Merge branch 'for-linville' of git://github.com/kvalo/ath

drivers/net/wireless/ath/ath10k/ce.c

@@ -266,7 +266,7 @@ static inline void ath10k_ce_engine_int_status_clear(struct ath10k *ar,
  * ath10k_ce_sendlist_send.
  * The caller takes responsibility for any needed locking.
  */
-static int ath10k_ce_send_nolock(struct ath10k_ce_pipe *ce_state,
+int ath10k_ce_send_nolock(struct ath10k_ce_pipe *ce_state,
 			  void *per_transfer_context,
 			  u32 buffer,
 			  unsigned int nbytes,
@@ -1067,9 +1067,9 @@ struct ath10k_ce_pipe *ath10k_ce_init(struct ath10k *ar,
 	 *
 	 * For the lack of a better place do the check here.
 	 */
-	BUILD_BUG_ON(TARGET_NUM_MSDU_DESC >
+	BUILD_BUG_ON(2*TARGET_NUM_MSDU_DESC >
 		     (CE_HTT_H2T_MSG_SRC_NENTRIES - 1));
-	BUILD_BUG_ON(TARGET_10X_NUM_MSDU_DESC >
+	BUILD_BUG_ON(2*TARGET_10X_NUM_MSDU_DESC >
 		     (CE_HTT_H2T_MSG_SRC_NENTRIES - 1));
 
 	ret = ath10k_pci_wake(ar);

drivers/net/wireless/ath/ath10k/ce.h

@@ -23,7 +23,7 @@
 
 /* Maximum number of Copy Engine's supported */
 #define CE_COUNT_MAX 8
-#define CE_HTT_H2T_MSG_SRC_NENTRIES 2048
+#define CE_HTT_H2T_MSG_SRC_NENTRIES 4096
 
 /* Descriptor rings must be aligned to this boundary */
 #define CE_DESC_RING_ALIGN	8
@@ -152,6 +152,13 @@ int ath10k_ce_send(struct ath10k_ce_pipe *ce_state,
 		   unsigned int transfer_id,
 		   unsigned int flags);
 
+int ath10k_ce_send_nolock(struct ath10k_ce_pipe *ce_state,
+			  void *per_transfer_context,
+			  u32 buffer,
+			  unsigned int nbytes,
+			  unsigned int transfer_id,
+			  unsigned int flags);
+
 void ath10k_ce_send_cb_register(struct ath10k_ce_pipe *ce_state,
 				void (*send_cb)(struct ath10k_ce_pipe *),
 				int disable_interrupts);

drivers/net/wireless/ath/ath10k/core.h

@@ -62,16 +62,13 @@ struct ath10k;
 
 struct ath10k_skb_cb {
 	dma_addr_t paddr;
-	bool is_mapped;
-	bool is_aborted;
 	u8 vdev_id;
 
 	struct {
 		u8 tid;
 		bool is_offchan;
-
-		u8 frag_len;
-		u8 pad_len;
+		struct ath10k_htt_txbuf *txbuf;
+		u32 txbuf_paddr;
 	} __packed htt;
 
 	struct {
@@ -87,32 +84,6 @@ static inline struct ath10k_skb_cb *ATH10K_SKB_CB(struct sk_buff *skb)
 	return (struct ath10k_skb_cb *)&IEEE80211_SKB_CB(skb)->driver_data;
 }
 
-static inline int ath10k_skb_map(struct device *dev, struct sk_buff *skb)
-{
-	if (ATH10K_SKB_CB(skb)->is_mapped)
-		return -EINVAL;
-
-	ATH10K_SKB_CB(skb)->paddr = dma_map_single(dev, skb->data, skb->len,
-						   DMA_TO_DEVICE);
-
-	if (unlikely(dma_mapping_error(dev, ATH10K_SKB_CB(skb)->paddr)))
-		return -EIO;
-
-	ATH10K_SKB_CB(skb)->is_mapped = true;
-	return 0;
-}
-
-static inline int ath10k_skb_unmap(struct device *dev, struct sk_buff *skb)
-{
-	if (!ATH10K_SKB_CB(skb)->is_mapped)
-		return -EINVAL;
-
-	dma_unmap_single(dev, ATH10K_SKB_CB(skb)->paddr, skb->len,
-			 DMA_TO_DEVICE);
-	ATH10K_SKB_CB(skb)->is_mapped = false;
-	return 0;
-}
-
 static inline u32 host_interest_item_address(u32 item_offset)
 {
 	return QCA988X_HOST_INTEREST_ADDRESS + item_offset;
@@ -288,6 +259,7 @@ struct ath10k_vif {
 
 	u8 fixed_rate;
 	u8 fixed_nss;
+	u8 force_sgi;
 };
 
 struct ath10k_vif_iter {

drivers/net/wireless/ath/ath10k/hif.h

@@ -21,6 +21,14 @@
 #include <linux/kernel.h>
 #include "core.h"
 
+struct ath10k_hif_sg_item {
+	u16 transfer_id;
+	void *transfer_context; /* NULL = tx completion callback not called */
+	void *vaddr; /* for debugging mostly */
+	u32 paddr;
+	u16 len;
+};
+
 struct ath10k_hif_cb {
 	int (*tx_completion)(struct ath10k *ar,
 			     struct sk_buff *wbuf,
@@ -31,11 +39,9 @@ struct ath10k_hif_cb {
 };
 
 struct ath10k_hif_ops {
-	/* Send the head of a buffer to HIF for transmission to the target. */
-	int (*send_head)(struct ath10k *ar, u8 pipe_id,
-			 unsigned int transfer_id,
-			 unsigned int nbytes,
-			 struct sk_buff *buf);
+	/* send a scatter-gather list to the target */
+	int (*tx_sg)(struct ath10k *ar, u8 pipe_id,
+		     struct ath10k_hif_sg_item *items, int n_items);
 
 	/*
 	 * API to handle HIF-specific BMI message exchanges, this API is
@@ -86,12 +92,11 @@ struct ath10k_hif_ops {
 };
 
 
-static inline int ath10k_hif_send_head(struct ath10k *ar, u8 pipe_id,
-				       unsigned int transfer_id,
-				       unsigned int nbytes,
-				       struct sk_buff *buf)
+static inline int ath10k_hif_tx_sg(struct ath10k *ar, u8 pipe_id,
+				   struct ath10k_hif_sg_item *items,
+				   int n_items)
 {
-	return ar->hif.ops->send_head(ar, pipe_id, transfer_id, nbytes, buf);
+	return ar->hif.ops->tx_sg(ar, pipe_id, items, n_items);
 }
 
 static inline int ath10k_hif_exchange_bmi_msg(struct ath10k *ar,

drivers/net/wireless/ath/ath10k/htc.c

@@ -63,7 +63,9 @@ static struct sk_buff *ath10k_htc_build_tx_ctrl_skb(void *ar)
 static inline void ath10k_htc_restore_tx_skb(struct ath10k_htc *htc,
 					     struct sk_buff *skb)
 {
-	ath10k_skb_unmap(htc->ar->dev, skb);
+	struct ath10k_skb_cb *skb_cb = ATH10K_SKB_CB(skb);
+
+	dma_unmap_single(htc->ar->dev, skb_cb->paddr, skb->len, DMA_TO_DEVICE);
 	skb_pull(skb, sizeof(struct ath10k_htc_hdr));
 }
 
@@ -122,6 +124,9 @@ int ath10k_htc_send(struct ath10k_htc *htc,
 		    struct sk_buff *skb)
 {
 	struct ath10k_htc_ep *ep = &htc->endpoint[eid];
+	struct ath10k_skb_cb *skb_cb = ATH10K_SKB_CB(skb);
+	struct ath10k_hif_sg_item sg_item;
+	struct device *dev = htc->ar->dev;
 	int credits = 0;
 	int ret;
 
@@ -157,19 +162,25 @@ int ath10k_htc_send(struct ath10k_htc *htc,
 
 	ath10k_htc_prepare_tx_skb(ep, skb);
 
-	ret = ath10k_skb_map(htc->ar->dev, skb);
+	skb_cb->paddr = dma_map_single(dev, skb->data, skb->len, DMA_TO_DEVICE);
+	ret = dma_mapping_error(dev, skb_cb->paddr);
 	if (ret)
 		goto err_credits;
 
-	ret = ath10k_hif_send_head(htc->ar, ep->ul_pipe_id, ep->eid,
-				   skb->len, skb);
+	sg_item.transfer_id = ep->eid;
+	sg_item.transfer_context = skb;
+	sg_item.vaddr = skb->data;
+	sg_item.paddr = skb_cb->paddr;
+	sg_item.len = skb->len;
+
+	ret = ath10k_hif_tx_sg(htc->ar, ep->ul_pipe_id, &sg_item, 1);
 	if (ret)
 		goto err_unmap;
 
 	return 0;
 
 err_unmap:
-	ath10k_skb_unmap(htc->ar->dev, skb);
+	dma_unmap_single(dev, skb_cb->paddr, skb->len, DMA_TO_DEVICE);
 err_credits:
 	if (ep->tx_credit_flow_enabled) {
 		spin_lock_bh(&htc->tx_lock);
@@ -191,10 +202,8 @@ static int ath10k_htc_tx_completion_handler(struct ath10k *ar,
 	struct ath10k_htc *htc = &ar->htc;
 	struct ath10k_htc_ep *ep = &htc->endpoint[eid];
 
-	if (!skb) {
-		ath10k_warn("invalid sk_buff completion - NULL pointer. firmware crashed?\n");
+	if (WARN_ON_ONCE(!skb))
 		return 0;
-	}
 
 	ath10k_htc_notify_tx_completion(ep, skb);
 	/* the skb now belongs to the completion handler */

drivers/net/wireless/ath/ath10k/htt.h

@@ -20,6 +20,7 @@
 
 #include <linux/bug.h>
 #include <linux/interrupt.h>
+#include <linux/dmapool.h>
 
 #include "htc.h"
 #include "rx_desc.h"
@@ -1181,11 +1182,20 @@ struct htt_rx_info {
 		u32 info1;
 		u32 info2;
 	} rate;
+
+	u32 tsf;
 	bool fcs_err;
 	bool amsdu_more;
 	bool mic_err;
 };
 
+struct ath10k_htt_txbuf {
+	struct htt_data_tx_desc_frag frags[2];
+	struct ath10k_htc_hdr htc_hdr;
+	struct htt_cmd_hdr cmd_hdr;
+	struct htt_data_tx_desc cmd_tx;
+} __packed;
+
 struct ath10k_htt {
 	struct ath10k *ar;
 	enum ath10k_htc_ep_id eid;
@@ -1267,11 +1277,18 @@ struct ath10k_htt {
 	struct sk_buff **pending_tx;
 	unsigned long *used_msdu_ids; /* bitmap */
 	wait_queue_head_t empty_tx_wq;
+	struct dma_pool *tx_pool;
 
 	/* set if host-fw communication goes haywire
 	 * used to avoid further failures */
 	bool rx_confused;
 	struct tasklet_struct rx_replenish_task;
+
+	/* This is used to group tx/rx completions separately and process them
+	 * in batches to reduce cache stalls */
+	struct tasklet_struct txrx_compl_task;
+	struct sk_buff_head tx_compl_q;
+	struct sk_buff_head rx_compl_q;
 };
 
 #define RX_HTT_HDR_STATUS_LEN 64
@@ -1343,4 +1360,5 @@ int ath10k_htt_tx_alloc_msdu_id(struct ath10k_htt *htt);
 void ath10k_htt_tx_free_msdu_id(struct ath10k_htt *htt, u16 msdu_id);
 int ath10k_htt_mgmt_tx(struct ath10k_htt *htt, struct sk_buff *);
 int ath10k_htt_tx(struct ath10k_htt *htt, struct sk_buff *);
+
 #endif

drivers/net/wireless/ath/ath10k/htt_tx.c

@@ -109,6 +109,14 @@ int ath10k_htt_tx_attach(struct ath10k_htt *htt)
 		return -ENOMEM;
 	}
 
+	htt->tx_pool = dma_pool_create("ath10k htt tx pool", htt->ar->dev,
+				       sizeof(struct ath10k_htt_txbuf), 4, 0);
+	if (!htt->tx_pool) {
+		kfree(htt->used_msdu_ids);
+		kfree(htt->pending_tx);
+		return -ENOMEM;
+	}
+
 	return 0;
 }
 
@@ -117,9 +125,7 @@ static void ath10k_htt_tx_cleanup_pending(struct ath10k_htt *htt)
 	struct htt_tx_done tx_done = {0};
 	int msdu_id;
 
-	/* No locks needed. Called after communication with the device has
-	 * been stopped. */
-
+	spin_lock_bh(&htt->tx_lock);
 	for (msdu_id = 0; msdu_id < htt->max_num_pending_tx; msdu_id++) {
 		if (!test_bit(msdu_id, htt->used_msdu_ids))
 			continue;
@@ -132,6 +138,7 @@ static void ath10k_htt_tx_cleanup_pending(struct ath10k_htt *htt)
 
 		ath10k_txrx_tx_unref(htt, &tx_done);
 	}
+	spin_unlock_bh(&htt->tx_lock);
 }
 
 void ath10k_htt_tx_detach(struct ath10k_htt *htt)
@@ -139,6 +146,7 @@ void ath10k_htt_tx_detach(struct ath10k_htt *htt)
 	ath10k_htt_tx_cleanup_pending(htt);
 	kfree(htt->pending_tx);
 	kfree(htt->used_msdu_ids);
+	dma_pool_destroy(htt->tx_pool);
 	return;
 }
 
@@ -334,7 +342,9 @@ int ath10k_htt_mgmt_tx(struct ath10k_htt *htt, struct sk_buff *msdu)
 		goto err_free_msdu_id;
 	}
 
-	res = ath10k_skb_map(dev, msdu);
+	skb_cb->paddr = dma_map_single(dev, msdu->data, msdu->len,
+				       DMA_TO_DEVICE);
+	res = dma_mapping_error(dev, skb_cb->paddr);
 	if (res)
 		goto err_free_txdesc;
 
@@ -348,8 +358,7 @@ int ath10k_htt_mgmt_tx(struct ath10k_htt *htt, struct sk_buff *msdu)
 	memcpy(cmd->mgmt_tx.hdr, msdu->data,
 	       min_t(int, msdu->len, HTT_MGMT_FRM_HDR_DOWNLOAD_LEN));
 
-	skb_cb->htt.frag_len = 0;
-	skb_cb->htt.pad_len = 0;
+	skb_cb->htt.txbuf = NULL;
 
 	res = ath10k_htc_send(&htt->ar->htc, htt->eid, txdesc);
 	if (res)
@@ -358,7 +367,7 @@ int ath10k_htt_mgmt_tx(struct ath10k_htt *htt, struct sk_buff *msdu)
 	return 0;
 
 err_unmap_msdu:
-	ath10k_skb_unmap(dev, msdu);
+	dma_unmap_single(dev, skb_cb->paddr, msdu->len, DMA_TO_DEVICE);
 err_free_txdesc:
 	dev_kfree_skb_any(txdesc);
 err_free_msdu_id:
@@ -375,19 +384,19 @@ int ath10k_htt_mgmt_tx(struct ath10k_htt *htt, struct sk_buff *msdu)
 int ath10k_htt_tx(struct ath10k_htt *htt, struct sk_buff *msdu)
 {
 	struct device *dev = htt->ar->dev;
-	struct htt_cmd *cmd;
-	struct htt_data_tx_desc_frag *tx_frags;
 	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)msdu->data;
 	struct ath10k_skb_cb *skb_cb = ATH10K_SKB_CB(msdu);
-	struct sk_buff *txdesc = NULL;
-	bool use_frags;
-	u8 vdev_id = ATH10K_SKB_CB(msdu)->vdev_id;
-	u8 tid;
-	int prefetch_len, desc_len;
-	int msdu_id = -1;
+	struct ath10k_hif_sg_item sg_items[2];
+	struct htt_data_tx_desc_frag *frags;
+	u8 vdev_id = skb_cb->vdev_id;
+	u8 tid = skb_cb->htt.tid;
+	int prefetch_len;
 	int res;
-	u8 flags0;
-	u16 flags1;
+	u8 flags0 = 0;
+	u16 msdu_id, flags1 = 0;
+	dma_addr_t paddr;
+	u32 frags_paddr;
+	bool use_frags;
 
 	res = ath10k_htt_tx_inc_pending(htt);
 	if (res)
@@ -406,114 +415,120 @@ int ath10k_htt_tx(struct ath10k_htt *htt, struct sk_buff *msdu)
 	prefetch_len = min(htt->prefetch_len, msdu->len);
 	prefetch_len = roundup(prefetch_len, 4);
 
-	desc_len = sizeof(cmd->hdr) + sizeof(cmd->data_tx) + prefetch_len;
-
-	txdesc = ath10k_htc_alloc_skb(desc_len);
-	if (!txdesc) {
-		res = -ENOMEM;
-		goto err_free_msdu_id;
-	}
-
 	/* Since HTT 3.0 there is no separate mgmt tx command. However in case
 	 * of mgmt tx using TX_FRM there is not tx fragment list. Instead of tx
 	 * fragment list host driver specifies directly frame pointer. */
 	use_frags = htt->target_version_major < 3 ||
 		    !ieee80211_is_mgmt(hdr->frame_control);
 
-	if (!IS_ALIGNED((unsigned long)txdesc->data, 4)) {
-		ath10k_warn("htt alignment check failed. dropping packet.\n");
-		res = -EIO;
-		goto err_free_txdesc;
-	}
-
-	if (use_frags) {
-		skb_cb->htt.frag_len = sizeof(*tx_frags) * 2;
-		skb_cb->htt.pad_len = (unsigned long)msdu->data -
-				      round_down((unsigned long)msdu->data, 4);
-
-		skb_push(msdu, skb_cb->htt.frag_len + skb_cb->htt.pad_len);
-	} else {
-		skb_cb->htt.frag_len = 0;
-		skb_cb->htt.pad_len = 0;
-	}
+	skb_cb->htt.txbuf = dma_pool_alloc(htt->tx_pool, GFP_ATOMIC,
+					   &paddr);
+	if (!skb_cb->htt.txbuf)
+		goto err_free_msdu_id;
+	skb_cb->htt.txbuf_paddr = paddr;
 
-	res = ath10k_skb_map(dev, msdu);
+	skb_cb->paddr = dma_map_single(dev, msdu->data, msdu->len,
+				       DMA_TO_DEVICE);
+	res = dma_mapping_error(dev, skb_cb->paddr);
 	if (res)
-		goto err_pull_txfrag;
+		goto err_free_txbuf;
 
-	if (use_frags) {
-		dma_sync_single_for_cpu(dev, skb_cb->paddr, msdu->len,
-					DMA_TO_DEVICE);
+	if (likely(use_frags)) {
+		frags = skb_cb->htt.txbuf->frags;
 
-		/* tx fragment list must be terminated with zero-entry */
-		tx_frags = (struct htt_data_tx_desc_frag *)msdu->data;
-		tx_frags[0].paddr = __cpu_to_le32(skb_cb->paddr +
-						  skb_cb->htt.frag_len +
-						  skb_cb->htt.pad_len);
-		tx_frags[0].len   = __cpu_to_le32(msdu->len -
-						  skb_cb->htt.frag_len -
-						  skb_cb->htt.pad_len);
-		tx_frags[1].paddr = __cpu_to_le32(0);
-		tx_frags[1].len   = __cpu_to_le32(0);
-
-		dma_sync_single_for_device(dev, skb_cb->paddr, msdu->len,
-					   DMA_TO_DEVICE);
-	}
+		frags[0].paddr = __cpu_to_le32(skb_cb->paddr);
+		frags[0].len = __cpu_to_le32(msdu->len);
+		frags[1].paddr = 0;
+		frags[1].len = 0;
 
-	ath10k_dbg(ATH10K_DBG_HTT, "tx-msdu 0x%llx\n",
-		   (unsigned long long) ATH10K_SKB_CB(msdu)->paddr);
-	ath10k_dbg_dump(ATH10K_DBG_HTT_DUMP, NULL, "tx-msdu: ",
-			msdu->data, msdu->len);
+		flags0 |= SM(ATH10K_HW_TXRX_NATIVE_WIFI,
+			     HTT_DATA_TX_DESC_FLAGS0_PKT_TYPE);
 
-	skb_put(txdesc, desc_len);
-	cmd = (struct htt_cmd *)txdesc->data;
+		frags_paddr = skb_cb->htt.txbuf_paddr;
+	} else {
+		flags0 |= SM(ATH10K_HW_TXRX_MGMT,
+			     HTT_DATA_TX_DESC_FLAGS0_PKT_TYPE);
 
-	tid = ATH10K_SKB_CB(msdu)->htt.tid;
+		frags_paddr = skb_cb->paddr;
+	}
 
-	ath10k_dbg(ATH10K_DBG_HTT, "htt data tx using tid %hhu\n", tid);
+	/* Normally all commands go through HTC which manages tx credits for
+	 * each endpoint and notifies when tx is completed.
+	 *
+	 * HTT endpoint is creditless so there's no need to care about HTC
+	 * flags. In that case it is trivial to fill the HTC header here.
+	 *
+	 * MSDU transmission is considered completed upon HTT event. This
+	 * implies no relevant resources can be freed until after the event is
+	 * received. That's why HTC tx completion handler itself is ignored by
+	 * setting NULL to transfer_context for all sg items.
+	 *
+	 * There is simply no point in pushing HTT TX_FRM through HTC tx path
+	 * as it's a waste of resources. By bypassing HTC it is possible to
+	 * avoid extra memory allocations, compress data structures and thus
+	 * improve performance. */
+
+	skb_cb->htt.txbuf->htc_hdr.eid = htt->eid;
+	skb_cb->htt.txbuf->htc_hdr.len = __cpu_to_le16(
+			sizeof(skb_cb->htt.txbuf->cmd_hdr) +
+			sizeof(skb_cb->htt.txbuf->cmd_tx) +
+			prefetch_len);
+	skb_cb->htt.txbuf->htc_hdr.flags = 0;
 
-	flags0  = 0;
 	if (!ieee80211_has_protected(hdr->frame_control))
 		flags0 |= HTT_DATA_TX_DESC_FLAGS0_NO_ENCRYPT;
-	flags0 |= HTT_DATA_TX_DESC_FLAGS0_MAC_HDR_PRESENT;
 
-	if (use_frags)
-		flags0 |= SM(ATH10K_HW_TXRX_NATIVE_WIFI,
-			     HTT_DATA_TX_DESC_FLAGS0_PKT_TYPE);
-	else
-		flags0 |= SM(ATH10K_HW_TXRX_MGMT,
-			     HTT_DATA_TX_DESC_FLAGS0_PKT_TYPE);
+	flags0 |= HTT_DATA_TX_DESC_FLAGS0_MAC_HDR_PRESENT;
 
-	flags1  = 0;
 	flags1 |= SM((u16)vdev_id, HTT_DATA_TX_DESC_FLAGS1_VDEV_ID);
 	flags1 |= SM((u16)tid, HTT_DATA_TX_DESC_FLAGS1_EXT_TID);
 	flags1 |= HTT_DATA_TX_DESC_FLAGS1_CKSUM_L3_OFFLOAD;
 	flags1 |= HTT_DATA_TX_DESC_FLAGS1_CKSUM_L4_OFFLOAD;
 
-	cmd->hdr.msg_type        = HTT_H2T_MSG_TYPE_TX_FRM;
-	cmd->data_tx.flags0      = flags0;
-	cmd->data_tx.flags1      = __cpu_to_le16(flags1);
-	cmd->data_tx.len         = __cpu_to_le16(msdu->len -
-						 skb_cb->htt.frag_len -
-						 skb_cb->htt.pad_len);
-	cmd->data_tx.id          = __cpu_to_le16(msdu_id);
-	cmd->data_tx.frags_paddr = __cpu_to_le32(skb_cb->paddr);
-	cmd->data_tx.peerid      = __cpu_to_le32(HTT_INVALID_PEERID);
-
-	memcpy(cmd->data_tx.prefetch, hdr, prefetch_len);
+	skb_cb->htt.txbuf->cmd_hdr.msg_type = HTT_H2T_MSG_TYPE_TX_FRM;
+	skb_cb->htt.txbuf->cmd_tx.flags0 = flags0;
+	skb_cb->htt.txbuf->cmd_tx.flags1 = __cpu_to_le16(flags1);
+	skb_cb->htt.txbuf->cmd_tx.len = __cpu_to_le16(msdu->len);
+	skb_cb->htt.txbuf->cmd_tx.id = __cpu_to_le16(msdu_id);
+	skb_cb->htt.txbuf->cmd_tx.frags_paddr = __cpu_to_le32(frags_paddr);
+	skb_cb->htt.txbuf->cmd_tx.peerid = __cpu_to_le32(HTT_INVALID_PEERID);
+
+	ath10k_dbg(ATH10K_DBG_HTT,
+		   "htt tx flags0 %hhu flags1 %hu len %d id %hu frags_paddr %08x, msdu_paddr %08x vdev %hhu tid %hhu\n",
+		   flags0, flags1, msdu->len, msdu_id, frags_paddr,
+		   (u32)skb_cb->paddr, vdev_id, tid);
+	ath10k_dbg_dump(ATH10K_DBG_HTT_DUMP, NULL, "htt tx msdu: ",
+			msdu->data, msdu->len);
 
-	res = ath10k_htc_send(&htt->ar->htc, htt->eid, txdesc);
+	sg_items[0].transfer_id = 0;
+	sg_items[0].transfer_context = NULL;
+	sg_items[0].vaddr = &skb_cb->htt.txbuf->htc_hdr;
+	sg_items[0].paddr = skb_cb->htt.txbuf_paddr +
+			    sizeof(skb_cb->htt.txbuf->frags);
+	sg_items[0].len = sizeof(skb_cb->htt.txbuf->htc_hdr) +
+			  sizeof(skb_cb->htt.txbuf->cmd_hdr) +
+			  sizeof(skb_cb->htt.txbuf->cmd_tx);
+
+	sg_items[1].transfer_id = 0;
+	sg_items[1].transfer_context = NULL;
+	sg_items[1].vaddr = msdu->data;
+	sg_items[1].paddr = skb_cb->paddr;
+	sg_items[1].len = prefetch_len;
+
+	res = ath10k_hif_tx_sg(htt->ar,
+			       htt->ar->htc.endpoint[htt->eid].ul_pipe_id,
+			       sg_items, ARRAY_SIZE(sg_items));
 	if (res)
 		goto err_unmap_msdu;
 
 	return 0;
 
 err_unmap_msdu:
-	ath10k_skb_unmap(dev, msdu);
-err_pull_txfrag:
-	skb_pull(msdu, skb_cb->htt.frag_len + skb_cb->htt.pad_len);
-err_free_txdesc:
-	dev_kfree_skb_any(txdesc);
+	dma_unmap_single(dev, skb_cb->paddr, msdu->len, DMA_TO_DEVICE);
+err_free_txbuf:
+	dma_pool_free(htt->tx_pool,
+		      skb_cb->htt.txbuf,
+		      skb_cb->htt.txbuf_paddr);
 err_free_msdu_id:
 	spin_lock_bh(&htt->tx_lock);
 	htt->pending_tx[msdu_id] = NULL;

drivers/net/wireless/ath/ath10k/pci.h

@@ -43,23 +43,6 @@ struct bmi_xfer {
 	u32 resp_len;
 };
 
-enum ath10k_pci_compl_state {
-	ATH10K_PCI_COMPL_FREE = 0,
-	ATH10K_PCI_COMPL_SEND,
-	ATH10K_PCI_COMPL_RECV,
-};
-
-struct ath10k_pci_compl {
-	struct list_head list;
-	enum ath10k_pci_compl_state state;
-	struct ath10k_ce_pipe *ce_state;
-	struct ath10k_pci_pipe *pipe_info;
-	struct sk_buff *skb;
-	unsigned int nbytes;
-	unsigned int transfer_id;
-	unsigned int flags;
-};
-
 /*
  * PCI-specific Target state
  *
@@ -175,9 +158,6 @@ struct ath10k_pci_pipe {
 	/* protects compl_free and num_send_allowed */
 	spinlock_t pipe_lock;
 
-	/* List of free CE completion slots */
-	struct list_head compl_free;
-
 	struct ath10k_pci *ar_pci;
 	struct tasklet_struct intr;
 };
@@ -205,14 +185,6 @@ struct ath10k_pci {
 	atomic_t keep_awake_count;
 	bool verified_awake;
 
-	/* List of CE completions to be processed */
-	struct list_head compl_process;
-
-	/* protects compl_processing and compl_process */
-	spinlock_t compl_lock;
-
-	bool compl_processing;
-
 	struct ath10k_pci_pipe pipe_info[CE_COUNT_MAX];
 
 	struct ath10k_hif_cb msg_callbacks_current;

drivers/net/wireless/ath/ath10k/txrx.c

@@ -51,7 +51,8 @@ void ath10k_txrx_tx_unref(struct ath10k_htt *htt,
 	struct ieee80211_tx_info *info;
 	struct ath10k_skb_cb *skb_cb;
 	struct sk_buff *msdu;
-	int ret;
+
+	lockdep_assert_held(&htt->tx_lock);
 
 	ath10k_dbg(ATH10K_DBG_HTT, "htt tx completion msdu_id %u discard %d no_ack %d\n",
 		   tx_done->msdu_id, !!tx_done->discard, !!tx_done->no_ack);
@@ -65,12 +66,12 @@ void ath10k_txrx_tx_unref(struct ath10k_htt *htt,
 	msdu = htt->pending_tx[tx_done->msdu_id];
 	skb_cb = ATH10K_SKB_CB(msdu);
 
-	ret = ath10k_skb_unmap(dev, msdu);
-	if (ret)
-		ath10k_warn("data skb unmap failed (%d)\n", ret);
+	dma_unmap_single(dev, skb_cb->paddr, msdu->len, DMA_TO_DEVICE);
 
-	if (skb_cb->htt.frag_len)
-		skb_pull(msdu, skb_cb->htt.frag_len + skb_cb->htt.pad_len);
+	if (skb_cb->htt.txbuf)
+		dma_pool_free(htt->tx_pool,
+			      skb_cb->htt.txbuf,
+			      skb_cb->htt.txbuf_paddr);
 
 	ath10k_report_offchan_tx(htt->ar, msdu);
 
@@ -92,13 +93,11 @@ void ath10k_txrx_tx_unref(struct ath10k_htt *htt,
 	/* we do not own the msdu anymore */
 
 exit:
-	spin_lock_bh(&htt->tx_lock);
 	htt->pending_tx[tx_done->msdu_id] = NULL;
 	ath10k_htt_tx_free_msdu_id(htt, tx_done->msdu_id);
 	__ath10k_htt_tx_dec_pending(htt);
 	if (htt->num_pending_tx == 0)
 		wake_up(&htt->empty_tx_wq);
-	spin_unlock_bh(&htt->tx_lock);
 }
 
 static const u8 rx_legacy_rate_idx[] = {
@@ -258,6 +257,12 @@ void ath10k_process_rx(struct ath10k *ar, struct htt_rx_info *info)
 	status->band = ch->band;
 	status->freq = ch->center_freq;
 
+	if (info->rate.info0 & HTT_RX_INDICATION_INFO0_END_VALID) {
+		/* TSF available only in 32-bit */
+		status->mactime = info->tsf & 0xffffffff;
+		status->flag |= RX_FLAG_MACTIME_END;
+	}
+
 	ath10k_dbg(ATH10K_DBG_DATA,
 		   "rx skb %p len %u %s%s%s%s%s %srate_idx %u vht_nss %u freq %u band %u flag 0x%x fcs-err %i\n",
 		   info->skb,
@@ -378,7 +383,8 @@ void ath10k_peer_unmap_event(struct ath10k_htt *htt,
 	spin_lock_bh(&ar->data_lock);
 	peer = ath10k_peer_find_by_id(ar, ev->peer_id);
 	if (!peer) {
-		ath10k_warn("unknown peer id %d\n", ev->peer_id);
+		ath10k_warn("peer-unmap-event: unknown peer id %d\n",
+			    ev->peer_id);
 		goto exit;
 	}
 

drivers/net/wireless/ath/ath10k/wmi.c

@@ -1360,7 +1360,7 @@ static void ath10k_wmi_event_host_swba(struct ath10k *ar, struct sk_buff *skb)
 	struct wmi_bcn_info *bcn_info;
 	struct ath10k_vif *arvif;
 	struct sk_buff *bcn;
-	int vdev_id = 0;
+	int ret, vdev_id = 0;
 
 	ath10k_dbg(ATH10K_DBG_MGMT, "WMI_HOST_SWBA_EVENTID\n");
 
@@ -1435,16 +1435,27 @@ static void ath10k_wmi_event_host_swba(struct ath10k *ar, struct sk_buff *skb)
 				ath10k_warn("SWBA overrun on vdev %d\n",
 					    arvif->vdev_id);
 
-			ath10k_skb_unmap(ar->dev, arvif->beacon);
+			dma_unmap_single(arvif->ar->dev,
+					 ATH10K_SKB_CB(arvif->beacon)->paddr,
+					 arvif->beacon->len, DMA_TO_DEVICE);
 			dev_kfree_skb_any(arvif->beacon);
 		}
 
-		ath10k_skb_map(ar->dev, bcn);
+		ATH10K_SKB_CB(bcn)->paddr = dma_map_single(arvif->ar->dev,
+							   bcn->data, bcn->len,
+							   DMA_TO_DEVICE);
+		ret = dma_mapping_error(arvif->ar->dev,
+					ATH10K_SKB_CB(bcn)->paddr);
+		if (ret) {
+			ath10k_warn("failed to map beacon: %d\n", ret);
+			goto skip;
+		}
 
 		arvif->beacon = bcn;
 		arvif->beacon_sent = false;
 
 		ath10k_wmi_tx_beacon_nowait(arvif);
+skip:
 		spin_unlock_bh(&ar->data_lock);
 	}
 }
@@ -3382,7 +3393,6 @@ int ath10k_wmi_scan_chan_list(struct ath10k *ar,
 		ci->max_power         = ch->max_power;
 		ci->reg_power         = ch->max_reg_power;
 		ci->antenna_max       = ch->max_antenna_gain;
-		ci->antenna_max       = 0;
 
 		/* mode & flags share storage */
 		ci->mode              = ch->mode;