCH 0/7] ath10k: add copy engine fast path support

[Rajkumar Manoharan]

Below patchset adds fast path support for uplink traffic by bypassing
HTC layer processing. This is enabled by making use of unused copy
engine 5 to receive HTT messages directly from HIF layer. From initial
validation in VHT80/5G mode TCP UL is improved to 900Mbps from ~840Mbps
in conducted test.

-Rajkumar

Rajkumar Manoharan (7):
  ath10k: export htc tx rx handlers
  ath10k: register per copy engine send completion callbacks
  ath10k: register per copy engine receive callbacks
  ath10k: export htt tx rx handlers
  ath10k: Configure copy engine 5 for HTT messages
  ath10k: remove unused dl_is_polled
  ath10k: remove htc polling for tx completion

 drivers/net/wireless/ath/ath10k/ce.c     |   8 +-
 drivers/net/wireless/ath/ath10k/ce.h     |   7 +-
 drivers/net/wireless/ath/ath10k/hif.h    |  26 +------
 drivers/net/wireless/ath/ath10k/htc.c    |  51 ++----------
 drivers/net/wireless/ath/ath10k/htc.h    |   4 +-
 drivers/net/wireless/ath/ath10k/htt.h    |   1 +
 drivers/net/wireless/ath/ath10k/htt_rx.c |   1 +
 drivers/net/wireless/ath/ath10k/htt_tx.c |   6 ++
 drivers/net/wireless/ath/ath10k/pci.c    | 129 ++++++++++++++++++++-----------
 drivers/net/wireless/ath/ath10k/pci.h    |   2 -
 10 files changed, 108 insertions(+), 127 deletions(-)

-- 
2.5.2

[PATCH 1/7] ath10k: export htc tx rx handlers

[Rajkumar Manoharan]

Export HTC layer tx and rx handlers. This will be used by HIF layer
for per-CE data processing. Instead of callback mechanism, HIF will
call appropriate upper layers API directly.

Reviewed-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
---
 drivers/net/wireless/ath/ath10k/htc.c | 8 ++++----
 drivers/net/wireless/ath/ath10k/htc.h | 2 ++
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/htc.c b/drivers/net/wireless/ath/ath10k/htc.c
index 32d9ff1..97c24b2 100644
--- a/drivers/net/wireless/ath/ath10k/htc.c
+++ b/drivers/net/wireless/ath/ath10k/htc.c
@@ -181,8 +181,7 @@ err_pull:
 	return ret;
 }
 
-static int ath10k_htc_tx_completion_handler(struct ath10k *ar,
-					    struct sk_buff *skb)
+int ath10k_htc_tx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 {
 	struct ath10k_htc *htc = &ar->htc;
 	struct ath10k_skb_cb *skb_cb;
@@ -199,6 +198,7 @@ static int ath10k_htc_tx_completion_handler(struct ath10k *ar,
 
 	return 0;
 }
+EXPORT_SYMBOL(ath10k_htc_tx_completion_handler);
 
 /***********/
 /* Receive */
@@ -304,8 +304,7 @@ static int ath10k_htc_process_trailer(struct ath10k_htc *htc,
 	return status;
 }
 
-static int ath10k_htc_rx_completion_handler(struct ath10k *ar,
-					    struct sk_buff *skb)
+int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 {
 	int status = 0;
 	struct ath10k_htc *htc = &ar->htc;
@@ -442,6 +441,7 @@ out:
 
 	return status;
 }
+EXPORT_SYMBOL(ath10k_htc_rx_completion_handler);
 
 static void ath10k_htc_control_rx_complete(struct ath10k *ar,
 					   struct sk_buff *skb)
diff --git a/drivers/net/wireless/ath/ath10k/htc.h b/drivers/net/wireless/ath/ath10k/htc.h
index 527179c..d1fc8e7 100644
--- a/drivers/net/wireless/ath/ath10k/htc.h
+++ b/drivers/net/wireless/ath/ath10k/htc.h
@@ -355,5 +355,7 @@ int ath10k_htc_connect_service(struct ath10k_htc *htc,
 int ath10k_htc_send(struct ath10k_htc *htc, enum ath10k_htc_ep_id eid,
 		    struct sk_buff *packet);
 struct sk_buff *ath10k_htc_alloc_skb(struct ath10k *ar, int size);
+int ath10k_htc_tx_completion_handler(struct ath10k *ar, struct sk_buff *skb);
+int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb);
 
 #endif
-- 
2.5.2

[PATCH 2/7] ath10k: register per copy engine send completion callbacks

[Rajkumar Manoharan]

Register send completion callbacks for every copy engines (CE) separately
instead of having common completion handler. Since some of the copy
engines delivers different type of messages, per-CE callbacks help to
service them differently.

Reviewed-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
---
 drivers/net/wireless/ath/ath10k/ce.c  |  3 +--
 drivers/net/wireless/ath/ath10k/ce.h  |  3 ++-
 drivers/net/wireless/ath/ath10k/hif.h |  2 --
 drivers/net/wireless/ath/ath10k/htc.c |  7 ++-----
 drivers/net/wireless/ath/ath10k/htc.h |  2 +-
 drivers/net/wireless/ath/ath10k/pci.c | 15 +++++++++------
 6 files changed, 15 insertions(+), 17 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/ce.c b/drivers/net/wireless/ath/ath10k/ce.c
index cf28fbe..05cab4f 100644
--- a/drivers/net/wireless/ath/ath10k/ce.c
+++ b/drivers/net/wireless/ath/ath10k/ce.c
@@ -1077,7 +1077,6 @@ void ath10k_ce_deinit_pipe(struct ath10k *ar, unsigned int ce_id)
 
 int ath10k_ce_alloc_pipe(struct ath10k *ar, int ce_id,
 			 const struct ce_attr *attr,
-			 void (*send_cb)(struct ath10k_ce_pipe *),
 			 void (*recv_cb)(struct ath10k_ce_pipe *))
 {
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
@@ -1104,7 +1103,7 @@ int ath10k_ce_alloc_pipe(struct ath10k *ar, int ce_id,
 	ce_state->src_sz_max = attr->src_sz_max;
 
 	if (attr->src_nentries)
-		ce_state->send_cb = send_cb;
+		ce_state->send_cb = attr->send_cb;
 
 	if (attr->dest_nentries)
 		ce_state->recv_cb = recv_cb;
diff --git a/drivers/net/wireless/ath/ath10k/ce.h b/drivers/net/wireless/ath/ath10k/ce.h
index 5c903e15..3829e33 100644
--- a/drivers/net/wireless/ath/ath10k/ce.h
+++ b/drivers/net/wireless/ath/ath10k/ce.h
@@ -210,7 +210,6 @@ int ath10k_ce_init_pipe(struct ath10k *ar, unsigned int ce_id,
 void ath10k_ce_deinit_pipe(struct ath10k *ar, unsigned int ce_id);
 int ath10k_ce_alloc_pipe(struct ath10k *ar, int ce_id,
 			 const struct ce_attr *attr,
-			 void (*send_cb)(struct ath10k_ce_pipe *),
 			 void (*recv_cb)(struct ath10k_ce_pipe *));
 void ath10k_ce_free_pipe(struct ath10k *ar, int ce_id);
 
@@ -277,6 +276,8 @@ struct ce_attr {
 
 	/* #entries in destination ring - Must be a power of 2 */
 	unsigned int dest_nentries;
+
+	void (*send_cb)(struct ath10k_ce_pipe *);
 };
 
 #define SR_BA_ADDRESS		0x0000
diff --git a/drivers/net/wireless/ath/ath10k/hif.h b/drivers/net/wireless/ath/ath10k/hif.h
index 0c92e02..3808920 100644
--- a/drivers/net/wireless/ath/ath10k/hif.h
+++ b/drivers/net/wireless/ath/ath10k/hif.h
@@ -31,8 +31,6 @@ struct ath10k_hif_sg_item {
 };
 
 struct ath10k_hif_cb {
-	int (*tx_completion)(struct ath10k *ar,
-			     struct sk_buff *wbuf);
 	int (*rx_completion)(struct ath10k *ar,
 			     struct sk_buff *wbuf);
 };
diff --git a/drivers/net/wireless/ath/ath10k/htc.c b/drivers/net/wireless/ath/ath10k/htc.c
index 97c24b2..89c0e40 100644
--- a/drivers/net/wireless/ath/ath10k/htc.c
+++ b/drivers/net/wireless/ath/ath10k/htc.c
@@ -181,22 +181,20 @@ err_pull:
 	return ret;
 }
 
-int ath10k_htc_tx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
+void ath10k_htc_tx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 {
 	struct ath10k_htc *htc = &ar->htc;
 	struct ath10k_skb_cb *skb_cb;
 	struct ath10k_htc_ep *ep;
 
 	if (WARN_ON_ONCE(!skb))
-		return 0;
+		return;
 
 	skb_cb = ATH10K_SKB_CB(skb);
 	ep = &htc->endpoint[skb_cb->eid];
 
 	ath10k_htc_notify_tx_completion(ep, skb);
 	/* the skb now belongs to the completion handler */
-
-	return 0;
 }
 EXPORT_SYMBOL(ath10k_htc_tx_completion_handler);
 
@@ -851,7 +849,6 @@ int ath10k_htc_init(struct ath10k *ar)
 
 	/* setup HIF layer callbacks */
 	htc_callbacks.rx_completion = ath10k_htc_rx_completion_handler;
-	htc_callbacks.tx_completion = ath10k_htc_tx_completion_handler;
 	htc->ar = ar;
 
 	/* Get HIF default pipe for HTC message exchange */
diff --git a/drivers/net/wireless/ath/ath10k/htc.h b/drivers/net/wireless/ath/ath10k/htc.h
index d1fc8e7..77b669e 100644
--- a/drivers/net/wireless/ath/ath10k/htc.h
+++ b/drivers/net/wireless/ath/ath10k/htc.h
@@ -355,7 +355,7 @@ int ath10k_htc_connect_service(struct ath10k_htc *htc,
 int ath10k_htc_send(struct ath10k_htc *htc, enum ath10k_htc_ep_id eid,
 		    struct sk_buff *packet);
 struct sk_buff *ath10k_htc_alloc_skb(struct ath10k *ar, int size);
-int ath10k_htc_tx_completion_handler(struct ath10k *ar, struct sk_buff *skb);
+void ath10k_htc_tx_completion_handler(struct ath10k *ar, struct sk_buff *skb);
 int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb);
 
 #endif
diff --git a/drivers/net/wireless/ath/ath10k/pci.c b/drivers/net/wireless/ath/ath10k/pci.c
index 1046ab6..2655125 100644
--- a/drivers/net/wireless/ath/ath10k/pci.c
+++ b/drivers/net/wireless/ath/ath10k/pci.c
@@ -104,6 +104,7 @@ static int ath10k_pci_bmi_wait(struct ath10k_ce_pipe *tx_pipe,
 			       struct ath10k_ce_pipe *rx_pipe,
 			       struct bmi_xfer *xfer);
 static int ath10k_pci_qca99x0_chip_reset(struct ath10k *ar);
+static void ath10k_pci_htc_tx_cb(struct ath10k_ce_pipe *ce_state);
 
 static const struct ce_attr host_ce_config_wlan[] = {
 	/* CE0: host->target HTC control and raw streams */
@@ -112,6 +113,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_nentries = 16,
 		.src_sz_max = 256,
 		.dest_nentries = 0,
+		.send_cb = ath10k_pci_htc_tx_cb,
 	},
 
 	/* CE1: target->host HTT + HTC control */
@@ -120,6 +122,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_nentries = 0,
 		.src_sz_max = 2048,
 		.dest_nentries = 512,
+		.send_cb = ath10k_pci_htc_tx_cb,
 	},
 
 	/* CE2: target->host WMI */
@@ -128,6 +131,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_nentries = 0,
 		.src_sz_max = 2048,
 		.dest_nentries = 128,
+		.send_cb = ath10k_pci_htc_tx_cb,
 	},
 
 	/* CE3: host->target WMI */
@@ -136,6 +140,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_nentries = 32,
 		.src_sz_max = 2048,
 		.dest_nentries = 0,
+		.send_cb = ath10k_pci_htc_tx_cb,
 	},
 
 	/* CE4: host->target HTT */
@@ -144,6 +149,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_nentries = CE_HTT_H2T_MSG_SRC_NENTRIES,
 		.src_sz_max = 256,
 		.dest_nentries = 0,
+		.send_cb = ath10k_pci_htc_tx_cb,
 	},
 
 	/* CE5: unused */
@@ -1102,11 +1108,9 @@ static int ath10k_pci_diag_write32(struct ath10k *ar, u32 address, u32 value)
 }
 
 /* Called by lower (CE) layer when a send to Target completes. */
-static void ath10k_pci_ce_send_done(struct ath10k_ce_pipe *ce_state)
+static void ath10k_pci_htc_tx_cb(struct ath10k_ce_pipe *ce_state)
 {
 	struct ath10k *ar = ce_state->ar;
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	struct ath10k_hif_cb *cb = &ar_pci->msg_callbacks_current;
 	struct sk_buff_head list;
 	struct sk_buff *skb;
 	u32 ce_data;
@@ -1124,7 +1128,7 @@ static void ath10k_pci_ce_send_done(struct ath10k_ce_pipe *ce_state)
 	}
 
 	while ((skb = __skb_dequeue(&list)))
-		cb->tx_completion(ar, skb);
+		ath10k_htc_tx_completion_handler(ar, skb);
 }
 
 /* Called by lower (CE) layer when data is received from the Target. */
@@ -1579,7 +1583,7 @@ static void ath10k_pci_tx_pipe_cleanup(struct ath10k_pci_pipe *pci_pipe)
 
 		ce_ring->per_transfer_context[i] = NULL;
 
-		ar_pci->msg_callbacks_current.tx_completion(ar, skb);
+		ath10k_htc_tx_completion_handler(ar, skb);
 	}
 }
 
@@ -2000,7 +2004,6 @@ static int ath10k_pci_alloc_pipes(struct ath10k *ar)
 		pipe->hif_ce_state = ar;
 
 		ret = ath10k_ce_alloc_pipe(ar, i, &host_ce_config_wlan[i],
-					   ath10k_pci_ce_send_done,
 					   ath10k_pci_ce_recv_data);
 		if (ret) {
 			ath10k_err(ar, "failed to allocate copy engine pipe %d: %d\n",
-- 
2.5.2

[PATCH 3/7] ath10k: register per copy engine receive callbacks

[Rajkumar Manoharan]

Register receive callbacks for every copy engines (CE) separately
instead of having common receive handler. Some of the copy engines
receives different type of messages (i.e HTT/HTC/pktlog) from target.
Hence to service them accordingly, register per copy engine receive
callbacks.

Reviewed-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
---
 drivers/net/wireless/ath/ath10k/ce.c  |  5 ++---
 drivers/net/wireless/ath/ath10k/ce.h  |  4 ++--
 drivers/net/wireless/ath/ath10k/hif.h | 14 --------------
 drivers/net/wireless/ath/ath10k/htc.c | 13 +------------
 drivers/net/wireless/ath/ath10k/htc.h |  2 +-
 drivers/net/wireless/ath/ath10k/pci.c | 26 +++++++++-----------------
 drivers/net/wireless/ath/ath10k/pci.h |  2 --
 7 files changed, 15 insertions(+), 51 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/ce.c b/drivers/net/wireless/ath/ath10k/ce.c
index 05cab4f..e90840d 100644
--- a/drivers/net/wireless/ath/ath10k/ce.c
+++ b/drivers/net/wireless/ath/ath10k/ce.c
@@ -1076,8 +1076,7 @@ void ath10k_ce_deinit_pipe(struct ath10k *ar, unsigned int ce_id)
 }
 
 int ath10k_ce_alloc_pipe(struct ath10k *ar, int ce_id,
-			 const struct ce_attr *attr,
-			 void (*recv_cb)(struct ath10k_ce_pipe *))
+			 const struct ce_attr *attr)
 {
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
 	struct ath10k_ce_pipe *ce_state = &ar_pci->ce_states[ce_id];
@@ -1106,7 +1105,7 @@ int ath10k_ce_alloc_pipe(struct ath10k *ar, int ce_id,
 		ce_state->send_cb = attr->send_cb;
 
 	if (attr->dest_nentries)
-		ce_state->recv_cb = recv_cb;
+		ce_state->recv_cb = attr->recv_cb;
 
 	if (attr->src_nentries) {
 		ce_state->src_ring = ath10k_ce_alloc_src_ring(ar, ce_id, attr);
diff --git a/drivers/net/wireless/ath/ath10k/ce.h b/drivers/net/wireless/ath/ath10k/ce.h
index 3829e33..dbb94fd 100644
--- a/drivers/net/wireless/ath/ath10k/ce.h
+++ b/drivers/net/wireless/ath/ath10k/ce.h
@@ -209,8 +209,7 @@ int ath10k_ce_init_pipe(struct ath10k *ar, unsigned int ce_id,
 			const struct ce_attr *attr);
 void ath10k_ce_deinit_pipe(struct ath10k *ar, unsigned int ce_id);
 int ath10k_ce_alloc_pipe(struct ath10k *ar, int ce_id,
-			 const struct ce_attr *attr,
-			 void (*recv_cb)(struct ath10k_ce_pipe *));
+			 const struct ce_attr *attr);
 void ath10k_ce_free_pipe(struct ath10k *ar, int ce_id);
 
 /*==================CE Engine Shutdown=======================*/
@@ -278,6 +277,7 @@ struct ce_attr {
 	unsigned int dest_nentries;
 
 	void (*send_cb)(struct ath10k_ce_pipe *);
+	void (*recv_cb)(struct ath10k_ce_pipe *);
 };
 
 #define SR_BA_ADDRESS		0x0000
diff --git a/drivers/net/wireless/ath/ath10k/hif.h b/drivers/net/wireless/ath/ath10k/hif.h
index 3808920..6e826bb 100644
--- a/drivers/net/wireless/ath/ath10k/hif.h
+++ b/drivers/net/wireless/ath/ath10k/hif.h
@@ -30,11 +30,6 @@ struct ath10k_hif_sg_item {
 	u16 len;
 };
 
-struct ath10k_hif_cb {
-	int (*rx_completion)(struct ath10k *ar,
-			     struct sk_buff *wbuf);
-};
-
 struct ath10k_hif_ops {
 	/* send a scatter-gather list to the target */
 	int (*tx_sg)(struct ath10k *ar, u8 pipe_id,
@@ -78,9 +73,6 @@ struct ath10k_hif_ops {
 	 */
 	void (*send_complete_check)(struct ath10k *ar, u8 pipe_id, int force);
 
-	void (*set_callbacks)(struct ath10k *ar,
-			      struct ath10k_hif_cb *callbacks);
-
 	u16 (*get_free_queue_number)(struct ath10k *ar, u8 pipe_id);
 
 	u32 (*read32)(struct ath10k *ar, u32 address);
@@ -161,12 +153,6 @@ static inline void ath10k_hif_send_complete_check(struct ath10k *ar,
 	ar->hif.ops->send_complete_check(ar, pipe_id, force);
 }
 
-static inline void ath10k_hif_set_callbacks(struct ath10k *ar,
-					    struct ath10k_hif_cb *callbacks)
-{
-	ar->hif.ops->set_callbacks(ar, callbacks);
-}
-
 static inline u16 ath10k_hif_get_free_queue_number(struct ath10k *ar,
 						   u8 pipe_id)
 {
diff --git a/drivers/net/wireless/ath/ath10k/htc.c b/drivers/net/wireless/ath/ath10k/htc.c
index 89c0e40..13d0119 100644
--- a/drivers/net/wireless/ath/ath10k/htc.c
+++ b/drivers/net/wireless/ath/ath10k/htc.c
@@ -302,7 +302,7 @@ static int ath10k_htc_process_trailer(struct ath10k_htc *htc,
 	return status;
 }
 
-int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
+void ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 {
 	int status = 0;
 	struct ath10k_htc *htc = &ar->htc;
@@ -323,7 +323,6 @@ int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 		ath10k_warn(ar, "HTC Rx: invalid eid %d\n", eid);
 		ath10k_dbg_dump(ar, ATH10K_DBG_HTC, "htc bad header", "",
 				hdr, sizeof(*hdr));
-		status = -EINVAL;
 		goto out;
 	}
 
@@ -345,7 +344,6 @@ int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 			    payload_len + sizeof(*hdr));
 		ath10k_dbg_dump(ar, ATH10K_DBG_HTC, "htc bad rx pkt len", "",
 				hdr, sizeof(*hdr));
-		status = -EINVAL;
 		goto out;
 	}
 
@@ -355,7 +353,6 @@ int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 			   skb->len, payload_len);
 		ath10k_dbg_dump(ar, ATH10K_DBG_HTC, "htc bad rx pkt len",
 				"", hdr, sizeof(*hdr));
-		status = -EINVAL;
 		goto out;
 	}
 
@@ -371,7 +368,6 @@ int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 		    (trailer_len > payload_len)) {
 			ath10k_warn(ar, "Invalid trailer length: %d\n",
 				    trailer_len);
-			status = -EPROTO;
 			goto out;
 		}
 
@@ -404,7 +400,6 @@ int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 				 * sending unsolicited messages on the ep 0
 				 */
 				ath10k_warn(ar, "HTC rx ctrl still processing\n");
-				status = -EINVAL;
 				complete(&htc->ctl_resp);
 				goto out;
 			}
@@ -436,8 +431,6 @@ int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 	skb = NULL;
 out:
 	kfree_skb(skb);
-
-	return status;
 }
 EXPORT_SYMBOL(ath10k_htc_rx_completion_handler);
 
@@ -839,7 +832,6 @@ int ath10k_htc_start(struct ath10k_htc *htc)
 /* registered target arrival callback from the HIF layer */
 int ath10k_htc_init(struct ath10k *ar)
 {
-	struct ath10k_hif_cb htc_callbacks;
 	struct ath10k_htc_ep *ep = NULL;
 	struct ath10k_htc *htc = &ar->htc;
 
@@ -847,14 +839,11 @@ int ath10k_htc_init(struct ath10k *ar)
 
 	ath10k_htc_reset_endpoint_states(htc);
 
-	/* setup HIF layer callbacks */
-	htc_callbacks.rx_completion = ath10k_htc_rx_completion_handler;
 	htc->ar = ar;
 
 	/* Get HIF default pipe for HTC message exchange */
 	ep = &htc->endpoint[ATH10K_HTC_EP_0];
 
-	ath10k_hif_set_callbacks(ar, &htc_callbacks);
 	ath10k_hif_get_default_pipe(ar, &ep->ul_pipe_id, &ep->dl_pipe_id);
 
 	init_completion(&htc->ctl_resp);
diff --git a/drivers/net/wireless/ath/ath10k/htc.h b/drivers/net/wireless/ath/ath10k/htc.h
index 77b669e..aed3708 100644
--- a/drivers/net/wireless/ath/ath10k/htc.h
+++ b/drivers/net/wireless/ath/ath10k/htc.h
@@ -356,6 +356,6 @@ int ath10k_htc_send(struct ath10k_htc *htc, enum ath10k_htc_ep_id eid,
 		    struct sk_buff *packet);
 struct sk_buff *ath10k_htc_alloc_skb(struct ath10k *ar, int size);
 void ath10k_htc_tx_completion_handler(struct ath10k *ar, struct sk_buff *skb);
-int ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb);
+void ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb);
 
 #endif
diff --git a/drivers/net/wireless/ath/ath10k/pci.c b/drivers/net/wireless/ath/ath10k/pci.c
index 2655125..617728d 100644
--- a/drivers/net/wireless/ath/ath10k/pci.c
+++ b/drivers/net/wireless/ath/ath10k/pci.c
@@ -105,6 +105,7 @@ static int ath10k_pci_bmi_wait(struct ath10k_ce_pipe *tx_pipe,
 			       struct bmi_xfer *xfer);
 static int ath10k_pci_qca99x0_chip_reset(struct ath10k *ar);
 static void ath10k_pci_htc_tx_cb(struct ath10k_ce_pipe *ce_state);
+static void ath10k_pci_htc_rx_cb(struct ath10k_ce_pipe *ce_state);
 
 static const struct ce_attr host_ce_config_wlan[] = {
 	/* CE0: host->target HTC control and raw streams */
@@ -114,6 +115,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_sz_max = 256,
 		.dest_nentries = 0,
 		.send_cb = ath10k_pci_htc_tx_cb,
+		.recv_cb = ath10k_pci_htc_rx_cb,
 	},
 
 	/* CE1: target->host HTT + HTC control */
@@ -123,6 +125,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_sz_max = 2048,
 		.dest_nentries = 512,
 		.send_cb = ath10k_pci_htc_tx_cb,
+		.recv_cb = ath10k_pci_htc_rx_cb,
 	},
 
 	/* CE2: target->host WMI */
@@ -132,6 +135,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_sz_max = 2048,
 		.dest_nentries = 128,
 		.send_cb = ath10k_pci_htc_tx_cb,
+		.recv_cb = ath10k_pci_htc_rx_cb,
 	},
 
 	/* CE3: host->target WMI */
@@ -141,6 +145,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_sz_max = 2048,
 		.dest_nentries = 0,
 		.send_cb = ath10k_pci_htc_tx_cb,
+		.recv_cb = ath10k_pci_htc_rx_cb,
 	},
 
 	/* CE4: host->target HTT */
@@ -150,6 +155,7 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_sz_max = 256,
 		.dest_nentries = 0,
 		.send_cb = ath10k_pci_htc_tx_cb,
+		.recv_cb = ath10k_pci_htc_rx_cb,
 	},
 
 	/* CE5: unused */
@@ -1132,12 +1138,11 @@ static void ath10k_pci_htc_tx_cb(struct ath10k_ce_pipe *ce_state)
 }
 
 /* Called by lower (CE) layer when data is received from the Target. */
-static void ath10k_pci_ce_recv_data(struct ath10k_ce_pipe *ce_state)
+static void ath10k_pci_htc_rx_cb(struct ath10k_ce_pipe *ce_state)
 {
 	struct ath10k *ar = ce_state->ar;
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
 	struct ath10k_pci_pipe *pipe_info =  &ar_pci->pipe_info[ce_state->id];
-	struct ath10k_hif_cb *cb = &ar_pci->msg_callbacks_current;
 	struct sk_buff *skb;
 	struct sk_buff_head list;
 	void *transfer_context;
@@ -1172,7 +1177,7 @@ static void ath10k_pci_ce_recv_data(struct ath10k_ce_pipe *ce_state)
 		ath10k_dbg_dump(ar, ATH10K_DBG_PCI_DUMP, NULL, "pci rx: ",
 				skb->data, skb->len);
 
-		cb->rx_completion(ar, skb);
+		ath10k_htc_rx_completion_handler(ar, skb);
 	}
 
 	ath10k_pci_rx_post_pipe(pipe_info);
@@ -1347,17 +1352,6 @@ static void ath10k_pci_hif_send_complete_check(struct ath10k *ar, u8 pipe,
 	ath10k_ce_per_engine_service(ar, pipe);
 }
 
-static void ath10k_pci_hif_set_callbacks(struct ath10k *ar,
-					 struct ath10k_hif_cb *callbacks)
-{
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-
-	ath10k_dbg(ar, ATH10K_DBG_PCI, "pci hif set callbacks\n");
-
-	memcpy(&ar_pci->msg_callbacks_current, callbacks,
-	       sizeof(ar_pci->msg_callbacks_current));
-}
-
 static void ath10k_pci_kill_tasklet(struct ath10k *ar)
 {
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
@@ -2003,8 +1997,7 @@ static int ath10k_pci_alloc_pipes(struct ath10k *ar)
 		pipe->pipe_num = i;
 		pipe->hif_ce_state = ar;
 
-		ret = ath10k_ce_alloc_pipe(ar, i, &host_ce_config_wlan[i],
-					   ath10k_pci_ce_recv_data);
+		ret = ath10k_ce_alloc_pipe(ar, i, &host_ce_config_wlan[i]);
 		if (ret) {
 			ath10k_err(ar, "failed to allocate copy engine pipe %d: %d\n",
 				   i, ret);
@@ -2424,7 +2417,6 @@ static const struct ath10k_hif_ops ath10k_pci_hif_ops = {
 	.map_service_to_pipe	= ath10k_pci_hif_map_service_to_pipe,
 	.get_default_pipe	= ath10k_pci_hif_get_default_pipe,
 	.send_complete_check	= ath10k_pci_hif_send_complete_check,
-	.set_callbacks		= ath10k_pci_hif_set_callbacks,
 	.get_free_queue_number	= ath10k_pci_hif_get_free_queue_number,
 	.power_up		= ath10k_pci_hif_power_up,
 	.power_down		= ath10k_pci_hif_power_down,
diff --git a/drivers/net/wireless/ath/ath10k/pci.h b/drivers/net/wireless/ath/ath10k/pci.h
index 8d364fb..9c91640 100644
--- a/drivers/net/wireless/ath/ath10k/pci.h
+++ b/drivers/net/wireless/ath/ath10k/pci.h
@@ -175,8 +175,6 @@ struct ath10k_pci {
 
 	struct ath10k_pci_pipe pipe_info[CE_COUNT_MAX];
 
-	struct ath10k_hif_cb msg_callbacks_current;
-
 	/* Copy Engine used for Diagnostic Accesses */
 	struct ath10k_ce_pipe *ce_diag;
 
-- 
2.5.2

[PATCH 4/7] ath10k: export htt tx rx handlers

[Rajkumar Manoharan]

Some special copy engines delivers messages directly to HTT by
bypassing HTC layer. Hence exporting tx_completion and rx_handler
for delivering the data to HTT layer.

Reviewed-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
---
 drivers/net/wireless/ath/ath10k/htt.h    | 1 +
 drivers/net/wireless/ath/ath10k/htt_rx.c | 1 +
 drivers/net/wireless/ath/ath10k/htt_tx.c | 6 ++++++
 3 files changed, 8 insertions(+)

diff --git a/drivers/net/wireless/ath/ath10k/htt.h b/drivers/net/wireless/ath/ath10k/htt.h
index 5a8e4ea..3b9ddef 100644
--- a/drivers/net/wireless/ath/ath10k/htt.h
+++ b/drivers/net/wireless/ath/ath10k/htt.h
@@ -1587,6 +1587,7 @@ int ath10k_htt_send_rx_ring_cfg_ll(struct ath10k_htt *htt);
 int ath10k_htt_h2t_aggr_cfg_msg(struct ath10k_htt *htt,
 				u8 max_subfrms_ampdu,
 				u8 max_subfrms_amsdu);
+void ath10k_htt_hif_tx_complete(struct ath10k *ar, struct sk_buff *skb);
 
 void __ath10k_htt_tx_dec_pending(struct ath10k_htt *htt, bool limit_mgmt_desc);
 int ath10k_htt_tx_alloc_msdu_id(struct ath10k_htt *htt, struct sk_buff *skb);
diff --git a/drivers/net/wireless/ath/ath10k/htt_rx.c b/drivers/net/wireless/ath/ath10k/htt_rx.c
index 606c1a3..6060dda 100644
--- a/drivers/net/wireless/ath/ath10k/htt_rx.c
+++ b/drivers/net/wireless/ath/ath10k/htt_rx.c
@@ -2125,6 +2125,7 @@ void ath10k_htt_t2h_msg_handler(struct ath10k *ar, struct sk_buff *skb)
 	/* Free the indication buffer */
 	dev_kfree_skb_any(skb);
 }
+EXPORT_SYMBOL(ath10k_htt_t2h_msg_handler);
 
 static void ath10k_htt_txrx_compl_task(unsigned long ptr)
 {
diff --git a/drivers/net/wireless/ath/ath10k/htt_tx.c b/drivers/net/wireless/ath/ath10k/htt_tx.c
index eb5ba9b..64ba977 100644
--- a/drivers/net/wireless/ath/ath10k/htt_tx.c
+++ b/drivers/net/wireless/ath/ath10k/htt_tx.c
@@ -175,6 +175,12 @@ void ath10k_htt_htc_tx_complete(struct ath10k *ar, struct sk_buff *skb)
 	dev_kfree_skb_any(skb);
 }
 
+void ath10k_htt_hif_tx_complete(struct ath10k *ar, struct sk_buff *skb)
+{
+	dev_kfree_skb_any(skb);
+}
+EXPORT_SYMBOL(ath10k_htt_hif_tx_complete);
+
 int ath10k_htt_h2t_ver_req_msg(struct ath10k_htt *htt)
 {
 	struct ath10k *ar = htt->ar;
-- 
2.5.2

[PATCH 5/7] ath10k: Configure copy engine 5 for HTT messages

[Rajkumar Manoharan]

Currently target to host (T2H) HTT messages are received at copy engine 1.
These messages are processed by HTC layer in both host and target.
To avoid HTC level processing overhead in both host and target,
the unused copy engine 5 is being used for receiving HTT T2H messages.
This will speedup the receive data processing as well as htt tx completion.
Hence host and target copy engine configuration tables are updated
to enable CE5 pipe. The in-direction HTT mapping is now pointing to CE5
for all HTT T2H.

Moreover HTT send completion messages are polled from HTC handler
as CE 4 is not interrupt-driven. For faster tx completion, CE4 polling
needs to be done whenever CE pipe which transports HTT Rx (target->host)
is processed. This avoids overhead of polling HTT messages from HTC
layer. Servicing CE 4 faster is helping to solve "failed to transmit
packet, dropping: -105".

Reviewed-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
---
 drivers/net/wireless/ath/ath10k/pci.c | 75 ++++++++++++++++++++++++++++++-----
 1 file changed, 64 insertions(+), 11 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/pci.c b/drivers/net/wireless/ath/ath10k/pci.c
index 617728d..bd8ef0c 100644
--- a/drivers/net/wireless/ath/ath10k/pci.c
+++ b/drivers/net/wireless/ath/ath10k/pci.c
@@ -106,6 +106,8 @@ static int ath10k_pci_bmi_wait(struct ath10k_ce_pipe *tx_pipe,
 static int ath10k_pci_qca99x0_chip_reset(struct ath10k *ar);
 static void ath10k_pci_htc_tx_cb(struct ath10k_ce_pipe *ce_state);
 static void ath10k_pci_htc_rx_cb(struct ath10k_ce_pipe *ce_state);
+static void ath10k_pci_htt_tx_cb(struct ath10k_ce_pipe *ce_state);
+static void ath10k_pci_htt_rx_cb(struct ath10k_ce_pipe *ce_state);
 
 static const struct ce_attr host_ce_config_wlan[] = {
 	/* CE0: host->target HTC control and raw streams */
@@ -154,16 +156,18 @@ static const struct ce_attr host_ce_config_wlan[] = {
 		.src_nentries = CE_HTT_H2T_MSG_SRC_NENTRIES,
 		.src_sz_max = 256,
 		.dest_nentries = 0,
-		.send_cb = ath10k_pci_htc_tx_cb,
+		.send_cb = ath10k_pci_htt_tx_cb,
 		.recv_cb = ath10k_pci_htc_rx_cb,
 	},
 
-	/* CE5: unused */
+	/* CE5: target->host HTT (HIF->HTT) */
 	{
 		.flags = CE_ATTR_FLAGS,
 		.src_nentries = 0,
-		.src_sz_max = 0,
-		.dest_nentries = 0,
+		.src_sz_max = 512,
+		.dest_nentries = 512,
+		.send_cb = ath10k_pci_htt_tx_cb,
+		.recv_cb = ath10k_pci_htt_rx_cb,
 	},
 
 	/* CE6: target autonomous hif_memcpy */
@@ -269,12 +273,12 @@ static const struct ce_pipe_config target_ce_config_wlan[] = {
 
 	/* NB: 50% of src nentries, since tx has 2 frags */
 
-	/* CE5: unused */
+	/* CE5: target->host HTT (HIF->HTT) */
 	{
 		.pipenum = __cpu_to_le32(5),
-		.pipedir = __cpu_to_le32(PIPEDIR_OUT),
+		.pipedir = __cpu_to_le32(PIPEDIR_IN),
 		.nentries = __cpu_to_le32(32),
-		.nbytes_max = __cpu_to_le32(2048),
+		.nbytes_max = __cpu_to_le32(512),
 		.flags = __cpu_to_le32(CE_ATTR_FLAGS),
 		.reserved = __cpu_to_le32(0),
 	},
@@ -408,7 +412,7 @@ static const struct service_to_pipe target_service_to_ce_map_wlan[] = {
 	{
 		__cpu_to_le32(ATH10K_HTC_SVC_ID_HTT_DATA_MSG),
 		__cpu_to_le32(PIPEDIR_IN),	/* in = DL = target -> host */
-		__cpu_to_le32(1),
+		__cpu_to_le32(5),
 	},
 
 	/* (Additions here) */
@@ -1137,8 +1141,9 @@ static void ath10k_pci_htc_tx_cb(struct ath10k_ce_pipe *ce_state)
 		ath10k_htc_tx_completion_handler(ar, skb);
 }
 
-/* Called by lower (CE) layer when data is received from the Target. */
-static void ath10k_pci_htc_rx_cb(struct ath10k_ce_pipe *ce_state)
+static void ath10k_pci_process_rx_cb(struct ath10k_ce_pipe *ce_state,
+				     void (*callback)(struct ath10k *ar,
+						      struct sk_buff *skb))
 {
 	struct ath10k *ar = ce_state->ar;
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
@@ -1177,12 +1182,60 @@ static void ath10k_pci_htc_rx_cb(struct ath10k_ce_pipe *ce_state)
 		ath10k_dbg_dump(ar, ATH10K_DBG_PCI_DUMP, NULL, "pci rx: ",
 				skb->data, skb->len);
 
-		ath10k_htc_rx_completion_handler(ar, skb);
+		callback(ar, skb);
 	}
 
 	ath10k_pci_rx_post_pipe(pipe_info);
 }
 
+/* Called by lower (CE) layer when data is received from the Target. */
+static void ath10k_pci_htc_rx_cb(struct ath10k_ce_pipe *ce_state)
+{
+	ath10k_pci_process_rx_cb(ce_state, ath10k_htc_rx_completion_handler);
+}
+
+/* Called by lower (CE) layer when a send to HTT Target completes. */
+static void ath10k_pci_htt_tx_cb(struct ath10k_ce_pipe *ce_state)
+{
+	struct ath10k *ar = ce_state->ar;
+	struct sk_buff *skb;
+	u32 ce_data;
+	unsigned int nbytes;
+	unsigned int transfer_id;
+
+	while (ath10k_ce_completed_send_next(ce_state, (void **)&skb, &ce_data,
+					     &nbytes, &transfer_id) == 0) {
+		/* no need to call tx completion for NULL pointers */
+		if (!skb)
+			continue;
+
+		dma_unmap_single(ar->dev, ATH10K_SKB_RXCB(skb)->paddr,
+				 skb->len, DMA_TO_DEVICE);
+		ath10k_htt_hif_tx_complete(ar, skb);
+	}
+}
+
+static void ath10k_pci_htt_rx_deliver(struct ath10k *ar, struct sk_buff *skb)
+{
+	skb_pull(skb, sizeof(struct ath10k_htc_hdr));
+	ath10k_htt_t2h_msg_handler(ar, skb);
+}
+
+/* Called by lower (CE) layer when HTT data is received from the Target. */
+static void ath10k_pci_htt_rx_cb(struct ath10k_ce_pipe *ce_state)
+{
+	struct ath10k *ar = ce_state->ar;
+	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
+
+	ath10k_pci_process_rx_cb(ce_state, ath10k_pci_htt_rx_deliver);
+
+	/* CE4 polling needs to be done whenever CE pipe which transports
+	 * HTT Rx (target->host) is processed.
+	 */
+	ce_state = &ar_pci->ce_states[4];
+	ath10k_pci_htt_tx_cb(ce_state);
+}
+
 static int ath10k_pci_hif_tx_sg(struct ath10k *ar, u8 pipe_id,
 				struct ath10k_hif_sg_item *items, int n_items)
 {
-- 
2.5.2

[PATCH 6/7] ath10k: remove unused dl_is_polled

[Rajkumar Manoharan]

Since polling for received messages not supported, remove unused
dl_is_polled.

Reviewed-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
---
 drivers/net/wireless/ath/ath10k/hif.h |  7 +++----
 drivers/net/wireless/ath/ath10k/htc.c |  7 +++----
 drivers/net/wireless/ath/ath10k/htc.h |  1 -
 drivers/net/wireless/ath/ath10k/pci.c | 11 +++--------
 4 files changed, 9 insertions(+), 17 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/hif.h b/drivers/net/wireless/ath/ath10k/hif.h
index 6e826bb..633594c 100644
--- a/drivers/net/wireless/ath/ath10k/hif.h
+++ b/drivers/net/wireless/ath/ath10k/hif.h
@@ -59,7 +59,7 @@ struct ath10k_hif_ops {
 
 	int (*map_service_to_pipe)(struct ath10k *ar, u16 service_id,
 				   u8 *ul_pipe, u8 *dl_pipe,
-				   int *ul_is_polled, int *dl_is_polled);
+				   int *ul_is_polled);
 
 	void (*get_default_pipe)(struct ath10k *ar, u8 *ul_pipe, u8 *dl_pipe);
 
@@ -133,12 +133,11 @@ static inline void ath10k_hif_stop(struct ath10k *ar)
 static inline int ath10k_hif_map_service_to_pipe(struct ath10k *ar,
 						 u16 service_id,
 						 u8 *ul_pipe, u8 *dl_pipe,
-						 int *ul_is_polled,
-						 int *dl_is_polled)
+						 int *ul_is_polled)
 {
 	return ar->hif.ops->map_service_to_pipe(ar, service_id,
 						ul_pipe, dl_pipe,
-						ul_is_polled, dl_is_polled);
+						ul_is_polled);
 }
 
 static inline void ath10k_hif_get_default_pipe(struct ath10k *ar,
diff --git a/drivers/net/wireless/ath/ath10k/htc.c b/drivers/net/wireless/ath/ath10k/htc.c
index 13d0119..20e0c48 100644
--- a/drivers/net/wireless/ath/ath10k/htc.c
+++ b/drivers/net/wireless/ath/ath10k/htc.c
@@ -759,8 +759,7 @@ setup:
 						ep->service_id,
 						&ep->ul_pipe_id,
 						&ep->dl_pipe_id,
-						&ep->ul_is_polled,
-						&ep->dl_is_polled);
+						&ep->ul_is_polled);
 	if (status)
 		return status;
 
@@ -770,8 +769,8 @@ setup:
 		   ep->dl_pipe_id, ep->eid);
 
 	ath10k_dbg(ar, ATH10K_DBG_BOOT,
-		   "boot htc ep %d ul polled %d dl polled %d\n",
-		   ep->eid, ep->ul_is_polled, ep->dl_is_polled);
+		   "boot htc ep %d ul polled %d\n",
+		   ep->eid, ep->ul_is_polled);
 
 	if (disable_credit_flow_ctrl && ep->tx_credit_flow_enabled) {
 		ep->tx_credit_flow_enabled = false;
diff --git a/drivers/net/wireless/ath/ath10k/htc.h b/drivers/net/wireless/ath/ath10k/htc.h
index aed3708..2ddd41e 100644
--- a/drivers/net/wireless/ath/ath10k/htc.h
+++ b/drivers/net/wireless/ath/ath10k/htc.h
@@ -313,7 +313,6 @@ struct ath10k_htc_ep {
 	u8 ul_pipe_id;
 	u8 dl_pipe_id;
 	int ul_is_polled; /* call HIF to get tx completions */
-	int dl_is_polled; /* call HIF to fetch rx (not implemented) */
 
 	u8 seq_no; /* for debugging */
 	int tx_credits;
diff --git a/drivers/net/wireless/ath/ath10k/pci.c b/drivers/net/wireless/ath/ath10k/pci.c
index bd8ef0c..1f717a8 100644
--- a/drivers/net/wireless/ath/ath10k/pci.c
+++ b/drivers/net/wireless/ath/ath10k/pci.c
@@ -1421,8 +1421,7 @@ static void ath10k_pci_kill_tasklet(struct ath10k *ar)
 
 static int ath10k_pci_hif_map_service_to_pipe(struct ath10k *ar,
 					      u16 service_id, u8 *ul_pipe,
-					      u8 *dl_pipe, int *ul_is_polled,
-					      int *dl_is_polled)
+					      u8 *dl_pipe, int *ul_is_polled)
 {
 	const struct service_to_pipe *entry;
 	bool ul_set = false, dl_set = false;
@@ -1430,9 +1429,6 @@ static int ath10k_pci_hif_map_service_to_pipe(struct ath10k *ar,
 
 	ath10k_dbg(ar, ATH10K_DBG_PCI, "pci hif map service\n");
 
-	/* polling for received messages not supported */
-	*dl_is_polled = 0;
-
 	for (i = 0; i < ARRAY_SIZE(target_service_to_ce_map_wlan); i++) {
 		entry = &target_service_to_ce_map_wlan[i];
 
@@ -1475,7 +1471,7 @@ static int ath10k_pci_hif_map_service_to_pipe(struct ath10k *ar,
 static void ath10k_pci_hif_get_default_pipe(struct ath10k *ar,
 					    u8 *ul_pipe, u8 *dl_pipe)
 {
-	int ul_is_polled, dl_is_polled;
+	int ul_is_polled;
 
 	ath10k_dbg(ar, ATH10K_DBG_PCI, "pci hif get default pipe\n");
 
@@ -1483,8 +1479,7 @@ static void ath10k_pci_hif_get_default_pipe(struct ath10k *ar,
 						 ATH10K_HTC_SVC_ID_RSVD_CTRL,
 						 ul_pipe,
 						 dl_pipe,
-						 &ul_is_polled,
-						 &dl_is_polled);
+						 &ul_is_polled);
 }
 
 static void ath10k_pci_irq_msi_fw_mask(struct ath10k *ar)
-- 
2.5.2

[PATCH 7/7] ath10k: remove htc polling for tx completion

[Rajkumar Manoharan]

Since polling for tx completion is handled whenever target to host
messages are received, removing the unnecessary polling mechanism for
send completion at HTC level.

Reviewed-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
---
 drivers/net/wireless/ath/ath10k/hif.h |  9 +++------
 drivers/net/wireless/ath/ath10k/htc.c | 26 +-------------------------
 drivers/net/wireless/ath/ath10k/htc.h |  1 -
 drivers/net/wireless/ath/ath10k/pci.c | 14 +++-----------
 4 files changed, 7 insertions(+), 43 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/hif.h b/drivers/net/wireless/ath/ath10k/hif.h
index 633594c..89e7076 100644
--- a/drivers/net/wireless/ath/ath10k/hif.h
+++ b/drivers/net/wireless/ath/ath10k/hif.h
@@ -58,8 +58,7 @@ struct ath10k_hif_ops {
 	void (*stop)(struct ath10k *ar);
 
 	int (*map_service_to_pipe)(struct ath10k *ar, u16 service_id,
-				   u8 *ul_pipe, u8 *dl_pipe,
-				   int *ul_is_polled);
+				   u8 *ul_pipe, u8 *dl_pipe);
 
 	void (*get_default_pipe)(struct ath10k *ar, u8 *ul_pipe, u8 *dl_pipe);
 
@@ -132,12 +131,10 @@ static inline void ath10k_hif_stop(struct ath10k *ar)
 
 static inline int ath10k_hif_map_service_to_pipe(struct ath10k *ar,
 						 u16 service_id,
-						 u8 *ul_pipe, u8 *dl_pipe,
-						 int *ul_is_polled)
+						 u8 *ul_pipe, u8 *dl_pipe)
 {
 	return ar->hif.ops->map_service_to_pipe(ar, service_id,
-						ul_pipe, dl_pipe,
-						ul_is_polled);
+						ul_pipe, dl_pipe);
 }
 
 static inline void ath10k_hif_get_default_pipe(struct ath10k *ar,
diff --git a/drivers/net/wireless/ath/ath10k/htc.c b/drivers/net/wireless/ath/ath10k/htc.c
index 20e0c48..5b3c6bc 100644
--- a/drivers/net/wireless/ath/ath10k/htc.c
+++ b/drivers/net/wireless/ath/ath10k/htc.c
@@ -23,16 +23,6 @@
 /* Send */
 /********/
 
-static inline void ath10k_htc_send_complete_check(struct ath10k_htc_ep *ep,
-						  int force)
-{
-	/*
-	 * Check whether HIF has any prior sends that have finished,
-	 * have not had the post-processing done.
-	 */
-	ath10k_hif_send_complete_check(ep->htc->ar, ep->ul_pipe_id, force);
-}
-
 static void ath10k_htc_control_tx_complete(struct ath10k *ar,
 					   struct sk_buff *skb)
 {
@@ -328,15 +318,6 @@ void ath10k_htc_rx_completion_handler(struct ath10k *ar, struct sk_buff *skb)
 
 	ep = &htc->endpoint[eid];
 
-	/*
-	 * If this endpoint that received a message from the target has
-	 * a to-target HIF pipe whose send completions are polled rather
-	 * than interrupt-driven, this is a good point to ask HIF to check
-	 * whether it has any completed sends to handle.
-	 */
-	if (ep->ul_is_polled)
-		ath10k_htc_send_complete_check(ep, 1);
-
 	payload_len = __le16_to_cpu(hdr->len);
 
 	if (payload_len + sizeof(*hdr) > ATH10K_HTC_MAX_LEN) {
@@ -758,8 +739,7 @@ setup:
 	status = ath10k_hif_map_service_to_pipe(htc->ar,
 						ep->service_id,
 						&ep->ul_pipe_id,
-						&ep->dl_pipe_id,
-						&ep->ul_is_polled);
+						&ep->dl_pipe_id);
 	if (status)
 		return status;
 
@@ -768,10 +748,6 @@ setup:
 		   htc_service_name(ep->service_id), ep->ul_pipe_id,
 		   ep->dl_pipe_id, ep->eid);
 
-	ath10k_dbg(ar, ATH10K_DBG_BOOT,
-		   "boot htc ep %d ul polled %d\n",
-		   ep->eid, ep->ul_is_polled);
-
 	if (disable_credit_flow_ctrl && ep->tx_credit_flow_enabled) {
 		ep->tx_credit_flow_enabled = false;
 		ath10k_dbg(ar, ATH10K_DBG_BOOT,
diff --git a/drivers/net/wireless/ath/ath10k/htc.h b/drivers/net/wireless/ath/ath10k/htc.h
index 2ddd41e..e70aa38 100644
--- a/drivers/net/wireless/ath/ath10k/htc.h
+++ b/drivers/net/wireless/ath/ath10k/htc.h
@@ -312,7 +312,6 @@ struct ath10k_htc_ep {
 	int max_ep_message_len;
 	u8 ul_pipe_id;
 	u8 dl_pipe_id;
-	int ul_is_polled; /* call HIF to get tx completions */
 
 	u8 seq_no; /* for debugging */
 	int tx_credits;
diff --git a/drivers/net/wireless/ath/ath10k/pci.c b/drivers/net/wireless/ath/ath10k/pci.c
index 1f717a8..e03d31e 100644
--- a/drivers/net/wireless/ath/ath10k/pci.c
+++ b/drivers/net/wireless/ath/ath10k/pci.c
@@ -1419,9 +1419,8 @@ static void ath10k_pci_kill_tasklet(struct ath10k *ar)
 	del_timer_sync(&ar_pci->rx_post_retry);
 }
 
-static int ath10k_pci_hif_map_service_to_pipe(struct ath10k *ar,
-					      u16 service_id, u8 *ul_pipe,
-					      u8 *dl_pipe, int *ul_is_polled)
+static int ath10k_pci_hif_map_service_to_pipe(struct ath10k *ar, u16 service_id,
+					      u8 *ul_pipe, u8 *dl_pipe)
 {
 	const struct service_to_pipe *entry;
 	bool ul_set = false, dl_set = false;
@@ -1462,24 +1461,17 @@ static int ath10k_pci_hif_map_service_to_pipe(struct ath10k *ar,
 	if (WARN_ON(!ul_set || !dl_set))
 		return -ENOENT;
 
-	*ul_is_polled =
-		(host_ce_config_wlan[*ul_pipe].flags & CE_ATTR_DIS_INTR) != 0;
-
 	return 0;
 }
 
 static void ath10k_pci_hif_get_default_pipe(struct ath10k *ar,
 					    u8 *ul_pipe, u8 *dl_pipe)
 {
-	int ul_is_polled;
-
 	ath10k_dbg(ar, ATH10K_DBG_PCI, "pci hif get default pipe\n");
 
 	(void)ath10k_pci_hif_map_service_to_pipe(ar,
 						 ATH10K_HTC_SVC_ID_RSVD_CTRL,
-						 ul_pipe,
-						 dl_pipe,
-						 &ul_is_polled);
+						 ul_pipe, dl_pipe);
 }
 
 static void ath10k_pci_irq_msi_fw_mask(struct ath10k *ar)
-- 
2.5.2

Re: [PATCH 0/7] ath10k: add copy engine fast path support

[Kalle Valo]

Rajkumar Manoharan <rmanohar@qti.qualcomm.com> writes:

> Below patchset adds fast path support for uplink traffic by bypassing
> HTC layer processing. This is enabled by making use of unused copy
> engine 5 to receive HTT messages directly from HIF layer. From initial
> validation in VHT80/5G mode TCP UL is improved to 900Mbps from ~840Mbps
> in conducted test.
>
> -Rajkumar
>
> Rajkumar Manoharan (7):
>   ath10k: export htc tx rx handlers
>   ath10k: register per copy engine send completion callbacks
>   ath10k: register per copy engine receive callbacks
>   ath10k: export htt tx rx handlers
>   ath10k: Configure copy engine 5 for HTT messages
>   ath10k: remove unused dl_is_polled
>   ath10k: remove htc polling for tx completion

With these patches I'm seeing a DMA warning every time with qca988x
during firmware boot (tested three out of three times, rebooted after
every test):

[   60.458260] ------------[ cut here ]------------
[   60.458374] WARNING: CPU: 1 PID: 0 at lib/dma-debug.c:1090 check_unmap+0x815/0x940()
[   60.458457] ath10k_pci 0000:02:00.0: DMA-API: device driver tries to free an invalid DMA memory address
[   60.458557] Modules linked in: ath10k_pci ath10k_core ath mac80211 cfg80211 [last unloaded: cfg80211]
[   60.458941] CPU: 1 PID: 0 Comm: swapper/1 Not tainted 4.3.0-rc3-wl-ath+ #1082
[   60.458981] Hardware name: Hewlett-Packard HP ProBook 6540b/1722, BIOS 68CDD Ver. F.04 01/27/2010
[   60.459021]  00000000 00000000 f4ef3de4 c133dd94 00000000 f4ef3e14 c10562de c1b3e12c
[   60.459378]  f4ef3e40 00000000 c1b3d3b9 00000442 c13693c5 c13693c5 f435f840 f4ef3eb4
[   60.461026]  00000000 f4ef3e2c c10563c3 00000009 f4ef3e24 c1b3e12c f4ef3e40 f4ef3ea4
[   60.461528] Call Trace:
[   60.461592]  [<c133dd94>] dump_stack+0x48/0x64
[   60.461627] ath10k_pci 0000:02:00.0: qca988x hw2.0 (0x4100016c, 0x043202ff) fw 10.2.4.70.9-2 api 3 htt-ver 2.1 wmi-op 5 htt-op 2 cal otp max-sta 128 raw 0 hwcrypto 1 features no-p2p,raw-mode
[   60.461643] ath10k_pci 0000:02:00.0: debug 1 debugfs 1 tracing 1 dfs 1 testmode 1
[   60.461718]  [<c10562de>] warn_slowpath_common+0x8e/0xd0
[   60.461750]  [<c13693c5>] ? check_unmap+0x815/0x940
[   60.461781]  [<c13693c5>] ? check_unmap+0x815/0x940
[   60.461814]  [<c10563c3>] warn_slowpath_fmt+0x33/0x40
[   60.461844]  [<c13693c5>] check_unmap+0x815/0x940
[   60.461881]  [<c18c0c06>] ? _raw_spin_unlock_irqrestore+0x36/0x60
[   60.461915]  [<c10aac14>] ? trace_hardirqs_on_caller+0xa4/0x1c0
[   60.461954]  [<fa1cf15d>] ? ath10k_dbg+0x6d/0x140 [ath10k_core]
[   60.461989]  [<c1088e35>] ? local_clock+0x25/0x30
[   60.462020]  [<c136974c>] debug_dma_unmap_page+0x8c/0xa0
[   60.462053]  [<fa24dc2f>] ath10k_pci_htt_tx_cb+0x8f/0xb0 [ath10k_pci]
[   60.462085]  [<fa24dc70>] ath10k_pci_htt_rx_cb+0x20/0x30 [ath10k_pci]
[   60.462117]  [<fa2517fc>] ath10k_ce_per_engine_service+0x5c/0xa0 [ath10k_pci]
[   60.462162]  [<fa2518b7>] ath10k_ce_per_engine_service_any+0x77/0x90 [ath10k_pci]
[   60.462194]  [<fa24f85b>] ath10k_pci_tasklet+0x1b/0x50 [ath10k_pci]
[   60.462226]  [<c105bc4e>] tasklet_action+0x9e/0xb0
[   60.462257]  [<c105b17b>] __do_softirq+0xbb/0x3c0
[   60.462288]  [<c105b0c0>] ? trace_event_raw_event_irq_handler_entry+0xa0/0xa0
[   60.462321]  [<c10059b9>] do_softirq_own_stack+0x29/0x40
[   60.462353]  <IRQ>  [<c105b716>] irq_exit+0x86/0xb0
[   60.462427]  [<c18c2310>] do_IRQ+0x60/0x120
[   60.462493]  [<c10a4deb>] ? trace_hardirqs_off+0xb/0x10
[   60.462575]  [<c18c1931>] common_interrupt+0x31/0x38
[   60.462617]  [<c1694fa9>] ? cpuidle_enter_state+0xc9/0x350
[   60.462648]  [<c1696879>] ? menu_select+0x239/0x4b0
[   60.462682]  [<c1695264>] cpuidle_enter+0x14/0x20
[   60.462714]  [<c109becc>] call_cpuidle+0x3c/0x70
[   60.462757]  [<c109c0a9>] cpu_startup_entry+0x1a9/0x390
[   60.462789]  [<c1038b75>] start_secondary+0x105/0x150
[   60.462820] ---[ end trace 4bfd4799ecf8f81b ]---

-- 
Kalle Valo

Re: [PATCH 0/7] ath10k: add copy engine fast path support

[Manoharan, Rajkumar]

> > Below patchset adds fast path support for uplink traffic by bypassing
> > HTC layer processing. This is enabled by making use of unused copy
> > engine 5 to receive HTT messages directly from HIF layer. From initial
> > validation in VHT80/5G mode TCP UL is improved to 900Mbps from ~840Mbps
> > in conducted test.
> >
> > -Rajkumar
> >
> > Rajkumar Manoharan (7):
> >   ath10k: export htc tx rx handlers
> >   ath10k: register per copy engine send completion callbacks
> >   ath10k: register per copy engine receive callbacks
> >   ath10k: export htt tx rx handlers
> >   ath10k: Configure copy engine 5 for HTT messages
> >   ath10k: remove unused dl_is_polled
> >   ath10k: remove htc polling for tx completion
> 
> With these patches I'm seeing a DMA warning every time with qca988x
> during firmware boot (tested three out of three times, rebooted after
> every test):
> 
> [   60.458260] ------------[ cut here ]------------
> [   60.458374] WARNING: CPU: 1 PID: 0 at lib/dma-debug.c:1090 check_unmap+0x815/0x940()
> [   60.458457] ath10k_pci 0000:02:00.0: DMA-API: device driver tries to free an invalid DMA memory address
> [   60.458557] Modules linked in: ath10k_pci ath10k_core ath mac80211 cfg80211 [last unloaded: cfg80211]
> [   60.458941] CPU: 1 PID: 0 Comm: swapper/1 Not tainted 4.3.0-rc3-wl-ath+ #1082
> [   60.458981] Hardware name: Hewlett-Packard HP ProBook 6540b/1722, BIOS 68CDD Ver. F.04 01/27/2010

Thanks for reporting this issue. Will post revised patchset asap.

-Rajkumar