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

Pull networking updates from David Miller:
 "Highlights:

   1) Maintain the TCP retransmit queue using an rbtree, with 1GB
      windows at 100Gb this really has become necessary. From Eric
      Dumazet.

   2) Multi-program support for cgroup+bpf, from Alexei Starovoitov.

   3) Perform broadcast flooding in hardware in mv88e6xxx, from Andrew
      Lunn.

   4) Add meter action support to openvswitch, from Andy Zhou.

   5) Add a data meta pointer for BPF accessible packets, from Daniel
      Borkmann.

   6) Namespace-ify almost all TCP sysctl knobs, from Eric Dumazet.

   7) Turn on Broadcom Tags in b53 driver, from Florian Fainelli.

   8) More work to move the RTNL mutex down, from Florian Westphal.

   9) Add 'bpftool' utility, to help with bpf program introspection.
      From Jakub Kicinski.

  10) Add new 'cpumap' type for XDP_REDIRECT action, from Jesper
      Dangaard Brouer.

  11) Support 'blocks' of transformations in the packet scheduler which
      can span multiple network devices, from Jiri Pirko.

  12) TC flower offload support in cxgb4, from Kumar Sanghvi.

  13) Priority based stream scheduler for SCTP, from Marcelo Ricardo
      Leitner.

  14) Thunderbolt networking driver, from Amir Levy and Mika Westerberg.

  15) Add RED qdisc offloadability, and use it in mlxsw driver. From
      Nogah Frankel.

  16) eBPF based device controller for cgroup v2, from Roman Gushchin.

  17) Add some fundamental tracepoints for TCP, from Song Liu.

  18) Remove garbage collection from ipv6 route layer, this is a
      significant accomplishment. From Wei Wang.

  19) Add multicast route offload support to mlxsw, from Yotam Gigi"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (2177 commits)
  tcp: highest_sack fix
  geneve: fix fill_info when link down
  bpf: fix lockdep splat
  net: cdc_ncm: GetNtbFormat endian fix
  openvswitch: meter: fix NULL pointer dereference in ovs_meter_cmd_reply_start
  netem: remove unnecessary 64 bit modulus
  netem: use 64 bit divide by rate
  tcp: Namespace-ify sysctl_tcp_default_congestion_control
  net: Protect iterations over net::fib_notifier_ops in fib_seq_sum()
  ipv6: set all.accept_dad to 0 by default
  uapi: fix linux/tls.h userspace compilation error
  usbnet: ipheth: prevent TX queue timeouts when device not ready
  vhost_net: conditionally enable tx polling
  uapi: fix linux/rxrpc.h userspace compilation errors
  net: stmmac: fix LPI transitioning for dwmac4
  atm: horizon: Fix irq release error
  net-sysfs: trigger netlink notification on ifalias change via sysfs
  openvswitch: Using kfree_rcu() to simplify the code
  openvswitch: Make local function ovs_nsh_key_attr_size() static
  openvswitch: Fix return value check in ovs_meter_cmd_features()
  ...

1 files changed, 315 insertions, 97 deletions

diff --git a/drivers/thunderbolt/nhi.c b/drivers/thunderbolt/nhi.c
index 05af126a2435..419a7a90bce0 100644
--- a/drivers/thunderbolt/nhi.c
+++ b/drivers/thunderbolt/nhi.c
@@ -22,6 +22,14 @@
 #define RING_TYPE(ring) ((ring)->is_tx ? "TX ring" : "RX ring")
 
 /*
+ * Used to enable end-to-end workaround for missing RX packets. Do not
+ * use this ring for anything else.
+ */
+#define RING_E2E_UNUSED_HOPID	2
+/* HopIDs 0-7 are reserved by the Thunderbolt protocol */
+#define RING_FIRST_USABLE_HOPID	8
+
+/*
  * Minimal number of vectors when we use MSI-X. Two for control channel
  * Rx/Tx and the rest four are for cross domain DMA paths.
  */
@@ -206,8 +214,10 @@ static void ring_work(struct work_struct *work)
 	struct tb_ring *ring = container_of(work, typeof(*ring), work);
 	struct ring_frame *frame;
 	bool canceled = false;
+	unsigned long flags;
 	LIST_HEAD(done);
-	mutex_lock(&ring->lock);
+
+	spin_lock_irqsave(&ring->lock, flags);
 
 	if (!ring->running) {
 		/*  Move all frames to done and mark them as canceled. */
@@ -229,30 +239,14 @@ static void ring_work(struct work_struct *work)
 			frame->eof = ring->descriptors[ring->tail].eof;
 			frame->sof = ring->descriptors[ring->tail].sof;
 			frame->flags = ring->descriptors[ring->tail].flags;
-			if (frame->sof != 0)
-				dev_WARN(&ring->nhi->pdev->dev,
-					 "%s %d got unexpected SOF: %#x\n",
-					 RING_TYPE(ring), ring->hop,
-					 frame->sof);
-			/*
-			 * known flags:
-			 * raw not enabled, interupt not set: 0x2=0010
-			 * raw enabled: 0xa=1010
-			 * raw not enabled: 0xb=1011
-			 * partial frame (>MAX_FRAME_SIZE): 0xe=1110
-			 */
-			if (frame->flags != 0xa)
-				dev_WARN(&ring->nhi->pdev->dev,
-					 "%s %d got unexpected flags: %#x\n",
-					 RING_TYPE(ring), ring->hop,
-					 frame->flags);
 		}
 		ring->tail = (ring->tail + 1) % ring->size;
 	}
 	ring_write_descriptors(ring);
 
 invoke_callback:
-	mutex_unlock(&ring->lock); /* allow callbacks to schedule new work */
+	/* allow callbacks to schedule new work */
+	spin_unlock_irqrestore(&ring->lock, flags);
 	while (!list_empty(&done)) {
 		frame = list_first_entry(&done, typeof(*frame), list);
 		/*
@@ -260,29 +254,128 @@ invoke_callback:
 		 * Do not hold on to it.
 		 */
 		list_del_init(&frame->list);
-		frame->callback(ring, frame, canceled);
+		if (frame->callback)
+			frame->callback(ring, frame, canceled);
 	}
 }
 
-int __ring_enqueue(struct tb_ring *ring, struct ring_frame *frame)
+int __tb_ring_enqueue(struct tb_ring *ring, struct ring_frame *frame)
 {
+	unsigned long flags;
 	int ret = 0;
-	mutex_lock(&ring->lock);
+
+	spin_lock_irqsave(&ring->lock, flags);
 	if (ring->running) {
 		list_add_tail(&frame->list, &ring->queue);
 		ring_write_descriptors(ring);
 	} else {
 		ret = -ESHUTDOWN;
 	}
-	mutex_unlock(&ring->lock);
+	spin_unlock_irqrestore(&ring->lock, flags);
 	return ret;
 }
+EXPORT_SYMBOL_GPL(__tb_ring_enqueue);
+
+/**
+ * tb_ring_poll() - Poll one completed frame from the ring
+ * @ring: Ring to poll
+ *
+ * This function can be called when @start_poll callback of the @ring
+ * has been called. It will read one completed frame from the ring and
+ * return it to the caller. Returns %NULL if there is no more completed
+ * frames.
+ */
+struct ring_frame *tb_ring_poll(struct tb_ring *ring)
+{
+	struct ring_frame *frame = NULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ring->lock, flags);
+	if (!ring->running)
+		goto unlock;
+	if (ring_empty(ring))
+		goto unlock;
+
+	if (ring->descriptors[ring->tail].flags & RING_DESC_COMPLETED) {
+		frame = list_first_entry(&ring->in_flight, typeof(*frame),
+					 list);
+		list_del_init(&frame->list);
+
+		if (!ring->is_tx) {
+			frame->size = ring->descriptors[ring->tail].length;
+			frame->eof = ring->descriptors[ring->tail].eof;
+			frame->sof = ring->descriptors[ring->tail].sof;
+			frame->flags = ring->descriptors[ring->tail].flags;
+		}
+
+		ring->tail = (ring->tail + 1) % ring->size;
+	}
+
+unlock:
+	spin_unlock_irqrestore(&ring->lock, flags);
+	return frame;
+}
+EXPORT_SYMBOL_GPL(tb_ring_poll);
+
+static void __ring_interrupt_mask(struct tb_ring *ring, bool mask)
+{
+	int idx = ring_interrupt_index(ring);
+	int reg = REG_RING_INTERRUPT_BASE + idx / 32 * 4;
+	int bit = idx % 32;
+	u32 val;
+
+	val = ioread32(ring->nhi->iobase + reg);
+	if (mask)
+		val &= ~BIT(bit);
+	else
+		val |= BIT(bit);
+	iowrite32(val, ring->nhi->iobase + reg);
+}
+
+/* Both @nhi->lock and @ring->lock should be held */
+static void __ring_interrupt(struct tb_ring *ring)
+{
+	if (!ring->running)
+		return;
+
+	if (ring->start_poll) {
+		__ring_interrupt_mask(ring, false);
+		ring->start_poll(ring->poll_data);
+	} else {
+		schedule_work(&ring->work);
+	}
+}
+
+/**
+ * tb_ring_poll_complete() - Re-start interrupt for the ring
+ * @ring: Ring to re-start the interrupt
+ *
+ * This will re-start (unmask) the ring interrupt once the user is done
+ * with polling.
+ */
+void tb_ring_poll_complete(struct tb_ring *ring)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&ring->nhi->lock, flags);
+	spin_lock(&ring->lock);
+	if (ring->start_poll)
+		__ring_interrupt_mask(ring, false);
+	spin_unlock(&ring->lock);
+	spin_unlock_irqrestore(&ring->nhi->lock, flags);
+}
+EXPORT_SYMBOL_GPL(tb_ring_poll_complete);
 
 static irqreturn_t ring_msix(int irq, void *data)
 {
 	struct tb_ring *ring = data;
 
-	schedule_work(&ring->work);
+	spin_lock(&ring->nhi->lock);
+	spin_lock(&ring->lock);
+	__ring_interrupt(ring);
+	spin_unlock(&ring->lock);
+	spin_unlock(&ring->nhi->lock);
+
 	return IRQ_HANDLED;
 }
 
@@ -320,30 +413,81 @@ static void ring_release_msix(struct tb_ring *ring)
 	ring->irq = 0;
 }
 
-static struct tb_ring *ring_alloc(struct tb_nhi *nhi, u32 hop, int size,
-				  bool transmit, unsigned int flags)
+static int nhi_alloc_hop(struct tb_nhi *nhi, struct tb_ring *ring)
+{
+	int ret = 0;
+
+	spin_lock_irq(&nhi->lock);
+
+	if (ring->hop < 0) {
+		unsigned int i;
+
+		/*
+		 * Automatically allocate HopID from the non-reserved
+		 * range 8 .. hop_count - 1.
+		 */
+		for (i = RING_FIRST_USABLE_HOPID; i < nhi->hop_count; i++) {
+			if (ring->is_tx) {
+				if (!nhi->tx_rings[i]) {
+					ring->hop = i;
+					break;
+				}
+			} else {
+				if (!nhi->rx_rings[i]) {
+					ring->hop = i;
+					break;
+				}
+			}
+		}
+	}
+
+	if (ring->hop < 0 || ring->hop >= nhi->hop_count) {
+		dev_warn(&nhi->pdev->dev, "invalid hop: %d\n", ring->hop);
+		ret = -EINVAL;
+		goto err_unlock;
+	}
+	if (ring->is_tx && nhi->tx_rings[ring->hop]) {
+		dev_warn(&nhi->pdev->dev, "TX hop %d already allocated\n",
+			 ring->hop);
+		ret = -EBUSY;
+		goto err_unlock;
+	} else if (!ring->is_tx && nhi->rx_rings[ring->hop]) {
+		dev_warn(&nhi->pdev->dev, "RX hop %d already allocated\n",
+			 ring->hop);
+		ret = -EBUSY;
+		goto err_unlock;
+	}
+
+	if (ring->is_tx)
+		nhi->tx_rings[ring->hop] = ring;
+	else
+		nhi->rx_rings[ring->hop] = ring;
+
+err_unlock:
+	spin_unlock_irq(&nhi->lock);
+
+	return ret;
+}
+
+static struct tb_ring *tb_ring_alloc(struct tb_nhi *nhi, u32 hop, int size,
+				     bool transmit, unsigned int flags,
+				     u16 sof_mask, u16 eof_mask,
+				     void (*start_poll)(void *),
+				     void *poll_data)
 {
 	struct tb_ring *ring = NULL;
 	dev_info(&nhi->pdev->dev, "allocating %s ring %d of size %d\n",
 		 transmit ? "TX" : "RX", hop, size);
 
-	mutex_lock(&nhi->lock);
-	if (hop >= nhi->hop_count) {
-		dev_WARN(&nhi->pdev->dev, "invalid hop: %d\n", hop);
-		goto err;
-	}
-	if (transmit && nhi->tx_rings[hop]) {
-		dev_WARN(&nhi->pdev->dev, "TX hop %d already allocated\n", hop);
-		goto err;
-	} else if (!transmit && nhi->rx_rings[hop]) {
-		dev_WARN(&nhi->pdev->dev, "RX hop %d already allocated\n", hop);
-		goto err;
-	}
+	/* Tx Ring 2 is reserved for E2E workaround */
+	if (transmit && hop == RING_E2E_UNUSED_HOPID)
+		return NULL;
+
 	ring = kzalloc(sizeof(*ring), GFP_KERNEL);
 	if (!ring)
-		goto err;
+		return NULL;
 
-	mutex_init(&ring->lock);
+	spin_lock_init(&ring->lock);
 	INIT_LIST_HEAD(&ring->queue);
 	INIT_LIST_HEAD(&ring->in_flight);
 	INIT_WORK(&ring->work, ring_work);
@@ -353,55 +497,88 @@ static struct tb_ring *ring_alloc(struct tb_nhi *nhi, u32 hop, int size,
 	ring->is_tx = transmit;
 	ring->size = size;
 	ring->flags = flags;
+	ring->sof_mask = sof_mask;
+	ring->eof_mask = eof_mask;
 	ring->head = 0;
 	ring->tail = 0;
 	ring->running = false;
-
-	if (ring_request_msix(ring, flags & RING_FLAG_NO_SUSPEND))
-		goto err;
+	ring->start_poll = start_poll;
+	ring->poll_data = poll_data;
 
 	ring->descriptors = dma_alloc_coherent(&ring->nhi->pdev->dev,
 			size * sizeof(*ring->descriptors),
 			&ring->descriptors_dma, GFP_KERNEL | __GFP_ZERO);
 	if (!ring->descriptors)
-		goto err;
+		goto err_free_ring;
+
+	if (ring_request_msix(ring, flags & RING_FLAG_NO_SUSPEND))
+		goto err_free_descs;
+
+	if (nhi_alloc_hop(nhi, ring))
+		goto err_release_msix;
 
-	if (transmit)
-		nhi->tx_rings[hop] = ring;
-	else
-		nhi->rx_rings[hop] = ring;
-	mutex_unlock(&nhi->lock);
 	return ring;
 
-err:
-	if (ring)
-		mutex_destroy(&ring->lock);
+err_release_msix:
+	ring_release_msix(ring);
+err_free_descs:
+	dma_free_coherent(&ring->nhi->pdev->dev,
+			  ring->size * sizeof(*ring->descriptors),
+			  ring->descriptors, ring->descriptors_dma);
+err_free_ring:
 	kfree(ring);
-	mutex_unlock(&nhi->lock);
+
 	return NULL;
 }
 
-struct tb_ring *ring_alloc_tx(struct tb_nhi *nhi, int hop, int size,
-			      unsigned int flags)
+/**
+ * tb_ring_alloc_tx() - Allocate DMA ring for transmit
+ * @nhi: Pointer to the NHI the ring is to be allocated
+ * @hop: HopID (ring) to allocate
+ * @size: Number of entries in the ring
+ * @flags: Flags for the ring
+ */
+struct tb_ring *tb_ring_alloc_tx(struct tb_nhi *nhi, int hop, int size,
+				 unsigned int flags)
 {
-	return ring_alloc(nhi, hop, size, true, flags);
+	return tb_ring_alloc(nhi, hop, size, true, flags, 0, 0, NULL, NULL);
 }
+EXPORT_SYMBOL_GPL(tb_ring_alloc_tx);
 
-struct tb_ring *ring_alloc_rx(struct tb_nhi *nhi, int hop, int size,
-			      unsigned int flags)
+/**
+ * tb_ring_alloc_rx() - Allocate DMA ring for receive
+ * @nhi: Pointer to the NHI the ring is to be allocated
+ * @hop: HopID (ring) to allocate. Pass %-1 for automatic allocation.
+ * @size: Number of entries in the ring
+ * @flags: Flags for the ring
+ * @sof_mask: Mask of PDF values that start a frame
+ * @eof_mask: Mask of PDF values that end a frame
+ * @start_poll: If not %NULL the ring will call this function when an
+ *		interrupt is triggered and masked, instead of callback
+ *		in each Rx frame.
+ * @poll_data: Optional data passed to @start_poll
+ */
+struct tb_ring *tb_ring_alloc_rx(struct tb_nhi *nhi, int hop, int size,
+				 unsigned int flags, u16 sof_mask, u16 eof_mask,
+				 void (*start_poll)(void *), void *poll_data)
 {
-	return ring_alloc(nhi, hop, size, false, flags);
+	return tb_ring_alloc(nhi, hop, size, false, flags, sof_mask, eof_mask,
+			     start_poll, poll_data);
 }
+EXPORT_SYMBOL_GPL(tb_ring_alloc_rx);
 
 /**
- * ring_start() - enable a ring
+ * tb_ring_start() - enable a ring
  *
- * Must not be invoked in parallel with ring_stop().
+ * Must not be invoked in parallel with tb_ring_stop().
  */
-void ring_start(struct tb_ring *ring)
+void tb_ring_start(struct tb_ring *ring)
 {
-	mutex_lock(&ring->nhi->lock);
-	mutex_lock(&ring->lock);
+	u16 frame_size;
+	u32 flags;
+
+	spin_lock_irq(&ring->nhi->lock);
+	spin_lock(&ring->lock);
 	if (ring->nhi->going_away)
 		goto err;
 	if (ring->running) {
@@ -411,43 +588,65 @@ void ring_start(struct tb_ring *ring)
 	dev_info(&ring->nhi->pdev->dev, "starting %s %d\n",
 		 RING_TYPE(ring), ring->hop);
 
+	if (ring->flags & RING_FLAG_FRAME) {
+		/* Means 4096 */
+		frame_size = 0;
+		flags = RING_FLAG_ENABLE;
+	} else {
+		frame_size = TB_FRAME_SIZE;
+		flags = RING_FLAG_ENABLE | RING_FLAG_RAW;
+	}
+
+	if (ring->flags & RING_FLAG_E2E && !ring->is_tx) {
+		u32 hop;
+
+		/*
+		 * In order not to lose Rx packets we enable end-to-end
+		 * workaround which transfers Rx credits to an unused Tx
+		 * HopID.
+		 */
+		hop = RING_E2E_UNUSED_HOPID << REG_RX_OPTIONS_E2E_HOP_SHIFT;
+		hop &= REG_RX_OPTIONS_E2E_HOP_MASK;
+		flags |= hop | RING_FLAG_E2E_FLOW_CONTROL;
+	}
+
 	ring_iowrite64desc(ring, ring->descriptors_dma, 0);
 	if (ring->is_tx) {
 		ring_iowrite32desc(ring, ring->size, 12);
 		ring_iowrite32options(ring, 0, 4); /* time releated ? */
-		ring_iowrite32options(ring,
-				      RING_FLAG_ENABLE | RING_FLAG_RAW, 0);
+		ring_iowrite32options(ring, flags, 0);
 	} else {
-		ring_iowrite32desc(ring,
-				   (TB_FRAME_SIZE << 16) | ring->size, 12);
-		ring_iowrite32options(ring, 0xffffffff, 4); /* SOF EOF mask */
-		ring_iowrite32options(ring,
-				      RING_FLAG_ENABLE | RING_FLAG_RAW, 0);
+		u32 sof_eof_mask = ring->sof_mask << 16 | ring->eof_mask;
+
+		ring_iowrite32desc(ring, (frame_size << 16) | ring->size, 12);
+		ring_iowrite32options(ring, sof_eof_mask, 4);
+		ring_iowrite32options(ring, flags, 0);
 	}
 	ring_interrupt_active(ring, true);
 	ring->running = true;
 err:
-	mutex_unlock(&ring->lock);
-	mutex_unlock(&ring->nhi->lock);
+	spin_unlock(&ring->lock);
+	spin_unlock_irq(&ring->nhi->lock);
 }
-
+EXPORT_SYMBOL_GPL(tb_ring_start);
 
 /**
- * ring_stop() - shutdown a ring
+ * tb_ring_stop() - shutdown a ring
  *
  * Must not be invoked from a callback.
  *
- * This method will disable the ring. Further calls to ring_tx/ring_rx will
- * return -ESHUTDOWN until ring_stop has been called.
+ * This method will disable the ring. Further calls to
+ * tb_ring_tx/tb_ring_rx will return -ESHUTDOWN until ring_stop has been
+ * called.
  *
  * All enqueued frames will be canceled and their callbacks will be executed
  * with frame->canceled set to true (on the callback thread). This method
  * returns only after all callback invocations have finished.
  */
-void ring_stop(struct tb_ring *ring)
+void tb_ring_stop(struct tb_ring *ring)
 {
-	mutex_lock(&ring->nhi->lock);
-	mutex_lock(&ring->lock);
+	spin_lock_irq(&ring->nhi->lock);
+	spin_lock(&ring->lock);
 	dev_info(&ring->nhi->pdev->dev, "stopping %s %d\n",
 		 RING_TYPE(ring), ring->hop);
 	if (ring->nhi->going_away)
@@ -468,8 +667,8 @@ void ring_stop(struct tb_ring *ring)
 	ring->running = false;
 
 err:
-	mutex_unlock(&ring->lock);
-	mutex_unlock(&ring->nhi->lock);
+	spin_unlock(&ring->lock);
+	spin_unlock_irq(&ring->nhi->lock);
 
 	/*
 	 * schedule ring->work to invoke callbacks on all remaining frames.
@@ -477,9 +676,10 @@ err:
 	schedule_work(&ring->work);
 	flush_work(&ring->work);
 }
+EXPORT_SYMBOL_GPL(tb_ring_stop);
 
 /*
- * ring_free() - free ring
+ * tb_ring_free() - free ring
  *
  * When this method returns all invocations of ring->callback will have
  * finished.
@@ -488,9 +688,9 @@ err:
  *
  * Must NOT be called from ring_frame->callback!
  */
-void ring_free(struct tb_ring *ring)
+void tb_ring_free(struct tb_ring *ring)
 {
-	mutex_lock(&ring->nhi->lock);
+	spin_lock_irq(&ring->nhi->lock);
 	/*
 	 * Dissociate the ring from the NHI. This also ensures that
 	 * nhi_interrupt_work cannot reschedule ring->work.
@@ -504,6 +704,7 @@ void ring_free(struct tb_ring *ring)
 		dev_WARN(&ring->nhi->pdev->dev, "%s %d still running\n",
 			 RING_TYPE(ring), ring->hop);
 	}
+	spin_unlock_irq(&ring->nhi->lock);
 
 	ring_release_msix(ring);
 
@@ -520,16 +721,15 @@ void ring_free(struct tb_ring *ring)
 		 RING_TYPE(ring),
 		 ring->hop);
 
-	mutex_unlock(&ring->nhi->lock);
 	/**
 	 * ring->work can no longer be scheduled (it is scheduled only
 	 * by nhi_interrupt_work, ring_stop and ring_msix). Wait for it
 	 * to finish before freeing the ring.
 	 */
 	flush_work(&ring->work);
-	mutex_destroy(&ring->lock);
 	kfree(ring);
 }
+EXPORT_SYMBOL_GPL(tb_ring_free);
 
 /**
  * nhi_mailbox_cmd() - Send a command through NHI mailbox
@@ -595,7 +795,7 @@ static void nhi_interrupt_work(struct work_struct *work)
 	int type = 0; /* current interrupt type 0: TX, 1: RX, 2: RX overflow */
 	struct tb_ring *ring;
 
-	mutex_lock(&nhi->lock);
+	spin_lock_irq(&nhi->lock);
 
 	/*
 	 * Starting at REG_RING_NOTIFY_BASE there are three status bitfields
@@ -630,10 +830,12 @@ static void nhi_interrupt_work(struct work_struct *work)
 				 hop);
 			continue;
 		}
-		/* we do not check ring->running, this is done in ring->work */
-		schedule_work(&ring->work);
+
+		spin_lock(&ring->lock);
+		__ring_interrupt(ring);
+		spin_unlock(&ring->lock);
 	}
-	mutex_unlock(&nhi->lock);
+	spin_unlock_irq(&nhi->lock);
 }
 
 static irqreturn_t nhi_msi(int irq, void *data)
@@ -651,6 +853,22 @@ static int nhi_suspend_noirq(struct device *dev)
 	return tb_domain_suspend_noirq(tb);
 }
 
+static void nhi_enable_int_throttling(struct tb_nhi *nhi)
+{
+	/* Throttling is specified in 256ns increments */
+	u32 throttle = DIV_ROUND_UP(128 * NSEC_PER_USEC, 256);
+	unsigned int i;
+
+	/*
+	 * Configure interrupt throttling for all vectors even if we
+	 * only use few.
+	 */
+	for (i = 0; i < MSIX_MAX_VECS; i++) {
+		u32 reg = REG_INT_THROTTLING_RATE + i * 4;
+		iowrite32(throttle, nhi->iobase + reg);
+	}
+}
+
 static int nhi_resume_noirq(struct device *dev)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
@@ -663,6 +881,8 @@ static int nhi_resume_noirq(struct device *dev)
 	 */
 	if (!pci_device_is_present(pdev))
 		tb->nhi->going_away = true;
+	else
+		nhi_enable_int_throttling(tb->nhi);
 
 	return tb_domain_resume_noirq(tb);
 }
@@ -705,7 +925,6 @@ static void nhi_shutdown(struct tb_nhi *nhi)
 		devm_free_irq(&nhi->pdev->dev, nhi->pdev->irq, nhi);
 		flush_work(&nhi->interrupt_work);
 	}
-	mutex_destroy(&nhi->lock);
 	ida_destroy(&nhi->msix_ida);
 }
 
@@ -717,6 +936,8 @@ static int nhi_init_msi(struct tb_nhi *nhi)
 	/* In case someone left them on. */
 	nhi_disable_interrupts(nhi);
 
+	nhi_enable_int_throttling(nhi);
+
 	ida_init(&nhi->msix_ida);
 
 	/*
@@ -792,13 +1013,10 @@ static int nhi_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 		return res;
 	}
 
-	mutex_init(&nhi->lock);
+	spin_lock_init(&nhi->lock);
 
 	pci_set_master(pdev);
 
-	/* magic value - clock related? */
-	iowrite32(3906250 / 10000, nhi->iobase + 0x38c00);
-
 	tb = icm_probe(nhi);
 	if (!tb)
 		tb = tb_probe(nhi);
@@ -926,5 +1144,5 @@ static void __exit nhi_unload(void)
 	tb_domain_exit();
 }
 
-module_init(nhi_init);
+fs_initcall(nhi_init);
 module_exit(nhi_unload);