Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next

Pull networking updates from David Miller:

 1) Significantly shrink the core networking routing structures. Result
    of http://vger.kernel.org/~davem/seoul2017_netdev_keynote.pdf

 2) Add netdevsim driver for testing various offloads, from Jakub
    Kicinski.

 3) Support cross-chip FDB operations in DSA, from Vivien Didelot.

 4) Add a 2nd listener hash table for TCP, similar to what was done for
    UDP. From Martin KaFai Lau.

 5) Add eBPF based queue selection to tun, from Jason Wang.

 6) Lockless qdisc support, from John Fastabend.

 7) SCTP stream interleave support, from Xin Long.

 8) Smoother TCP receive autotuning, from Eric Dumazet.

 9) Lots of erspan tunneling enhancements, from William Tu.

10) Add true function call support to BPF, from Alexei Starovoitov.

11) Add explicit support for GRO HW offloading, from Michael Chan.

12) Support extack generation in more netlink subsystems. From Alexander
    Aring, Quentin Monnet, and Jakub Kicinski.

13) Add 1000BaseX, flow control, and EEE support to mvneta driver. From
    Russell King.

14) Add flow table abstraction to netfilter, from Pablo Neira Ayuso.

15) Many improvements and simplifications to the NFP driver bpf JIT,
    from Jakub Kicinski.

16) Support for ipv6 non-equal cost multipath routing, from Ido
    Schimmel.

17) Add resource abstration to devlink, from Arkadi Sharshevsky.

18) Packet scheduler classifier shared filter block support, from Jiri
    Pirko.

19) Avoid locking in act_csum, from Davide Caratti.

20) devinet_ioctl() simplifications from Al viro.

21) More TCP bpf improvements from Lawrence Brakmo.

22) Add support for onlink ipv6 route flag, similar to ipv4, from David
    Ahern.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1925 commits)
  tls: Add support for encryption using async offload accelerator
  ip6mr: fix stale iterator
  net/sched: kconfig: Remove blank help texts
  openvswitch: meter: Use 64-bit arithmetic instead of 32-bit
  tcp_nv: fix potential integer overflow in tcpnv_acked
  r8169: fix RTL8168EP take too long to complete driver initialization.
  qmi_wwan: Add support for Quectel EP06
  rtnetlink: enable IFLA_IF_NETNSID for RTM_NEWLINK
  ipmr: Fix ptrdiff_t print formatting
  ibmvnic: Wait for device response when changing MAC
  qlcnic: fix deadlock bug
  tcp: release sk_frag.page in tcp_disconnect
  ipv4: Get the address of interface correctly.
  net_sched: gen_estimator: fix lockdep splat
  net: macb: Handle HRESP error
  net/mlx5e: IPoIB, Fix copy-paste bug in flow steering refactoring
  ipv6: addrconf: break critical section in addrconf_verify_rtnl()
  ipv6: change route cache aging logic
  i40e/i40evf: Update DESC_NEEDED value to reflect larger value
  bnxt_en: cleanup DIM work on device shutdown
  ...

1 files changed, 540 insertions, 60 deletions

diff --git a/drivers/net/wireless/ath/ath10k/htt_tx.c b/drivers/net/wireless/ath/ath10k/htt_tx.c
index 685faac..d334b7b 100644
--- a/drivers/net/wireless/ath/ath10k/htt_tx.c
+++ b/drivers/net/wireless/ath/ath10k/htt_tx.c
@@ -1,6 +1,6 @@
 /*
  * Copyright (c) 2005-2011 Atheros Communications Inc.
- * Copyright (c) 2011-2013 Qualcomm Atheros, Inc.
+ * Copyright (c) 2011-2017 Qualcomm Atheros, Inc.
  *
  * Permission to use, copy, modify, and/or distribute this software for any
  * purpose with or without fee is hereby granted, provided that the above
@@ -229,50 +229,91 @@ void ath10k_htt_tx_free_msdu_id(struct ath10k_htt *htt, u16 msdu_id)
 	idr_remove(&htt->pending_tx, msdu_id);
 }
 
-static void ath10k_htt_tx_free_cont_txbuf(struct ath10k_htt *htt)
+static void ath10k_htt_tx_free_cont_txbuf_32(struct ath10k_htt *htt)
 {
 	struct ath10k *ar = htt->ar;
 	size_t size;
 
-	if (!htt->txbuf.vaddr)
+	if (!htt->txbuf.vaddr_txbuff_32)
 		return;
 
-	size = htt->max_num_pending_tx * sizeof(struct ath10k_htt_txbuf);
-	dma_free_coherent(ar->dev, size, htt->txbuf.vaddr, htt->txbuf.paddr);
-	htt->txbuf.vaddr = NULL;
+	size = htt->txbuf.size;
+	dma_free_coherent(ar->dev, size, htt->txbuf.vaddr_txbuff_32,
+			  htt->txbuf.paddr);
+	htt->txbuf.vaddr_txbuff_32 = NULL;
 }
 
-static int ath10k_htt_tx_alloc_cont_txbuf(struct ath10k_htt *htt)
+static int ath10k_htt_tx_alloc_cont_txbuf_32(struct ath10k_htt *htt)
 {
 	struct ath10k *ar = htt->ar;
 	size_t size;
 
-	size = htt->max_num_pending_tx * sizeof(struct ath10k_htt_txbuf);
-	htt->txbuf.vaddr = dma_alloc_coherent(ar->dev, size, &htt->txbuf.paddr,
-					      GFP_KERNEL);
-	if (!htt->txbuf.vaddr)
+	size = htt->max_num_pending_tx *
+			sizeof(struct ath10k_htt_txbuf_32);
+
+	htt->txbuf.vaddr_txbuff_32 = dma_alloc_coherent(ar->dev, size,
+							&htt->txbuf.paddr,
+							GFP_KERNEL);
+	if (!htt->txbuf.vaddr_txbuff_32)
 		return -ENOMEM;
 
+	htt->txbuf.size = size;
+
 	return 0;
 }
 
-static void ath10k_htt_tx_free_cont_frag_desc(struct ath10k_htt *htt)
+static void ath10k_htt_tx_free_cont_txbuf_64(struct ath10k_htt *htt)
 {
+	struct ath10k *ar = htt->ar;
 	size_t size;
 
-	if (!htt->frag_desc.vaddr)
+	if (!htt->txbuf.vaddr_txbuff_64)
 		return;
 
-	size = htt->max_num_pending_tx * sizeof(struct htt_msdu_ext_desc);
+	size = htt->txbuf.size;
+	dma_free_coherent(ar->dev, size, htt->txbuf.vaddr_txbuff_64,
+			  htt->txbuf.paddr);
+	htt->txbuf.vaddr_txbuff_64 = NULL;
+}
+
+static int ath10k_htt_tx_alloc_cont_txbuf_64(struct ath10k_htt *htt)
+{
+	struct ath10k *ar = htt->ar;
+	size_t size;
+
+	size = htt->max_num_pending_tx *
+			sizeof(struct ath10k_htt_txbuf_64);
+
+	htt->txbuf.vaddr_txbuff_64 = dma_alloc_coherent(ar->dev, size,
+							&htt->txbuf.paddr,
+							GFP_KERNEL);
+	if (!htt->txbuf.vaddr_txbuff_64)
+		return -ENOMEM;
+
+	htt->txbuf.size = size;
+
+	return 0;
+}
+
+static void ath10k_htt_tx_free_cont_frag_desc_32(struct ath10k_htt *htt)
+{
+	size_t size;
+
+	if (!htt->frag_desc.vaddr_desc_32)
+		return;
+
+	size = htt->max_num_pending_tx *
+			sizeof(struct htt_msdu_ext_desc);
 
 	dma_free_coherent(htt->ar->dev,
 			  size,
-			  htt->frag_desc.vaddr,
+			  htt->frag_desc.vaddr_desc_32,
 			  htt->frag_desc.paddr);
-	htt->frag_desc.vaddr = NULL;
+
+	htt->frag_desc.vaddr_desc_32 = NULL;
 }
 
-static int ath10k_htt_tx_alloc_cont_frag_desc(struct ath10k_htt *htt)
+static int ath10k_htt_tx_alloc_cont_frag_desc_32(struct ath10k_htt *htt)
 {
 	struct ath10k *ar = htt->ar;
 	size_t size;
@@ -280,12 +321,57 @@ static int ath10k_htt_tx_alloc_cont_frag_desc(struct ath10k_htt *htt)
 	if (!ar->hw_params.continuous_frag_desc)
 		return 0;
 
-	size = htt->max_num_pending_tx * sizeof(struct htt_msdu_ext_desc);
-	htt->frag_desc.vaddr = dma_alloc_coherent(ar->dev, size,
-						  &htt->frag_desc.paddr,
-						  GFP_KERNEL);
-	if (!htt->frag_desc.vaddr)
+	size = htt->max_num_pending_tx *
+			sizeof(struct htt_msdu_ext_desc);
+	htt->frag_desc.vaddr_desc_32 = dma_alloc_coherent(ar->dev, size,
+							  &htt->frag_desc.paddr,
+							  GFP_KERNEL);
+	if (!htt->frag_desc.vaddr_desc_32) {
+		ath10k_err(ar, "failed to alloc fragment desc memory\n");
 		return -ENOMEM;
+	}
+	htt->frag_desc.size = size;
+
+	return 0;
+}
+
+static void ath10k_htt_tx_free_cont_frag_desc_64(struct ath10k_htt *htt)
+{
+	size_t size;
+
+	if (!htt->frag_desc.vaddr_desc_64)
+		return;
+
+	size = htt->max_num_pending_tx *
+			sizeof(struct htt_msdu_ext_desc_64);
+
+	dma_free_coherent(htt->ar->dev,
+			  size,
+			  htt->frag_desc.vaddr_desc_64,
+			  htt->frag_desc.paddr);
+
+	htt->frag_desc.vaddr_desc_64 = NULL;
+}
+
+static int ath10k_htt_tx_alloc_cont_frag_desc_64(struct ath10k_htt *htt)
+{
+	struct ath10k *ar = htt->ar;
+	size_t size;
+
+	if (!ar->hw_params.continuous_frag_desc)
+		return 0;
+
+	size = htt->max_num_pending_tx *
+			sizeof(struct htt_msdu_ext_desc_64);
+
+	htt->frag_desc.vaddr_desc_64 = dma_alloc_coherent(ar->dev, size,
+							  &htt->frag_desc.paddr,
+							  GFP_KERNEL);
+	if (!htt->frag_desc.vaddr_desc_64) {
+		ath10k_err(ar, "failed to alloc fragment desc memory\n");
+		return -ENOMEM;
+	}
+	htt->frag_desc.size = size;
 
 	return 0;
 }
@@ -357,13 +443,13 @@ static int ath10k_htt_tx_alloc_buf(struct ath10k_htt *htt)
 	struct ath10k *ar = htt->ar;
 	int ret;
 
-	ret = ath10k_htt_tx_alloc_cont_txbuf(htt);
+	ret = htt->tx_ops->htt_alloc_txbuff(htt);
 	if (ret) {
 		ath10k_err(ar, "failed to alloc cont tx buffer: %d\n", ret);
 		return ret;
 	}
 
-	ret = ath10k_htt_tx_alloc_cont_frag_desc(htt);
+	ret = htt->tx_ops->htt_alloc_frag_desc(htt);
 	if (ret) {
 		ath10k_err(ar, "failed to alloc cont frag desc: %d\n", ret);
 		goto free_txbuf;
@@ -387,10 +473,10 @@ free_txq:
 	ath10k_htt_tx_free_txq(htt);
 
 free_frag_desc:
-	ath10k_htt_tx_free_cont_frag_desc(htt);
+	htt->tx_ops->htt_free_frag_desc(htt);
 
 free_txbuf:
-	ath10k_htt_tx_free_cont_txbuf(htt);
+	htt->tx_ops->htt_free_txbuff(htt);
 
 	return ret;
 }
@@ -444,9 +530,9 @@ void ath10k_htt_tx_destroy(struct ath10k_htt *htt)
 	if (!htt->tx_mem_allocated)
 		return;
 
-	ath10k_htt_tx_free_cont_txbuf(htt);
+	htt->tx_ops->htt_free_txbuff(htt);
 	ath10k_htt_tx_free_txq(htt);
-	ath10k_htt_tx_free_cont_frag_desc(htt);
+	htt->tx_ops->htt_free_frag_desc(htt);
 	ath10k_htt_tx_free_txdone_fifo(htt);
 	htt->tx_mem_allocated = false;
 }
@@ -545,12 +631,12 @@ int ath10k_htt_h2t_stats_req(struct ath10k_htt *htt, u8 mask, u64 cookie)
 	return 0;
 }
 
-int ath10k_htt_send_frag_desc_bank_cfg(struct ath10k_htt *htt)
+static int ath10k_htt_send_frag_desc_bank_cfg_32(struct ath10k_htt *htt)
 {
 	struct ath10k *ar = htt->ar;
 	struct sk_buff *skb;
 	struct htt_cmd *cmd;
-	struct htt_frag_desc_bank_cfg *cfg;
+	struct htt_frag_desc_bank_cfg32 *cfg;
 	int ret, size;
 	u8 info;
 
@@ -562,7 +648,7 @@ int ath10k_htt_send_frag_desc_bank_cfg(struct ath10k_htt *htt)
 		return -EINVAL;
 	}
 
-	size = sizeof(cmd->hdr) + sizeof(cmd->frag_desc_bank_cfg);
+	size = sizeof(cmd->hdr) + sizeof(cmd->frag_desc_bank_cfg32);
 	skb = ath10k_htc_alloc_skb(ar, size);
 	if (!skb)
 		return -ENOMEM;
@@ -579,7 +665,7 @@ int ath10k_htt_send_frag_desc_bank_cfg(struct ath10k_htt *htt)
 		     ar->running_fw->fw_file.fw_features))
 		info |= HTT_FRAG_DESC_BANK_CFG_INFO_Q_STATE_VALID;
 
-	cfg = &cmd->frag_desc_bank_cfg;
+	cfg = &cmd->frag_desc_bank_cfg32;
 	cfg->info = info;
 	cfg->num_banks = 1;
 	cfg->desc_size = sizeof(struct htt_msdu_ext_desc);
@@ -607,12 +693,112 @@ int ath10k_htt_send_frag_desc_bank_cfg(struct ath10k_htt *htt)
 	return 0;
 }
 
-int ath10k_htt_send_rx_ring_cfg_ll(struct ath10k_htt *htt)
+static int ath10k_htt_send_frag_desc_bank_cfg_64(struct ath10k_htt *htt)
 {
 	struct ath10k *ar = htt->ar;
 	struct sk_buff *skb;
 	struct htt_cmd *cmd;
-	struct htt_rx_ring_setup_ring *ring;
+	struct htt_frag_desc_bank_cfg64 *cfg;
+	int ret, size;
+	u8 info;
+
+	if (!ar->hw_params.continuous_frag_desc)
+		return 0;
+
+	if (!htt->frag_desc.paddr) {
+		ath10k_warn(ar, "invalid frag desc memory\n");
+		return -EINVAL;
+	}
+
+	size = sizeof(cmd->hdr) + sizeof(cmd->frag_desc_bank_cfg64);
+	skb = ath10k_htc_alloc_skb(ar, size);
+	if (!skb)
+		return -ENOMEM;
+
+	skb_put(skb, size);
+	cmd = (struct htt_cmd *)skb->data;
+	cmd->hdr.msg_type = HTT_H2T_MSG_TYPE_FRAG_DESC_BANK_CFG;
+
+	info = 0;
+	info |= SM(htt->tx_q_state.type,
+		   HTT_FRAG_DESC_BANK_CFG_INFO_Q_STATE_DEPTH_TYPE);
+
+	if (test_bit(ATH10K_FW_FEATURE_PEER_FLOW_CONTROL,
+		     ar->running_fw->fw_file.fw_features))
+		info |= HTT_FRAG_DESC_BANK_CFG_INFO_Q_STATE_VALID;
+
+	cfg = &cmd->frag_desc_bank_cfg64;
+	cfg->info = info;
+	cfg->num_banks = 1;
+	cfg->desc_size = sizeof(struct htt_msdu_ext_desc_64);
+	cfg->bank_base_addrs[0] =  __cpu_to_le64(htt->frag_desc.paddr);
+	cfg->bank_id[0].bank_min_id = 0;
+	cfg->bank_id[0].bank_max_id = __cpu_to_le16(htt->max_num_pending_tx -
+						    1);
+
+	cfg->q_state.paddr = cpu_to_le32(htt->tx_q_state.paddr);
+	cfg->q_state.num_peers = cpu_to_le16(htt->tx_q_state.num_peers);
+	cfg->q_state.num_tids = cpu_to_le16(htt->tx_q_state.num_tids);
+	cfg->q_state.record_size = HTT_TX_Q_STATE_ENTRY_SIZE;
+	cfg->q_state.record_multiplier = HTT_TX_Q_STATE_ENTRY_MULTIPLIER;
+
+	ath10k_dbg(ar, ATH10K_DBG_HTT, "htt frag desc bank cmd\n");
+
+	ret = ath10k_htc_send(&htt->ar->htc, htt->eid, skb);
+	if (ret) {
+		ath10k_warn(ar, "failed to send frag desc bank cfg request: %d\n",
+			    ret);
+		dev_kfree_skb_any(skb);
+		return ret;
+	}
+
+	return 0;
+}
+
+static void ath10k_htt_fill_rx_desc_offset_32(void *rx_ring)
+{
+	struct htt_rx_ring_setup_ring32 *ring =
+			(struct htt_rx_ring_setup_ring32 *)rx_ring;
+
+#define desc_offset(x) (offsetof(struct htt_rx_desc, x) / 4)
+	ring->mac80211_hdr_offset = __cpu_to_le16(desc_offset(rx_hdr_status));
+	ring->msdu_payload_offset = __cpu_to_le16(desc_offset(msdu_payload));
+	ring->ppdu_start_offset = __cpu_to_le16(desc_offset(ppdu_start));
+	ring->ppdu_end_offset = __cpu_to_le16(desc_offset(ppdu_end));
+	ring->mpdu_start_offset = __cpu_to_le16(desc_offset(mpdu_start));
+	ring->mpdu_end_offset = __cpu_to_le16(desc_offset(mpdu_end));
+	ring->msdu_start_offset = __cpu_to_le16(desc_offset(msdu_start));
+	ring->msdu_end_offset = __cpu_to_le16(desc_offset(msdu_end));
+	ring->rx_attention_offset = __cpu_to_le16(desc_offset(attention));
+	ring->frag_info_offset = __cpu_to_le16(desc_offset(frag_info));
+#undef desc_offset
+}
+
+static void ath10k_htt_fill_rx_desc_offset_64(void *rx_ring)
+{
+	struct htt_rx_ring_setup_ring64 *ring =
+			(struct htt_rx_ring_setup_ring64 *)rx_ring;
+
+#define desc_offset(x) (offsetof(struct htt_rx_desc, x) / 4)
+	ring->mac80211_hdr_offset = __cpu_to_le16(desc_offset(rx_hdr_status));
+	ring->msdu_payload_offset = __cpu_to_le16(desc_offset(msdu_payload));
+	ring->ppdu_start_offset = __cpu_to_le16(desc_offset(ppdu_start));
+	ring->ppdu_end_offset = __cpu_to_le16(desc_offset(ppdu_end));
+	ring->mpdu_start_offset = __cpu_to_le16(desc_offset(mpdu_start));
+	ring->mpdu_end_offset = __cpu_to_le16(desc_offset(mpdu_end));
+	ring->msdu_start_offset = __cpu_to_le16(desc_offset(msdu_start));
+	ring->msdu_end_offset = __cpu_to_le16(desc_offset(msdu_end));
+	ring->rx_attention_offset = __cpu_to_le16(desc_offset(attention));
+	ring->frag_info_offset = __cpu_to_le16(desc_offset(frag_info));
+#undef desc_offset
+}
+
+static int ath10k_htt_send_rx_ring_cfg_32(struct ath10k_htt *htt)
+{
+	struct ath10k *ar = htt->ar;
+	struct sk_buff *skb;
+	struct htt_cmd *cmd;
+	struct htt_rx_ring_setup_ring32 *ring;
 	const int num_rx_ring = 1;
 	u16 flags;
 	u32 fw_idx;
@@ -626,7 +812,7 @@ int ath10k_htt_send_rx_ring_cfg_ll(struct ath10k_htt *htt)
 	BUILD_BUG_ON(!IS_ALIGNED(HTT_RX_BUF_SIZE, 4));
 	BUILD_BUG_ON((HTT_RX_BUF_SIZE & HTT_MAX_CACHE_LINE_SIZE_MASK) != 0);
 
-	len = sizeof(cmd->hdr) + sizeof(cmd->rx_setup.hdr)
+	len = sizeof(cmd->hdr) + sizeof(cmd->rx_setup_32.hdr)
 	    + (sizeof(*ring) * num_rx_ring);
 	skb = ath10k_htc_alloc_skb(ar, len);
 	if (!skb)
@@ -635,10 +821,10 @@ int ath10k_htt_send_rx_ring_cfg_ll(struct ath10k_htt *htt)
 	skb_put(skb, len);
 
 	cmd = (struct htt_cmd *)skb->data;
-	ring = &cmd->rx_setup.rings[0];
+	ring = &cmd->rx_setup_32.rings[0];
 
 	cmd->hdr.msg_type = HTT_H2T_MSG_TYPE_RX_RING_CFG;
-	cmd->rx_setup.hdr.num_rings = 1;
+	cmd->rx_setup_32.hdr.num_rings = 1;
 
 	/* FIXME: do we need all of this? */
 	flags = 0;
@@ -669,21 +855,76 @@ int ath10k_htt_send_rx_ring_cfg_ll(struct ath10k_htt *htt)
 	ring->flags = __cpu_to_le16(flags);
 	ring->fw_idx_init_val = __cpu_to_le16(fw_idx);
 
-#define desc_offset(x) (offsetof(struct htt_rx_desc, x) / 4)
+	ath10k_htt_fill_rx_desc_offset_32(ring);
+	ret = ath10k_htc_send(&htt->ar->htc, htt->eid, skb);
+	if (ret) {
+		dev_kfree_skb_any(skb);
+		return ret;
+	}
 
-	ring->mac80211_hdr_offset = __cpu_to_le16(desc_offset(rx_hdr_status));
-	ring->msdu_payload_offset = __cpu_to_le16(desc_offset(msdu_payload));
-	ring->ppdu_start_offset = __cpu_to_le16(desc_offset(ppdu_start));
-	ring->ppdu_end_offset = __cpu_to_le16(desc_offset(ppdu_end));
-	ring->mpdu_start_offset = __cpu_to_le16(desc_offset(mpdu_start));
-	ring->mpdu_end_offset = __cpu_to_le16(desc_offset(mpdu_end));
-	ring->msdu_start_offset = __cpu_to_le16(desc_offset(msdu_start));
-	ring->msdu_end_offset = __cpu_to_le16(desc_offset(msdu_end));
-	ring->rx_attention_offset = __cpu_to_le16(desc_offset(attention));
-	ring->frag_info_offset = __cpu_to_le16(desc_offset(frag_info));
+	return 0;
+}
 
-#undef desc_offset
+static int ath10k_htt_send_rx_ring_cfg_64(struct ath10k_htt *htt)
+{
+	struct ath10k *ar = htt->ar;
+	struct sk_buff *skb;
+	struct htt_cmd *cmd;
+	struct htt_rx_ring_setup_ring64 *ring;
+	const int num_rx_ring = 1;
+	u16 flags;
+	u32 fw_idx;
+	int len;
+	int ret;
+
+	/* HW expects the buffer to be an integral number of 4-byte
+	 * "words"
+	 */
+	BUILD_BUG_ON(!IS_ALIGNED(HTT_RX_BUF_SIZE, 4));
+	BUILD_BUG_ON((HTT_RX_BUF_SIZE & HTT_MAX_CACHE_LINE_SIZE_MASK) != 0);
+
+	len = sizeof(cmd->hdr) + sizeof(cmd->rx_setup_64.hdr)
+	    + (sizeof(*ring) * num_rx_ring);
+	skb = ath10k_htc_alloc_skb(ar, len);
+	if (!skb)
+		return -ENOMEM;
+
+	skb_put(skb, len);
+
+	cmd = (struct htt_cmd *)skb->data;
+	ring = &cmd->rx_setup_64.rings[0];
+
+	cmd->hdr.msg_type = HTT_H2T_MSG_TYPE_RX_RING_CFG;
+	cmd->rx_setup_64.hdr.num_rings = 1;
+
+	flags = 0;
+	flags |= HTT_RX_RING_FLAGS_MAC80211_HDR;
+	flags |= HTT_RX_RING_FLAGS_MSDU_PAYLOAD;
+	flags |= HTT_RX_RING_FLAGS_PPDU_START;
+	flags |= HTT_RX_RING_FLAGS_PPDU_END;
+	flags |= HTT_RX_RING_FLAGS_MPDU_START;
+	flags |= HTT_RX_RING_FLAGS_MPDU_END;
+	flags |= HTT_RX_RING_FLAGS_MSDU_START;
+	flags |= HTT_RX_RING_FLAGS_MSDU_END;
+	flags |= HTT_RX_RING_FLAGS_RX_ATTENTION;
+	flags |= HTT_RX_RING_FLAGS_FRAG_INFO;
+	flags |= HTT_RX_RING_FLAGS_UNICAST_RX;
+	flags |= HTT_RX_RING_FLAGS_MULTICAST_RX;
+	flags |= HTT_RX_RING_FLAGS_CTRL_RX;
+	flags |= HTT_RX_RING_FLAGS_MGMT_RX;
+	flags |= HTT_RX_RING_FLAGS_NULL_RX;
+	flags |= HTT_RX_RING_FLAGS_PHY_DATA_RX;
 
+	fw_idx = __le32_to_cpu(*htt->rx_ring.alloc_idx.vaddr);
+
+	ring->fw_idx_shadow_reg_paddr = __cpu_to_le64(htt->rx_ring.alloc_idx.paddr);
+	ring->rx_ring_base_paddr = __cpu_to_le64(htt->rx_ring.base_paddr);
+	ring->rx_ring_len = __cpu_to_le16(htt->rx_ring.size);
+	ring->rx_ring_bufsize = __cpu_to_le16(HTT_RX_BUF_SIZE);
+	ring->flags = __cpu_to_le16(flags);
+	ring->fw_idx_init_val = __cpu_to_le16(fw_idx);
+
+	ath10k_htt_fill_rx_desc_offset_64(ring);
 	ret = ath10k_htc_send(&htt->ar->htc, htt->eid, skb);
 	if (ret) {
 		dev_kfree_skb_any(skb);
@@ -895,8 +1136,9 @@ err:
 	return res;
 }
 
-int ath10k_htt_tx(struct ath10k_htt *htt, enum ath10k_hw_txrx_mode txmode,
-		  struct sk_buff *msdu)
+static int ath10k_htt_tx_32(struct ath10k_htt *htt,
+			    enum ath10k_hw_txrx_mode txmode,
+			    struct sk_buff *msdu)
 {
 	struct ath10k *ar = htt->ar;
 	struct device *dev = ar->dev;
@@ -904,7 +1146,7 @@ int ath10k_htt_tx(struct ath10k_htt *htt, enum ath10k_hw_txrx_mode txmode,
 	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(msdu);
 	struct ath10k_skb_cb *skb_cb = ATH10K_SKB_CB(msdu);
 	struct ath10k_hif_sg_item sg_items[2];
-	struct ath10k_htt_txbuf *txbuf;
+	struct ath10k_htt_txbuf_32 *txbuf;
 	struct htt_data_tx_desc_frag *frags;
 	bool is_eth = (txmode == ATH10K_HW_TXRX_ETHERNET);
 	u8 vdev_id = ath10k_htt_tx_get_vdev_id(ar, msdu);
@@ -917,6 +1159,7 @@ int ath10k_htt_tx(struct ath10k_htt *htt, enum ath10k_hw_txrx_mode txmode,
 	u32 frags_paddr = 0;
 	u32 txbuf_paddr;
 	struct htt_msdu_ext_desc *ext_desc = NULL;
+	struct htt_msdu_ext_desc *ext_desc_t = NULL;
 
 	spin_lock_bh(&htt->tx_lock);
 	res = ath10k_htt_tx_alloc_msdu_id(htt, msdu);
@@ -929,9 +1172,9 @@ int ath10k_htt_tx(struct ath10k_htt *htt, enum ath10k_hw_txrx_mode txmode,
 	prefetch_len = min(htt->prefetch_len, msdu->len);
 	prefetch_len = roundup(prefetch_len, 4);
 
-	txbuf = &htt->txbuf.vaddr[msdu_id];
+	txbuf = htt->txbuf.vaddr_txbuff_32 + msdu_id;
 	txbuf_paddr = htt->txbuf.paddr +
-		      (sizeof(struct ath10k_htt_txbuf) * msdu_id);
+		      (sizeof(struct ath10k_htt_txbuf_32) * msdu_id);
 
 	if ((ieee80211_is_action(hdr->frame_control) ||
 	     ieee80211_is_deauth(hdr->frame_control) ||
@@ -962,11 +1205,12 @@ int ath10k_htt_tx(struct ath10k_htt *htt, enum ath10k_hw_txrx_mode txmode,
 		/* pass through */
 	case ATH10K_HW_TXRX_ETHERNET:
 		if (ar->hw_params.continuous_frag_desc) {
-			memset(&htt->frag_desc.vaddr[msdu_id], 0,
+			ext_desc_t = htt->frag_desc.vaddr_desc_32;
+			memset(&ext_desc_t[msdu_id], 0,
 			       sizeof(struct htt_msdu_ext_desc));
 			frags = (struct htt_data_tx_desc_frag *)
-				&htt->frag_desc.vaddr[msdu_id].frags;
-			ext_desc = &htt->frag_desc.vaddr[msdu_id];
+				&ext_desc_t[msdu_id].frags;
+			ext_desc = &ext_desc_t[msdu_id];
 			frags[0].tword_addr.paddr_lo =
 				__cpu_to_le32(skb_cb->paddr);
 			frags[0].tword_addr.paddr_hi = 0;
@@ -1055,9 +1299,9 @@ int ath10k_htt_tx(struct ath10k_htt *htt, enum ath10k_hw_txrx_mode txmode,
 
 	trace_ath10k_htt_tx(ar, msdu_id, msdu->len, vdev_id, tid);
 	ath10k_dbg(ar, ATH10K_DBG_HTT,
-		   "htt tx flags0 %hhu flags1 %hu len %d id %hu frags_paddr %08x, msdu_paddr %08x vdev %hhu tid %hhu freq %hu\n",
-		   flags0, flags1, msdu->len, msdu_id, frags_paddr,
-		   (u32)skb_cb->paddr, vdev_id, tid, freq);
+		   "htt tx flags0 %hhu flags1 %hu len %d id %hu frags_paddr %pad, msdu_paddr %pad vdev %hhu tid %hhu freq %hu\n",
+		   flags0, flags1, msdu->len, msdu_id, &frags_paddr,
+		   &skb_cb->paddr, vdev_id, tid, freq);
 	ath10k_dbg_dump(ar, ATH10K_DBG_HTT_DUMP, NULL, "htt tx msdu: ",
 			msdu->data, msdu->len);
 	trace_ath10k_tx_hdr(ar, msdu->data, msdu->len);
@@ -1093,3 +1337,239 @@ err_free_msdu_id:
 err:
 	return res;
 }
+
+static int ath10k_htt_tx_64(struct ath10k_htt *htt,
+			    enum ath10k_hw_txrx_mode txmode,
+			    struct sk_buff *msdu)
+{
+	struct ath10k *ar = htt->ar;
+	struct device *dev = ar->dev;
+	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)msdu->data;
+	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(msdu);
+	struct ath10k_skb_cb *skb_cb = ATH10K_SKB_CB(msdu);
+	struct ath10k_hif_sg_item sg_items[2];
+	struct ath10k_htt_txbuf_64 *txbuf;
+	struct htt_data_tx_desc_frag *frags;
+	bool is_eth = (txmode == ATH10K_HW_TXRX_ETHERNET);
+	u8 vdev_id = ath10k_htt_tx_get_vdev_id(ar, msdu);
+	u8 tid = ath10k_htt_tx_get_tid(msdu, is_eth);
+	int prefetch_len;
+	int res;
+	u8 flags0 = 0;
+	u16 msdu_id, flags1 = 0;
+	u16 freq = 0;
+	dma_addr_t frags_paddr = 0;
+	u32 txbuf_paddr;
+	struct htt_msdu_ext_desc_64 *ext_desc = NULL;
+	struct htt_msdu_ext_desc_64 *ext_desc_t = NULL;
+
+	spin_lock_bh(&htt->tx_lock);
+	res = ath10k_htt_tx_alloc_msdu_id(htt, msdu);
+	spin_unlock_bh(&htt->tx_lock);
+	if (res < 0)
+		goto err;
+
+	msdu_id = res;
+
+	prefetch_len = min(htt->prefetch_len, msdu->len);
+	prefetch_len = roundup(prefetch_len, 4);
+
+	txbuf = htt->txbuf.vaddr_txbuff_64 + msdu_id;
+	txbuf_paddr = htt->txbuf.paddr +
+		      (sizeof(struct ath10k_htt_txbuf_64) * msdu_id);
+
+	if ((ieee80211_is_action(hdr->frame_control) ||
+	     ieee80211_is_deauth(hdr->frame_control) ||
+	     ieee80211_is_disassoc(hdr->frame_control)) &&
+	     ieee80211_has_protected(hdr->frame_control)) {
+		skb_put(msdu, IEEE80211_CCMP_MIC_LEN);
+	} else if (!(skb_cb->flags & ATH10K_SKB_F_NO_HWCRYPT) &&
+		   txmode == ATH10K_HW_TXRX_RAW &&
+		   ieee80211_has_protected(hdr->frame_control)) {
+		skb_put(msdu, IEEE80211_CCMP_MIC_LEN);
+	}
+
+	skb_cb->paddr = dma_map_single(dev, msdu->data, msdu->len,
+				       DMA_TO_DEVICE);
+	res = dma_mapping_error(dev, skb_cb->paddr);
+	if (res) {
+		res = -EIO;
+		goto err_free_msdu_id;
+	}
+
+	if (unlikely(info->flags & IEEE80211_TX_CTL_TX_OFFCHAN))
+		freq = ar->scan.roc_freq;
+
+	switch (txmode) {
+	case ATH10K_HW_TXRX_RAW:
+	case ATH10K_HW_TXRX_NATIVE_WIFI:
+		flags0 |= HTT_DATA_TX_DESC_FLAGS0_MAC_HDR_PRESENT;
+		/* pass through */
+	case ATH10K_HW_TXRX_ETHERNET:
+		if (ar->hw_params.continuous_frag_desc) {
+			ext_desc_t = htt->frag_desc.vaddr_desc_64;
+			memset(&ext_desc_t[msdu_id], 0,
+			       sizeof(struct htt_msdu_ext_desc_64));
+			frags = (struct htt_data_tx_desc_frag *)
+				&ext_desc_t[msdu_id].frags;
+			ext_desc = &ext_desc_t[msdu_id];
+			frags[0].tword_addr.paddr_lo =
+				__cpu_to_le32(skb_cb->paddr);
+			frags[0].tword_addr.paddr_hi =
+				__cpu_to_le16(upper_32_bits(skb_cb->paddr));
+			frags[0].tword_addr.len_16 = __cpu_to_le16(msdu->len);
+
+			frags_paddr =  htt->frag_desc.paddr +
+			   (sizeof(struct htt_msdu_ext_desc_64) * msdu_id);
+		} else {
+			frags = txbuf->frags;
+			frags[0].tword_addr.paddr_lo =
+						__cpu_to_le32(skb_cb->paddr);
+			frags[0].tword_addr.paddr_hi =
+				__cpu_to_le16(upper_32_bits(skb_cb->paddr));
+			frags[0].tword_addr.len_16 = __cpu_to_le16(msdu->len);
+			frags[1].tword_addr.paddr_lo = 0;
+			frags[1].tword_addr.paddr_hi = 0;
+			frags[1].tword_addr.len_16 = 0;
+		}
+		flags0 |= SM(txmode, HTT_DATA_TX_DESC_FLAGS0_PKT_TYPE);
+		break;
+	case ATH10K_HW_TXRX_MGMT:
+		flags0 |= SM(ATH10K_HW_TXRX_MGMT,
+			     HTT_DATA_TX_DESC_FLAGS0_PKT_TYPE);
+		flags0 |= HTT_DATA_TX_DESC_FLAGS0_MAC_HDR_PRESENT;
+
+		frags_paddr = skb_cb->paddr;
+		break;
+	}
+
+	/* 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.
+	 */
+
+	txbuf->htc_hdr.eid = htt->eid;
+	txbuf->htc_hdr.len = __cpu_to_le16(sizeof(txbuf->cmd_hdr) +
+					   sizeof(txbuf->cmd_tx) +
+					   prefetch_len);
+	txbuf->htc_hdr.flags = 0;
+
+	if (skb_cb->flags & ATH10K_SKB_F_NO_HWCRYPT)
+		flags0 |= HTT_DATA_TX_DESC_FLAGS0_NO_ENCRYPT;
+
+	flags1 |= SM((u16)vdev_id, HTT_DATA_TX_DESC_FLAGS1_VDEV_ID);
+	flags1 |= SM((u16)tid, HTT_DATA_TX_DESC_FLAGS1_EXT_TID);
+	if (msdu->ip_summed == CHECKSUM_PARTIAL &&
+	    !test_bit(ATH10K_FLAG_RAW_MODE, &ar->dev_flags)) {
+		flags1 |= HTT_DATA_TX_DESC_FLAGS1_CKSUM_L3_OFFLOAD;
+		flags1 |= HTT_DATA_TX_DESC_FLAGS1_CKSUM_L4_OFFLOAD;
+		if (ar->hw_params.continuous_frag_desc)
+			ext_desc->flags |= HTT_MSDU_CHECKSUM_ENABLE;
+	}
+
+	/* Prevent firmware from sending up tx inspection requests. There's
+	 * nothing ath10k can do with frames requested for inspection so force
+	 * it to simply rely a regular tx completion with discard status.
+	 */
+	flags1 |= HTT_DATA_TX_DESC_FLAGS1_POSTPONED;
+
+	txbuf->cmd_hdr.msg_type = HTT_H2T_MSG_TYPE_TX_FRM;
+	txbuf->cmd_tx.flags0 = flags0;
+	txbuf->cmd_tx.flags1 = __cpu_to_le16(flags1);
+	txbuf->cmd_tx.len = __cpu_to_le16(msdu->len);
+	txbuf->cmd_tx.id = __cpu_to_le16(msdu_id);
+
+	/* fill fragment descriptor */
+	txbuf->cmd_tx.frags_paddr = __cpu_to_le64(frags_paddr);
+	if (ath10k_mac_tx_frm_has_freq(ar)) {
+		txbuf->cmd_tx.offchan_tx.peerid =
+				__cpu_to_le16(HTT_INVALID_PEERID);
+		txbuf->cmd_tx.offchan_tx.freq =
+				__cpu_to_le16(freq);
+	} else {
+		txbuf->cmd_tx.peerid =
+				__cpu_to_le32(HTT_INVALID_PEERID);
+	}
+
+	trace_ath10k_htt_tx(ar, msdu_id, msdu->len, vdev_id, tid);
+	ath10k_dbg(ar, ATH10K_DBG_HTT,
+		   "htt tx flags0 %hhu flags1 %hu len %d id %hu frags_paddr %pad, msdu_paddr %pad vdev %hhu tid %hhu freq %hu\n",
+		   flags0, flags1, msdu->len, msdu_id, &frags_paddr,
+		   &skb_cb->paddr, vdev_id, tid, freq);
+	ath10k_dbg_dump(ar, ATH10K_DBG_HTT_DUMP, NULL, "htt tx msdu: ",
+			msdu->data, msdu->len);
+	trace_ath10k_tx_hdr(ar, msdu->data, msdu->len);
+	trace_ath10k_tx_payload(ar, msdu->data, msdu->len);
+
+	sg_items[0].transfer_id = 0;
+	sg_items[0].transfer_context = NULL;
+	sg_items[0].vaddr = &txbuf->htc_hdr;
+	sg_items[0].paddr = txbuf_paddr +
+			    sizeof(txbuf->frags);
+	sg_items[0].len = sizeof(txbuf->htc_hdr) +
+			  sizeof(txbuf->cmd_hdr) +
+			  sizeof(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:
+	dma_unmap_single(dev, skb_cb->paddr, msdu->len, DMA_TO_DEVICE);
+err_free_msdu_id:
+	ath10k_htt_tx_free_msdu_id(htt, msdu_id);
+err:
+	return res;
+}
+
+static const struct ath10k_htt_tx_ops htt_tx_ops_32 = {
+	.htt_send_rx_ring_cfg = ath10k_htt_send_rx_ring_cfg_32,
+	.htt_send_frag_desc_bank_cfg = ath10k_htt_send_frag_desc_bank_cfg_32,
+	.htt_alloc_frag_desc = ath10k_htt_tx_alloc_cont_frag_desc_32,
+	.htt_free_frag_desc = ath10k_htt_tx_free_cont_frag_desc_32,
+	.htt_tx = ath10k_htt_tx_32,
+	.htt_alloc_txbuff = ath10k_htt_tx_alloc_cont_txbuf_32,
+	.htt_free_txbuff = ath10k_htt_tx_free_cont_txbuf_32,
+};
+
+static const struct ath10k_htt_tx_ops htt_tx_ops_64 = {
+	.htt_send_rx_ring_cfg = ath10k_htt_send_rx_ring_cfg_64,
+	.htt_send_frag_desc_bank_cfg = ath10k_htt_send_frag_desc_bank_cfg_64,
+	.htt_alloc_frag_desc = ath10k_htt_tx_alloc_cont_frag_desc_64,
+	.htt_free_frag_desc = ath10k_htt_tx_free_cont_frag_desc_64,
+	.htt_tx = ath10k_htt_tx_64,
+	.htt_alloc_txbuff = ath10k_htt_tx_alloc_cont_txbuf_64,
+	.htt_free_txbuff = ath10k_htt_tx_free_cont_txbuf_64,
+};
+
+void ath10k_htt_set_tx_ops(struct ath10k_htt *htt)
+{
+	struct ath10k *ar = htt->ar;
+
+	if (ar->hw_params.target_64bit)
+		htt->tx_ops = &htt_tx_ops_64;
+	else
+		htt->tx_ops = &htt_tx_ops_32;
+}