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

Pull networking updates from David Miller:
 "Most notable changes in here:

   1) By far the biggest accomplishment, thanks to a large range of
      contributors, is the addition of multi-send for transmit.  This is
      the result of discussions back in Chicago, and the hard work of
      several individuals.

      Now, when the ->ndo_start_xmit() method of a driver sees
      skb->xmit_more as true, it can choose to defer the doorbell
      telling the driver to start processing the new TX queue entires.

      skb->xmit_more means that the generic networking is guaranteed to
      call the driver immediately with another SKB to send.

      There is logic added to the qdisc layer to dequeue multiple
      packets at a time, and the handling mis-predicted offloads in
      software is now done with no locks held.

      Finally, pktgen is extended to have a "burst" parameter that can
      be used to test a multi-send implementation.

      Several drivers have xmit_more support: i40e, igb, ixgbe, mlx4,
      virtio_net

      Adding support is almost trivial, so export more drivers to
      support this optimization soon.

      I want to thank, in no particular or implied order, Jesper
      Dangaard Brouer, Eric Dumazet, Alexander Duyck, Tom Herbert, Jamal
      Hadi Salim, John Fastabend, Florian Westphal, Daniel Borkmann,
      David Tat, Hannes Frederic Sowa, and Rusty Russell.

   2) PTP and timestamping support in bnx2x, from Michal Kalderon.

   3) Allow adjusting the rx_copybreak threshold for a driver via
      ethtool, and add rx_copybreak support to enic driver.  From
      Govindarajulu Varadarajan.

   4) Significant enhancements to the generic PHY layer and the bcm7xxx
      driver in particular (EEE support, auto power down, etc.) from
      Florian Fainelli.

   5) Allow raw buffers to be used for flow dissection, allowing drivers
      to determine the optimal "linear pull" size for devices that DMA
      into pools of pages.  The objective is to get exactly the
      necessary amount of headers into the linear SKB area pre-pulled,
      but no more.  The new interface drivers use is eth_get_headlen().
      From WANG Cong, with driver conversions (several had their own
      by-hand duplicated implementations) by Alexander Duyck and Eric
      Dumazet.

   6) Support checksumming more smoothly and efficiently for
      encapsulations, and add "foo over UDP" facility.  From Tom
      Herbert.

   7) Add Broadcom SF2 switch driver to DSA layer, from Florian
      Fainelli.

   8) eBPF now can load programs via a system call and has an extensive
      testsuite.  Alexei Starovoitov and Daniel Borkmann.

   9) Major overhaul of the packet scheduler to use RCU in several major
      areas such as the classifiers and rate estimators.  From John
      Fastabend.

  10) Add driver for Intel FM10000 Ethernet Switch, from Alexander
      Duyck.

  11) Rearrange TCP_SKB_CB() to reduce cache line misses, from Eric
      Dumazet.

  12) Add Datacenter TCP congestion control algorithm support, From
      Florian Westphal.

  13) Reorganize sk_buff so that __copy_skb_header() is significantly
      faster.  From Eric Dumazet"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1558 commits)
  netlabel: directly return netlbl_unlabel_genl_init()
  net: add netdev_txq_bql_{enqueue, complete}_prefetchw() helpers
  net: description of dma_cookie cause make xmldocs warning
  cxgb4: clean up a type issue
  cxgb4: potential shift wrapping bug
  i40e: skb->xmit_more support
  net: fs_enet: Add NAPI TX
  net: fs_enet: Remove non NAPI RX
  r8169:add support for RTL8168EP
  net_sched: copy exts->type in tcf_exts_change()
  wimax: convert printk to pr_foo()
  af_unix: remove 0 assignment on static
  ipv6: Do not warn for informational ICMP messages, regardless of type.
  Update Intel Ethernet Driver maintainers list
  bridge: Save frag_max_size between PRE_ROUTING and POST_ROUTING
  tipc: fix bug in multicast congestion handling
  net: better IFF_XMIT_DST_RELEASE support
  net/mlx4_en: remove NETDEV_TX_BUSY
  3c59x: fix bad split of cpu_to_le32(pci_map_single())
  net: bcmgenet: fix Tx ring priority programming
  ...

1 files changed, 696 insertions, 940 deletions

diff --git a/drivers/net/wireless/ath/ath10k/pci.c b/drivers/net/wireless/ath/ath10k/pci.c
index 3376963a4862..59e0ea83be50 100644
--- a/drivers/net/wireless/ath/ath10k/pci.c
+++ b/drivers/net/wireless/ath/ath10k/pci.c
@@ -44,13 +44,9 @@ enum ath10k_pci_reset_mode {
 	ATH10K_PCI_RESET_WARM_ONLY = 1,
 };
 
-static unsigned int ath10k_pci_target_ps;
 static unsigned int ath10k_pci_irq_mode = ATH10K_PCI_IRQ_AUTO;
 static unsigned int ath10k_pci_reset_mode = ATH10K_PCI_RESET_AUTO;
 
-module_param_named(target_ps, ath10k_pci_target_ps, uint, 0644);
-MODULE_PARM_DESC(target_ps, "Enable ath10k Target (SoC) PS option");
-
 module_param_named(irq_mode, ath10k_pci_irq_mode, uint, 0644);
 MODULE_PARM_DESC(irq_mode, "0: auto, 1: legacy, 2: msi (default: 0)");
 
@@ -68,13 +64,7 @@ static const struct pci_device_id ath10k_pci_id_table[] = {
 	{0}
 };
 
-static int ath10k_pci_diag_read_access(struct ath10k *ar, u32 address,
-				       u32 *data);
-
-static int ath10k_pci_post_rx(struct ath10k *ar);
-static int ath10k_pci_post_rx_pipe(struct ath10k_pci_pipe *pipe_info,
-					     int num);
-static void ath10k_pci_rx_pipe_cleanup(struct ath10k_pci_pipe *pipe_info);
+static void ath10k_pci_buffer_cleanup(struct ath10k *ar);
 static int ath10k_pci_cold_reset(struct ath10k *ar);
 static int ath10k_pci_warm_reset(struct ath10k *ar);
 static int ath10k_pci_wait_for_target_init(struct ath10k *ar);
@@ -156,79 +146,175 @@ static const struct ce_attr host_ce_config_wlan[] = {
 static const struct ce_pipe_config target_ce_config_wlan[] = {
 	/* CE0: host->target HTC control and raw streams */
 	{
-		.pipenum = 0,
-		.pipedir = PIPEDIR_OUT,
-		.nentries = 32,
-		.nbytes_max = 256,
-		.flags = CE_ATTR_FLAGS,
-		.reserved = 0,
+		.pipenum = __cpu_to_le32(0),
+		.pipedir = __cpu_to_le32(PIPEDIR_OUT),
+		.nentries = __cpu_to_le32(32),
+		.nbytes_max = __cpu_to_le32(256),
+		.flags = __cpu_to_le32(CE_ATTR_FLAGS),
+		.reserved = __cpu_to_le32(0),
 	},
 
 	/* CE1: target->host HTT + HTC control */
 	{
-		.pipenum = 1,
-		.pipedir = PIPEDIR_IN,
-		.nentries = 32,
-		.nbytes_max = 512,
-		.flags = CE_ATTR_FLAGS,
-		.reserved = 0,
+		.pipenum = __cpu_to_le32(1),
+		.pipedir = __cpu_to_le32(PIPEDIR_IN),
+		.nentries = __cpu_to_le32(32),
+		.nbytes_max = __cpu_to_le32(512),
+		.flags = __cpu_to_le32(CE_ATTR_FLAGS),
+		.reserved = __cpu_to_le32(0),
 	},
 
 	/* CE2: target->host WMI */
 	{
-		.pipenum = 2,
-		.pipedir = PIPEDIR_IN,
-		.nentries = 32,
-		.nbytes_max = 2048,
-		.flags = CE_ATTR_FLAGS,
-		.reserved = 0,
+		.pipenum = __cpu_to_le32(2),
+		.pipedir = __cpu_to_le32(PIPEDIR_IN),
+		.nentries = __cpu_to_le32(32),
+		.nbytes_max = __cpu_to_le32(2048),
+		.flags = __cpu_to_le32(CE_ATTR_FLAGS),
+		.reserved = __cpu_to_le32(0),
 	},
 
 	/* CE3: host->target WMI */
 	{
-		.pipenum = 3,
-		.pipedir = PIPEDIR_OUT,
-		.nentries = 32,
-		.nbytes_max = 2048,
-		.flags = CE_ATTR_FLAGS,
-		.reserved = 0,
+		.pipenum = __cpu_to_le32(3),
+		.pipedir = __cpu_to_le32(PIPEDIR_OUT),
+		.nentries = __cpu_to_le32(32),
+		.nbytes_max = __cpu_to_le32(2048),
+		.flags = __cpu_to_le32(CE_ATTR_FLAGS),
+		.reserved = __cpu_to_le32(0),
 	},
 
 	/* CE4: host->target HTT */
 	{
-		.pipenum = 4,
-		.pipedir = PIPEDIR_OUT,
-		.nentries = 256,
-		.nbytes_max = 256,
-		.flags = CE_ATTR_FLAGS,
-		.reserved = 0,
+		.pipenum = __cpu_to_le32(4),
+		.pipedir = __cpu_to_le32(PIPEDIR_OUT),
+		.nentries = __cpu_to_le32(256),
+		.nbytes_max = __cpu_to_le32(256),
+		.flags = __cpu_to_le32(CE_ATTR_FLAGS),
+		.reserved = __cpu_to_le32(0),
 	},
 
 	/* NB: 50% of src nentries, since tx has 2 frags */
 
 	/* CE5: unused */
 	{
-		.pipenum = 5,
-		.pipedir = PIPEDIR_OUT,
-		.nentries = 32,
-		.nbytes_max = 2048,
-		.flags = CE_ATTR_FLAGS,
-		.reserved = 0,
+		.pipenum = __cpu_to_le32(5),
+		.pipedir = __cpu_to_le32(PIPEDIR_OUT),
+		.nentries = __cpu_to_le32(32),
+		.nbytes_max = __cpu_to_le32(2048),
+		.flags = __cpu_to_le32(CE_ATTR_FLAGS),
+		.reserved = __cpu_to_le32(0),
 	},
 
 	/* CE6: Reserved for target autonomous hif_memcpy */
 	{
-		.pipenum = 6,
-		.pipedir = PIPEDIR_INOUT,
-		.nentries = 32,
-		.nbytes_max = 4096,
-		.flags = CE_ATTR_FLAGS,
-		.reserved = 0,
+		.pipenum = __cpu_to_le32(6),
+		.pipedir = __cpu_to_le32(PIPEDIR_INOUT),
+		.nentries = __cpu_to_le32(32),
+		.nbytes_max = __cpu_to_le32(4096),
+		.flags = __cpu_to_le32(CE_ATTR_FLAGS),
+		.reserved = __cpu_to_le32(0),
 	},
 
 	/* CE7 used only by Host */
 };
 
+/*
+ * Map from service/endpoint to Copy Engine.
+ * This table is derived from the CE_PCI TABLE, above.
+ * It is passed to the Target at startup for use by firmware.
+ */
+static const struct service_to_pipe target_service_to_ce_map_wlan[] = {
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_DATA_VO),
+		__cpu_to_le32(PIPEDIR_OUT),	/* out = UL = host -> target */
+		__cpu_to_le32(3),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_DATA_VO),
+		__cpu_to_le32(PIPEDIR_IN),	/* in = DL = target -> host */
+		__cpu_to_le32(2),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_DATA_BK),
+		__cpu_to_le32(PIPEDIR_OUT),	/* out = UL = host -> target */
+		__cpu_to_le32(3),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_DATA_BK),
+		__cpu_to_le32(PIPEDIR_IN),	/* in = DL = target -> host */
+		__cpu_to_le32(2),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_DATA_BE),
+		__cpu_to_le32(PIPEDIR_OUT),	/* out = UL = host -> target */
+		__cpu_to_le32(3),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_DATA_BE),
+		__cpu_to_le32(PIPEDIR_IN),	/* in = DL = target -> host */
+		__cpu_to_le32(2),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_DATA_VI),
+		__cpu_to_le32(PIPEDIR_OUT),	/* out = UL = host -> target */
+		__cpu_to_le32(3),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_DATA_VI),
+		__cpu_to_le32(PIPEDIR_IN),	/* in = DL = target -> host */
+		__cpu_to_le32(2),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_CONTROL),
+		__cpu_to_le32(PIPEDIR_OUT),	/* out = UL = host -> target */
+		__cpu_to_le32(3),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_WMI_CONTROL),
+		__cpu_to_le32(PIPEDIR_IN),	/* in = DL = target -> host */
+		__cpu_to_le32(2),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_RSVD_CTRL),
+		__cpu_to_le32(PIPEDIR_OUT),	/* out = UL = host -> target */
+		__cpu_to_le32(0),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_RSVD_CTRL),
+		__cpu_to_le32(PIPEDIR_IN),	/* in = DL = target -> host */
+		__cpu_to_le32(1),
+	},
+	{ /* not used */
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_TEST_RAW_STREAMS),
+		__cpu_to_le32(PIPEDIR_OUT),	/* out = UL = host -> target */
+		__cpu_to_le32(0),
+	},
+	{ /* not used */
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_TEST_RAW_STREAMS),
+		__cpu_to_le32(PIPEDIR_IN),	/* in = DL = target -> host */
+		__cpu_to_le32(1),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_HTT_DATA_MSG),
+		__cpu_to_le32(PIPEDIR_OUT),	/* out = UL = host -> target */
+		__cpu_to_le32(4),
+	},
+	{
+		__cpu_to_le32(ATH10K_HTC_SVC_ID_HTT_DATA_MSG),
+		__cpu_to_le32(PIPEDIR_IN),	/* in = DL = target -> host */
+		__cpu_to_le32(1),
+	},
+
+	/* (Additions here) */
+
+	{ /* must be last */
+		__cpu_to_le32(0),
+		__cpu_to_le32(0),
+		__cpu_to_le32(0),
+	},
+};
+
 static bool ath10k_pci_irq_pending(struct ath10k *ar)
 {
 	u32 cause;
@@ -254,8 +340,8 @@ static void ath10k_pci_disable_and_clear_legacy_irq(struct ath10k *ar)
 
 	/* IMPORTANT: this extra read transaction is required to
 	 * flush the posted write buffer. */
-	(void) ath10k_pci_read32(ar, SOC_CORE_BASE_ADDRESS +
-				 PCIE_INTR_ENABLE_ADDRESS);
+	(void)ath10k_pci_read32(ar, SOC_CORE_BASE_ADDRESS +
+				PCIE_INTR_ENABLE_ADDRESS);
 }
 
 static void ath10k_pci_enable_legacy_irq(struct ath10k *ar)
@@ -266,48 +352,116 @@ static void ath10k_pci_enable_legacy_irq(struct ath10k *ar)
 
 	/* IMPORTANT: this extra read transaction is required to
 	 * flush the posted write buffer. */
-	(void) ath10k_pci_read32(ar, SOC_CORE_BASE_ADDRESS +
-				 PCIE_INTR_ENABLE_ADDRESS);
+	(void)ath10k_pci_read32(ar, SOC_CORE_BASE_ADDRESS +
+				PCIE_INTR_ENABLE_ADDRESS);
 }
 
-static irqreturn_t ath10k_pci_early_irq_handler(int irq, void *arg)
+static inline const char *ath10k_pci_get_irq_method(struct ath10k *ar)
 {
-	struct ath10k *ar = arg;
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
 
-	if (ar_pci->num_msi_intrs == 0) {
-		if (!ath10k_pci_irq_pending(ar))
-			return IRQ_NONE;
-
-		ath10k_pci_disable_and_clear_legacy_irq(ar);
-	}
+	if (ar_pci->num_msi_intrs > 1)
+		return "msi-x";
 
-	tasklet_schedule(&ar_pci->early_irq_tasklet);
+	if (ar_pci->num_msi_intrs == 1)
+		return "msi";
 
-	return IRQ_HANDLED;
+	return "legacy";
 }
 
-static int ath10k_pci_request_early_irq(struct ath10k *ar)
+static int __ath10k_pci_rx_post_buf(struct ath10k_pci_pipe *pipe)
 {
+	struct ath10k *ar = pipe->hif_ce_state;
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
+	struct ath10k_ce_pipe *ce_pipe = pipe->ce_hdl;
+	struct sk_buff *skb;
+	dma_addr_t paddr;
 	int ret;
 
-	/* Regardless whether MSI-X/MSI/legacy irqs have been set up the first
-	 * interrupt from irq vector is triggered in all cases for FW
-	 * indication/errors */
-	ret = request_irq(ar_pci->pdev->irq, ath10k_pci_early_irq_handler,
-			  IRQF_SHARED, "ath10k_pci (early)", ar);
+	lockdep_assert_held(&ar_pci->ce_lock);
+
+	skb = dev_alloc_skb(pipe->buf_sz);
+	if (!skb)
+		return -ENOMEM;
+
+	WARN_ONCE((unsigned long)skb->data & 3, "unaligned skb");
+
+	paddr = dma_map_single(ar->dev, skb->data,
+			       skb->len + skb_tailroom(skb),
+			       DMA_FROM_DEVICE);
+	if (unlikely(dma_mapping_error(ar->dev, paddr))) {
+		ath10k_warn(ar, "failed to dma map pci rx buf\n");
+		dev_kfree_skb_any(skb);
+		return -EIO;
+	}
+
+	ATH10K_SKB_CB(skb)->paddr = paddr;
+
+	ret = __ath10k_ce_rx_post_buf(ce_pipe, skb, paddr);
 	if (ret) {
-		ath10k_warn("failed to request early irq: %d\n", ret);
+		ath10k_warn(ar, "failed to post pci rx buf: %d\n", ret);
+		dma_unmap_single(ar->dev, paddr, skb->len + skb_tailroom(skb),
+				 DMA_FROM_DEVICE);
+		dev_kfree_skb_any(skb);
 		return ret;
 	}
 
 	return 0;
 }
 
-static void ath10k_pci_free_early_irq(struct ath10k *ar)
+static void __ath10k_pci_rx_post_pipe(struct ath10k_pci_pipe *pipe)
+{
+	struct ath10k *ar = pipe->hif_ce_state;
+	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
+	struct ath10k_ce_pipe *ce_pipe = pipe->ce_hdl;
+	int ret, num;
+
+	lockdep_assert_held(&ar_pci->ce_lock);
+
+	if (pipe->buf_sz == 0)
+		return;
+
+	if (!ce_pipe->dest_ring)
+		return;
+
+	num = __ath10k_ce_rx_num_free_bufs(ce_pipe);
+	while (num--) {
+		ret = __ath10k_pci_rx_post_buf(pipe);
+		if (ret) {
+			ath10k_warn(ar, "failed to post pci rx buf: %d\n", ret);
+			mod_timer(&ar_pci->rx_post_retry, jiffies +
+				  ATH10K_PCI_RX_POST_RETRY_MS);
+			break;
+		}
+	}
+}
+
+static void ath10k_pci_rx_post_pipe(struct ath10k_pci_pipe *pipe)
+{
+	struct ath10k *ar = pipe->hif_ce_state;
+	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
+
+	spin_lock_bh(&ar_pci->ce_lock);
+	__ath10k_pci_rx_post_pipe(pipe);
+	spin_unlock_bh(&ar_pci->ce_lock);
+}
+
+static void ath10k_pci_rx_post(struct ath10k *ar)
+{
+	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
+	int i;
+
+	spin_lock_bh(&ar_pci->ce_lock);
+	for (i = 0; i < CE_COUNT; i++)
+		__ath10k_pci_rx_post_pipe(&ar_pci->pipe_info[i]);
+	spin_unlock_bh(&ar_pci->ce_lock);
+}
+
+static void ath10k_pci_rx_replenish_retry(unsigned long ptr)
 {
-	free_irq(ath10k_pci_priv(ar)->pdev->irq, ar);
+	struct ath10k *ar = (void *)ptr;
+
+	ath10k_pci_rx_post(ar);
 }
 
 /*
@@ -331,25 +485,6 @@ static int ath10k_pci_diag_read_mem(struct ath10k *ar, u32 address, void *data,
 	void *data_buf = NULL;
 	int i;
 
-	/*
-	 * This code cannot handle reads to non-memory space. Redirect to the
-	 * register read fn but preserve the multi word read capability of
-	 * this fn
-	 */
-	if (address < DRAM_BASE_ADDRESS) {
-		if (!IS_ALIGNED(address, 4) ||
-		    !IS_ALIGNED((unsigned long)data, 4))
-			return -EIO;
-
-		while ((nbytes >= 4) &&  ((ret = ath10k_pci_diag_read_access(
-					   ar, address, (u32 *)data)) == 0)) {
-			nbytes -= sizeof(u32);
-			address += sizeof(u32);
-			data += sizeof(u32);
-		}
-		return ret;
-	}
-
 	ce_diag = ar_pci->ce_diag;
 
 	/*
@@ -376,7 +511,7 @@ static int ath10k_pci_diag_read_mem(struct ath10k *ar, u32 address, void *data,
 		nbytes = min_t(unsigned int, remaining_bytes,
 			       DIAG_TRANSFER_LIMIT);
 
-		ret = ath10k_ce_recv_buf_enqueue(ce_diag, NULL, ce_data);
+		ret = ath10k_ce_rx_post_buf(ce_diag, NULL, ce_data);
 		if (ret != 0)
 			goto done;
 
@@ -389,13 +524,11 @@ static int ath10k_pci_diag_read_mem(struct ath10k *ar, u32 address, void *data,
 		 * convert it from Target CPU virtual address space
 		 * to CE address space
 		 */
-		ath10k_pci_wake(ar);
 		address = TARG_CPU_SPACE_TO_CE_SPACE(ar, ar_pci->mem,
 						     address);
-		ath10k_pci_sleep(ar);
 
 		ret = ath10k_ce_send(ce_diag, NULL, (u32)address, nbytes, 0,
-				 0);
+				     0);
 		if (ret)
 			goto done;
 
@@ -415,7 +548,7 @@ static int ath10k_pci_diag_read_mem(struct ath10k *ar, u32 address, void *data,
 			goto done;
 		}
 
-		if (buf != (u32) address) {
+		if (buf != (u32)address) {
 			ret = -EIO;
 			goto done;
 		}
@@ -448,15 +581,10 @@ static int ath10k_pci_diag_read_mem(struct ath10k *ar, u32 address, void *data,
 	}
 
 done:
-	if (ret == 0) {
-		/* Copy data from allocated DMA buf to caller's buf */
-		WARN_ON_ONCE(orig_nbytes & 3);
-		for (i = 0; i < orig_nbytes / sizeof(__le32); i++) {
-			((u32 *)data)[i] =
-				__le32_to_cpu(((__le32 *)data_buf)[i]);
-		}
-	} else
-		ath10k_warn("failed to read diag value at 0x%x: %d\n",
+	if (ret == 0)
+		memcpy(data, data_buf, orig_nbytes);
+	else
+		ath10k_warn(ar, "failed to read diag value at 0x%x: %d\n",
 			    address, ret);
 
 	if (data_buf)
@@ -466,20 +594,45 @@ done:
 	return ret;
 }
 
-/* Read 4-byte aligned data from Target memory or register */
-static int ath10k_pci_diag_read_access(struct ath10k *ar, u32 address,
-				       u32 *data)
+static int ath10k_pci_diag_read32(struct ath10k *ar, u32 address, u32 *value)
 {
-	/* Assume range doesn't cross this boundary */
-	if (address >= DRAM_BASE_ADDRESS)
-		return ath10k_pci_diag_read_mem(ar, address, data, sizeof(u32));
+	__le32 val = 0;
+	int ret;
+
+	ret = ath10k_pci_diag_read_mem(ar, address, &val, sizeof(val));
+	*value = __le32_to_cpu(val);
+
+	return ret;
+}
+
+static int __ath10k_pci_diag_read_hi(struct ath10k *ar, void *dest,
+				     u32 src, u32 len)
+{
+	u32 host_addr, addr;
+	int ret;
+
+	host_addr = host_interest_item_address(src);
+
+	ret = ath10k_pci_diag_read32(ar, host_addr, &addr);
+	if (ret != 0) {
+		ath10k_warn(ar, "failed to get memcpy hi address for firmware address %d: %d\n",
+			    src, ret);
+		return ret;
+	}
+
+	ret = ath10k_pci_diag_read_mem(ar, addr, dest, len);
+	if (ret != 0) {
+		ath10k_warn(ar, "failed to memcpy firmware memory from %d (%d B): %d\n",
+			    addr, len, ret);
+		return ret;
+	}
 
-	ath10k_pci_wake(ar);
-	*data = ath10k_pci_read32(ar, address);
-	ath10k_pci_sleep(ar);
 	return 0;
 }
 
+#define ath10k_pci_diag_read_hi(ar, dest, src, len)		\
+	__ath10k_pci_diag_read_hi(ar, dest, HI_ITEM(src), len)
+
 static int ath10k_pci_diag_write_mem(struct ath10k *ar, u32 address,
 				     const void *data, int nbytes)
 {
@@ -514,9 +667,7 @@ static int ath10k_pci_diag_write_mem(struct ath10k *ar, u32 address,
 	}
 
 	/* Copy caller's data to allocated DMA buf */
-	WARN_ON_ONCE(orig_nbytes & 3);
-	for (i = 0; i < orig_nbytes / sizeof(__le32); i++)
-		((__le32 *)data_buf)[i] = __cpu_to_le32(((u32 *)data)[i]);
+	memcpy(data_buf, data, orig_nbytes);
 
 	/*
 	 * The address supplied by the caller is in the
@@ -528,9 +679,7 @@ static int ath10k_pci_diag_write_mem(struct ath10k *ar, u32 address,
 	 * to
 	 *    CE address space
 	 */
-	ath10k_pci_wake(ar);
 	address = TARG_CPU_SPACE_TO_CE_SPACE(ar, ar_pci->mem, address);
-	ath10k_pci_sleep(ar);
 
 	remaining_bytes = orig_nbytes;
 	ce_data = ce_data_base;
@@ -539,7 +688,7 @@ static int ath10k_pci_diag_write_mem(struct ath10k *ar, u32 address,
 		nbytes = min_t(int, remaining_bytes, DIAG_TRANSFER_LIMIT);
 
 		/* Set up to receive directly into Target(!) address */
-		ret = ath10k_ce_recv_buf_enqueue(ce_diag, NULL, address);
+		ret = ath10k_ce_rx_post_buf(ce_diag, NULL, address);
 		if (ret != 0)
 			goto done;
 
@@ -547,7 +696,7 @@ static int ath10k_pci_diag_write_mem(struct ath10k *ar, u32 address,
 		 * Request CE to send caller-supplied data that
 		 * was copied to bounce buffer to Target(!) address.
 		 */
-		ret = ath10k_ce_send(ce_diag, NULL, (u32) ce_data,
+		ret = ath10k_ce_send(ce_diag, NULL, (u32)ce_data,
 				     nbytes, 0, 0);
 		if (ret != 0)
 			goto done;
@@ -608,66 +757,34 @@ done:
 	}
 
 	if (ret != 0)
-		ath10k_warn("failed to write diag value at 0x%x: %d\n",
+		ath10k_warn(ar, "failed to write diag value at 0x%x: %d\n",
 			    address, ret);
 
 	return ret;
 }
 
-/* Write 4B data to Target memory or register */
-static int ath10k_pci_diag_write_access(struct ath10k *ar, u32 address,
-					u32 data)
+static int ath10k_pci_diag_write32(struct ath10k *ar, u32 address, u32 value)
 {
-	/* Assume range doesn't cross this boundary */
-	if (address >= DRAM_BASE_ADDRESS)
-		return ath10k_pci_diag_write_mem(ar, address, &data,
-						 sizeof(u32));
+	__le32 val = __cpu_to_le32(value);
 
-	ath10k_pci_wake(ar);
-	ath10k_pci_write32(ar, address, data);
-	ath10k_pci_sleep(ar);
-	return 0;
+	return ath10k_pci_diag_write_mem(ar, address, &val, sizeof(val));
 }
 
-static bool ath10k_pci_target_is_awake(struct ath10k *ar)
+static bool ath10k_pci_is_awake(struct ath10k *ar)
 {
-	void __iomem *mem = ath10k_pci_priv(ar)->mem;
-	u32 val;
-	val = ioread32(mem + PCIE_LOCAL_BASE_ADDRESS +
-		       RTC_STATE_ADDRESS);
-	return (RTC_STATE_V_GET(val) == RTC_STATE_V_ON);
+	u32 val = ath10k_pci_reg_read32(ar, RTC_STATE_ADDRESS);
+
+	return RTC_STATE_V_GET(val) == RTC_STATE_V_ON;
 }
 
-int ath10k_do_pci_wake(struct ath10k *ar)
+static int ath10k_pci_wake_wait(struct ath10k *ar)
 {
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	void __iomem *pci_addr = ar_pci->mem;
 	int tot_delay = 0;
 	int curr_delay = 5;
 
-	if (atomic_read(&ar_pci->keep_awake_count) == 0) {
-		/* Force AWAKE */
-		iowrite32(PCIE_SOC_WAKE_V_MASK,
-			  pci_addr + PCIE_LOCAL_BASE_ADDRESS +
-			  PCIE_SOC_WAKE_ADDRESS);
-	}
-	atomic_inc(&ar_pci->keep_awake_count);
-
-	if (ar_pci->verified_awake)
-		return 0;
-
-	for (;;) {
-		if (ath10k_pci_target_is_awake(ar)) {
-			ar_pci->verified_awake = true;
+	while (tot_delay < PCIE_WAKE_TIMEOUT) {
+		if (ath10k_pci_is_awake(ar))
 			return 0;
-		}
-
-		if (tot_delay > PCIE_WAKE_TIMEOUT) {
-			ath10k_warn("target took longer %d us to wake up (awake count %d)\n",
-				    PCIE_WAKE_TIMEOUT,
-				    atomic_read(&ar_pci->keep_awake_count));
-			return -ETIMEDOUT;
-		}
 
 		udelay(curr_delay);
 		tot_delay += curr_delay;
@@ -675,20 +792,21 @@ int ath10k_do_pci_wake(struct ath10k *ar)
 		if (curr_delay < 50)
 			curr_delay += 5;
 	}
+
+	return -ETIMEDOUT;
 }
 
-void ath10k_do_pci_sleep(struct ath10k *ar)
+static int ath10k_pci_wake(struct ath10k *ar)
 {
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	void __iomem *pci_addr = ar_pci->mem;
+	ath10k_pci_reg_write32(ar, PCIE_SOC_WAKE_ADDRESS,
+			       PCIE_SOC_WAKE_V_MASK);
+	return ath10k_pci_wake_wait(ar);
+}
 
-	if (atomic_dec_and_test(&ar_pci->keep_awake_count)) {
-		/* Allow sleep */
-		ar_pci->verified_awake = false;
-		iowrite32(PCIE_SOC_WAKE_RESET,
-			  pci_addr + PCIE_LOCAL_BASE_ADDRESS +
-			  PCIE_SOC_WAKE_ADDRESS);
-	}
+static void ath10k_pci_sleep(struct ath10k *ar)
+{
+	ath10k_pci_reg_write32(ar, PCIE_SOC_WAKE_ADDRESS,
+			       PCIE_SOC_WAKE_RESET);
 }
 
 /* Called by lower (CE) layer when a send to Target completes. */
@@ -726,19 +844,17 @@ static void ath10k_pci_ce_recv_data(struct ath10k_ce_pipe *ce_state)
 	unsigned int nbytes, max_nbytes;
 	unsigned int transfer_id;
 	unsigned int flags;
-	int err, num_replenish = 0;
 
 	while (ath10k_ce_completed_recv_next(ce_state, &transfer_context,
 					     &ce_data, &nbytes, &transfer_id,
 					     &flags) == 0) {
-		num_replenish++;
 		skb = transfer_context;
 		max_nbytes = skb->len + skb_tailroom(skb);
 		dma_unmap_single(ar->dev, ATH10K_SKB_CB(skb)->paddr,
 				 max_nbytes, DMA_FROM_DEVICE);
 
 		if (unlikely(max_nbytes < nbytes)) {
-			ath10k_warn("rxed more than expected (nbytes %d, max %d)",
+			ath10k_warn(ar, "rxed more than expected (nbytes %d, max %d)",
 				    nbytes, max_nbytes);
 			dev_kfree_skb_any(skb);
 			continue;
@@ -748,12 +864,7 @@ static void ath10k_pci_ce_recv_data(struct ath10k_ce_pipe *ce_state)
 		cb->rx_completion(ar, skb, pipe_info->pipe_num);
 	}
 
-	err = ath10k_pci_post_rx_pipe(pipe_info, num_replenish);
-	if (unlikely(err)) {
-		/* FIXME: retry */
-		ath10k_warn("failed to replenish CE rx ring %d (%d bufs): %d\n",
-			    pipe_info->pipe_num, num_replenish, err);
-	}
+	ath10k_pci_rx_post_pipe(pipe_info);
 }
 
 static int ath10k_pci_hif_tx_sg(struct ath10k *ar, u8 pipe_id,
@@ -781,10 +892,10 @@ static int ath10k_pci_hif_tx_sg(struct ath10k *ar, u8 pipe_id,
 	}
 
 	for (i = 0; i < n_items - 1; i++) {
-		ath10k_dbg(ATH10K_DBG_PCI,
+		ath10k_dbg(ar, ATH10K_DBG_PCI,
 			   "pci tx item %d paddr 0x%08x len %d n_items %d\n",
 			   i, items[i].paddr, items[i].len, n_items);
-		ath10k_dbg_dump(ATH10K_DBG_PCI_DUMP, NULL, "item data: ",
+		ath10k_dbg_dump(ar, ATH10K_DBG_PCI_DUMP, NULL, "pci tx data: ",
 				items[i].vaddr, items[i].len);
 
 		err = ath10k_ce_send_nolock(ce_pipe,
@@ -799,10 +910,10 @@ static int ath10k_pci_hif_tx_sg(struct ath10k *ar, u8 pipe_id,
 
 	/* `i` is equal to `n_items -1` after for() */
 
-	ath10k_dbg(ATH10K_DBG_PCI,
+	ath10k_dbg(ar, ATH10K_DBG_PCI,
 		   "pci tx item %d paddr 0x%08x len %d n_items %d\n",
 		   i, items[i].paddr, items[i].len, n_items);
-	ath10k_dbg_dump(ATH10K_DBG_PCI_DUMP, NULL, "item data: ",
+	ath10k_dbg_dump(ar, ATH10K_DBG_PCI_DUMP, NULL, "pci tx data: ",
 			items[i].vaddr, items[i].len);
 
 	err = ath10k_ce_send_nolock(ce_pipe,
@@ -829,52 +940,64 @@ static u16 ath10k_pci_hif_get_free_queue_number(struct ath10k *ar, u8 pipe)
 {
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
 
-	ath10k_dbg(ATH10K_DBG_PCI, "pci hif get free queue number\n");
+	ath10k_dbg(ar, ATH10K_DBG_PCI, "pci hif get free queue number\n");
 
 	return ath10k_ce_num_free_src_entries(ar_pci->pipe_info[pipe].ce_hdl);
 }
 
-static void ath10k_pci_hif_dump_area(struct ath10k *ar)
+static void ath10k_pci_dump_registers(struct ath10k *ar,
+				      struct ath10k_fw_crash_data *crash_data)
 {
-	u32 reg_dump_area = 0;
-	u32 reg_dump_values[REG_DUMP_COUNT_QCA988X] = {};
-	u32 host_addr;
-	int ret;
-	u32 i;
+	__le32 reg_dump_values[REG_DUMP_COUNT_QCA988X] = {};
+	int i, ret;
 
-	ath10k_err("firmware crashed!\n");
-	ath10k_err("hardware name %s version 0x%x\n",
-		   ar->hw_params.name, ar->target_version);
-	ath10k_err("firmware version: %s\n", ar->hw->wiphy->fw_version);
+	lockdep_assert_held(&ar->data_lock);
 
-	host_addr = host_interest_item_address(HI_ITEM(hi_failure_state));
-	ret = ath10k_pci_diag_read_mem(ar, host_addr,
-				       &reg_dump_area, sizeof(u32));
+	ret = ath10k_pci_diag_read_hi(ar, &reg_dump_values[0],
+				      hi_failure_state,
+				      REG_DUMP_COUNT_QCA988X * sizeof(__le32));
 	if (ret) {
-		ath10k_err("failed to read FW dump area address: %d\n", ret);
-		return;
-	}
-
-	ath10k_err("target register Dump Location: 0x%08X\n", reg_dump_area);
-
-	ret = ath10k_pci_diag_read_mem(ar, reg_dump_area,
-				       &reg_dump_values[0],
-				       REG_DUMP_COUNT_QCA988X * sizeof(u32));
-	if (ret != 0) {
-		ath10k_err("failed to read FW dump area: %d\n", ret);
+		ath10k_err(ar, "failed to read firmware dump area: %d\n", ret);
 		return;
 	}
 
 	BUILD_BUG_ON(REG_DUMP_COUNT_QCA988X % 4);
 
-	ath10k_err("target Register Dump\n");
+	ath10k_err(ar, "firmware register dump:\n");
 	for (i = 0; i < REG_DUMP_COUNT_QCA988X; i += 4)
-		ath10k_err("[%02d]: 0x%08X 0x%08X 0x%08X 0x%08X\n",
+		ath10k_err(ar, "[%02d]: 0x%08X 0x%08X 0x%08X 0x%08X\n",
 			   i,
-			   reg_dump_values[i],
-			   reg_dump_values[i + 1],
-			   reg_dump_values[i + 2],
-			   reg_dump_values[i + 3]);
+			   __le32_to_cpu(reg_dump_values[i]),
+			   __le32_to_cpu(reg_dump_values[i + 1]),
+			   __le32_to_cpu(reg_dump_values[i + 2]),
+			   __le32_to_cpu(reg_dump_values[i + 3]));
+
+	if (!crash_data)
+		return;
+
+	for (i = 0; i < REG_DUMP_COUNT_QCA988X; i++)
+		crash_data->registers[i] = reg_dump_values[i];
+}
+
+static void ath10k_pci_fw_crashed_dump(struct ath10k *ar)
+{
+	struct ath10k_fw_crash_data *crash_data;
+	char uuid[50];
+
+	spin_lock_bh(&ar->data_lock);
+
+	crash_data = ath10k_debug_get_new_fw_crash_data(ar);
+
+	if (crash_data)
+		scnprintf(uuid, sizeof(uuid), "%pUl", &crash_data->uuid);
+	else
+		scnprintf(uuid, sizeof(uuid), "n/a");
+
+	ath10k_err(ar, "firmware crashed! (uuid %s)\n", uuid);
+	ath10k_print_driver_info(ar);
+	ath10k_pci_dump_registers(ar, crash_data);
+
+	spin_unlock_bh(&ar->data_lock);
 
 	queue_work(ar->workqueue, &ar->restart_work);
 }
@@ -882,7 +1005,7 @@ static void ath10k_pci_hif_dump_area(struct ath10k *ar)
 static void ath10k_pci_hif_send_complete_check(struct ath10k *ar, u8 pipe,
 					       int force)
 {
-	ath10k_dbg(ATH10K_DBG_PCI, "pci hif send complete check\n");
+	ath10k_dbg(ar, ATH10K_DBG_PCI, "pci hif send complete check\n");
 
 	if (!force) {
 		int resources;
@@ -910,43 +1033,12 @@ static void ath10k_pci_hif_set_callbacks(struct ath10k *ar,
 {
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
 
-	ath10k_dbg(ATH10K_DBG_PCI, "pci hif set callbacks\n");
+	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 int ath10k_pci_setup_ce_irq(struct ath10k *ar)
-{
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	const struct ce_attr *attr;
-	struct ath10k_pci_pipe *pipe_info;
-	int pipe_num, disable_interrupts;
-
-	for (pipe_num = 0; pipe_num < CE_COUNT; pipe_num++) {
-		pipe_info = &ar_pci->pipe_info[pipe_num];
-
-		/* Handle Diagnostic CE specially */
-		if (pipe_info->ce_hdl == ar_pci->ce_diag)
-			continue;
-
-		attr = &host_ce_config_wlan[pipe_num];
-
-		if (attr->src_nentries) {
-			disable_interrupts = attr->flags & CE_ATTR_DIS_INTR;
-			ath10k_ce_send_cb_register(pipe_info->ce_hdl,
-						   ath10k_pci_ce_send_done,
-						   disable_interrupts);
-		}
-
-		if (attr->dest_nentries)
-			ath10k_ce_recv_cb_register(pipe_info->ce_hdl,
-						   ath10k_pci_ce_recv_data);
-	}
-
-	return 0;
-}
-
 static void ath10k_pci_kill_tasklet(struct ath10k *ar)
 {
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
@@ -954,82 +1046,72 @@ static void ath10k_pci_kill_tasklet(struct ath10k *ar)
 
 	tasklet_kill(&ar_pci->intr_tq);
 	tasklet_kill(&ar_pci->msi_fw_err);
-	tasklet_kill(&ar_pci->early_irq_tasklet);
 
 	for (i = 0; i < CE_COUNT; i++)
 		tasklet_kill(&ar_pci->pipe_info[i].intr);
+
+	del_timer_sync(&ar_pci->rx_post_retry);
 }
 
-/* TODO - temporary mapping while we have too few CE's */
 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)
 {
-	int ret = 0;
+	const struct service_to_pipe *entry;
+	bool ul_set = false, dl_set = false;
+	int i;
 
-	ath10k_dbg(ATH10K_DBG_PCI, "pci hif map service\n");
+	ath10k_dbg(ar, ATH10K_DBG_PCI, "pci hif map service\n");
 
 	/* polling for received messages not supported */
 	*dl_is_polled = 0;
 
-	switch (service_id) {
-	case ATH10K_HTC_SVC_ID_HTT_DATA_MSG:
-		/*
-		 * Host->target HTT gets its own pipe, so it can be polled
-		 * while other pipes are interrupt driven.
-		 */
-		*ul_pipe = 4;
-		/*
-		 * Use the same target->host pipe for HTC ctrl, HTC raw
-		 * streams, and HTT.
-		 */
-		*dl_pipe = 1;
-		break;
-
-	case ATH10K_HTC_SVC_ID_RSVD_CTRL:
-	case ATH10K_HTC_SVC_ID_TEST_RAW_STREAMS:
-		/*
-		 * Note: HTC_RAW_STREAMS_SVC is currently unused, and
-		 * HTC_CTRL_RSVD_SVC could share the same pipe as the
-		 * WMI services.  So, if another CE is needed, change
-		 * this to *ul_pipe = 3, which frees up CE 0.
-		 */
-		/* *ul_pipe = 3; */
-		*ul_pipe = 0;
-		*dl_pipe = 1;
-		break;
+	for (i = 0; i < ARRAY_SIZE(target_service_to_ce_map_wlan); i++) {
+		entry = &target_service_to_ce_map_wlan[i];
 
-	case ATH10K_HTC_SVC_ID_WMI_DATA_BK:
-	case ATH10K_HTC_SVC_ID_WMI_DATA_BE:
-	case ATH10K_HTC_SVC_ID_WMI_DATA_VI:
-	case ATH10K_HTC_SVC_ID_WMI_DATA_VO:
+		if (__le32_to_cpu(entry->service_id) != service_id)
+			continue;
 
-	case ATH10K_HTC_SVC_ID_WMI_CONTROL:
-		*ul_pipe = 3;
-		*dl_pipe = 2;
-		break;
+		switch (__le32_to_cpu(entry->pipedir)) {
+		case PIPEDIR_NONE:
+			break;
+		case PIPEDIR_IN:
+			WARN_ON(dl_set);
+			*dl_pipe = __le32_to_cpu(entry->pipenum);
+			dl_set = true;
+			break;
+		case PIPEDIR_OUT:
+			WARN_ON(ul_set);
+			*ul_pipe = __le32_to_cpu(entry->pipenum);
+			ul_set = true;
+			break;
+		case PIPEDIR_INOUT:
+			WARN_ON(dl_set);
+			WARN_ON(ul_set);
+			*dl_pipe = __le32_to_cpu(entry->pipenum);
+			*ul_pipe = __le32_to_cpu(entry->pipenum);
+			dl_set = true;
+			ul_set = true;
+			break;
+		}
+	}
 
-		/* pipe 5 unused   */
-		/* pipe 6 reserved */
-		/* pipe 7 reserved */
+	if (WARN_ON(!ul_set || !dl_set))
+		return -ENOENT;
 
-	default:
-		ret = -1;
-		break;
-	}
 	*ul_is_polled =
 		(host_ce_config_wlan[*ul_pipe].flags & CE_ATTR_DIS_INTR) != 0;
 
-	return ret;
+	return 0;
 }
 
 static void ath10k_pci_hif_get_default_pipe(struct ath10k *ar,
-						u8 *ul_pipe, u8 *dl_pipe)
+					    u8 *ul_pipe, u8 *dl_pipe)
 {
 	int ul_is_polled, dl_is_polled;
 
-	ath10k_dbg(ATH10K_DBG_PCI, "pci hif get default pipe\n");
+	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,
@@ -1039,141 +1121,34 @@ static void ath10k_pci_hif_get_default_pipe(struct ath10k *ar,
 						 &dl_is_polled);
 }
 
-static int ath10k_pci_post_rx_pipe(struct ath10k_pci_pipe *pipe_info,
-				   int num)
+static void ath10k_pci_irq_disable(struct ath10k *ar)
 {
-	struct ath10k *ar = pipe_info->hif_ce_state;
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	struct ath10k_ce_pipe *ce_state = pipe_info->ce_hdl;
-	struct sk_buff *skb;
-	dma_addr_t ce_data;
-	int i, ret = 0;
-
-	if (pipe_info->buf_sz == 0)
-		return 0;
-
-	for (i = 0; i < num; i++) {
-		skb = dev_alloc_skb(pipe_info->buf_sz);
-		if (!skb) {
-			ath10k_warn("failed to allocate skbuff for pipe %d\n",
-				    num);
-			ret = -ENOMEM;
-			goto err;
-		}
-
-		WARN_ONCE((unsigned long)skb->data & 3, "unaligned skb");
-
-		ce_data = dma_map_single(ar->dev, skb->data,
-					 skb->len + skb_tailroom(skb),
-					 DMA_FROM_DEVICE);
-
-		if (unlikely(dma_mapping_error(ar->dev, ce_data))) {
-			ath10k_warn("failed to DMA map sk_buff\n");
-			dev_kfree_skb_any(skb);
-			ret = -EIO;
-			goto err;
-		}
-
-		ATH10K_SKB_CB(skb)->paddr = ce_data;
-
-		pci_dma_sync_single_for_device(ar_pci->pdev, ce_data,
-					       pipe_info->buf_sz,
-					       PCI_DMA_FROMDEVICE);
-
-		ret = ath10k_ce_recv_buf_enqueue(ce_state, (void *)skb,
-						 ce_data);
-		if (ret) {
-			ath10k_warn("failed to enqueue to pipe %d: %d\n",
-				    num, ret);
-			goto err;
-		}
-	}
+	int i;
 
-	return ret;
+	ath10k_ce_disable_interrupts(ar);
+	ath10k_pci_disable_and_clear_legacy_irq(ar);
+	/* FIXME: How to mask all MSI interrupts? */
 
-err:
-	ath10k_pci_rx_pipe_cleanup(pipe_info);
-	return ret;
+	for (i = 0; i < max(1, ar_pci->num_msi_intrs); i++)
+		synchronize_irq(ar_pci->pdev->irq + i);
 }
 
-static int ath10k_pci_post_rx(struct ath10k *ar)
+static void ath10k_pci_irq_enable(struct ath10k *ar)
 {
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	struct ath10k_pci_pipe *pipe_info;
-	const struct ce_attr *attr;
-	int pipe_num, ret = 0;
-
-	for (pipe_num = 0; pipe_num < CE_COUNT; pipe_num++) {
-		pipe_info = &ar_pci->pipe_info[pipe_num];
-		attr = &host_ce_config_wlan[pipe_num];
-
-		if (attr->dest_nentries == 0)
-			continue;
-
-		ret = ath10k_pci_post_rx_pipe(pipe_info,
-					      attr->dest_nentries - 1);
-		if (ret) {
-			ath10k_warn("failed to post RX buffer for pipe %d: %d\n",
-				    pipe_num, ret);
-
-			for (; pipe_num >= 0; pipe_num--) {
-				pipe_info = &ar_pci->pipe_info[pipe_num];
-				ath10k_pci_rx_pipe_cleanup(pipe_info);
-			}
-			return ret;
-		}
-	}
-
-	return 0;
+	ath10k_ce_enable_interrupts(ar);
+	ath10k_pci_enable_legacy_irq(ar);
+	/* FIXME: How to unmask all MSI interrupts? */
 }
 
 static int ath10k_pci_hif_start(struct ath10k *ar)
 {
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	int ret, ret_early;
-
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot hif start\n");
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot hif start\n");
 
-	ath10k_pci_free_early_irq(ar);
-	ath10k_pci_kill_tasklet(ar);
+	ath10k_pci_irq_enable(ar);
+	ath10k_pci_rx_post(ar);
 
-	ret = ath10k_pci_request_irq(ar);
-	if (ret) {
-		ath10k_warn("failed to post RX buffers for all pipes: %d\n",
-			    ret);
-		goto err_early_irq;
-	}
-
-	ret = ath10k_pci_setup_ce_irq(ar);
-	if (ret) {
-		ath10k_warn("failed to setup CE interrupts: %d\n", ret);
-		goto err_stop;
-	}
-
-	/* Post buffers once to start things off. */
-	ret = ath10k_pci_post_rx(ar);
-	if (ret) {
-		ath10k_warn("failed to post RX buffers for all pipes: %d\n",
-			    ret);
-		goto err_stop;
-	}
-
-	ar_pci->started = 1;
 	return 0;
-
-err_stop:
-	ath10k_ce_disable_interrupts(ar);
-	ath10k_pci_free_irq(ar);
-	ath10k_pci_kill_tasklet(ar);
-err_early_irq:
-	/* Though there should be no interrupts (device was reset)
-	 * power_down() expects the early IRQ to be installed as per the
-	 * driver lifecycle. */
-	ret_early = ath10k_pci_request_early_irq(ar);
-	if (ret_early)
-		ath10k_warn("failed to re-enable early irq: %d\n", ret_early);
-
-	return ret;
 }
 
 static void ath10k_pci_rx_pipe_cleanup(struct ath10k_pci_pipe *pipe_info)
@@ -1193,10 +1168,6 @@ static void ath10k_pci_rx_pipe_cleanup(struct ath10k_pci_pipe *pipe_info)
 
 	ar = pipe_info->hif_ce_state;
 	ar_pci = ath10k_pci_priv(ar);
-
-	if (!ar_pci->started)
-		return;
-
 	ce_hdl = pipe_info->ce_hdl;
 
 	while (ath10k_ce_revoke_recv_next(ce_hdl, (void **)&netbuf,
@@ -1227,10 +1198,6 @@ static void ath10k_pci_tx_pipe_cleanup(struct ath10k_pci_pipe *pipe_info)
 
 	ar = pipe_info->hif_ce_state;
 	ar_pci = ath10k_pci_priv(ar);
-
-	if (!ar_pci->started)
-		return;
-
 	ce_hdl = pipe_info->ce_hdl;
 
 	while (ath10k_ce_cancel_send_next(ce_hdl, (void **)&netbuf,
@@ -1275,41 +1242,31 @@ static void ath10k_pci_ce_deinit(struct ath10k *ar)
 		ath10k_ce_deinit_pipe(ar, i);
 }
 
-static void ath10k_pci_hif_stop(struct ath10k *ar)
+static void ath10k_pci_flush(struct ath10k *ar)
 {
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	int ret;
-
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot hif stop\n");
-
-	if (WARN_ON(!ar_pci->started))
-		return;
-
-	ret = ath10k_ce_disable_interrupts(ar);
-	if (ret)
-		ath10k_warn("failed to disable CE interrupts: %d\n", ret);
-
-	ath10k_pci_free_irq(ar);
 	ath10k_pci_kill_tasklet(ar);
-
-	ret = ath10k_pci_request_early_irq(ar);
-	if (ret)
-		ath10k_warn("failed to re-enable early irq: %d\n", ret);
-
-	/* At this point, asynchronous threads are stopped, the target should
-	 * not DMA nor interrupt. We process the leftovers and then free
-	 * everything else up. */
-
 	ath10k_pci_buffer_cleanup(ar);
+}
 
-	/* Make the sure the device won't access any structures on the host by
-	 * resetting it. The device was fed with PCI CE ringbuffer
-	 * configuration during init. If ringbuffers are freed and the device
-	 * were to access them this could lead to memory corruption on the
-	 * host. */
+static void ath10k_pci_hif_stop(struct ath10k *ar)
+{
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot hif stop\n");
+
+	/* Most likely the device has HTT Rx ring configured. The only way to
+	 * prevent the device from accessing (and possible corrupting) host
+	 * memory is to reset the chip now.
+	 *
+	 * There's also no known way of masking MSI interrupts on the device.
+	 * For ranged MSI the CE-related interrupts can be masked. However
+	 * regardless how many MSI interrupts are assigned the first one
+	 * is always used for firmware indications (crashes) and cannot be
+	 * masked. To prevent the device from asserting the interrupt reset it
+	 * before proceeding with cleanup.
+	 */
 	ath10k_pci_warm_reset(ar);
 
-	ar_pci->started = 0;
+	ath10k_pci_irq_disable(ar);
+	ath10k_pci_flush(ar);
 }
 
 static int ath10k_pci_hif_exchange_bmi_msg(struct ath10k *ar,
@@ -1360,7 +1317,7 @@ static int ath10k_pci_hif_exchange_bmi_msg(struct ath10k *ar,
 		xfer.wait_for_resp = true;
 		xfer.resp_len = 0;
 
-		ath10k_ce_recv_buf_enqueue(ce_rx, &xfer, resp_paddr);
+		ath10k_ce_rx_post_buf(ce_rx, &xfer, resp_paddr);
 	}
 
 	ret = ath10k_ce_send(ce_tx, &xfer, req_paddr, req_len, -1, 0);
@@ -1418,6 +1375,7 @@ static void ath10k_pci_bmi_send_done(struct ath10k_ce_pipe *ce_state)
 
 static void ath10k_pci_bmi_recv_data(struct ath10k_ce_pipe *ce_state)
 {
+	struct ath10k *ar = ce_state->ar;
 	struct bmi_xfer *xfer;
 	u32 ce_data;
 	unsigned int nbytes;
@@ -1429,7 +1387,7 @@ static void ath10k_pci_bmi_recv_data(struct ath10k_ce_pipe *ce_state)
 		return;
 
 	if (!xfer->wait_for_resp) {
-		ath10k_warn("unexpected: BMI data received; ignoring\n");
+		ath10k_warn(ar, "unexpected: BMI data received; ignoring\n");
 		return;
 	}
 
@@ -1457,129 +1415,17 @@ static int ath10k_pci_bmi_wait(struct ath10k_ce_pipe *tx_pipe,
 }
 
 /*
- * Map from service/endpoint to Copy Engine.
- * This table is derived from the CE_PCI TABLE, above.
- * It is passed to the Target at startup for use by firmware.
- */
-static const struct service_to_pipe target_service_to_ce_map_wlan[] = {
-	{
-		 ATH10K_HTC_SVC_ID_WMI_DATA_VO,
-		 PIPEDIR_OUT,		/* out = UL = host -> target */
-		 3,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_WMI_DATA_VO,
-		 PIPEDIR_IN,		/* in = DL = target -> host */
-		 2,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_WMI_DATA_BK,
-		 PIPEDIR_OUT,		/* out = UL = host -> target */
-		 3,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_WMI_DATA_BK,
-		 PIPEDIR_IN,		/* in = DL = target -> host */
-		 2,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_WMI_DATA_BE,
-		 PIPEDIR_OUT,		/* out = UL = host -> target */
-		 3,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_WMI_DATA_BE,
-		 PIPEDIR_IN,		/* in = DL = target -> host */
-		 2,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_WMI_DATA_VI,
-		 PIPEDIR_OUT,		/* out = UL = host -> target */
-		 3,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_WMI_DATA_VI,
-		 PIPEDIR_IN,		/* in = DL = target -> host */
-		 2,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_WMI_CONTROL,
-		 PIPEDIR_OUT,		/* out = UL = host -> target */
-		 3,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_WMI_CONTROL,
-		 PIPEDIR_IN,		/* in = DL = target -> host */
-		 2,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_RSVD_CTRL,
-		 PIPEDIR_OUT,		/* out = UL = host -> target */
-		 0,		/* could be moved to 3 (share with WMI) */
-	},
-	{
-		 ATH10K_HTC_SVC_ID_RSVD_CTRL,
-		 PIPEDIR_IN,		/* in = DL = target -> host */
-		 1,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_TEST_RAW_STREAMS,	/* not currently used */
-		 PIPEDIR_OUT,		/* out = UL = host -> target */
-		 0,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_TEST_RAW_STREAMS,	/* not currently used */
-		 PIPEDIR_IN,		/* in = DL = target -> host */
-		 1,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_HTT_DATA_MSG,
-		 PIPEDIR_OUT,		/* out = UL = host -> target */
-		 4,
-	},
-	{
-		 ATH10K_HTC_SVC_ID_HTT_DATA_MSG,
-		 PIPEDIR_IN,		/* in = DL = target -> host */
-		 1,
-	},
-
-	/* (Additions here) */
-
-	{				/* Must be last */
-		 0,
-		 0,
-		 0,
-	},
-};
-
-/*
  * Send an interrupt to the device to wake up the Target CPU
  * so it has an opportunity to notice any changed state.
  */
 static int ath10k_pci_wake_target_cpu(struct ath10k *ar)
 {
-	int ret;
-	u32 core_ctrl;
+	u32 addr, val;
 
-	ret = ath10k_pci_diag_read_access(ar, SOC_CORE_BASE_ADDRESS |
-					      CORE_CTRL_ADDRESS,
-					  &core_ctrl);
-	if (ret) {
-		ath10k_warn("failed to read core_ctrl: %d\n", ret);
-		return ret;
-	}
-
-	/* A_INUM_FIRMWARE interrupt to Target CPU */
-	core_ctrl |= CORE_CTRL_CPU_INTR_MASK;
-
-	ret = ath10k_pci_diag_write_access(ar, SOC_CORE_BASE_ADDRESS |
-					       CORE_CTRL_ADDRESS,
-					   core_ctrl);
-	if (ret) {
-		ath10k_warn("failed to set target CPU interrupt mask: %d\n",
-			    ret);
-		return ret;
-	}
+	addr = SOC_CORE_BASE_ADDRESS | CORE_CTRL_ADDRESS;
+	val = ath10k_pci_read32(ar, addr);
+	val |= CORE_CTRL_CPU_INTR_MASK;
+	ath10k_pci_write32(ar, addr, val);
 
 	return 0;
 }
@@ -1602,92 +1448,92 @@ static int ath10k_pci_init_config(struct ath10k *ar)
 		host_interest_item_address(HI_ITEM(hi_interconnect_state));
 
 	/* Supply Target-side CE configuration */
-	ret = ath10k_pci_diag_read_access(ar, interconnect_targ_addr,
-					  &pcie_state_targ_addr);
+	ret = ath10k_pci_diag_read32(ar, interconnect_targ_addr,
+				     &pcie_state_targ_addr);
 	if (ret != 0) {
-		ath10k_err("Failed to get pcie state addr: %d\n", ret);
+		ath10k_err(ar, "Failed to get pcie state addr: %d\n", ret);
 		return ret;
 	}
 
 	if (pcie_state_targ_addr == 0) {
 		ret = -EIO;
-		ath10k_err("Invalid pcie state addr\n");
+		ath10k_err(ar, "Invalid pcie state addr\n");
 		return ret;
 	}
 
-	ret = ath10k_pci_diag_read_access(ar, pcie_state_targ_addr +
+	ret = ath10k_pci_diag_read32(ar, (pcie_state_targ_addr +
 					  offsetof(struct pcie_state,
-						   pipe_cfg_addr),
-					  &pipe_cfg_targ_addr);
+						   pipe_cfg_addr)),
+				     &pipe_cfg_targ_addr);
 	if (ret != 0) {
-		ath10k_err("Failed to get pipe cfg addr: %d\n", ret);
+		ath10k_err(ar, "Failed to get pipe cfg addr: %d\n", ret);
 		return ret;
 	}
 
 	if (pipe_cfg_targ_addr == 0) {
 		ret = -EIO;
-		ath10k_err("Invalid pipe cfg addr\n");
+		ath10k_err(ar, "Invalid pipe cfg addr\n");
 		return ret;
 	}
 
 	ret = ath10k_pci_diag_write_mem(ar, pipe_cfg_targ_addr,
-				 target_ce_config_wlan,
-				 sizeof(target_ce_config_wlan));
+					target_ce_config_wlan,
+					sizeof(target_ce_config_wlan));
 
 	if (ret != 0) {
-		ath10k_err("Failed to write pipe cfg: %d\n", ret);
+		ath10k_err(ar, "Failed to write pipe cfg: %d\n", ret);
 		return ret;
 	}
 
-	ret = ath10k_pci_diag_read_access(ar, pcie_state_targ_addr +
+	ret = ath10k_pci_diag_read32(ar, (pcie_state_targ_addr +
 					  offsetof(struct pcie_state,
-						   svc_to_pipe_map),
-					  &svc_to_pipe_map);
+						   svc_to_pipe_map)),
+				     &svc_to_pipe_map);
 	if (ret != 0) {
-		ath10k_err("Failed to get svc/pipe map: %d\n", ret);
+		ath10k_err(ar, "Failed to get svc/pipe map: %d\n", ret);
 		return ret;
 	}
 
 	if (svc_to_pipe_map == 0) {
 		ret = -EIO;
-		ath10k_err("Invalid svc_to_pipe map\n");
+		ath10k_err(ar, "Invalid svc_to_pipe map\n");
 		return ret;
 	}
 
 	ret = ath10k_pci_diag_write_mem(ar, svc_to_pipe_map,
-				 target_service_to_ce_map_wlan,
-				 sizeof(target_service_to_ce_map_wlan));
+					target_service_to_ce_map_wlan,
+					sizeof(target_service_to_ce_map_wlan));
 	if (ret != 0) {
-		ath10k_err("Failed to write svc/pipe map: %d\n", ret);
+		ath10k_err(ar, "Failed to write svc/pipe map: %d\n", ret);
 		return ret;
 	}
 
-	ret = ath10k_pci_diag_read_access(ar, pcie_state_targ_addr +
+	ret = ath10k_pci_diag_read32(ar, (pcie_state_targ_addr +
 					  offsetof(struct pcie_state,
-						   config_flags),
-					  &pcie_config_flags);
+						   config_flags)),
+				     &pcie_config_flags);
 	if (ret != 0) {
-		ath10k_err("Failed to get pcie config_flags: %d\n", ret);
+		ath10k_err(ar, "Failed to get pcie config_flags: %d\n", ret);
 		return ret;
 	}
 
 	pcie_config_flags &= ~PCIE_CONFIG_FLAG_ENABLE_L1;
 
-	ret = ath10k_pci_diag_write_mem(ar, pcie_state_targ_addr +
-				 offsetof(struct pcie_state, config_flags),
-				 &pcie_config_flags,
-				 sizeof(pcie_config_flags));
+	ret = ath10k_pci_diag_write32(ar, (pcie_state_targ_addr +
+					   offsetof(struct pcie_state,
+						    config_flags)),
+				      pcie_config_flags);
 	if (ret != 0) {
-		ath10k_err("Failed to write pcie config_flags: %d\n", ret);
+		ath10k_err(ar, "Failed to write pcie config_flags: %d\n", ret);
 		return ret;
 	}
 
 	/* configure early allocation */
 	ealloc_targ_addr = host_interest_item_address(HI_ITEM(hi_early_alloc));
 
-	ret = ath10k_pci_diag_read_access(ar, ealloc_targ_addr, &ealloc_value);
+	ret = ath10k_pci_diag_read32(ar, ealloc_targ_addr, &ealloc_value);
 	if (ret != 0) {
-		ath10k_err("Faile to get early alloc val: %d\n", ret);
+		ath10k_err(ar, "Faile to get early alloc val: %d\n", ret);
 		return ret;
 	}
 
@@ -1697,26 +1543,26 @@ static int ath10k_pci_init_config(struct ath10k *ar)
 	ealloc_value |= ((1 << HI_EARLY_ALLOC_IRAM_BANKS_SHIFT) &
 			 HI_EARLY_ALLOC_IRAM_BANKS_MASK);
 
-	ret = ath10k_pci_diag_write_access(ar, ealloc_targ_addr, ealloc_value);
+	ret = ath10k_pci_diag_write32(ar, ealloc_targ_addr, ealloc_value);
 	if (ret != 0) {
-		ath10k_err("Failed to set early alloc val: %d\n", ret);
+		ath10k_err(ar, "Failed to set early alloc val: %d\n", ret);
 		return ret;
 	}
 
 	/* Tell Target to proceed with initialization */
 	flag2_targ_addr = host_interest_item_address(HI_ITEM(hi_option_flag2));
 
-	ret = ath10k_pci_diag_read_access(ar, flag2_targ_addr, &flag2_value);
+	ret = ath10k_pci_diag_read32(ar, flag2_targ_addr, &flag2_value);
 	if (ret != 0) {
-		ath10k_err("Failed to get option val: %d\n", ret);
+		ath10k_err(ar, "Failed to get option val: %d\n", ret);
 		return ret;
 	}
 
 	flag2_value |= HI_OPTION_EARLY_CFG_DONE;
 
-	ret = ath10k_pci_diag_write_access(ar, flag2_targ_addr, flag2_value);
+	ret = ath10k_pci_diag_write32(ar, flag2_targ_addr, flag2_value);
 	if (ret != 0) {
-		ath10k_err("Failed to set option val: %d\n", ret);
+		ath10k_err(ar, "Failed to set option val: %d\n", ret);
 		return ret;
 	}
 
@@ -1730,7 +1576,7 @@ static int ath10k_pci_alloc_ce(struct ath10k *ar)
 	for (i = 0; i < CE_COUNT; i++) {
 		ret = ath10k_ce_alloc_pipe(ar, i, &host_ce_config_wlan[i]);
 		if (ret) {
-			ath10k_err("failed to allocate copy engine pipe %d: %d\n",
+			ath10k_err(ar, "failed to allocate copy engine pipe %d: %d\n",
 				   i, ret);
 			return ret;
 		}
@@ -1761,9 +1607,11 @@ static int ath10k_pci_ce_init(struct ath10k *ar)
 		pipe_info->hif_ce_state = ar;
 		attr = &host_ce_config_wlan[pipe_num];
 
-		ret = ath10k_ce_init_pipe(ar, pipe_num, attr);
+		ret = ath10k_ce_init_pipe(ar, pipe_num, attr,
+					  ath10k_pci_ce_send_done,
+					  ath10k_pci_ce_recv_data);
 		if (ret) {
-			ath10k_err("failed to initialize copy engine pipe %d: %d\n",
+			ath10k_err(ar, "failed to initialize copy engine pipe %d: %d\n",
 				   pipe_num, ret);
 			return ret;
 		}
@@ -1777,38 +1625,25 @@ static int ath10k_pci_ce_init(struct ath10k *ar)
 			continue;
 		}
 
-		pipe_info->buf_sz = (size_t) (attr->src_sz_max);
+		pipe_info->buf_sz = (size_t)(attr->src_sz_max);
 	}
 
 	return 0;
 }
 
-static void ath10k_pci_fw_interrupt_handler(struct ath10k *ar)
+static bool ath10k_pci_has_fw_crashed(struct ath10k *ar)
 {
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	u32 fw_indicator;
-
-	ath10k_pci_wake(ar);
-
-	fw_indicator = ath10k_pci_read32(ar, FW_INDICATOR_ADDRESS);
-
-	if (fw_indicator & FW_IND_EVENT_PENDING) {
-		/* ACK: clear Target-side pending event */
-		ath10k_pci_write32(ar, FW_INDICATOR_ADDRESS,
-				   fw_indicator & ~FW_IND_EVENT_PENDING);
+	return ath10k_pci_read32(ar, FW_INDICATOR_ADDRESS) &
+	       FW_IND_EVENT_PENDING;
+}
 
-		if (ar_pci->started) {
-			ath10k_pci_hif_dump_area(ar);
-		} else {
-			/*
-			 * Probable Target failure before we're prepared
-			 * to handle it.  Generally unexpected.
-			 */
-			ath10k_warn("early firmware event indicated\n");
-		}
-	}
+static void ath10k_pci_fw_crashed_clear(struct ath10k *ar)
+{
+	u32 val;
 
-	ath10k_pci_sleep(ar);
+	val = ath10k_pci_read32(ar, FW_INDICATOR_ADDRESS);
+	val &= ~FW_IND_EVENT_PENDING;
+	ath10k_pci_write32(ar, FW_INDICATOR_ADDRESS, val);
 }
 
 /* this function effectively clears target memory controller assert line */
@@ -1833,25 +1668,19 @@ static void ath10k_pci_warm_reset_si0(struct ath10k *ar)
 
 static int ath10k_pci_warm_reset(struct ath10k *ar)
 {
-	int ret = 0;
 	u32 val;
 
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot warm reset\n");
-
-	ret = ath10k_do_pci_wake(ar);
-	if (ret) {
-		ath10k_err("failed to wake up target: %d\n", ret);
-		return ret;
-	}
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot warm reset\n");
 
 	/* debug */
 	val = ath10k_pci_read32(ar, SOC_CORE_BASE_ADDRESS +
 				PCIE_INTR_CAUSE_ADDRESS);
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot host cpu intr cause: 0x%08x\n", val);
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot host cpu intr cause: 0x%08x\n",
+		   val);
 
 	val = ath10k_pci_read32(ar, SOC_CORE_BASE_ADDRESS +
 				CPU_INTR_ADDRESS);
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot target cpu intr cause: 0x%08x\n",
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot target cpu intr cause: 0x%08x\n",
 		   val);
 
 	/* disable pending irqs */
@@ -1894,11 +1723,12 @@ static int ath10k_pci_warm_reset(struct ath10k *ar)
 	/* debug */
 	val = ath10k_pci_read32(ar, SOC_CORE_BASE_ADDRESS +
 				PCIE_INTR_CAUSE_ADDRESS);
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot host cpu intr cause: 0x%08x\n", val);
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot host cpu intr cause: 0x%08x\n",
+		   val);
 
 	val = ath10k_pci_read32(ar, SOC_CORE_BASE_ADDRESS +
 				CPU_INTR_ADDRESS);
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot target cpu intr cause: 0x%08x\n",
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot target cpu intr cause: 0x%08x\n",
 		   val);
 
 	/* CPU warm reset */
@@ -1909,20 +1739,18 @@ static int ath10k_pci_warm_reset(struct ath10k *ar)
 
 	val = ath10k_pci_read32(ar, RTC_SOC_BASE_ADDRESS +
 				SOC_RESET_CONTROL_ADDRESS);
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot target reset state: 0x%08x\n", val);
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot target reset state: 0x%08x\n",
+		   val);
 
 	msleep(100);
 
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot warm reset complete\n");
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot warm reset complete\n");
 
-	ath10k_do_pci_sleep(ar);
-	return ret;
+	return 0;
 }
 
 static int __ath10k_pci_hif_power_up(struct ath10k *ar, bool cold_reset)
 {
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	const char *irq_mode;
 	int ret;
 
 	/*
@@ -1941,80 +1769,39 @@ static int __ath10k_pci_hif_power_up(struct ath10k *ar, bool cold_reset)
 		ret = ath10k_pci_warm_reset(ar);
 
 	if (ret) {
-		ath10k_err("failed to reset target: %d\n", ret);
+		ath10k_err(ar, "failed to reset target: %d\n", ret);
 		goto err;
 	}
 
-	if (!test_bit(ATH10K_PCI_FEATURE_SOC_POWER_SAVE, ar_pci->features))
-		/* Force AWAKE forever */
-		ath10k_do_pci_wake(ar);
-
 	ret = ath10k_pci_ce_init(ar);
 	if (ret) {
-		ath10k_err("failed to initialize CE: %d\n", ret);
-		goto err_ps;
-	}
-
-	ret = ath10k_ce_disable_interrupts(ar);
-	if (ret) {
-		ath10k_err("failed to disable CE interrupts: %d\n", ret);
-		goto err_ce;
-	}
-
-	ret = ath10k_pci_init_irq(ar);
-	if (ret) {
-		ath10k_err("failed to init irqs: %d\n", ret);
-		goto err_ce;
-	}
-
-	ret = ath10k_pci_request_early_irq(ar);
-	if (ret) {
-		ath10k_err("failed to request early irq: %d\n", ret);
-		goto err_deinit_irq;
+		ath10k_err(ar, "failed to initialize CE: %d\n", ret);
+		goto err;
 	}
 
 	ret = ath10k_pci_wait_for_target_init(ar);
 	if (ret) {
-		ath10k_err("failed to wait for target to init: %d\n", ret);
-		goto err_free_early_irq;
+		ath10k_err(ar, "failed to wait for target to init: %d\n", ret);
+		goto err_ce;
 	}
 
 	ret = ath10k_pci_init_config(ar);
 	if (ret) {
-		ath10k_err("failed to setup init config: %d\n", ret);
-		goto err_free_early_irq;
+		ath10k_err(ar, "failed to setup init config: %d\n", ret);
+		goto err_ce;
 	}
 
 	ret = ath10k_pci_wake_target_cpu(ar);
 	if (ret) {
-		ath10k_err("could not wake up target CPU: %d\n", ret);
-		goto err_free_early_irq;
+		ath10k_err(ar, "could not wake up target CPU: %d\n", ret);
+		goto err_ce;
 	}
 
-	if (ar_pci->num_msi_intrs > 1)
-		irq_mode = "MSI-X";
-	else if (ar_pci->num_msi_intrs == 1)
-		irq_mode = "MSI";
-	else
-		irq_mode = "legacy";
-
-	if (!test_bit(ATH10K_FLAG_FIRST_BOOT_DONE, &ar->dev_flags))
-		ath10k_info("pci irq %s irq_mode %d reset_mode %d\n",
-			    irq_mode, ath10k_pci_irq_mode,
-			    ath10k_pci_reset_mode);
-
 	return 0;
 
-err_free_early_irq:
-	ath10k_pci_free_early_irq(ar);
-err_deinit_irq:
-	ath10k_pci_deinit_irq(ar);
 err_ce:
 	ath10k_pci_ce_deinit(ar);
 	ath10k_pci_warm_reset(ar);
-err_ps:
-	if (!test_bit(ATH10K_PCI_FEATURE_SOC_POWER_SAVE, ar_pci->features))
-		ath10k_do_pci_sleep(ar);
 err:
 	return ret;
 }
@@ -2034,7 +1821,7 @@ static int ath10k_pci_hif_power_up_warm(struct ath10k *ar)
 		if (ret == 0)
 			break;
 
-		ath10k_warn("failed to warm reset (attempt %d out of %d): %d\n",
+		ath10k_warn(ar, "failed to warm reset (attempt %d out of %d): %d\n",
 			    i + 1, ATH10K_PCI_NUM_WARM_RESET_ATTEMPTS, ret);
 	}
 
@@ -2045,7 +1832,7 @@ static int ath10k_pci_hif_power_up(struct ath10k *ar)
 {
 	int ret;
 
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot hif power up\n");
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot hif power up\n");
 
 	/*
 	 * Hardware CUS232 version 2 has some issues with cold reset and the
@@ -2057,17 +1844,17 @@ static int ath10k_pci_hif_power_up(struct ath10k *ar)
 	 */
 	ret = ath10k_pci_hif_power_up_warm(ar);
 	if (ret) {
-		ath10k_warn("failed to power up target using warm reset: %d\n",
+		ath10k_warn(ar, "failed to power up target using warm reset: %d\n",
 			    ret);
 
 		if (ath10k_pci_reset_mode == ATH10K_PCI_RESET_WARM_ONLY)
 			return ret;
 
-		ath10k_warn("trying cold reset\n");
+		ath10k_warn(ar, "trying cold reset\n");
 
 		ret = __ath10k_pci_hif_power_up(ar, true);
 		if (ret) {
-			ath10k_err("failed to power up target using cold reset too (%d)\n",
+			ath10k_err(ar, "failed to power up target using cold reset too (%d)\n",
 				   ret);
 			return ret;
 		}
@@ -2078,18 +1865,9 @@ static int ath10k_pci_hif_power_up(struct ath10k *ar)
 
 static void ath10k_pci_hif_power_down(struct ath10k *ar)
 {
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot hif power down\n");
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot hif power down\n");
 
-	ath10k_pci_free_early_irq(ar);
-	ath10k_pci_kill_tasklet(ar);
-	ath10k_pci_deinit_irq(ar);
-	ath10k_pci_ce_deinit(ar);
 	ath10k_pci_warm_reset(ar);
-
-	if (!test_bit(ATH10K_PCI_FEATURE_SOC_POWER_SAVE, ar_pci->features))
-		ath10k_do_pci_sleep(ar);
 }
 
 #ifdef CONFIG_PM
@@ -2171,7 +1949,13 @@ static void ath10k_msi_err_tasklet(unsigned long data)
 {
 	struct ath10k *ar = (struct ath10k *)data;
 
-	ath10k_pci_fw_interrupt_handler(ar);
+	if (!ath10k_pci_has_fw_crashed(ar)) {
+		ath10k_warn(ar, "received unsolicited fw crash interrupt\n");
+		return;
+	}
+
+	ath10k_pci_fw_crashed_clear(ar);
+	ath10k_pci_fw_crashed_dump(ar);
 }
 
 /*
@@ -2185,7 +1969,8 @@ static irqreturn_t ath10k_pci_per_engine_handler(int irq, void *arg)
 	int ce_id = irq - ar_pci->pdev->irq - MSI_ASSIGN_CE_INITIAL;
 
 	if (ce_id < 0 || ce_id >= ARRAY_SIZE(ar_pci->pipe_info)) {
-		ath10k_warn("unexpected/invalid irq %d ce_id %d\n", irq, ce_id);
+		ath10k_warn(ar, "unexpected/invalid irq %d ce_id %d\n", irq,
+			    ce_id);
 		return IRQ_HANDLED;
 	}
 
@@ -2232,36 +2017,17 @@ static irqreturn_t ath10k_pci_interrupt_handler(int irq, void *arg)
 	return IRQ_HANDLED;
 }
 
-static void ath10k_pci_early_irq_tasklet(unsigned long data)
+static void ath10k_pci_tasklet(unsigned long data)
 {
 	struct ath10k *ar = (struct ath10k *)data;
-	u32 fw_ind;
-	int ret;
+	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
 
-	ret = ath10k_pci_wake(ar);
-	if (ret) {
-		ath10k_warn("failed to wake target in early irq tasklet: %d\n",
-			    ret);
+	if (ath10k_pci_has_fw_crashed(ar)) {
+		ath10k_pci_fw_crashed_clear(ar);
+		ath10k_pci_fw_crashed_dump(ar);
 		return;
 	}
 
-	fw_ind = ath10k_pci_read32(ar, FW_INDICATOR_ADDRESS);
-	if (fw_ind & FW_IND_EVENT_PENDING) {
-		ath10k_pci_write32(ar, FW_INDICATOR_ADDRESS,
-				   fw_ind & ~FW_IND_EVENT_PENDING);
-		ath10k_pci_hif_dump_area(ar);
-	}
-
-	ath10k_pci_sleep(ar);
-	ath10k_pci_enable_legacy_irq(ar);
-}
-
-static void ath10k_pci_tasklet(unsigned long data)
-{
-	struct ath10k *ar = (struct ath10k *)data;
-	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-
-	ath10k_pci_fw_interrupt_handler(ar); /* FIXME: Handle FW error */
 	ath10k_ce_per_engine_service_any(ar);
 
 	/* Re-enable legacy irq that was disabled in the irq handler */
@@ -2278,7 +2044,7 @@ static int ath10k_pci_request_irq_msix(struct ath10k *ar)
 			  ath10k_pci_msi_fw_handler,
 			  IRQF_SHARED, "ath10k_pci", ar);
 	if (ret) {
-		ath10k_warn("failed to request MSI-X fw irq %d: %d\n",
+		ath10k_warn(ar, "failed to request MSI-X fw irq %d: %d\n",
 			    ar_pci->pdev->irq + MSI_ASSIGN_FW, ret);
 		return ret;
 	}
@@ -2288,7 +2054,7 @@ static int ath10k_pci_request_irq_msix(struct ath10k *ar)
 				  ath10k_pci_per_engine_handler,
 				  IRQF_SHARED, "ath10k_pci", ar);
 		if (ret) {
-			ath10k_warn("failed to request MSI-X ce irq %d: %d\n",
+			ath10k_warn(ar, "failed to request MSI-X ce irq %d: %d\n",
 				    ar_pci->pdev->irq + i, ret);
 
 			for (i--; i >= MSI_ASSIGN_CE_INITIAL; i--)
@@ -2311,7 +2077,7 @@ static int ath10k_pci_request_irq_msi(struct ath10k *ar)
 			  ath10k_pci_interrupt_handler,
 			  IRQF_SHARED, "ath10k_pci", ar);
 	if (ret) {
-		ath10k_warn("failed to request MSI irq %d: %d\n",
+		ath10k_warn(ar, "failed to request MSI irq %d: %d\n",
 			    ar_pci->pdev->irq, ret);
 		return ret;
 	}
@@ -2328,7 +2094,7 @@ static int ath10k_pci_request_irq_legacy(struct ath10k *ar)
 			  ath10k_pci_interrupt_handler,
 			  IRQF_SHARED, "ath10k_pci", ar);
 	if (ret) {
-		ath10k_warn("failed to request legacy irq %d: %d\n",
+		ath10k_warn(ar, "failed to request legacy irq %d: %d\n",
 			    ar_pci->pdev->irq, ret);
 		return ret;
 	}
@@ -2349,7 +2115,7 @@ static int ath10k_pci_request_irq(struct ath10k *ar)
 		return ath10k_pci_request_irq_msix(ar);
 	}
 
-	ath10k_warn("unknown irq configuration upon request\n");
+	ath10k_warn(ar, "unknown irq configuration upon request\n");
 	return -EINVAL;
 }
 
@@ -2372,8 +2138,6 @@ static void ath10k_pci_init_irq_tasklets(struct ath10k *ar)
 	tasklet_init(&ar_pci->intr_tq, ath10k_pci_tasklet, (unsigned long)ar);
 	tasklet_init(&ar_pci->msi_fw_err, ath10k_msi_err_tasklet,
 		     (unsigned long)ar);
-	tasklet_init(&ar_pci->early_irq_tasklet, ath10k_pci_early_irq_tasklet,
-		     (unsigned long)ar);
 
 	for (i = 0; i < CE_COUNT; i++) {
 		ar_pci->pipe_info[i].ar_pci = ar_pci;
@@ -2385,21 +2149,19 @@ static void ath10k_pci_init_irq_tasklets(struct ath10k *ar)
 static int ath10k_pci_init_irq(struct ath10k *ar)
 {
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
-	bool msix_supported = test_bit(ATH10K_PCI_FEATURE_MSI_X,
-				       ar_pci->features);
 	int ret;
 
 	ath10k_pci_init_irq_tasklets(ar);
 
-	if (ath10k_pci_irq_mode != ATH10K_PCI_IRQ_AUTO &&
-	    !test_bit(ATH10K_FLAG_FIRST_BOOT_DONE, &ar->dev_flags))
-		ath10k_info("limiting irq mode to: %d\n", ath10k_pci_irq_mode);
+	if (ath10k_pci_irq_mode != ATH10K_PCI_IRQ_AUTO)
+		ath10k_info(ar, "limiting irq mode to: %d\n",
+			    ath10k_pci_irq_mode);
 
 	/* Try MSI-X */
-	if (ath10k_pci_irq_mode == ATH10K_PCI_IRQ_AUTO && msix_supported) {
+	if (ath10k_pci_irq_mode == ATH10K_PCI_IRQ_AUTO) {
 		ar_pci->num_msi_intrs = MSI_NUM_REQUEST;
 		ret = pci_enable_msi_range(ar_pci->pdev, ar_pci->num_msi_intrs,
-							 ar_pci->num_msi_intrs);
+					   ar_pci->num_msi_intrs);
 		if (ret > 0)
 			return 0;
 
@@ -2426,34 +2188,16 @@ static int ath10k_pci_init_irq(struct ath10k *ar)
 	 * synchronization checking. */
 	ar_pci->num_msi_intrs = 0;
 
-	ret = ath10k_pci_wake(ar);
-	if (ret) {
-		ath10k_warn("failed to wake target: %d\n", ret);
-		return ret;
-	}
-
 	ath10k_pci_write32(ar, SOC_CORE_BASE_ADDRESS + PCIE_INTR_ENABLE_ADDRESS,
 			   PCIE_INTR_FIRMWARE_MASK | PCIE_INTR_CE_MASK_ALL);
-	ath10k_pci_sleep(ar);
 
 	return 0;
 }
 
-static int ath10k_pci_deinit_irq_legacy(struct ath10k *ar)
+static void ath10k_pci_deinit_irq_legacy(struct ath10k *ar)
 {
-	int ret;
-
-	ret = ath10k_pci_wake(ar);
-	if (ret) {
-		ath10k_warn("failed to wake target: %d\n", ret);
-		return ret;
-	}
-
 	ath10k_pci_write32(ar, SOC_CORE_BASE_ADDRESS + PCIE_INTR_ENABLE_ADDRESS,
 			   0);
-	ath10k_pci_sleep(ar);
-
-	return 0;
 }
 
 static int ath10k_pci_deinit_irq(struct ath10k *ar)
@@ -2462,7 +2206,8 @@ static int ath10k_pci_deinit_irq(struct ath10k *ar)
 
 	switch (ar_pci->num_msi_intrs) {
 	case 0:
-		return ath10k_pci_deinit_irq_legacy(ar);
+		ath10k_pci_deinit_irq_legacy(ar);
+		return 0;
 	case 1:
 		/* fall-through */
 	case MSI_NUM_REQUEST:
@@ -2472,7 +2217,7 @@ static int ath10k_pci_deinit_irq(struct ath10k *ar)
 		pci_disable_msi(ar_pci->pdev);
 	}
 
-	ath10k_warn("unknown irq configuration upon deinit\n");
+	ath10k_warn(ar, "unknown irq configuration upon deinit\n");
 	return -EINVAL;
 }
 
@@ -2480,23 +2225,17 @@ static int ath10k_pci_wait_for_target_init(struct ath10k *ar)
 {
 	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
 	unsigned long timeout;
-	int ret;
 	u32 val;
 
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot waiting target to initialise\n");
-
-	ret = ath10k_pci_wake(ar);
-	if (ret) {
-		ath10k_err("failed to wake up target for init: %d\n", ret);
-		return ret;
-	}
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot waiting target to initialise\n");
 
 	timeout = jiffies + msecs_to_jiffies(ATH10K_PCI_TARGET_WAIT);
 
 	do {
 		val = ath10k_pci_read32(ar, FW_INDICATOR_ADDRESS);
 
-		ath10k_dbg(ATH10K_DBG_BOOT, "boot target indicator %x\n", val);
+		ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot target indicator %x\n",
+			   val);
 
 		/* target should never return this */
 		if (val == 0xffffffff)
@@ -2511,55 +2250,42 @@ static int ath10k_pci_wait_for_target_init(struct ath10k *ar)
 
 		if (ar_pci->num_msi_intrs == 0)
 			/* Fix potential race by repeating CORE_BASE writes */
-			ath10k_pci_soc_write32(ar, PCIE_INTR_ENABLE_ADDRESS,
-					       PCIE_INTR_FIRMWARE_MASK |
-					       PCIE_INTR_CE_MASK_ALL);
+			ath10k_pci_write32(ar, SOC_CORE_BASE_ADDRESS +
+					   PCIE_INTR_ENABLE_ADDRESS,
+					   PCIE_INTR_FIRMWARE_MASK |
+					   PCIE_INTR_CE_MASK_ALL);
 
 		mdelay(10);
 	} while (time_before(jiffies, timeout));
 
 	if (val == 0xffffffff) {
-		ath10k_err("failed to read device register, device is gone\n");
-		ret = -EIO;
-		goto out;
+		ath10k_err(ar, "failed to read device register, device is gone\n");
+		return -EIO;
 	}
 
 	if (val & FW_IND_EVENT_PENDING) {
-		ath10k_warn("device has crashed during init\n");
-		ath10k_pci_write32(ar, FW_INDICATOR_ADDRESS,
-				   val & ~FW_IND_EVENT_PENDING);
-		ath10k_pci_hif_dump_area(ar);
-		ret = -ECOMM;
-		goto out;
+		ath10k_warn(ar, "device has crashed during init\n");
+		ath10k_pci_fw_crashed_clear(ar);
+		ath10k_pci_fw_crashed_dump(ar);
+		return -ECOMM;
 	}
 
 	if (!(val & FW_IND_INITIALIZED)) {
-		ath10k_err("failed to receive initialized event from target: %08x\n",
+		ath10k_err(ar, "failed to receive initialized event from target: %08x\n",
 			   val);
-		ret = -ETIMEDOUT;
-		goto out;
+		return -ETIMEDOUT;
 	}
 
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot target initialised\n");
-
-out:
-	ath10k_pci_sleep(ar);
-	return ret;
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot target initialised\n");
+	return 0;
 }
 
 static int ath10k_pci_cold_reset(struct ath10k *ar)
 {
-	int i, ret;
+	int i;
 	u32 val;
 
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot cold reset\n");
-
-	ret = ath10k_do_pci_wake(ar);
-	if (ret) {
-		ath10k_err("failed to wake up target: %d\n",
-			   ret);
-		return ret;
-	}
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot cold reset\n");
 
 	/* Put Target, including PCIe, into RESET. */
 	val = ath10k_pci_reg_read32(ar, SOC_GLOBAL_RESET_ADDRESS);
@@ -2584,169 +2310,199 @@ static int ath10k_pci_cold_reset(struct ath10k *ar)
 		msleep(1);
 	}
 
-	ath10k_do_pci_sleep(ar);
-
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot cold reset complete\n");
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot cold reset complete\n");
 
 	return 0;
 }
 
-static void ath10k_pci_dump_features(struct ath10k_pci *ar_pci)
-{
-	int i;
-
-	for (i = 0; i < ATH10K_PCI_FEATURE_COUNT; i++) {
-		if (!test_bit(i, ar_pci->features))
-			continue;
-
-		switch (i) {
-		case ATH10K_PCI_FEATURE_MSI_X:
-			ath10k_dbg(ATH10K_DBG_BOOT, "device supports MSI-X\n");
-			break;
-		case ATH10K_PCI_FEATURE_SOC_POWER_SAVE:
-			ath10k_dbg(ATH10K_DBG_BOOT, "QCA98XX SoC power save enabled\n");
-			break;
-		}
-	}
-}
-
-static int ath10k_pci_probe(struct pci_dev *pdev,
-			    const struct pci_device_id *pci_dev)
+static int ath10k_pci_claim(struct ath10k *ar)
 {
-	void __iomem *mem;
-	int ret = 0;
-	struct ath10k *ar;
-	struct ath10k_pci *ar_pci;
-	u32 lcr_val, chip_id;
-
-	ath10k_dbg(ATH10K_DBG_PCI, "pci probe\n");
-
-	ar_pci = kzalloc(sizeof(*ar_pci), GFP_KERNEL);
-	if (ar_pci == NULL)
-		return -ENOMEM;
-
-	ar_pci->pdev = pdev;
-	ar_pci->dev = &pdev->dev;
-
-	switch (pci_dev->device) {
-	case QCA988X_2_0_DEVICE_ID:
-		set_bit(ATH10K_PCI_FEATURE_MSI_X, ar_pci->features);
-		break;
-	default:
-		ret = -ENODEV;
-		ath10k_err("Unknown device ID: %d\n", pci_dev->device);
-		goto err_ar_pci;
-	}
-
-	if (ath10k_pci_target_ps)
-		set_bit(ATH10K_PCI_FEATURE_SOC_POWER_SAVE, ar_pci->features);
-
-	ath10k_pci_dump_features(ar_pci);
-
-	ar = ath10k_core_create(ar_pci, ar_pci->dev, &ath10k_pci_hif_ops);
-	if (!ar) {
-		ath10k_err("failed to create driver core\n");
-		ret = -EINVAL;
-		goto err_ar_pci;
-	}
-
-	ar_pci->ar = ar;
-	atomic_set(&ar_pci->keep_awake_count, 0);
+	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
+	struct pci_dev *pdev = ar_pci->pdev;
+	u32 lcr_val;
+	int ret;
 
 	pci_set_drvdata(pdev, ar);
 
 	ret = pci_enable_device(pdev);
 	if (ret) {
-		ath10k_err("failed to enable PCI device: %d\n", ret);
-		goto err_ar;
+		ath10k_err(ar, "failed to enable pci device: %d\n", ret);
+		return ret;
 	}
 
-	/* Request MMIO resources */
 	ret = pci_request_region(pdev, BAR_NUM, "ath");
 	if (ret) {
-		ath10k_err("failed to request MMIO region: %d\n", ret);
+		ath10k_err(ar, "failed to request region BAR%d: %d\n", BAR_NUM,
+			   ret);
 		goto err_device;
 	}
 
-	/*
-	 * Target structures have a limit of 32 bit DMA pointers.
-	 * DMA pointers can be wider than 32 bits by default on some systems.
-	 */
+	/* Target expects 32 bit DMA. Enforce it. */
 	ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
 	if (ret) {
-		ath10k_err("failed to set DMA mask to 32-bit: %d\n", ret);
+		ath10k_err(ar, "failed to set dma mask to 32-bit: %d\n", ret);
 		goto err_region;
 	}
 
 	ret = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
 	if (ret) {
-		ath10k_err("failed to set consistent DMA mask to 32-bit\n");
+		ath10k_err(ar, "failed to set consistent dma mask to 32-bit: %d\n",
+			   ret);
 		goto err_region;
 	}
 
-	/* Set bus master bit in PCI_COMMAND to enable DMA */
 	pci_set_master(pdev);
 
-	/*
-	 * Temporary FIX: disable ASPM
-	 * Will be removed after the OTP is programmed
-	 */
+	/* Workaround: Disable ASPM */
 	pci_read_config_dword(pdev, 0x80, &lcr_val);
 	pci_write_config_dword(pdev, 0x80, (lcr_val & 0xffffff00));
 
 	/* Arrange for access to Target SoC registers. */
-	mem = pci_iomap(pdev, BAR_NUM, 0);
-	if (!mem) {
-		ath10k_err("failed to perform IOMAP for BAR%d\n", BAR_NUM);
+	ar_pci->mem = pci_iomap(pdev, BAR_NUM, 0);
+	if (!ar_pci->mem) {
+		ath10k_err(ar, "failed to iomap BAR%d\n", BAR_NUM);
 		ret = -EIO;
 		goto err_master;
 	}
 
-	ar_pci->mem = mem;
+	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot pci_mem 0x%p\n", ar_pci->mem);
+	return 0;
+
+err_master:
+	pci_clear_master(pdev);
+
+err_region:
+	pci_release_region(pdev, BAR_NUM);
+
+err_device:
+	pci_disable_device(pdev);
+
+	return ret;
+}
+
+static void ath10k_pci_release(struct ath10k *ar)
+{
+	struct ath10k_pci *ar_pci = ath10k_pci_priv(ar);
+	struct pci_dev *pdev = ar_pci->pdev;
+
+	pci_iounmap(pdev, ar_pci->mem);
+	pci_release_region(pdev, BAR_NUM);
+	pci_clear_master(pdev);
+	pci_disable_device(pdev);
+}
+
+static int ath10k_pci_probe(struct pci_dev *pdev,
+			    const struct pci_device_id *pci_dev)
+{
+	int ret = 0;
+	struct ath10k *ar;
+	struct ath10k_pci *ar_pci;
+	u32 chip_id;
+
+	ar = ath10k_core_create(sizeof(*ar_pci), &pdev->dev,
+				&ath10k_pci_hif_ops);
+	if (!ar) {
+		dev_err(&pdev->dev, "failed to allocate core\n");
+		return -ENOMEM;
+	}
+
+	ath10k_dbg(ar, ATH10K_DBG_PCI, "pci probe\n");
+
+	ar_pci = ath10k_pci_priv(ar);
+	ar_pci->pdev = pdev;
+	ar_pci->dev = &pdev->dev;
+	ar_pci->ar = ar;
 
 	spin_lock_init(&ar_pci->ce_lock);
+	setup_timer(&ar_pci->rx_post_retry, ath10k_pci_rx_replenish_retry,
+		    (unsigned long)ar);
 
-	ret = ath10k_do_pci_wake(ar);
+	ret = ath10k_pci_claim(ar);
 	if (ret) {
-		ath10k_err("Failed to get chip id: %d\n", ret);
-		goto err_iomap;
+		ath10k_err(ar, "failed to claim device: %d\n", ret);
+		goto err_core_destroy;
 	}
 
-	chip_id = ath10k_pci_soc_read32(ar, SOC_CHIP_ID_ADDRESS);
+	ret = ath10k_pci_wake(ar);
+	if (ret) {
+		ath10k_err(ar, "failed to wake up: %d\n", ret);
+		goto err_release;
+	}
 
-	ath10k_do_pci_sleep(ar);
+	chip_id = ath10k_pci_soc_read32(ar, SOC_CHIP_ID_ADDRESS);
+	if (chip_id == 0xffffffff) {
+		ath10k_err(ar, "failed to get chip id\n");
+		goto err_sleep;
+	}
 
 	ret = ath10k_pci_alloc_ce(ar);
 	if (ret) {
-		ath10k_err("failed to allocate copy engine pipes: %d\n", ret);
-		goto err_iomap;
+		ath10k_err(ar, "failed to allocate copy engine pipes: %d\n",
+			   ret);
+		goto err_sleep;
 	}
 
-	ath10k_dbg(ATH10K_DBG_BOOT, "boot pci_mem 0x%p\n", ar_pci->mem);
+	ath10k_pci_ce_deinit(ar);
 
-	ret = ath10k_core_register(ar, chip_id);
+	ret = ath10k_ce_disable_interrupts(ar);
 	if (ret) {
-		ath10k_err("failed to register driver core: %d\n", ret);
+		ath10k_err(ar, "failed to disable copy engine interrupts: %d\n",
+			   ret);
 		goto err_free_ce;
 	}
 
+	/* Workaround: There's no known way to mask all possible interrupts via
+	 * device CSR. The only way to make sure device doesn't assert
+	 * interrupts is to reset it. Interrupts are then disabled on host
+	 * after handlers are registered.
+	 */
+	ath10k_pci_warm_reset(ar);
+
+	ret = ath10k_pci_init_irq(ar);
+	if (ret) {
+		ath10k_err(ar, "failed to init irqs: %d\n", ret);
+		goto err_free_ce;
+	}
+
+	ath10k_info(ar, "pci irq %s interrupts %d irq_mode %d reset_mode %d\n",
+		    ath10k_pci_get_irq_method(ar), ar_pci->num_msi_intrs,
+		    ath10k_pci_irq_mode, ath10k_pci_reset_mode);
+
+	ret = ath10k_pci_request_irq(ar);
+	if (ret) {
+		ath10k_warn(ar, "failed to request irqs: %d\n", ret);
+		goto err_deinit_irq;
+	}
+
+	/* This shouldn't race as the device has been reset above. */
+	ath10k_pci_irq_disable(ar);
+
+	ret = ath10k_core_register(ar, chip_id);
+	if (ret) {
+		ath10k_err(ar, "failed to register driver core: %d\n", ret);
+		goto err_free_irq;
+	}
+
 	return 0;
 
+err_free_irq:
+	ath10k_pci_free_irq(ar);
+	ath10k_pci_kill_tasklet(ar);
+
+err_deinit_irq:
+	ath10k_pci_deinit_irq(ar);
+
 err_free_ce:
 	ath10k_pci_free_ce(ar);
-err_iomap:
-	pci_iounmap(pdev, mem);
-err_master:
-	pci_clear_master(pdev);
-err_region:
-	pci_release_region(pdev, BAR_NUM);
-err_device:
-	pci_disable_device(pdev);
-err_ar:
+
+err_sleep:
+	ath10k_pci_sleep(ar);
+
+err_release:
+	ath10k_pci_release(ar);
+
+err_core_destroy:
 	ath10k_core_destroy(ar);
-err_ar_pci:
-	/* call HIF PCI free here */
-	kfree(ar_pci);
 
 	return ret;
 }
@@ -2756,7 +2512,7 @@ static void ath10k_pci_remove(struct pci_dev *pdev)
 	struct ath10k *ar = pci_get_drvdata(pdev);
 	struct ath10k_pci *ar_pci;
 
-	ath10k_dbg(ATH10K_DBG_PCI, "pci remove\n");
+	ath10k_dbg(ar, ATH10K_DBG_PCI, "pci remove\n");
 
 	if (!ar)
 		return;
@@ -2767,15 +2523,14 @@ static void ath10k_pci_remove(struct pci_dev *pdev)
 		return;
 
 	ath10k_core_unregister(ar);
+	ath10k_pci_free_irq(ar);
+	ath10k_pci_kill_tasklet(ar);
+	ath10k_pci_deinit_irq(ar);
+	ath10k_pci_ce_deinit(ar);
 	ath10k_pci_free_ce(ar);
-
-	pci_iounmap(pdev, ar_pci->mem);
-	pci_release_region(pdev, BAR_NUM);
-	pci_clear_master(pdev);
-	pci_disable_device(pdev);
-
+	ath10k_pci_sleep(ar);
+	ath10k_pci_release(ar);
 	ath10k_core_destroy(ar);
-	kfree(ar_pci);
 }
 
 MODULE_DEVICE_TABLE(pci, ath10k_pci_id_table);
@@ -2793,7 +2548,8 @@ static int __init ath10k_pci_init(void)
 
 	ret = pci_register_driver(&ath10k_pci_driver);
 	if (ret)
-		ath10k_err("failed to register PCI driver: %d\n", ret);
+		printk(KERN_ERR "failed to register ath10k pci driver: %d\n",
+		       ret);
 
 	return ret;
 }
@@ -2809,5 +2565,5 @@ module_exit(ath10k_pci_exit);
 MODULE_AUTHOR("Qualcomm Atheros");
 MODULE_DESCRIPTION("Driver support for Atheros QCA988X PCIe devices");
 MODULE_LICENSE("Dual BSD/GPL");
-MODULE_FIRMWARE(QCA988X_HW_2_0_FW_DIR "/" QCA988X_HW_2_0_FW_2_FILE);
+MODULE_FIRMWARE(QCA988X_HW_2_0_FW_DIR "/" QCA988X_HW_2_0_FW_3_FILE);
 MODULE_FIRMWARE(QCA988X_HW_2_0_FW_DIR "/" QCA988X_HW_2_0_BOARD_DATA_FILE);