Merge branch 'tracing/core' into tracing/hw-breakpoints

Conflicts:
	arch/Kconfig
	kernel/trace/trace.h

Merge reason: resolve the conflicts, plus adopt to the new
              ring-buffer APIs.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

23 files changed, 1054 insertions, 468 deletions

diff --git a/drivers/usb/host/Kconfig b/drivers/usb/host/Kconfig
index 1576a0520ad..1a920c70b5a 100644
--- a/drivers/usb/host/Kconfig
+++ b/drivers/usb/host/Kconfig
@@ -181,26 +181,27 @@ config USB_OHCI_HCD_PPC_SOC
 	  Enables support for the USB controller on the MPC52xx or
 	  STB03xxx processor chip.  If unsure, say Y.
 
-config USB_OHCI_HCD_PPC_OF
-	bool "OHCI support for PPC USB controller on OF platform bus"
-	depends on USB_OHCI_HCD && PPC_OF
-	default y
-	---help---
-	  Enables support for the USB controller PowerPC present on the
-	  OpenFirmware platform bus.
-
 config USB_OHCI_HCD_PPC_OF_BE
-	bool "Support big endian HC"
-	depends on USB_OHCI_HCD_PPC_OF
-	default y
+	bool "OHCI support for OF platform bus (big endian)"
+	depends on USB_OHCI_HCD && PPC_OF
 	select USB_OHCI_BIG_ENDIAN_DESC
 	select USB_OHCI_BIG_ENDIAN_MMIO
+	---help---
+	  Enables support for big-endian USB controllers present on the
+	  OpenFirmware platform bus.
 
 config USB_OHCI_HCD_PPC_OF_LE
-	bool "Support little endian HC"
-	depends on USB_OHCI_HCD_PPC_OF
-	default n
+	bool "OHCI support for OF platform bus (little endian)"
+	depends on USB_OHCI_HCD && PPC_OF
 	select USB_OHCI_LITTLE_ENDIAN
+	---help---
+	  Enables support for little-endian USB controllers present on the
+	  OpenFirmware platform bus.
+
+config USB_OHCI_HCD_PPC_OF
+	bool
+	depends on USB_OHCI_HCD && PPC_OF
+	default USB_OHCI_HCD_PPC_OF_BE || USB_OHCI_HCD_PPC_OF_LE
 
 config USB_OHCI_HCD_PCI
 	bool "OHCI support for PCI-bus USB controllers"
@@ -337,10 +338,10 @@ config USB_R8A66597_HCD
 
 config SUPERH_ON_CHIP_R8A66597
 	boolean "Enable SuperH on-chip R8A66597 USB"
-	depends on USB_R8A66597_HCD && (CPU_SUBTYPE_SH7366 || CPU_SUBTYPE_SH7723)
+	depends on USB_R8A66597_HCD && (CPU_SUBTYPE_SH7366 || CPU_SUBTYPE_SH7723 || CPU_SUBTYPE_SH7724)
 	help
 	   This driver enables support for the on-chip R8A66597 in the
-	   SH7366 and SH7723 processors.
+	   SH7366, SH7723 and SH7724 processors.
 
 config USB_WHCI_HCD
 	tristate "Wireless USB Host Controller Interface (WHCI) driver (EXPERIMENTAL)"
diff --git a/drivers/usb/host/ehci-au1xxx.c b/drivers/usb/host/ehci-au1xxx.c
index c3a778bd359..59d208d94d4 100644
--- a/drivers/usb/host/ehci-au1xxx.c
+++ b/drivers/usb/host/ehci-au1xxx.c
@@ -113,6 +113,8 @@ static const struct hc_driver ehci_au1xxx_hc_driver = {
 	.bus_resume		= ehci_bus_resume,
 	.relinquish_port	= ehci_relinquish_port,
 	.port_handed_over	= ehci_port_handed_over,
+
+	.clear_tt_buffer_complete	= ehci_clear_tt_buffer_complete,
 };
 
 static int ehci_hcd_au1xxx_drv_probe(struct platform_device *pdev)
diff --git a/drivers/usb/host/ehci-fsl.c b/drivers/usb/host/ehci-fsl.c
index bf86809c512..991174937db 100644
--- a/drivers/usb/host/ehci-fsl.c
+++ b/drivers/usb/host/ehci-fsl.c
@@ -325,6 +325,8 @@ static const struct hc_driver ehci_fsl_hc_driver = {
 	.bus_resume = ehci_bus_resume,
 	.relinquish_port = ehci_relinquish_port,
 	.port_handed_over = ehci_port_handed_over,
+
+	.clear_tt_buffer_complete = ehci_clear_tt_buffer_complete,
 };
 
 static int ehci_fsl_drv_probe(struct platform_device *pdev)
diff --git a/drivers/usb/host/ehci-hcd.c b/drivers/usb/host/ehci-hcd.c
index 2b72473544d..11c627ce602 100644
--- a/drivers/usb/host/ehci-hcd.c
+++ b/drivers/usb/host/ehci-hcd.c
@@ -903,7 +903,8 @@ static int ehci_urb_dequeue(struct usb_hcd *hcd, struct urb *urb, int status)
 			/* already started */
 			break;
 		case QH_STATE_IDLE:
-			WARN_ON(1);
+			/* QH might be waiting for a Clear-TT-Buffer */
+			qh_completions(ehci, qh);
 			break;
 		}
 		break;
@@ -1003,6 +1004,8 @@ idle_timeout:
 		schedule_timeout_uninterruptible(1);
 		goto rescan;
 	case QH_STATE_IDLE:		/* fully unlinked */
+		if (qh->clearing_tt)
+			goto idle_timeout;
 		if (list_empty (&qh->qtd_list)) {
 			qh_put (qh);
 			break;
@@ -1030,12 +1033,14 @@ ehci_endpoint_reset(struct usb_hcd *hcd, struct usb_host_endpoint *ep)
 	struct ehci_hcd		*ehci = hcd_to_ehci(hcd);
 	struct ehci_qh		*qh;
 	int			eptype = usb_endpoint_type(&ep->desc);
+	int			epnum = usb_endpoint_num(&ep->desc);
+	int			is_out = usb_endpoint_dir_out(&ep->desc);
+	unsigned long		flags;
 
 	if (eptype != USB_ENDPOINT_XFER_BULK && eptype != USB_ENDPOINT_XFER_INT)
 		return;
 
- rescan:
-	spin_lock_irq(&ehci->lock);
+	spin_lock_irqsave(&ehci->lock, flags);
 	qh = ep->hcpriv;
 
 	/* For Bulk and Interrupt endpoints we maintain the toggle state
@@ -1044,29 +1049,24 @@ ehci_endpoint_reset(struct usb_hcd *hcd, struct usb_host_endpoint *ep)
 	 * the toggle bit in the QH.
 	 */
 	if (qh) {
+		usb_settoggle(qh->dev, epnum, is_out, 0);
 		if (!list_empty(&qh->qtd_list)) {
 			WARN_ONCE(1, "clear_halt for a busy endpoint\n");
-		} else if (qh->qh_state == QH_STATE_IDLE) {
-			qh->hw_token &= ~cpu_to_hc32(ehci, QTD_TOGGLE);
-		} else {
-			/* It's not safe to write into the overlay area
-			 * while the QH is active.  Unlink it first and
-			 * wait for the unlink to complete.
+		} else if (qh->qh_state == QH_STATE_LINKED) {
+
+			/* The toggle value in the QH can't be updated
+			 * while the QH is active.  Unlink it now;
+			 * re-linking will call qh_refresh().
 			 */
-			if (qh->qh_state == QH_STATE_LINKED) {
-				if (eptype == USB_ENDPOINT_XFER_BULK) {
-					unlink_async(ehci, qh);
-				} else {
-					intr_deschedule(ehci, qh);
-					(void) qh_schedule(ehci, qh);
-				}
+			if (eptype == USB_ENDPOINT_XFER_BULK) {
+				unlink_async(ehci, qh);
+			} else {
+				intr_deschedule(ehci, qh);
+				(void) qh_schedule(ehci, qh);
 			}
-			spin_unlock_irq(&ehci->lock);
-			schedule_timeout_uninterruptible(1);
-			goto rescan;
 		}
 	}
-	spin_unlock_irq(&ehci->lock);
+	spin_unlock_irqrestore(&ehci->lock, flags);
 }
 
 static int ehci_get_frame (struct usb_hcd *hcd)
diff --git a/drivers/usb/host/ehci-ixp4xx.c b/drivers/usb/host/ehci-ixp4xx.c
index a44bb4a9495..89b7c70c6ed 100644
--- a/drivers/usb/host/ehci-ixp4xx.c
+++ b/drivers/usb/host/ehci-ixp4xx.c
@@ -61,6 +61,8 @@ static const struct hc_driver ixp4xx_ehci_hc_driver = {
 #endif
 	.relinquish_port	= ehci_relinquish_port,
 	.port_handed_over	= ehci_port_handed_over,
+
+	.clear_tt_buffer_complete	= ehci_clear_tt_buffer_complete,
 };
 
 static int ixp4xx_ehci_probe(struct platform_device *pdev)
diff --git a/drivers/usb/host/ehci-orion.c b/drivers/usb/host/ehci-orion.c
index 770dd9aba62..1d283e1b2b8 100644
--- a/drivers/usb/host/ehci-orion.c
+++ b/drivers/usb/host/ehci-orion.c
@@ -105,6 +105,7 @@ static int ehci_orion_setup(struct usb_hcd *hcd)
 	struct ehci_hcd *ehci = hcd_to_ehci(hcd);
 	int retval;
 
+	ehci_reset(ehci);
 	retval = ehci_halt(ehci);
 	if (retval)
 		return retval;
@@ -118,7 +119,6 @@ static int ehci_orion_setup(struct usb_hcd *hcd)
 
 	hcd->has_tt = 1;
 
-	ehci_reset(ehci);
 	ehci_port_power(ehci, 0);
 
 	return retval;
@@ -165,6 +165,8 @@ static const struct hc_driver ehci_orion_hc_driver = {
 	.bus_resume = ehci_bus_resume,
 	.relinquish_port = ehci_relinquish_port,
 	.port_handed_over = ehci_port_handed_over,
+
+	.clear_tt_buffer_complete = ehci_clear_tt_buffer_complete,
 };
 
 static void __init
diff --git a/drivers/usb/host/ehci-pci.c b/drivers/usb/host/ehci-pci.c
index f3683e1da16..c2f1b7df918 100644
--- a/drivers/usb/host/ehci-pci.c
+++ b/drivers/usb/host/ehci-pci.c
@@ -404,6 +404,8 @@ static const struct hc_driver ehci_pci_hc_driver = {
 	.bus_resume =		ehci_bus_resume,
 	.relinquish_port =	ehci_relinquish_port,
 	.port_handed_over =	ehci_port_handed_over,
+
+	.clear_tt_buffer_complete	= ehci_clear_tt_buffer_complete,
 };
 
 /*-------------------------------------------------------------------------*/
diff --git a/drivers/usb/host/ehci-ppc-of.c b/drivers/usb/host/ehci-ppc-of.c
index fbd272288fc..36f96da129f 100644
--- a/drivers/usb/host/ehci-ppc-of.c
+++ b/drivers/usb/host/ehci-ppc-of.c
@@ -79,6 +79,8 @@ static const struct hc_driver ehci_ppc_of_hc_driver = {
 #endif
 	.relinquish_port	= ehci_relinquish_port,
 	.port_handed_over	= ehci_port_handed_over,
+
+	.clear_tt_buffer_complete	= ehci_clear_tt_buffer_complete,
 };
 
 
diff --git a/drivers/usb/host/ehci-ps3.c b/drivers/usb/host/ehci-ps3.c
index eecd2a0680a..1dee33b9139 100644
--- a/drivers/usb/host/ehci-ps3.c
+++ b/drivers/usb/host/ehci-ps3.c
@@ -75,9 +75,11 @@ static const struct hc_driver ps3_ehci_hc_driver = {
 #endif
 	.relinquish_port	= ehci_relinquish_port,
 	.port_handed_over	= ehci_port_handed_over,
+
+	.clear_tt_buffer_complete	= ehci_clear_tt_buffer_complete,
 };
 
-static int ps3_ehci_probe(struct ps3_system_bus_device *dev)
+static int __devinit ps3_ehci_probe(struct ps3_system_bus_device *dev)
 {
 	int result;
 	struct usb_hcd *hcd;
@@ -225,7 +227,7 @@ static int ps3_ehci_remove(struct ps3_system_bus_device *dev)
 	return 0;
 }
 
-static int ps3_ehci_driver_register(struct ps3_system_bus_driver *drv)
+static int __init ps3_ehci_driver_register(struct ps3_system_bus_driver *drv)
 {
 	return firmware_has_feature(FW_FEATURE_PS3_LV1)
 		? ps3_system_bus_driver_register(drv)
diff --git a/drivers/usb/host/ehci-q.c b/drivers/usb/host/ehci-q.c
index 3192f683f80..7673554fa64 100644
--- a/drivers/usb/host/ehci-q.c
+++ b/drivers/usb/host/ehci-q.c
@@ -93,6 +93,22 @@ qh_update (struct ehci_hcd *ehci, struct ehci_qh *qh, struct ehci_qtd *qtd)
 	qh->hw_qtd_next = QTD_NEXT(ehci, qtd->qtd_dma);
 	qh->hw_alt_next = EHCI_LIST_END(ehci);
 
+	/* Except for control endpoints, we make hardware maintain data
+	 * toggle (like OHCI) ... here (re)initialize the toggle in the QH,
+	 * and set the pseudo-toggle in udev. Only usb_clear_halt() will
+	 * ever clear it.
+	 */
+	if (!(qh->hw_info1 & cpu_to_hc32(ehci, 1 << 14))) {
+		unsigned	is_out, epnum;
+
+		is_out = !(qtd->hw_token & cpu_to_hc32(ehci, 1 << 8));
+		epnum = (hc32_to_cpup(ehci, &qh->hw_info1) >> 8) & 0x0f;
+		if (unlikely (!usb_gettoggle (qh->dev, epnum, is_out))) {
+			qh->hw_token &= ~cpu_to_hc32(ehci, QTD_TOGGLE);
+			usb_settoggle (qh->dev, epnum, is_out, 1);
+		}
+	}
+
 	/* HC must see latest qtd and qh data before we clear ACTIVE+HALT */
 	wmb ();
 	qh->hw_token &= cpu_to_hc32(ehci, QTD_TOGGLE | QTD_STS_PING);
@@ -123,6 +139,55 @@ qh_refresh (struct ehci_hcd *ehci, struct ehci_qh *qh)
 
 /*-------------------------------------------------------------------------*/
 
+static void qh_link_async(struct ehci_hcd *ehci, struct ehci_qh *qh);
+
+static void ehci_clear_tt_buffer_complete(struct usb_hcd *hcd,
+		struct usb_host_endpoint *ep)
+{
+	struct ehci_hcd		*ehci = hcd_to_ehci(hcd);
+	struct ehci_qh		*qh = ep->hcpriv;
+	unsigned long		flags;
+
+	spin_lock_irqsave(&ehci->lock, flags);
+	qh->clearing_tt = 0;
+	if (qh->qh_state == QH_STATE_IDLE && !list_empty(&qh->qtd_list)
+			&& HC_IS_RUNNING(hcd->state))
+		qh_link_async(ehci, qh);
+	spin_unlock_irqrestore(&ehci->lock, flags);
+}
+
+static void ehci_clear_tt_buffer(struct ehci_hcd *ehci, struct ehci_qh *qh,
+		struct urb *urb, u32 token)
+{
+
+	/* If an async split transaction gets an error or is unlinked,
+	 * the TT buffer may be left in an indeterminate state.  We
+	 * have to clear the TT buffer.
+	 *
+	 * Note: this routine is never called for Isochronous transfers.
+	 */
+	if (urb->dev->tt && !usb_pipeint(urb->pipe) && !qh->clearing_tt) {
+#ifdef DEBUG
+		struct usb_device *tt = urb->dev->tt->hub;
+		dev_dbg(&tt->dev,
+			"clear tt buffer port %d, a%d ep%d t%08x\n",
+			urb->dev->ttport, urb->dev->devnum,
+			usb_pipeendpoint(urb->pipe), token);
+#endif /* DEBUG */
+		if (!ehci_is_TDI(ehci)
+				|| urb->dev->tt->hub !=
+				   ehci_to_hcd(ehci)->self.root_hub) {
+			if (usb_hub_clear_tt_buffer(urb) == 0)
+				qh->clearing_tt = 1;
+		} else {
+
+			/* REVISIT ARC-derived cores don't clear the root
+			 * hub TT buffer in this way...
+			 */
+		}
+	}
+}
+
 static int qtd_copy_status (
 	struct ehci_hcd *ehci,
 	struct urb *urb,
@@ -149,6 +214,14 @@ static int qtd_copy_status (
 		if (token & QTD_STS_BABBLE) {
 			/* FIXME "must" disable babbling device's port too */
 			status = -EOVERFLOW;
+		/* CERR nonzero + halt --> stall */
+		} else if (QTD_CERR(token)) {
+			status = -EPIPE;
+
+		/* In theory, more than one of the following bits can be set
+		 * since they are sticky and the transaction is retried.
+		 * Which to test first is rather arbitrary.
+		 */
 		} else if (token & QTD_STS_MMF) {
 			/* fs/ls interrupt xfer missed the complete-split */
 			status = -EPROTO;
@@ -157,21 +230,15 @@ static int qtd_copy_status (
 				? -ENOSR  /* hc couldn't read data */
 				: -ECOMM; /* hc couldn't write data */
 		} else if (token & QTD_STS_XACT) {
-			/* timeout, bad crc, wrong PID, etc; retried */
-			if (QTD_CERR (token))
-				status = -EPIPE;
-			else {
-				ehci_dbg (ehci, "devpath %s ep%d%s 3strikes\n",
-					urb->dev->devpath,
-					usb_pipeendpoint (urb->pipe),
-					usb_pipein (urb->pipe) ? "in" : "out");
-				status = -EPROTO;
-			}
-		/* CERR nonzero + no errors + halt --> stall */
-		} else if (QTD_CERR (token))
-			status = -EPIPE;
-		else	/* unknown */
+			/* timeout, bad CRC, wrong PID, etc */
+			ehci_dbg(ehci, "devpath %s ep%d%s 3strikes\n",
+				urb->dev->devpath,
+				usb_pipeendpoint(urb->pipe),
+				usb_pipein(urb->pipe) ? "in" : "out");
 			status = -EPROTO;
+		} else {	/* unknown */
+			status = -EPROTO;
+		}
 
 		ehci_vdbg (ehci,
 			"dev%d ep%d%s qtd token %08x --> status %d\n",
@@ -179,28 +246,6 @@ static int qtd_copy_status (
 			usb_pipeendpoint (urb->pipe),
 			usb_pipein (urb->pipe) ? "in" : "out",
 			token, status);
-
-		/* if async CSPLIT failed, try cleaning out the TT buffer */
-		if (status != -EPIPE
-				&& urb->dev->tt
-				&& !usb_pipeint(urb->pipe)
-				&& ((token & QTD_STS_MMF) != 0
-					|| QTD_CERR(token) == 0)
-				&& (!ehci_is_TDI(ehci)
-			                || urb->dev->tt->hub !=
-					   ehci_to_hcd(ehci)->self.root_hub)) {
-#ifdef DEBUG
-			struct usb_device *tt = urb->dev->tt->hub;
-			dev_dbg (&tt->dev,
-				"clear tt buffer port %d, a%d ep%d t%08x\n",
-				urb->dev->ttport, urb->dev->devnum,
-				usb_pipeendpoint (urb->pipe), token);
-#endif /* DEBUG */
-			/* REVISIT ARC-derived cores don't clear the root
-			 * hub TT buffer in this way...
-			 */
-			usb_hub_tt_clear_buffer (urb->dev, urb->pipe);
-		}
 	}
 
 	return status;
@@ -330,12 +375,11 @@ qh_completions (struct ehci_hcd *ehci, struct ehci_qh *qh)
 				 */
 				if ((token & QTD_STS_XACT) &&
 						QTD_CERR(token) == 0 &&
-						--qh->xacterrs > 0 &&
+						++qh->xacterrs < QH_XACTERR_MAX &&
 						!urb->unlinked) {
 					ehci_dbg(ehci,
 	"detected XactErr len %zu/%zu retry %d\n",
-	qtd->length - QTD_LENGTH(token), qtd->length,
-	QH_XACTERR_MAX - qh->xacterrs);
+	qtd->length - QTD_LENGTH(token), qtd->length, qh->xacterrs);
 
 					/* reset the token in the qtd and the
 					 * qh overlay (which still contains
@@ -391,9 +435,16 @@ qh_completions (struct ehci_hcd *ehci, struct ehci_qh *qh)
 			/* qh unlinked; token in overlay may be most current */
 			if (state == QH_STATE_IDLE
 					&& cpu_to_hc32(ehci, qtd->qtd_dma)
-						== qh->hw_current)
+						== qh->hw_current) {
 				token = hc32_to_cpu(ehci, qh->hw_token);
 
+				/* An unlink may leave an incomplete
+				 * async transaction in the TT buffer.
+				 * We have to clear it.
+				 */
+				ehci_clear_tt_buffer(ehci, qh, urb, token);
+			}
+
 			/* force halt for unlinked or blocked qh, so we'll
 			 * patch the qh later and so that completions can't
 			 * activate it while we "know" it's stopped.
@@ -419,6 +470,13 @@ halt:
 					&& (qtd->hw_alt_next
 						& EHCI_LIST_END(ehci)))
 				last_status = -EINPROGRESS;
+
+			/* As part of low/full-speed endpoint-halt processing
+			 * we must clear the TT buffer (11.17.5).
+			 */
+			if (unlikely(last_status != -EINPROGRESS &&
+					last_status != -EREMOTEIO))
+				ehci_clear_tt_buffer(ehci, qh, urb, token);
 		}
 
 		/* if we're removing something not at the queue head,
@@ -435,7 +493,7 @@ halt:
 		last = qtd;
 
 		/* reinit the xacterr counter for the next qtd */
-		qh->xacterrs = QH_XACTERR_MAX;
+		qh->xacterrs = 0;
 	}
 
 	/* last urb's completion might still need calling */
@@ -834,6 +892,7 @@ done:
 	qh->qh_state = QH_STATE_IDLE;
 	qh->hw_info1 = cpu_to_hc32(ehci, info1);
 	qh->hw_info2 = cpu_to_hc32(ehci, info2);
+	usb_settoggle (urb->dev, usb_pipeendpoint (urb->pipe), !is_input, 1);
 	qh_refresh (ehci, qh);
 	return qh;
 }
@@ -847,6 +906,10 @@ static void qh_link_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
 	__hc32		dma = QH_NEXT(ehci, qh->qh_dma);
 	struct ehci_qh	*head;
 
+	/* Don't link a QH if there's a Clear-TT-Buffer pending */
+	if (unlikely(qh->clearing_tt))
+		return;
+
 	/* (re)start the async schedule? */
 	head = ehci->async;
 	timer_action_done (ehci, TIMER_ASYNC_OFF);
@@ -864,7 +927,7 @@ static void qh_link_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
 		}
 	}
 
-	/* clear halt and maybe recover from silicon quirk */
+	/* clear halt and/or toggle; and maybe recover from silicon quirk */
 	if (qh->qh_state == QH_STATE_IDLE)
 		qh_refresh (ehci, qh);
 
@@ -876,7 +939,8 @@ static void qh_link_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
 	head->qh_next.qh = qh;
 	head->hw_next = dma;
 
-	qh->xacterrs = QH_XACTERR_MAX;
+	qh_get(qh);
+	qh->xacterrs = 0;
 	qh->qh_state = QH_STATE_LINKED;
 	/* qtd completions reported later by interrupt */
 }
@@ -1016,7 +1080,7 @@ submit_async (
 	 * the HC and TT handle it when the TT has a buffer ready.
 	 */
 	if (likely (qh->qh_state == QH_STATE_IDLE))
-		qh_link_async (ehci, qh_get (qh));
+		qh_link_async(ehci, qh);
  done:
 	spin_unlock_irqrestore (&ehci->lock, flags);
 	if (unlikely (qh == NULL))
@@ -1051,8 +1115,6 @@ static void end_unlink_async (struct ehci_hcd *ehci)
 			&& HC_IS_RUNNING (ehci_to_hcd(ehci)->state))
 		qh_link_async (ehci, qh);
 	else {
-		qh_put (qh);		// refcount from async list
-
 		/* it's not free to turn the async schedule on/off; leave it
 		 * active but idle for a while once it empties.
 		 */
@@ -1060,6 +1122,7 @@ static void end_unlink_async (struct ehci_hcd *ehci)
 				&& ehci->async->qh_next.qh == NULL)
 			timer_action (ehci, TIMER_ASYNC_OFF);
 	}
+	qh_put(qh);			/* refcount from async list */
 
 	if (next) {
 		ehci->reclaim = NULL;
diff --git a/drivers/usb/host/ehci-sched.c b/drivers/usb/host/ehci-sched.c
index 9d1babc7ff6..edd61ee9032 100644
--- a/drivers/usb/host/ehci-sched.c
+++ b/drivers/usb/host/ehci-sched.c
@@ -542,6 +542,7 @@ static int qh_link_periodic (struct ehci_hcd *ehci, struct ehci_qh *qh)
 		}
 	}
 	qh->qh_state = QH_STATE_LINKED;
+	qh->xacterrs = 0;
 	qh_get (qh);
 
 	/* update per-qh bandwidth for usbfs */
@@ -1619,11 +1620,14 @@ itd_complete (
 				desc->status = -EPROTO;
 
 			/* HC need not update length with this error */
-			if (!(t & EHCI_ISOC_BABBLE))
-				desc->actual_length = EHCI_ITD_LENGTH (t);
+			if (!(t & EHCI_ISOC_BABBLE)) {
+				desc->actual_length = EHCI_ITD_LENGTH(t);
+				urb->actual_length += desc->actual_length;
+			}
 		} else if (likely ((t & EHCI_ISOC_ACTIVE) == 0)) {
 			desc->status = 0;
-			desc->actual_length = EHCI_ITD_LENGTH (t);
+			desc->actual_length = EHCI_ITD_LENGTH(t);
+			urb->actual_length += desc->actual_length;
 		} else {
 			/* URB was too late */
 			desc->status = -EXDEV;
@@ -2014,7 +2018,8 @@ sitd_complete (
 			desc->status = -EPROTO;
 	} else {
 		desc->status = 0;
-		desc->actual_length = desc->length - SITD_LENGTH (t);
+		desc->actual_length = desc->length - SITD_LENGTH(t);
+		urb->actual_length += desc->actual_length;
 	}
 	stream->depth -= stream->interval << 3;
 
diff --git a/drivers/usb/host/ehci.h b/drivers/usb/host/ehci.h
index 90ad3395bb2..2bfff30f470 100644
--- a/drivers/usb/host/ehci.h
+++ b/drivers/usb/host/ehci.h
@@ -354,7 +354,9 @@ struct ehci_qh {
 	unsigned short		period;		/* polling interval */
 	unsigned short		start;		/* where polling starts */
 #define NO_FRAME ((unsigned short)~0)			/* pick new start */
+
 	struct usb_device	*dev;		/* access to TT */
+	unsigned		clearing_tt:1;	/* Clear-TT-Buf in progress */
 } __attribute__ ((aligned (32)));
 
 /*-------------------------------------------------------------------------*/
diff --git a/drivers/usb/host/fhci-sched.c b/drivers/usb/host/fhci-sched.c
index bb63b68ddb7..62a226b6167 100644
--- a/drivers/usb/host/fhci-sched.c
+++ b/drivers/usb/host/fhci-sched.c
@@ -576,9 +576,7 @@ irqreturn_t fhci_irq(struct usb_hcd *hcd)
 			out_be16(&usb->fhci->regs->usb_event,
 				 usb->saved_msk);
 		} else if (usb->port_status == FHCI_PORT_DISABLED) {
-			if (fhci_ioports_check_bus_state(fhci) == 1 &&
-					usb->port_status != FHCI_PORT_LOW &&
-					usb->port_status != FHCI_PORT_FULL)
+			if (fhci_ioports_check_bus_state(fhci) == 1)
 				fhci_device_connected_interrupt(fhci);
 		}
 		usb_er &= ~USB_E_RESET_MASK;
@@ -605,9 +603,7 @@ irqreturn_t fhci_irq(struct usb_hcd *hcd)
 	}
 
 	if (usb_er & USB_E_IDLE_MASK) {
-		if (usb->port_status == FHCI_PORT_DISABLED &&
-				usb->port_status != FHCI_PORT_LOW &&
-				usb->port_status != FHCI_PORT_FULL) {
+		if (usb->port_status == FHCI_PORT_DISABLED) {
 			usb_er &= ~USB_E_RESET_MASK;
 			fhci_device_connected_interrupt(fhci);
 		} else if (usb->port_status ==
diff --git a/drivers/usb/host/isp1760-if.c b/drivers/usb/host/isp1760-if.c
index 3fa3a170279..d4feebfc63b 100644
--- a/drivers/usb/host/isp1760-if.c
+++ b/drivers/usb/host/isp1760-if.c
@@ -361,7 +361,7 @@ static int __devexit isp1760_plat_remove(struct platform_device *pdev)
 
 static struct platform_driver isp1760_plat_driver = {
 	.probe	= isp1760_plat_probe,
-	.remove	= isp1760_plat_remove,
+	.remove	= __devexit_p(isp1760_plat_remove),
 	.driver	= {
 		.name	= "isp1760",
 	},
diff --git a/drivers/usb/host/ohci-omap.c b/drivers/usb/host/ohci-omap.c
index f3aaba35e91..83cbecd2a1e 100644
--- a/drivers/usb/host/ohci-omap.c
+++ b/drivers/usb/host/ohci-omap.c
@@ -282,6 +282,7 @@ static int ohci_omap_init(struct usb_hcd *hcd)
 static void ohci_omap_stop(struct usb_hcd *hcd)
 {
 	dev_dbg(hcd->self.controller, "stopping USB Controller\n");
+	ohci_stop(hcd);
 	omap_ohci_clock_power(0);
 }
 
diff --git a/drivers/usb/host/ohci-ps3.c b/drivers/usb/host/ohci-ps3.c
index 1d56259c5db..700950455f4 100644
--- a/drivers/usb/host/ohci-ps3.c
+++ b/drivers/usb/host/ohci-ps3.c
@@ -75,7 +75,7 @@ static const struct hc_driver ps3_ohci_hc_driver = {
 #endif
 };
 
-static int ps3_ohci_probe(struct ps3_system_bus_device *dev)
+static int __devinit ps3_ohci_probe(struct ps3_system_bus_device *dev)
 {
 	int result;
 	struct usb_hcd *hcd;
@@ -224,7 +224,7 @@ static int ps3_ohci_remove(struct ps3_system_bus_device *dev)
 	return 0;
 }
 
-static int ps3_ohci_driver_register(struct ps3_system_bus_driver *drv)
+static int __init ps3_ohci_driver_register(struct ps3_system_bus_driver *drv)
 {
 	return firmware_has_feature(FW_FEATURE_PS3_LV1)
 		? ps3_system_bus_driver_register(drv)
diff --git a/drivers/usb/host/r8a66597-hcd.c b/drivers/usb/host/r8a66597-hcd.c
index 56976cc0352..e18f74946e6 100644
--- a/drivers/usb/host/r8a66597-hcd.c
+++ b/drivers/usb/host/r8a66597-hcd.c
@@ -26,7 +26,6 @@
 #include <linux/module.h>
 #include <linux/kernel.h>
 #include <linux/sched.h>
-#include <linux/smp_lock.h>
 #include <linux/errno.h>
 #include <linux/init.h>
 #include <linux/timer.h>
diff --git a/drivers/usb/host/xhci-dbg.c b/drivers/usb/host/xhci-dbg.c
index 2501c571f85..705e3432415 100644
--- a/drivers/usb/host/xhci-dbg.c
+++ b/drivers/usb/host/xhci-dbg.c
@@ -173,6 +173,7 @@ void xhci_print_ir_set(struct xhci_hcd *xhci, struct xhci_intr_reg *ir_set, int
 {
 	void *addr;
 	u32 temp;
+	u64 temp_64;
 
 	addr = &ir_set->irq_pending;
 	temp = xhci_readl(xhci, addr);
@@ -200,25 +201,15 @@ void xhci_print_ir_set(struct xhci_hcd *xhci, struct xhci_intr_reg *ir_set, int
 		xhci_dbg(xhci, "  WARN: %p: ir_set.rsvd = 0x%x\n",
 				addr, (unsigned int)temp);
 
-	addr = &ir_set->erst_base[0];
-	temp = xhci_readl(xhci, addr);
-	xhci_dbg(xhci, "  %p: ir_set.erst_base[0] = 0x%x\n",
-			addr, (unsigned int) temp);
-
-	addr = &ir_set->erst_base[1];
-	temp = xhci_readl(xhci, addr);
-	xhci_dbg(xhci, "  %p: ir_set.erst_base[1] = 0x%x\n",
-			addr, (unsigned int) temp);
+	addr = &ir_set->erst_base;
+	temp_64 = xhci_read_64(xhci, addr);
+	xhci_dbg(xhci, "  %p: ir_set.erst_base = @%08llx\n",
+			addr, temp_64);
 
-	addr = &ir_set->erst_dequeue[0];
-	temp = xhci_readl(xhci, addr);
-	xhci_dbg(xhci, "  %p: ir_set.erst_dequeue[0] = 0x%x\n",
-			addr, (unsigned int) temp);
-
-	addr = &ir_set->erst_dequeue[1];
-	temp = xhci_readl(xhci, addr);
-	xhci_dbg(xhci, "  %p: ir_set.erst_dequeue[1] = 0x%x\n",
-			addr, (unsigned int) temp);
+	addr = &ir_set->erst_dequeue;
+	temp_64 = xhci_read_64(xhci, addr);
+	xhci_dbg(xhci, "  %p: ir_set.erst_dequeue = @%08llx\n",
+			addr, temp_64);
 }
 
 void xhci_print_run_regs(struct xhci_hcd *xhci)
@@ -268,8 +259,7 @@ void xhci_debug_trb(struct xhci_hcd *xhci, union xhci_trb *trb)
 		xhci_dbg(xhci, "Link TRB:\n");
 		xhci_print_trb_offsets(xhci, trb);
 
-		address = trb->link.segment_ptr[0] +
-			(((u64) trb->link.segment_ptr[1]) << 32);
+		address = trb->link.segment_ptr;
 		xhci_dbg(xhci, "Next ring segment DMA address = 0x%llx\n", address);
 
 		xhci_dbg(xhci, "Interrupter target = 0x%x\n",
@@ -282,8 +272,7 @@ void xhci_debug_trb(struct xhci_hcd *xhci, union xhci_trb *trb)
 				(unsigned int) (trb->link.control & TRB_NO_SNOOP));
 		break;
 	case TRB_TYPE(TRB_TRANSFER):
-		address = trb->trans_event.buffer[0] +
-			(((u64) trb->trans_event.buffer[1]) << 32);
+		address = trb->trans_event.buffer;
 		/*
 		 * FIXME: look at flags to figure out if it's an address or if
 		 * the data is directly in the buffer field.
@@ -291,8 +280,7 @@ void xhci_debug_trb(struct xhci_hcd *xhci, union xhci_trb *trb)
 		xhci_dbg(xhci, "DMA address or buffer contents= %llu\n", address);
 		break;
 	case TRB_TYPE(TRB_COMPLETION):
-		address = trb->event_cmd.cmd_trb[0] +
-			(((u64) trb->event_cmd.cmd_trb[1]) << 32);
+		address = trb->event_cmd.cmd_trb;
 		xhci_dbg(xhci, "Command TRB pointer = %llu\n", address);
 		xhci_dbg(xhci, "Completion status = %u\n",
 				(unsigned int) GET_COMP_CODE(trb->event_cmd.status));
@@ -328,8 +316,8 @@ void xhci_debug_segment(struct xhci_hcd *xhci, struct xhci_segment *seg)
 	for (i = 0; i < TRBS_PER_SEGMENT; ++i) {
 		trb = &seg->trbs[i];
 		xhci_dbg(xhci, "@%08x %08x %08x %08x %08x\n", addr,
-				(unsigned int) trb->link.segment_ptr[0],
-				(unsigned int) trb->link.segment_ptr[1],
+				lower_32_bits(trb->link.segment_ptr),
+				upper_32_bits(trb->link.segment_ptr),
 				(unsigned int) trb->link.intr_target,
 				(unsigned int) trb->link.control);
 		addr += sizeof(*trb);
@@ -386,8 +374,8 @@ void xhci_dbg_erst(struct xhci_hcd *xhci, struct xhci_erst *erst)
 		entry = &erst->entries[i];
 		xhci_dbg(xhci, "@%08x %08x %08x %08x %08x\n",
 				(unsigned int) addr,
-				(unsigned int) entry->seg_addr[0],
-				(unsigned int) entry->seg_addr[1],
+				lower_32_bits(entry->seg_addr),
+				upper_32_bits(entry->seg_addr),
 				(unsigned int) entry->seg_size,
 				(unsigned int) entry->rsvd);
 		addr += sizeof(*entry);
@@ -396,90 +384,147 @@ void xhci_dbg_erst(struct xhci_hcd *xhci, struct xhci_erst *erst)
 
 void xhci_dbg_cmd_ptrs(struct xhci_hcd *xhci)
 {
-	u32 val;
+	u64 val;
 
-	val = xhci_readl(xhci, &xhci->op_regs->cmd_ring[0]);
-	xhci_dbg(xhci, "// xHC command ring deq ptr low bits + flags = 0x%x\n", val);
-	val = xhci_readl(xhci, &xhci->op_regs->cmd_ring[1]);
-	xhci_dbg(xhci, "// xHC command ring deq ptr high bits = 0x%x\n", val);
+	val = xhci_read_64(xhci, &xhci->op_regs->cmd_ring);
+	xhci_dbg(xhci, "// xHC command ring deq ptr low bits + flags = @%08x\n",
+			lower_32_bits(val));
+	xhci_dbg(xhci, "// xHC command ring deq ptr high bits = @%08x\n",
+			upper_32_bits(val));
 }
 
-void xhci_dbg_ctx(struct xhci_hcd *xhci, struct xhci_device_control *ctx, dma_addr_t dma, unsigned int last_ep)
+/* Print the last 32 bytes for 64-byte contexts */
+static void dbg_rsvd64(struct xhci_hcd *xhci, u64 *ctx, dma_addr_t dma)
+{
+	int i;
+	for (i = 0; i < 4; ++i) {
+		xhci_dbg(xhci, "@%p (virt) @%08llx "
+			 "(dma) %#08llx - rsvd64[%d]\n",
+			 &ctx[4 + i], (unsigned long long)dma,
+			 ctx[4 + i], i);
+		dma += 8;
+	}
+}
+
+void xhci_dbg_slot_ctx(struct xhci_hcd *xhci, struct xhci_container_ctx *ctx)
 {
-	int i, j;
-	int last_ep_ctx = 31;
 	/* Fields are 32 bits wide, DMA addresses are in bytes */
 	int field_size = 32 / 8;
+	int i;
 
-	xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - drop flags\n",
-			&ctx->drop_flags, (unsigned long long)dma,
-			ctx->drop_flags);
-	dma += field_size;
-	xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - add flags\n",
-			&ctx->add_flags, (unsigned long long)dma,
-			ctx->add_flags);
-	dma += field_size;
-	for (i = 0; i > 6; ++i) {
-		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - rsvd[%d]\n",
-				&ctx->rsvd[i], (unsigned long long)dma,
-				ctx->rsvd[i], i);
-		dma += field_size;
-	}
+	struct xhci_slot_ctx *slot_ctx = xhci_get_slot_ctx(xhci, ctx);
+	dma_addr_t dma = ctx->dma + ((unsigned long)slot_ctx - (unsigned long)ctx);
+	int csz = HCC_64BYTE_CONTEXT(xhci->hcc_params);
 
 	xhci_dbg(xhci, "Slot Context:\n");
 	xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - dev_info\n",
-			&ctx->slot.dev_info,
-			(unsigned long long)dma, ctx->slot.dev_info);
+			&slot_ctx->dev_info,
+			(unsigned long long)dma, slot_ctx->dev_info);
 	dma += field_size;
 	xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - dev_info2\n",
-			&ctx->slot.dev_info2,
-			(unsigned long long)dma, ctx->slot.dev_info2);
+			&slot_ctx->dev_info2,
+			(unsigned long long)dma, slot_ctx->dev_info2);
 	dma += field_size;
 	xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - tt_info\n",
-			&ctx->slot.tt_info,
-			(unsigned long long)dma, ctx->slot.tt_info);
+			&slot_ctx->tt_info,
+			(unsigned long long)dma, slot_ctx->tt_info);
 	dma += field_size;
 	xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - dev_state\n",
-			&ctx->slot.dev_state,
-			(unsigned long long)dma, ctx->slot.dev_state);
+			&slot_ctx->dev_state,
+			(unsigned long long)dma, slot_ctx->dev_state);
 	dma += field_size;
-	for (i = 0; i > 4; ++i) {
+	for (i = 0; i < 4; ++i) {
 		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - rsvd[%d]\n",
-				&ctx->slot.reserved[i], (unsigned long long)dma,
-				ctx->slot.reserved[i], i);
+				&slot_ctx->reserved[i], (unsigned long long)dma,
+				slot_ctx->reserved[i], i);
 		dma += field_size;
 	}
 
+	if (csz)
+		dbg_rsvd64(xhci, (u64 *)slot_ctx, dma);
+}
+
+void xhci_dbg_ep_ctx(struct xhci_hcd *xhci,
+		     struct xhci_container_ctx *ctx,
+		     unsigned int last_ep)
+{
+	int i, j;
+	int last_ep_ctx = 31;
+	/* Fields are 32 bits wide, DMA addresses are in bytes */
+	int field_size = 32 / 8;
+	int csz = HCC_64BYTE_CONTEXT(xhci->hcc_params);
+
 	if (last_ep < 31)
 		last_ep_ctx = last_ep + 1;
 	for (i = 0; i < last_ep_ctx; ++i) {
+		struct xhci_ep_ctx *ep_ctx = xhci_get_ep_ctx(xhci, ctx, i);
+		dma_addr_t dma = ctx->dma +
+			((unsigned long)ep_ctx - (unsigned long)ctx);
+
 		xhci_dbg(xhci, "Endpoint %02d Context:\n", i);
 		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - ep_info\n",
-				&ctx->ep[i].ep_info,
-				(unsigned long long)dma, ctx->ep[i].ep_info);
+				&ep_ctx->ep_info,
+				(unsigned long long)dma, ep_ctx->ep_info);
 		dma += field_size;
 		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - ep_info2\n",
-				&ctx->ep[i].ep_info2,
-				(unsigned long long)dma, ctx->ep[i].ep_info2);
-		dma += field_size;
-		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - deq[0]\n",
-				&ctx->ep[i].deq[0],
-				(unsigned long long)dma, ctx->ep[i].deq[0]);
-		dma += field_size;
-		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - deq[1]\n",
-				&ctx->ep[i].deq[1],
-				(unsigned long long)dma, ctx->ep[i].deq[1]);
+				&ep_ctx->ep_info2,
+				(unsigned long long)dma, ep_ctx->ep_info2);
 		dma += field_size;
+		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08llx - deq\n",
+				&ep_ctx->deq,
+				(unsigned long long)dma, ep_ctx->deq);
+		dma += 2*field_size;
 		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - tx_info\n",
-				&ctx->ep[i].tx_info,
-				(unsigned long long)dma, ctx->ep[i].tx_info);
+				&ep_ctx->tx_info,
+				(unsigned long long)dma, ep_ctx->tx_info);
 		dma += field_size;
 		for (j = 0; j < 3; ++j) {
 			xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - rsvd[%d]\n",
-					&ctx->ep[i].reserved[j],
+					&ep_ctx->reserved[j],
 					(unsigned long long)dma,
-					ctx->ep[i].reserved[j], j);
+					ep_ctx->reserved[j], j);
+			dma += field_size;
+		}
+
+		if (csz)
+			dbg_rsvd64(xhci, (u64 *)ep_ctx, dma);
+	}
+}
+
+void xhci_dbg_ctx(struct xhci_hcd *xhci,
+		  struct xhci_container_ctx *ctx,
+		  unsigned int last_ep)
+{
+	int i;
+	/* Fields are 32 bits wide, DMA addresses are in bytes */
+	int field_size = 32 / 8;
+	struct xhci_slot_ctx *slot_ctx;
+	dma_addr_t dma = ctx->dma;
+	int csz = HCC_64BYTE_CONTEXT(xhci->hcc_params);
+
+	if (ctx->type == XHCI_CTX_TYPE_INPUT) {
+		struct xhci_input_control_ctx *ctrl_ctx =
+			xhci_get_input_control_ctx(xhci, ctx);
+		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - drop flags\n",
+			 &ctrl_ctx->drop_flags, (unsigned long long)dma,
+			 ctrl_ctx->drop_flags);
+		dma += field_size;
+		xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - add flags\n",
+			 &ctrl_ctx->add_flags, (unsigned long long)dma,
+			 ctrl_ctx->add_flags);
+		dma += field_size;
+		for (i = 0; i < 6; ++i) {
+			xhci_dbg(xhci, "@%p (virt) @%08llx (dma) %#08x - rsvd2[%d]\n",
+				 &ctrl_ctx->rsvd2[i], (unsigned long long)dma,
+				 ctrl_ctx->rsvd2[i], i);
 			dma += field_size;
 		}
+
+		if (csz)
+			dbg_rsvd64(xhci, (u64 *)ctrl_ctx, dma);
 	}
+
+	slot_ctx = xhci_get_slot_ctx(xhci, ctx);
+	xhci_dbg_slot_ctx(xhci, ctx);
+	xhci_dbg_ep_ctx(xhci, ctx, last_ep);
 }
diff --git a/drivers/usb/host/xhci-hcd.c b/drivers/usb/host/xhci-hcd.c
index dba3e07ccd0..816c39caca1 100644
--- a/drivers/usb/host/xhci-hcd.c
+++ b/drivers/usb/host/xhci-hcd.c
@@ -103,7 +103,10 @@ int xhci_reset(struct xhci_hcd *xhci)
 	u32 state;
 
 	state = xhci_readl(xhci, &xhci->op_regs->status);
-	BUG_ON((state & STS_HALT) == 0);
+	if ((state & STS_HALT) == 0) {
+		xhci_warn(xhci, "Host controller not halted, aborting reset.\n");
+		return 0;
+	}
 
 	xhci_dbg(xhci, "// Reset the HC\n");
 	command = xhci_readl(xhci, &xhci->op_regs->command);
@@ -226,6 +229,7 @@ int xhci_init(struct usb_hcd *hcd)
 static void xhci_work(struct xhci_hcd *xhci)
 {
 	u32 temp;
+	u64 temp_64;
 
 	/*
 	 * Clear the op reg interrupt status first,
@@ -248,9 +252,9 @@ static void xhci_work(struct xhci_hcd *xhci)
 	/* FIXME this should be a delayed service routine that clears the EHB */
 	xhci_handle_event(xhci);
 
-	/* Clear the event handler busy flag; the event ring should be empty. */
-	temp = xhci_readl(xhci, &xhci->ir_set->erst_dequeue[0]);
-	xhci_writel(xhci, temp & ~ERST_EHB, &xhci->ir_set->erst_dequeue[0]);
+	/* Clear the event handler busy flag (RW1C); the event ring should be empty. */
+	temp_64 = xhci_read_64(xhci, &xhci->ir_set->erst_dequeue);
+	xhci_write_64(xhci, temp_64 | ERST_EHB, &xhci->ir_set->erst_dequeue);
 	/* Flush posted writes -- FIXME is this necessary? */
 	xhci_readl(xhci, &xhci->ir_set->irq_pending);
 }
@@ -266,19 +270,34 @@ irqreturn_t xhci_irq(struct usb_hcd *hcd)
 {
 	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
 	u32 temp, temp2;
+	union xhci_trb *trb;
 
 	spin_lock(&xhci->lock);
+	trb = xhci->event_ring->dequeue;
 	/* Check if the xHC generated the interrupt, or the irq is shared */
 	temp = xhci_readl(xhci, &xhci->op_regs->status);
 	temp2 = xhci_readl(xhci, &xhci->ir_set->irq_pending);
+	if (temp == 0xffffffff && temp2 == 0xffffffff)
+		goto hw_died;
+
 	if (!(temp & STS_EINT) && !ER_IRQ_PENDING(temp2)) {
 		spin_unlock(&xhci->lock);
 		return IRQ_NONE;
 	}
+	xhci_dbg(xhci, "op reg status = %08x\n", temp);
+	xhci_dbg(xhci, "ir set irq_pending = %08x\n", temp2);
+	xhci_dbg(xhci, "Event ring dequeue ptr:\n");
+	xhci_dbg(xhci, "@%llx %08x %08x %08x %08x\n",
+			(unsigned long long)xhci_trb_virt_to_dma(xhci->event_ring->deq_seg, trb),
+			lower_32_bits(trb->link.segment_ptr),
+			upper_32_bits(trb->link.segment_ptr),
+			(unsigned int) trb->link.intr_target,
+			(unsigned int) trb->link.control);
 
 	if (temp & STS_FATAL) {
 		xhci_warn(xhci, "WARNING: Host System Error\n");
 		xhci_halt(xhci);
+hw_died:
 		xhci_to_hcd(xhci)->state = HC_STATE_HALT;
 		spin_unlock(&xhci->lock);
 		return -ESHUTDOWN;
@@ -295,6 +314,7 @@ void xhci_event_ring_work(unsigned long arg)
 {
 	unsigned long flags;
 	int temp;
+	u64 temp_64;
 	struct xhci_hcd *xhci = (struct xhci_hcd *) arg;
 	int i, j;
 
@@ -311,9 +331,9 @@ void xhci_event_ring_work(unsigned long arg)
 	xhci_dbg(xhci, "Event ring:\n");
 	xhci_debug_segment(xhci, xhci->event_ring->deq_seg);
 	xhci_dbg_ring_ptrs(xhci, xhci->event_ring);
-	temp = xhci_readl(xhci, &xhci->ir_set->erst_dequeue[0]);
-	temp &= ERST_PTR_MASK;
-	xhci_dbg(xhci, "ERST deq = 0x%x\n", temp);
+	temp_64 = xhci_read_64(xhci, &xhci->ir_set->erst_dequeue);
+	temp_64 &= ~ERST_PTR_MASK;
+	xhci_dbg(xhci, "ERST deq = 64'h%0lx\n", (long unsigned int) temp_64);
 	xhci_dbg(xhci, "Command ring:\n");
 	xhci_debug_segment(xhci, xhci->cmd_ring->deq_seg);
 	xhci_dbg_ring_ptrs(xhci, xhci->cmd_ring);
@@ -356,6 +376,7 @@ void xhci_event_ring_work(unsigned long arg)
 int xhci_run(struct usb_hcd *hcd)
 {
 	u32 temp;
+	u64 temp_64;
 	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
 	void (*doorbell)(struct xhci_hcd *) = NULL;
 
@@ -382,6 +403,20 @@ int xhci_run(struct usb_hcd *hcd)
 	add_timer(&xhci->event_ring_timer);
 #endif
 
+	xhci_dbg(xhci, "Command ring memory map follows:\n");
+	xhci_debug_ring(xhci, xhci->cmd_ring);
+	xhci_dbg_ring_ptrs(xhci, xhci->cmd_ring);
+	xhci_dbg_cmd_ptrs(xhci);
+
+	xhci_dbg(xhci, "ERST memory map follows:\n");
+	xhci_dbg_erst(xhci, &xhci->erst);
+	xhci_dbg(xhci, "Event ring:\n");
+	xhci_debug_ring(xhci, xhci->event_ring);
+	xhci_dbg_ring_ptrs(xhci, xhci->event_ring);
+	temp_64 = xhci_read_64(xhci, &xhci->ir_set->erst_dequeue);
+	temp_64 &= ~ERST_PTR_MASK;
+	xhci_dbg(xhci, "ERST deq = 64'h%0lx\n", (long unsigned int) temp_64);
+
 	xhci_dbg(xhci, "// Set the interrupt modulation register\n");
 	temp = xhci_readl(xhci, &xhci->ir_set->irq_control);
 	temp &= ~ER_IRQ_INTERVAL_MASK;
@@ -406,22 +441,6 @@ int xhci_run(struct usb_hcd *hcd)
 	if (NUM_TEST_NOOPS > 0)
 		doorbell = xhci_setup_one_noop(xhci);
 
-	xhci_dbg(xhci, "Command ring memory map follows:\n");
-	xhci_debug_ring(xhci, xhci->cmd_ring);
-	xhci_dbg_ring_ptrs(xhci, xhci->cmd_ring);
-	xhci_dbg_cmd_ptrs(xhci);
-
-	xhci_dbg(xhci, "ERST memory map follows:\n");
-	xhci_dbg_erst(xhci, &xhci->erst);
-	xhci_dbg(xhci, "Event ring:\n");
-	xhci_debug_ring(xhci, xhci->event_ring);
-	xhci_dbg_ring_ptrs(xhci, xhci->event_ring);
-	temp = xhci_readl(xhci, &xhci->ir_set->erst_dequeue[0]);
-	temp &= ERST_PTR_MASK;
-	xhci_dbg(xhci, "ERST deq = 0x%x\n", temp);
-	temp = xhci_readl(xhci, &xhci->ir_set->erst_dequeue[1]);
-	xhci_dbg(xhci, "ERST deq upper = 0x%x\n", temp);
-
 	temp = xhci_readl(xhci, &xhci->op_regs->command);
 	temp |= (CMD_RUN);
 	xhci_dbg(xhci, "// Turn on HC, cmd = 0x%x.\n",
@@ -601,10 +620,13 @@ int xhci_urb_enqueue(struct usb_hcd *hcd, struct urb *urb, gfp_t mem_flags)
 		goto exit;
 	}
 	if (usb_endpoint_xfer_control(&urb->ep->desc))
-		ret = xhci_queue_ctrl_tx(xhci, mem_flags, urb,
+		/* We have a spinlock and interrupts disabled, so we must pass
+		 * atomic context to this function, which may allocate memory.
+		 */
+		ret = xhci_queue_ctrl_tx(xhci, GFP_ATOMIC, urb,
 				slot_id, ep_index);
 	else if (usb_endpoint_xfer_bulk(&urb->ep->desc))
-		ret = xhci_queue_bulk_tx(xhci, mem_flags, urb,
+		ret = xhci_queue_bulk_tx(xhci, GFP_ATOMIC, urb,
 				slot_id, ep_index);
 	else
 		ret = -EINVAL;
@@ -661,8 +683,12 @@ int xhci_urb_dequeue(struct usb_hcd *hcd, struct urb *urb, int status)
 		goto done;
 
 	xhci_dbg(xhci, "Cancel URB %p\n", urb);
+	xhci_dbg(xhci, "Event ring:\n");
+	xhci_debug_ring(xhci, xhci->event_ring);
 	ep_index = xhci_get_endpoint_index(&urb->ep->desc);
 	ep_ring = xhci->devs[urb->dev->slot_id]->ep_rings[ep_index];
+	xhci_dbg(xhci, "Endpoint ring:\n");
+	xhci_debug_ring(xhci, ep_ring);
 	td = (struct xhci_td *) urb->hcpriv;
 
 	ep_ring->cancels_pending++;
@@ -696,7 +722,9 @@ int xhci_drop_endpoint(struct usb_hcd *hcd, struct usb_device *udev,
 		struct usb_host_endpoint *ep)
 {
 	struct xhci_hcd *xhci;
-	struct xhci_device_control *in_ctx;
+	struct xhci_container_ctx *in_ctx, *out_ctx;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	struct xhci_slot_ctx *slot_ctx;
 	unsigned int last_ctx;
 	unsigned int ep_index;
 	struct xhci_ep_ctx *ep_ctx;
@@ -724,31 +752,34 @@ int xhci_drop_endpoint(struct usb_hcd *hcd, struct usb_device *udev,
 	}
 
 	in_ctx = xhci->devs[udev->slot_id]->in_ctx;
+	out_ctx = xhci->devs[udev->slot_id]->out_ctx;
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, in_ctx);
 	ep_index = xhci_get_endpoint_index(&ep->desc);
-	ep_ctx = &xhci->devs[udev->slot_id]->out_ctx->ep[ep_index];
+	ep_ctx = xhci_get_ep_ctx(xhci, out_ctx, ep_index);
 	/* If the HC already knows the endpoint is disabled,
 	 * or the HCD has noted it is disabled, ignore this request
 	 */
 	if ((ep_ctx->ep_info & EP_STATE_MASK) == EP_STATE_DISABLED ||
-			in_ctx->drop_flags & xhci_get_endpoint_flag(&ep->desc)) {
+			ctrl_ctx->drop_flags & xhci_get_endpoint_flag(&ep->desc)) {
 		xhci_warn(xhci, "xHCI %s called with disabled ep %p\n",
 				__func__, ep);
 		return 0;
 	}
 
-	in_ctx->drop_flags |= drop_flag;
-	new_drop_flags = in_ctx->drop_flags;
+	ctrl_ctx->drop_flags |= drop_flag;
+	new_drop_flags = ctrl_ctx->drop_flags;
 
-	in_ctx->add_flags = ~drop_flag;
-	new_add_flags = in_ctx->add_flags;
+	ctrl_ctx->add_flags = ~drop_flag;
+	new_add_flags = ctrl_ctx->add_flags;
 
-	last_ctx = xhci_last_valid_endpoint(in_ctx->add_flags);
+	last_ctx = xhci_last_valid_endpoint(ctrl_ctx->add_flags);
+	slot_ctx = xhci_get_slot_ctx(xhci, in_ctx);
 	/* Update the last valid endpoint context, if we deleted the last one */
-	if ((in_ctx->slot.dev_info & LAST_CTX_MASK) > LAST_CTX(last_ctx)) {
-		in_ctx->slot.dev_info &= ~LAST_CTX_MASK;
-		in_ctx->slot.dev_info |= LAST_CTX(last_ctx);
+	if ((slot_ctx->dev_info & LAST_CTX_MASK) > LAST_CTX(last_ctx)) {
+		slot_ctx->dev_info &= ~LAST_CTX_MASK;
+		slot_ctx->dev_info |= LAST_CTX(last_ctx);
 	}
-	new_slot_info = in_ctx->slot.dev_info;
+	new_slot_info = slot_ctx->dev_info;
 
 	xhci_endpoint_zero(xhci, xhci->devs[udev->slot_id], ep);
 
@@ -778,17 +809,22 @@ int xhci_add_endpoint(struct usb_hcd *hcd, struct usb_device *udev,
 		struct usb_host_endpoint *ep)
 {
 	struct xhci_hcd *xhci;
-	struct xhci_device_control *in_ctx;
+	struct xhci_container_ctx *in_ctx, *out_ctx;
 	unsigned int ep_index;
 	struct xhci_ep_ctx *ep_ctx;
+	struct xhci_slot_ctx *slot_ctx;
+	struct xhci_input_control_ctx *ctrl_ctx;
 	u32 added_ctxs;
 	unsigned int last_ctx;
 	u32 new_add_flags, new_drop_flags, new_slot_info;
 	int ret = 0;
 
 	ret = xhci_check_args(hcd, udev, ep, 1, __func__);
-	if (ret <= 0)
+	if (ret <= 0) {
+		/* So we won't queue a reset ep command for a root hub */
+		ep->hcpriv = NULL;
 		return ret;
+	}
 	xhci = hcd_to_xhci(hcd);
 
 	added_ctxs = xhci_get_endpoint_flag(&ep->desc);
@@ -810,12 +846,14 @@ int xhci_add_endpoint(struct usb_hcd *hcd, struct usb_device *udev,
 	}
 
 	in_ctx = xhci->devs[udev->slot_id]->in_ctx;
+	out_ctx = xhci->devs[udev->slot_id]->out_ctx;
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, in_ctx);
 	ep_index = xhci_get_endpoint_index(&ep->desc);
-	ep_ctx = &xhci->devs[udev->slot_id]->out_ctx->ep[ep_index];
+	ep_ctx = xhci_get_ep_ctx(xhci, out_ctx, ep_index);
 	/* If the HCD has already noted the endpoint is enabled,
 	 * ignore this request.
 	 */
-	if (in_ctx->add_flags & xhci_get_endpoint_flag(&ep->desc)) {
+	if (ctrl_ctx->add_flags & xhci_get_endpoint_flag(&ep->desc)) {
 		xhci_warn(xhci, "xHCI %s called with enabled ep %p\n",
 				__func__, ep);
 		return 0;
@@ -833,8 +871,8 @@ int xhci_add_endpoint(struct usb_hcd *hcd, struct usb_device *udev,
 		return -ENOMEM;
 	}
 
-	in_ctx->add_flags |= added_ctxs;
-	new_add_flags = in_ctx->add_flags;
+	ctrl_ctx->add_flags |= added_ctxs;
+	new_add_flags = ctrl_ctx->add_flags;
 
 	/* If xhci_endpoint_disable() was called for this endpoint, but the
 	 * xHC hasn't been notified yet through the check_bandwidth() call,
@@ -842,14 +880,18 @@ int xhci_add_endpoint(struct usb_hcd *hcd, struct usb_device *udev,
 	 * descriptors.  We must drop and re-add this endpoint, so we leave the
 	 * drop flags alone.
 	 */
-	new_drop_flags = in_ctx->drop_flags;
+	new_drop_flags = ctrl_ctx->drop_flags;
 
+	slot_ctx = xhci_get_slot_ctx(xhci, in_ctx);
 	/* Update the last valid endpoint context, if we just added one past */
-	if ((in_ctx->slot.dev_info & LAST_CTX_MASK) < LAST_CTX(last_ctx)) {
-		in_ctx->slot.dev_info &= ~LAST_CTX_MASK;
-		in_ctx->slot.dev_info |= LAST_CTX(last_ctx);
+	if ((slot_ctx->dev_info & LAST_CTX_MASK) < LAST_CTX(last_ctx)) {
+		slot_ctx->dev_info &= ~LAST_CTX_MASK;
+		slot_ctx->dev_info |= LAST_CTX(last_ctx);
 	}
-	new_slot_info = in_ctx->slot.dev_info;
+	new_slot_info = slot_ctx->dev_info;
+
+	/* Store the usb_device pointer for later use */
+	ep->hcpriv = udev;
 
 	xhci_dbg(xhci, "add ep 0x%x, slot id %d, new drop flags = %#x, new add flags = %#x, new slot info = %#x\n",
 			(unsigned int) ep->desc.bEndpointAddress,
@@ -860,9 +902,11 @@ int xhci_add_endpoint(struct usb_hcd *hcd, struct usb_device *udev,
 	return 0;
 }
 
-static void xhci_zero_in_ctx(struct xhci_virt_device *virt_dev)
+static void xhci_zero_in_ctx(struct xhci_hcd *xhci, struct xhci_virt_device *virt_dev)
 {
+	struct xhci_input_control_ctx *ctrl_ctx;
 	struct xhci_ep_ctx *ep_ctx;
+	struct xhci_slot_ctx *slot_ctx;
 	int i;
 
 	/* When a device's add flag and drop flag are zero, any subsequent
@@ -870,17 +914,18 @@ static void xhci_zero_in_ctx(struct xhci_virt_device *virt_dev)
 	 * untouched.  Make sure we don't leave any old state in the input
 	 * endpoint contexts.
 	 */
-	virt_dev->in_ctx->drop_flags = 0;
-	virt_dev->in_ctx->add_flags = 0;
-	virt_dev->in_ctx->slot.dev_info &= ~LAST_CTX_MASK;
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, virt_dev->in_ctx);
+	ctrl_ctx->drop_flags = 0;
+	ctrl_ctx->add_flags = 0;
+	slot_ctx = xhci_get_slot_ctx(xhci, virt_dev->in_ctx);
+	slot_ctx->dev_info &= ~LAST_CTX_MASK;
 	/* Endpoint 0 is always valid */
-	virt_dev->in_ctx->slot.dev_info |= LAST_CTX(1);
+	slot_ctx->dev_info |= LAST_CTX(1);
 	for (i = 1; i < 31; ++i) {
-		ep_ctx = &virt_dev->in_ctx->ep[i];
+		ep_ctx = xhci_get_ep_ctx(xhci, virt_dev->in_ctx, i);
 		ep_ctx->ep_info = 0;
 		ep_ctx->ep_info2 = 0;
-		ep_ctx->deq[0] = 0;
-		ep_ctx->deq[1] = 0;
+		ep_ctx->deq = 0;
 		ep_ctx->tx_info = 0;
 	}
 }
@@ -903,6 +948,8 @@ int xhci_check_bandwidth(struct usb_hcd *hcd, struct usb_device *udev)
 	unsigned long flags;
 	struct xhci_hcd *xhci;
 	struct xhci_virt_device	*virt_dev;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	struct xhci_slot_ctx *slot_ctx;
 
 	ret = xhci_check_args(hcd, udev, NULL, 0, __func__);
 	if (ret <= 0)
@@ -918,16 +965,18 @@ int xhci_check_bandwidth(struct usb_hcd *hcd, struct usb_device *udev)
 	virt_dev = xhci->devs[udev->slot_id];
 
 	/* See section 4.6.6 - A0 = 1; A1 = D0 = D1 = 0 */
-	virt_dev->in_ctx->add_flags |= SLOT_FLAG;
-	virt_dev->in_ctx->add_flags &= ~EP0_FLAG;
-	virt_dev->in_ctx->drop_flags &= ~SLOT_FLAG;
-	virt_dev->in_ctx->drop_flags &= ~EP0_FLAG;
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, virt_dev->in_ctx);
+	ctrl_ctx->add_flags |= SLOT_FLAG;
+	ctrl_ctx->add_flags &= ~EP0_FLAG;
+	ctrl_ctx->drop_flags &= ~SLOT_FLAG;
+	ctrl_ctx->drop_flags &= ~EP0_FLAG;
 	xhci_dbg(xhci, "New Input Control Context:\n");
-	xhci_dbg_ctx(xhci, virt_dev->in_ctx, virt_dev->in_ctx_dma,
-			LAST_CTX_TO_EP_NUM(virt_dev->in_ctx->slot.dev_info));
+	slot_ctx = xhci_get_slot_ctx(xhci, virt_dev->in_ctx);
+	xhci_dbg_ctx(xhci, virt_dev->in_ctx,
+			LAST_CTX_TO_EP_NUM(slot_ctx->dev_info));
 
 	spin_lock_irqsave(&xhci->lock, flags);
-	ret = xhci_queue_configure_endpoint(xhci, virt_dev->in_ctx_dma,
+	ret = xhci_queue_configure_endpoint(xhci, virt_dev->in_ctx->dma,
 			udev->slot_id);
 	if (ret < 0) {
 		spin_unlock_irqrestore(&xhci->lock, flags);
@@ -982,10 +1031,10 @@ int xhci_check_bandwidth(struct usb_hcd *hcd, struct usb_device *udev)
 	}
 
 	xhci_dbg(xhci, "Output context after successful config ep cmd:\n");
-	xhci_dbg_ctx(xhci, virt_dev->out_ctx, virt_dev->out_ctx_dma,
-			LAST_CTX_TO_EP_NUM(virt_dev->in_ctx->slot.dev_info));
+	xhci_dbg_ctx(xhci, virt_dev->out_ctx,
+			LAST_CTX_TO_EP_NUM(slot_ctx->dev_info));
 
-	xhci_zero_in_ctx(virt_dev);
+	xhci_zero_in_ctx(xhci, virt_dev);
 	/* Free any old rings */
 	for (i = 1; i < 31; ++i) {
 		if (virt_dev->new_ep_rings[i]) {
@@ -1023,7 +1072,67 @@ void xhci_reset_bandwidth(struct usb_hcd *hcd, struct usb_device *udev)
 			virt_dev->new_ep_rings[i] = NULL;
 		}
 	}
-	xhci_zero_in_ctx(virt_dev);
+	xhci_zero_in_ctx(xhci, virt_dev);
+}
+
+/* Deal with stalled endpoints.  The core should have sent the control message
+ * to clear the halt condition.  However, we need to make the xHCI hardware
+ * reset its sequence number, since a device will expect a sequence number of
+ * zero after the halt condition is cleared.
+ * Context: in_interrupt
+ */
+void xhci_endpoint_reset(struct usb_hcd *hcd,
+		struct usb_host_endpoint *ep)
+{
+	struct xhci_hcd *xhci;
+	struct usb_device *udev;
+	unsigned int ep_index;
+	unsigned long flags;
+	int ret;
+	struct xhci_dequeue_state deq_state;
+	struct xhci_ring *ep_ring;
+
+	xhci = hcd_to_xhci(hcd);
+	udev = (struct usb_device *) ep->hcpriv;
+	/* Called with a root hub endpoint (or an endpoint that wasn't added
+	 * with xhci_add_endpoint()
+	 */
+	if (!ep->hcpriv)
+		return;
+	ep_index = xhci_get_endpoint_index(&ep->desc);
+	ep_ring = xhci->devs[udev->slot_id]->ep_rings[ep_index];
+	if (!ep_ring->stopped_td) {
+		xhci_dbg(xhci, "Endpoint 0x%x not halted, refusing to reset.\n",
+				ep->desc.bEndpointAddress);
+		return;
+	}
+
+	xhci_dbg(xhci, "Queueing reset endpoint command\n");
+	spin_lock_irqsave(&xhci->lock, flags);
+	ret = xhci_queue_reset_ep(xhci, udev->slot_id, ep_index);
+	/*
+	 * Can't change the ring dequeue pointer until it's transitioned to the
+	 * stopped state, which is only upon a successful reset endpoint
+	 * command.  Better hope that last command worked!
+	 */
+	if (!ret) {
+		xhci_dbg(xhci, "Cleaning up stalled endpoint ring\n");
+		/* We need to move the HW's dequeue pointer past this TD,
+		 * or it will attempt to resend it on the next doorbell ring.
+		 */
+		xhci_find_new_dequeue_state(xhci, udev->slot_id,
+				ep_index, ep_ring->stopped_td, &deq_state);
+		xhci_dbg(xhci, "Queueing new dequeue state\n");
+		xhci_queue_new_dequeue_state(xhci, ep_ring,
+				udev->slot_id,
+				ep_index, &deq_state);
+		kfree(ep_ring->stopped_td);
+		xhci_ring_cmd_db(xhci);
+	}
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	if (ret)
+		xhci_warn(xhci, "FIXME allocate a new ring segment\n");
 }
 
 /*
@@ -1120,7 +1229,9 @@ int xhci_address_device(struct usb_hcd *hcd, struct usb_device *udev)
 	struct xhci_virt_device *virt_dev;
 	int ret = 0;
 	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
-	u32 temp;
+	struct xhci_slot_ctx *slot_ctx;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	u64 temp_64;
 
 	if (!udev->slot_id) {
 		xhci_dbg(xhci, "Bad Slot ID %d\n", udev->slot_id);
@@ -1133,10 +1244,12 @@ int xhci_address_device(struct usb_hcd *hcd, struct usb_device *udev)
 	if (!udev->config)
 		xhci_setup_addressable_virt_dev(xhci, udev);
 	/* Otherwise, assume the core has the device configured how it wants */
+	xhci_dbg(xhci, "Slot ID %d Input Context:\n", udev->slot_id);
+	xhci_dbg_ctx(xhci, virt_dev->in_ctx, 2);
 
 	spin_lock_irqsave(&xhci->lock, flags);
-	ret = xhci_queue_address_device(xhci, virt_dev->in_ctx_dma,
-			udev->slot_id);
+	ret = xhci_queue_address_device(xhci, virt_dev->in_ctx->dma,
+					udev->slot_id);
 	if (ret) {
 		spin_unlock_irqrestore(&xhci->lock, flags);
 		xhci_dbg(xhci, "FIXME: allocate a command ring segment\n");
@@ -1176,41 +1289,37 @@ int xhci_address_device(struct usb_hcd *hcd, struct usb_device *udev)
 	default:
 		xhci_err(xhci, "ERROR: unexpected command completion "
 				"code 0x%x.\n", virt_dev->cmd_status);
+		xhci_dbg(xhci, "Slot ID %d Output Context:\n", udev->slot_id);
+		xhci_dbg_ctx(xhci, virt_dev->out_ctx, 2);
 		ret = -EINVAL;
 		break;
 	}
 	if (ret) {
 		return ret;
 	}
-	temp = xhci_readl(xhci, &xhci->op_regs->dcbaa_ptr[0]);
-	xhci_dbg(xhci, "Op regs DCBAA ptr[0] = %#08x\n", temp);
-	temp = xhci_readl(xhci, &xhci->op_regs->dcbaa_ptr[1]);
-	xhci_dbg(xhci, "Op regs DCBAA ptr[1] = %#08x\n", temp);
-	xhci_dbg(xhci, "Slot ID %d dcbaa entry[0] @%p = %#08x\n",
-			udev->slot_id,
-			&xhci->dcbaa->dev_context_ptrs[2*udev->slot_id],
-			xhci->dcbaa->dev_context_ptrs[2*udev->slot_id]);
-	xhci_dbg(xhci, "Slot ID %d dcbaa entry[1] @%p = %#08x\n",
+	temp_64 = xhci_read_64(xhci, &xhci->op_regs->dcbaa_ptr);
+	xhci_dbg(xhci, "Op regs DCBAA ptr = %#016llx\n", temp_64);
+	xhci_dbg(xhci, "Slot ID %d dcbaa entry @%p = %#016llx\n",
 			udev->slot_id,
-			&xhci->dcbaa->dev_context_ptrs[2*udev->slot_id+1],
-			xhci->dcbaa->dev_context_ptrs[2*udev->slot_id+1]);
+			&xhci->dcbaa->dev_context_ptrs[udev->slot_id],
+			(unsigned long long)
+				xhci->dcbaa->dev_context_ptrs[udev->slot_id]);
 	xhci_dbg(xhci, "Output Context DMA address = %#08llx\n",
-			(unsigned long long)virt_dev->out_ctx_dma);
+			(unsigned long long)virt_dev->out_ctx->dma);
 	xhci_dbg(xhci, "Slot ID %d Input Context:\n", udev->slot_id);
-	xhci_dbg_ctx(xhci, virt_dev->in_ctx, virt_dev->in_ctx_dma, 2);
+	xhci_dbg_ctx(xhci, virt_dev->in_ctx, 2);
 	xhci_dbg(xhci, "Slot ID %d Output Context:\n", udev->slot_id);
-	xhci_dbg_ctx(xhci, virt_dev->out_ctx, virt_dev->out_ctx_dma, 2);
+	xhci_dbg_ctx(xhci, virt_dev->out_ctx, 2);
 	/*
 	 * USB core uses address 1 for the roothubs, so we add one to the
 	 * address given back to us by the HC.
 	 */
-	udev->devnum = (virt_dev->out_ctx->slot.dev_state & DEV_ADDR_MASK) + 1;
+	slot_ctx = xhci_get_slot_ctx(xhci, virt_dev->out_ctx);
+	udev->devnum = (slot_ctx->dev_state & DEV_ADDR_MASK) + 1;
 	/* Zero the input context control for later use */
-	virt_dev->in_ctx->add_flags = 0;
-	virt_dev->in_ctx->drop_flags = 0;
-	/* Mirror flags in the output context for future ep enable/disable */
-	virt_dev->out_ctx->add_flags = SLOT_FLAG | EP0_FLAG;
-	virt_dev->out_ctx->drop_flags = 0;
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, virt_dev->in_ctx);
+	ctrl_ctx->add_flags = 0;
+	ctrl_ctx->drop_flags = 0;
 
 	xhci_dbg(xhci, "Device address = %d\n", udev->devnum);
 	/* XXX Meh, not sure if anyone else but choose_address uses this. */
@@ -1252,7 +1361,6 @@ static int __init xhci_hcd_init(void)
 	/* xhci_device_control has eight fields, and also
 	 * embeds one xhci_slot_ctx and 31 xhci_ep_ctx
 	 */
-	BUILD_BUG_ON(sizeof(struct xhci_device_control) != (8+8+8*31)*32/8);
 	BUILD_BUG_ON(sizeof(struct xhci_stream_ctx) != 4*32/8);
 	BUILD_BUG_ON(sizeof(union xhci_trb) != 4*32/8);
 	BUILD_BUG_ON(sizeof(struct xhci_erst_entry) != 4*32/8);
diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
index c8a72de1c50..e6b9a1c6002 100644
--- a/drivers/usb/host/xhci-mem.c
+++ b/drivers/usb/host/xhci-mem.c
@@ -88,7 +88,7 @@ static void xhci_link_segments(struct xhci_hcd *xhci, struct xhci_segment *prev,
 		return;
 	prev->next = next;
 	if (link_trbs) {
-		prev->trbs[TRBS_PER_SEGMENT-1].link.segment_ptr[0] = next->dma;
+		prev->trbs[TRBS_PER_SEGMENT-1].link.segment_ptr = next->dma;
 
 		/* Set the last TRB in the segment to have a TRB type ID of Link TRB */
 		val = prev->trbs[TRBS_PER_SEGMENT-1].link.control;
@@ -189,6 +189,63 @@ fail:
 	return 0;
 }
 
+#define CTX_SIZE(_hcc) (HCC_64BYTE_CONTEXT(_hcc) ? 64 : 32)
+
+struct xhci_container_ctx *xhci_alloc_container_ctx(struct xhci_hcd *xhci,
+						    int type, gfp_t flags)
+{
+	struct xhci_container_ctx *ctx = kzalloc(sizeof(*ctx), flags);
+	if (!ctx)
+		return NULL;
+
+	BUG_ON((type != XHCI_CTX_TYPE_DEVICE) && (type != XHCI_CTX_TYPE_INPUT));
+	ctx->type = type;
+	ctx->size = HCC_64BYTE_CONTEXT(xhci->hcc_params) ? 2048 : 1024;
+	if (type == XHCI_CTX_TYPE_INPUT)
+		ctx->size += CTX_SIZE(xhci->hcc_params);
+
+	ctx->bytes = dma_pool_alloc(xhci->device_pool, flags, &ctx->dma);
+	memset(ctx->bytes, 0, ctx->size);
+	return ctx;
+}
+
+void xhci_free_container_ctx(struct xhci_hcd *xhci,
+			     struct xhci_container_ctx *ctx)
+{
+	dma_pool_free(xhci->device_pool, ctx->bytes, ctx->dma);
+	kfree(ctx);
+}
+
+struct xhci_input_control_ctx *xhci_get_input_control_ctx(struct xhci_hcd *xhci,
+					      struct xhci_container_ctx *ctx)
+{
+	BUG_ON(ctx->type != XHCI_CTX_TYPE_INPUT);
+	return (struct xhci_input_control_ctx *)ctx->bytes;
+}
+
+struct xhci_slot_ctx *xhci_get_slot_ctx(struct xhci_hcd *xhci,
+					struct xhci_container_ctx *ctx)
+{
+	if (ctx->type == XHCI_CTX_TYPE_DEVICE)
+		return (struct xhci_slot_ctx *)ctx->bytes;
+
+	return (struct xhci_slot_ctx *)
+		(ctx->bytes + CTX_SIZE(xhci->hcc_params));
+}
+
+struct xhci_ep_ctx *xhci_get_ep_ctx(struct xhci_hcd *xhci,
+				    struct xhci_container_ctx *ctx,
+				    unsigned int ep_index)
+{
+	/* increment ep index by offset of start of ep ctx array */
+	ep_index++;
+	if (ctx->type == XHCI_CTX_TYPE_INPUT)
+		ep_index++;
+
+	return (struct xhci_ep_ctx *)
+		(ctx->bytes + (ep_index * CTX_SIZE(xhci->hcc_params)));
+}
+
 /* All the xhci_tds in the ring's TD list should be freed at this point */
 void xhci_free_virt_device(struct xhci_hcd *xhci, int slot_id)
 {
@@ -200,8 +257,7 @@ void xhci_free_virt_device(struct xhci_hcd *xhci, int slot_id)
 		return;
 
 	dev = xhci->devs[slot_id];
-	xhci->dcbaa->dev_context_ptrs[2*slot_id] = 0;
-	xhci->dcbaa->dev_context_ptrs[2*slot_id + 1] = 0;
+	xhci->dcbaa->dev_context_ptrs[slot_id] = 0;
 	if (!dev)
 		return;
 
@@ -210,11 +266,10 @@ void xhci_free_virt_device(struct xhci_hcd *xhci, int slot_id)
 			xhci_ring_free(xhci, dev->ep_rings[i]);
 
 	if (dev->in_ctx)
-		dma_pool_free(xhci->device_pool,
-				dev->in_ctx, dev->in_ctx_dma);
+		xhci_free_container_ctx(xhci, dev->in_ctx);
 	if (dev->out_ctx)
-		dma_pool_free(xhci->device_pool,
-				dev->out_ctx, dev->out_ctx_dma);
+		xhci_free_container_ctx(xhci, dev->out_ctx);
+
 	kfree(xhci->devs[slot_id]);
 	xhci->devs[slot_id] = 0;
 }
@@ -222,7 +277,6 @@ void xhci_free_virt_device(struct xhci_hcd *xhci, int slot_id)
 int xhci_alloc_virt_device(struct xhci_hcd *xhci, int slot_id,
 		struct usb_device *udev, gfp_t flags)
 {
-	dma_addr_t	dma;
 	struct xhci_virt_device *dev;
 
 	/* Slot ID 0 is reserved */
@@ -236,23 +290,21 @@ int xhci_alloc_virt_device(struct xhci_hcd *xhci, int slot_id,
 		return 0;
 	dev = xhci->devs[slot_id];
 
-	/* Allocate the (output) device context that will be used in the HC */
-	dev->out_ctx = dma_pool_alloc(xhci->device_pool, flags, &dma);
+	/* Allocate the (output) device context that will be used in the HC. */
+	dev->out_ctx = xhci_alloc_container_ctx(xhci, XHCI_CTX_TYPE_DEVICE, flags);
 	if (!dev->out_ctx)
 		goto fail;
-	dev->out_ctx_dma = dma;
+
 	xhci_dbg(xhci, "Slot %d output ctx = 0x%llx (dma)\n", slot_id,
-			(unsigned long long)dma);
-	memset(dev->out_ctx, 0, sizeof(*dev->out_ctx));
+			(unsigned long long)dev->out_ctx->dma);
 
 	/* Allocate the (input) device context for address device command */
-	dev->in_ctx = dma_pool_alloc(xhci->device_pool, flags, &dma);
+	dev->in_ctx = xhci_alloc_container_ctx(xhci, XHCI_CTX_TYPE_INPUT, flags);
 	if (!dev->in_ctx)
 		goto fail;
-	dev->in_ctx_dma = dma;
+
 	xhci_dbg(xhci, "Slot %d input ctx = 0x%llx (dma)\n", slot_id,
-			(unsigned long long)dma);
-	memset(dev->in_ctx, 0, sizeof(*dev->in_ctx));
+			(unsigned long long)dev->in_ctx->dma);
 
 	/* Allocate endpoint 0 ring */
 	dev->ep_rings[0] = xhci_ring_alloc(xhci, 1, true, flags);
@@ -261,17 +313,12 @@ int xhci_alloc_virt_device(struct xhci_hcd *xhci, int slot_id,
 
 	init_completion(&dev->cmd_completion);
 
-	/*
-	 * Point to output device context in dcbaa; skip the output control
-	 * context, which is eight 32 bit fields (or 32 bytes long)
-	 */
-	xhci->dcbaa->dev_context_ptrs[2*slot_id] =
-		(u32) dev->out_ctx_dma + (32);
+	/* Point to output device context in dcbaa. */
+	xhci->dcbaa->dev_context_ptrs[slot_id] = dev->out_ctx->dma;
 	xhci_dbg(xhci, "Set slot id %d dcbaa entry %p to 0x%llx\n",
 			slot_id,
-			&xhci->dcbaa->dev_context_ptrs[2*slot_id],
-			(unsigned long long)dev->out_ctx_dma);
-	xhci->dcbaa->dev_context_ptrs[2*slot_id + 1] = 0;
+			&xhci->dcbaa->dev_context_ptrs[slot_id],
+			(unsigned long long) xhci->dcbaa->dev_context_ptrs[slot_id]);
 
 	return 1;
 fail:
@@ -285,6 +332,8 @@ int xhci_setup_addressable_virt_dev(struct xhci_hcd *xhci, struct usb_device *ud
 	struct xhci_virt_device *dev;
 	struct xhci_ep_ctx	*ep0_ctx;
 	struct usb_device	*top_dev;
+	struct xhci_slot_ctx    *slot_ctx;
+	struct xhci_input_control_ctx *ctrl_ctx;
 
 	dev = xhci->devs[udev->slot_id];
 	/* Slot ID 0 is reserved */
@@ -293,27 +342,29 @@ int xhci_setup_addressable_virt_dev(struct xhci_hcd *xhci, struct usb_device *ud
 				udev->slot_id);
 		return -EINVAL;
 	}
-	ep0_ctx = &dev->in_ctx->ep[0];
+	ep0_ctx = xhci_get_ep_ctx(xhci, dev->in_ctx, 0);
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, dev->in_ctx);
+	slot_ctx = xhci_get_slot_ctx(xhci, dev->in_ctx);
 
 	/* 2) New slot context and endpoint 0 context are valid*/
-	dev->in_ctx->add_flags = SLOT_FLAG | EP0_FLAG;
+	ctrl_ctx->add_flags = SLOT_FLAG | EP0_FLAG;
 
 	/* 3) Only the control endpoint is valid - one endpoint context */
-	dev->in_ctx->slot.dev_info |= LAST_CTX(1);
+	slot_ctx->dev_info |= LAST_CTX(1);
 
 	switch (udev->speed) {
 	case USB_SPEED_SUPER:
-		dev->in_ctx->slot.dev_info |= (u32) udev->route;
-		dev->in_ctx->slot.dev_info |= (u32) SLOT_SPEED_SS;
+		slot_ctx->dev_info |= (u32) udev->route;
+		slot_ctx->dev_info |= (u32) SLOT_SPEED_SS;
 		break;
 	case USB_SPEED_HIGH:
-		dev->in_ctx->slot.dev_info |= (u32) SLOT_SPEED_HS;
+		slot_ctx->dev_info |= (u32) SLOT_SPEED_HS;
 		break;
 	case USB_SPEED_FULL:
-		dev->in_ctx->slot.dev_info |= (u32) SLOT_SPEED_FS;
+		slot_ctx->dev_info |= (u32) SLOT_SPEED_FS;
 		break;
 	case USB_SPEED_LOW:
-		dev->in_ctx->slot.dev_info |= (u32) SLOT_SPEED_LS;
+		slot_ctx->dev_info |= (u32) SLOT_SPEED_LS;
 		break;
 	case USB_SPEED_VARIABLE:
 		xhci_dbg(xhci, "FIXME xHCI doesn't support wireless speeds\n");
@@ -327,7 +378,7 @@ int xhci_setup_addressable_virt_dev(struct xhci_hcd *xhci, struct usb_device *ud
 	for (top_dev = udev; top_dev->parent && top_dev->parent->parent;
 			top_dev = top_dev->parent)
 		/* Found device below root hub */;
-	dev->in_ctx->slot.dev_info2 |= (u32) ROOT_HUB_PORT(top_dev->portnum);
+	slot_ctx->dev_info2 |= (u32) ROOT_HUB_PORT(top_dev->portnum);
 	xhci_dbg(xhci, "Set root hub portnum to %d\n", top_dev->portnum);
 
 	/* Is this a LS/FS device under a HS hub? */
@@ -337,8 +388,8 @@ int xhci_setup_addressable_virt_dev(struct xhci_hcd *xhci, struct usb_device *ud
 	 */
 	if ((udev->speed == USB_SPEED_LOW || udev->speed == USB_SPEED_FULL) &&
 			udev->tt) {
-		dev->in_ctx->slot.tt_info = udev->tt->hub->slot_id;
-		dev->in_ctx->slot.tt_info |= udev->ttport << 8;
+		slot_ctx->tt_info = udev->tt->hub->slot_id;
+		slot_ctx->tt_info |= udev->ttport << 8;
 	}
 	xhci_dbg(xhci, "udev->tt = %p\n", udev->tt);
 	xhci_dbg(xhci, "udev->ttport = 0x%x\n", udev->ttport);
@@ -360,10 +411,9 @@ int xhci_setup_addressable_virt_dev(struct xhci_hcd *xhci, struct usb_device *ud
 	ep0_ctx->ep_info2 |= MAX_BURST(0);
 	ep0_ctx->ep_info2 |= ERROR_COUNT(3);
 
-	ep0_ctx->deq[0] =
+	ep0_ctx->deq =
 		dev->ep_rings[0]->first_seg->dma;
-	ep0_ctx->deq[0] |= dev->ep_rings[0]->cycle_state;
-	ep0_ctx->deq[1] = 0;
+	ep0_ctx->deq |= dev->ep_rings[0]->cycle_state;
 
 	/* Steps 7 and 8 were done in xhci_alloc_virt_device() */
 
@@ -470,25 +520,26 @@ int xhci_endpoint_init(struct xhci_hcd *xhci,
 	unsigned int max_burst;
 
 	ep_index = xhci_get_endpoint_index(&ep->desc);
-	ep_ctx = &virt_dev->in_ctx->ep[ep_index];
+	ep_ctx = xhci_get_ep_ctx(xhci, virt_dev->in_ctx, ep_index);
 
 	/* Set up the endpoint ring */
 	virt_dev->new_ep_rings[ep_index] = xhci_ring_alloc(xhci, 1, true, mem_flags);
 	if (!virt_dev->new_ep_rings[ep_index])
 		return -ENOMEM;
 	ep_ring = virt_dev->new_ep_rings[ep_index];
-	ep_ctx->deq[0] = ep_ring->first_seg->dma | ep_ring->cycle_state;
-	ep_ctx->deq[1] = 0;
+	ep_ctx->deq = ep_ring->first_seg->dma | ep_ring->cycle_state;
 
 	ep_ctx->ep_info = xhci_get_endpoint_interval(udev, ep);
 
 	/* FIXME dig Mult and streams info out of ep companion desc */
 
-	/* Allow 3 retries for everything but isoc */
+	/* Allow 3 retries for everything but isoc;
+	 * error count = 0 means infinite retries.
+	 */
 	if (!usb_endpoint_xfer_isoc(&ep->desc))
 		ep_ctx->ep_info2 = ERROR_COUNT(3);
 	else
-		ep_ctx->ep_info2 = ERROR_COUNT(0);
+		ep_ctx->ep_info2 = ERROR_COUNT(1);
 
 	ep_ctx->ep_info2 |= xhci_get_endpoint_type(udev, ep);
 
@@ -498,7 +549,12 @@ int xhci_endpoint_init(struct xhci_hcd *xhci,
 		max_packet = ep->desc.wMaxPacketSize;
 		ep_ctx->ep_info2 |= MAX_PACKET(max_packet);
 		/* dig out max burst from ep companion desc */
-		max_packet = ep->ss_ep_comp->desc.bMaxBurst;
+		if (!ep->ss_ep_comp) {
+			xhci_warn(xhci, "WARN no SS endpoint companion descriptor.\n");
+			max_packet = 0;
+		} else {
+			max_packet = ep->ss_ep_comp->desc.bMaxBurst;
+		}
 		ep_ctx->ep_info2 |= MAX_BURST(max_packet);
 		break;
 	case USB_SPEED_HIGH:
@@ -531,18 +587,114 @@ void xhci_endpoint_zero(struct xhci_hcd *xhci,
 	struct xhci_ep_ctx *ep_ctx;
 
 	ep_index = xhci_get_endpoint_index(&ep->desc);
-	ep_ctx = &virt_dev->in_ctx->ep[ep_index];
+	ep_ctx = xhci_get_ep_ctx(xhci, virt_dev->in_ctx, ep_index);
 
 	ep_ctx->ep_info = 0;
 	ep_ctx->ep_info2 = 0;
-	ep_ctx->deq[0] = 0;
-	ep_ctx->deq[1] = 0;
+	ep_ctx->deq = 0;
 	ep_ctx->tx_info = 0;
 	/* Don't free the endpoint ring until the set interface or configuration
 	 * request succeeds.
 	 */
 }
 
+/* Set up the scratchpad buffer array and scratchpad buffers, if needed. */
+static int scratchpad_alloc(struct xhci_hcd *xhci, gfp_t flags)
+{
+	int i;
+	struct device *dev = xhci_to_hcd(xhci)->self.controller;
+	int num_sp = HCS_MAX_SCRATCHPAD(xhci->hcs_params2);
+
+	xhci_dbg(xhci, "Allocating %d scratchpad buffers\n", num_sp);
+
+	if (!num_sp)
+		return 0;
+
+	xhci->scratchpad = kzalloc(sizeof(*xhci->scratchpad), flags);
+	if (!xhci->scratchpad)
+		goto fail_sp;
+
+	xhci->scratchpad->sp_array =
+		pci_alloc_consistent(to_pci_dev(dev),
+				     num_sp * sizeof(u64),
+				     &xhci->scratchpad->sp_dma);
+	if (!xhci->scratchpad->sp_array)
+		goto fail_sp2;
+
+	xhci->scratchpad->sp_buffers = kzalloc(sizeof(void *) * num_sp, flags);
+	if (!xhci->scratchpad->sp_buffers)
+		goto fail_sp3;
+
+	xhci->scratchpad->sp_dma_buffers =
+		kzalloc(sizeof(dma_addr_t) * num_sp, flags);
+
+	if (!xhci->scratchpad->sp_dma_buffers)
+		goto fail_sp4;
+
+	xhci->dcbaa->dev_context_ptrs[0] = xhci->scratchpad->sp_dma;
+	for (i = 0; i < num_sp; i++) {
+		dma_addr_t dma;
+		void *buf = pci_alloc_consistent(to_pci_dev(dev),
+						 xhci->page_size, &dma);
+		if (!buf)
+			goto fail_sp5;
+
+		xhci->scratchpad->sp_array[i] = dma;
+		xhci->scratchpad->sp_buffers[i] = buf;
+		xhci->scratchpad->sp_dma_buffers[i] = dma;
+	}
+
+	return 0;
+
+ fail_sp5:
+	for (i = i - 1; i >= 0; i--) {
+		pci_free_consistent(to_pci_dev(dev), xhci->page_size,
+				    xhci->scratchpad->sp_buffers[i],
+				    xhci->scratchpad->sp_dma_buffers[i]);
+	}
+	kfree(xhci->scratchpad->sp_dma_buffers);
+
+ fail_sp4:
+	kfree(xhci->scratchpad->sp_buffers);
+
+ fail_sp3:
+	pci_free_consistent(to_pci_dev(dev), num_sp * sizeof(u64),
+			    xhci->scratchpad->sp_array,
+			    xhci->scratchpad->sp_dma);
+
+ fail_sp2:
+	kfree(xhci->scratchpad);
+	xhci->scratchpad = NULL;
+
+ fail_sp:
+	return -ENOMEM;
+}
+
+static void scratchpad_free(struct xhci_hcd *xhci)
+{
+	int num_sp;
+	int i;
+	struct pci_dev	*pdev = to_pci_dev(xhci_to_hcd(xhci)->self.controller);
+
+	if (!xhci->scratchpad)
+		return;
+
+	num_sp = HCS_MAX_SCRATCHPAD(xhci->hcs_params2);
+
+	for (i = 0; i < num_sp; i++) {
+		pci_free_consistent(pdev, xhci->page_size,
+				    xhci->scratchpad->sp_buffers[i],
+				    xhci->scratchpad->sp_dma_buffers[i]);
+	}
+	kfree(xhci->scratchpad->sp_dma_buffers);
+	kfree(xhci->scratchpad->sp_buffers);
+	pci_free_consistent(pdev, num_sp * sizeof(u64),
+			    xhci->scratchpad->sp_array,
+			    xhci->scratchpad->sp_dma);
+	kfree(xhci->scratchpad);
+	xhci->scratchpad = NULL;
+}
+
 void xhci_mem_cleanup(struct xhci_hcd *xhci)
 {
 	struct pci_dev	*pdev = to_pci_dev(xhci_to_hcd(xhci)->self.controller);
@@ -551,10 +703,8 @@ void xhci_mem_cleanup(struct xhci_hcd *xhci)
 
 	/* Free the Event Ring Segment Table and the actual Event Ring */
 	xhci_writel(xhci, 0, &xhci->ir_set->erst_size);
-	xhci_writel(xhci, 0, &xhci->ir_set->erst_base[0]);
-	xhci_writel(xhci, 0, &xhci->ir_set->erst_base[1]);
-	xhci_writel(xhci, 0, &xhci->ir_set->erst_dequeue[0]);
-	xhci_writel(xhci, 0, &xhci->ir_set->erst_dequeue[1]);
+	xhci_write_64(xhci, 0, &xhci->ir_set->erst_base);
+	xhci_write_64(xhci, 0, &xhci->ir_set->erst_dequeue);
 	size = sizeof(struct xhci_erst_entry)*(xhci->erst.num_entries);
 	if (xhci->erst.entries)
 		pci_free_consistent(pdev, size,
@@ -566,8 +716,7 @@ void xhci_mem_cleanup(struct xhci_hcd *xhci)
 	xhci->event_ring = NULL;
 	xhci_dbg(xhci, "Freed event ring\n");
 
-	xhci_writel(xhci, 0, &xhci->op_regs->cmd_ring[0]);
-	xhci_writel(xhci, 0, &xhci->op_regs->cmd_ring[1]);
+	xhci_write_64(xhci, 0, &xhci->op_regs->cmd_ring);
 	if (xhci->cmd_ring)
 		xhci_ring_free(xhci, xhci->cmd_ring);
 	xhci->cmd_ring = NULL;
@@ -586,8 +735,7 @@ void xhci_mem_cleanup(struct xhci_hcd *xhci)
 	xhci->device_pool = NULL;
 	xhci_dbg(xhci, "Freed device context pool\n");
 
-	xhci_writel(xhci, 0, &xhci->op_regs->dcbaa_ptr[0]);
-	xhci_writel(xhci, 0, &xhci->op_regs->dcbaa_ptr[1]);
+	xhci_write_64(xhci, 0, &xhci->op_regs->dcbaa_ptr);
 	if (xhci->dcbaa)
 		pci_free_consistent(pdev, sizeof(*xhci->dcbaa),
 				xhci->dcbaa, xhci->dcbaa->dma);
@@ -595,6 +743,7 @@ void xhci_mem_cleanup(struct xhci_hcd *xhci)
 
 	xhci->page_size = 0;
 	xhci->page_shift = 0;
+	scratchpad_free(xhci);
 }
 
 int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
@@ -602,6 +751,7 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	dma_addr_t	dma;
 	struct device	*dev = xhci_to_hcd(xhci)->self.controller;
 	unsigned int	val, val2;
+	u64		val_64;
 	struct xhci_segment	*seg;
 	u32 page_size;
 	int i;
@@ -647,8 +797,7 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	xhci->dcbaa->dma = dma;
 	xhci_dbg(xhci, "// Device context base array address = 0x%llx (DMA), %p (virt)\n",
 			(unsigned long long)xhci->dcbaa->dma, xhci->dcbaa);
-	xhci_writel(xhci, dma, &xhci->op_regs->dcbaa_ptr[0]);
-	xhci_writel(xhci, (u32) 0, &xhci->op_regs->dcbaa_ptr[1]);
+	xhci_write_64(xhci, dma, &xhci->op_regs->dcbaa_ptr);
 
 	/*
 	 * Initialize the ring segment pool.  The ring must be a contiguous
@@ -658,11 +807,10 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	 */
 	xhci->segment_pool = dma_pool_create("xHCI ring segments", dev,
 			SEGMENT_SIZE, 64, xhci->page_size);
+
 	/* See Table 46 and Note on Figure 55 */
-	/* FIXME support 64-byte contexts */
 	xhci->device_pool = dma_pool_create("xHCI input/output contexts", dev,
-			sizeof(struct xhci_device_control),
-			64, xhci->page_size);
+			2112, 64, xhci->page_size);
 	if (!xhci->segment_pool || !xhci->device_pool)
 		goto fail;
 
@@ -675,14 +823,12 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 			(unsigned long long)xhci->cmd_ring->first_seg->dma);
 
 	/* Set the address in the Command Ring Control register */
-	val = xhci_readl(xhci, &xhci->op_regs->cmd_ring[0]);
-	val = (val & ~CMD_RING_ADDR_MASK) |
-		(xhci->cmd_ring->first_seg->dma & CMD_RING_ADDR_MASK) |
+	val_64 = xhci_read_64(xhci, &xhci->op_regs->cmd_ring);
+	val_64 = (val_64 & (u64) CMD_RING_RSVD_BITS) |
+		(xhci->cmd_ring->first_seg->dma & (u64) ~CMD_RING_RSVD_BITS) |
 		xhci->cmd_ring->cycle_state;
-	xhci_dbg(xhci, "// Setting command ring address low bits to 0x%x\n", val);
-	xhci_writel(xhci, val, &xhci->op_regs->cmd_ring[0]);
-	xhci_dbg(xhci, "// Setting command ring address high bits to 0x0\n");
-	xhci_writel(xhci, (u32) 0, &xhci->op_regs->cmd_ring[1]);
+	xhci_dbg(xhci, "// Setting command ring address to 0x%x\n", val);
+	xhci_write_64(xhci, val_64, &xhci->op_regs->cmd_ring);
 	xhci_dbg_cmd_ptrs(xhci);
 
 	val = xhci_readl(xhci, &xhci->cap_regs->db_off);
@@ -722,8 +868,7 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	/* set ring base address and size for each segment table entry */
 	for (val = 0, seg = xhci->event_ring->first_seg; val < ERST_NUM_SEGS; val++) {
 		struct xhci_erst_entry *entry = &xhci->erst.entries[val];
-		entry->seg_addr[0] = seg->dma;
-		entry->seg_addr[1] = 0;
+		entry->seg_addr = seg->dma;
 		entry->seg_size = TRBS_PER_SEGMENT;
 		entry->rsvd = 0;
 		seg = seg->next;
@@ -741,11 +886,10 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	/* set the segment table base address */
 	xhci_dbg(xhci, "// Set ERST base address for ir_set 0 = 0x%llx\n",
 			(unsigned long long)xhci->erst.erst_dma_addr);
-	val = xhci_readl(xhci, &xhci->ir_set->erst_base[0]);
-	val &= ERST_PTR_MASK;
-	val |= (xhci->erst.erst_dma_addr & ~ERST_PTR_MASK);
-	xhci_writel(xhci, val, &xhci->ir_set->erst_base[0]);
-	xhci_writel(xhci, 0, &xhci->ir_set->erst_base[1]);
+	val_64 = xhci_read_64(xhci, &xhci->ir_set->erst_base);
+	val_64 &= ERST_PTR_MASK;
+	val_64 |= (xhci->erst.erst_dma_addr & (u64) ~ERST_PTR_MASK);
+	xhci_write_64(xhci, val_64, &xhci->ir_set->erst_base);
 
 	/* Set the event ring dequeue address */
 	xhci_set_hc_event_deq(xhci);
@@ -761,7 +905,11 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	for (i = 0; i < MAX_HC_SLOTS; ++i)
 		xhci->devs[i] = 0;
 
+	if (scratchpad_alloc(xhci, flags))
+		goto fail;
+
 	return 0;
+
 fail:
 	xhci_warn(xhci, "Couldn't initialize memory\n");
 	xhci_mem_cleanup(xhci);
diff --git a/drivers/usb/host/xhci-pci.c b/drivers/usb/host/xhci-pci.c
index 1462709e26c..592fe7e623f 100644
--- a/drivers/usb/host/xhci-pci.c
+++ b/drivers/usb/host/xhci-pci.c
@@ -117,6 +117,7 @@ static const struct hc_driver xhci_pci_hc_driver = {
 	.free_dev =		xhci_free_dev,
 	.add_endpoint =		xhci_add_endpoint,
 	.drop_endpoint =	xhci_drop_endpoint,
+	.endpoint_reset =	xhci_endpoint_reset,
 	.check_bandwidth =	xhci_check_bandwidth,
 	.reset_bandwidth =	xhci_reset_bandwidth,
 	.address_device =	xhci_address_device,
diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 02d81985c45..aa88a067148 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -135,6 +135,7 @@ static void next_trb(struct xhci_hcd *xhci,
 static void inc_deq(struct xhci_hcd *xhci, struct xhci_ring *ring, bool consumer)
 {
 	union xhci_trb *next = ++(ring->dequeue);
+	unsigned long long addr;
 
 	ring->deq_updates++;
 	/* Update the dequeue pointer further if that was a link TRB or we're at
@@ -152,6 +153,13 @@ static void inc_deq(struct xhci_hcd *xhci, struct xhci_ring *ring, bool consumer
 		ring->dequeue = ring->deq_seg->trbs;
 		next = ring->dequeue;
 	}
+	addr = (unsigned long long) xhci_trb_virt_to_dma(ring->deq_seg, ring->dequeue);
+	if (ring == xhci->event_ring)
+		xhci_dbg(xhci, "Event ring deq = 0x%llx (DMA)\n", addr);
+	else if (ring == xhci->cmd_ring)
+		xhci_dbg(xhci, "Command ring deq = 0x%llx (DMA)\n", addr);
+	else
+		xhci_dbg(xhci, "Ring deq = 0x%llx (DMA)\n", addr);
 }
 
 /*
@@ -171,6 +179,7 @@ static void inc_enq(struct xhci_hcd *xhci, struct xhci_ring *ring, bool consumer
 {
 	u32 chain;
 	union xhci_trb *next;
+	unsigned long long addr;
 
 	chain = ring->enqueue->generic.field[3] & TRB_CHAIN;
 	next = ++(ring->enqueue);
@@ -204,6 +213,13 @@ static void inc_enq(struct xhci_hcd *xhci, struct xhci_ring *ring, bool consumer
 		ring->enqueue = ring->enq_seg->trbs;
 		next = ring->enqueue;
 	}
+	addr = (unsigned long long) xhci_trb_virt_to_dma(ring->enq_seg, ring->enqueue);
+	if (ring == xhci->event_ring)
+		xhci_dbg(xhci, "Event ring enq = 0x%llx (DMA)\n", addr);
+	else if (ring == xhci->cmd_ring)
+		xhci_dbg(xhci, "Command ring enq = 0x%llx (DMA)\n", addr);
+	else
+		xhci_dbg(xhci, "Ring enq = 0x%llx (DMA)\n", addr);
 }
 
 /*
@@ -237,7 +253,7 @@ static int room_on_ring(struct xhci_hcd *xhci, struct xhci_ring *ring,
 
 void xhci_set_hc_event_deq(struct xhci_hcd *xhci)
 {
-	u32 temp;
+	u64 temp;
 	dma_addr_t deq;
 
 	deq = xhci_trb_virt_to_dma(xhci->event_ring->deq_seg,
@@ -246,13 +262,15 @@ void xhci_set_hc_event_deq(struct xhci_hcd *xhci)
 		xhci_warn(xhci, "WARN something wrong with SW event ring "
 				"dequeue ptr.\n");
 	/* Update HC event ring dequeue pointer */
-	temp = xhci_readl(xhci, &xhci->ir_set->erst_dequeue[0]);
+	temp = xhci_read_64(xhci, &xhci->ir_set->erst_dequeue);
 	temp &= ERST_PTR_MASK;
-	if (!in_interrupt())
-		xhci_dbg(xhci, "// Write event ring dequeue pointer\n");
-	xhci_writel(xhci, 0, &xhci->ir_set->erst_dequeue[1]);
-	xhci_writel(xhci, (deq & ~ERST_PTR_MASK) | temp,
-			&xhci->ir_set->erst_dequeue[0]);
+	/* Don't clear the EHB bit (which is RW1C) because
+	 * there might be more events to service.
+	 */
+	temp &= ~ERST_EHB;
+	xhci_dbg(xhci, "// Write event ring dequeue pointer, preserving EHB bit\n");
+	xhci_write_64(xhci, ((u64) deq & (u64) ~ERST_PTR_MASK) | temp,
+			&xhci->ir_set->erst_dequeue);
 }
 
 /* Ring the host controller doorbell after placing a command on the ring */
@@ -279,7 +297,8 @@ static void ring_ep_doorbell(struct xhci_hcd *xhci,
 	/* Don't ring the doorbell for this endpoint if there are pending
 	 * cancellations because the we don't want to interrupt processing.
 	 */
-	if (!ep_ring->cancels_pending && !(ep_ring->state & SET_DEQ_PENDING)) {
+	if (!ep_ring->cancels_pending && !(ep_ring->state & SET_DEQ_PENDING)
+			&& !(ep_ring->state & EP_HALTED)) {
 		field = xhci_readl(xhci, db_addr) & DB_MASK;
 		xhci_writel(xhci, field | EPI_TO_DB(ep_index), db_addr);
 		/* Flush PCI posted writes - FIXME Matthew Wilcox says this
@@ -316,12 +335,6 @@ static struct xhci_segment *find_trb_seg(
 	return cur_seg;
 }
 
-struct dequeue_state {
-	struct xhci_segment *new_deq_seg;
-	union xhci_trb *new_deq_ptr;
-	int new_cycle_state;
-};
-
 /*
  * Move the xHC's endpoint ring dequeue pointer past cur_td.
  * Record the new state of the xHC's endpoint ring dequeue segment,
@@ -336,24 +349,30 @@ struct dequeue_state {
  *  - Finally we move the dequeue state one TRB further, toggling the cycle bit
  *    if we've moved it past a link TRB with the toggle cycle bit set.
  */
-static void find_new_dequeue_state(struct xhci_hcd *xhci,
+void xhci_find_new_dequeue_state(struct xhci_hcd *xhci,
 		unsigned int slot_id, unsigned int ep_index,
-		struct xhci_td *cur_td, struct dequeue_state *state)
+		struct xhci_td *cur_td, struct xhci_dequeue_state *state)
 {
 	struct xhci_virt_device *dev = xhci->devs[slot_id];
 	struct xhci_ring *ep_ring = dev->ep_rings[ep_index];
 	struct xhci_generic_trb *trb;
+	struct xhci_ep_ctx *ep_ctx;
+	dma_addr_t addr;
 
 	state->new_cycle_state = 0;
+	xhci_dbg(xhci, "Finding segment containing stopped TRB.\n");
 	state->new_deq_seg = find_trb_seg(cur_td->start_seg,
 			ep_ring->stopped_trb,
 			&state->new_cycle_state);
 	if (!state->new_deq_seg)
 		BUG();
 	/* Dig out the cycle state saved by the xHC during the stop ep cmd */
-	state->new_cycle_state = 0x1 & dev->out_ctx->ep[ep_index].deq[0];
+	xhci_dbg(xhci, "Finding endpoint context\n");
+	ep_ctx = xhci_get_ep_ctx(xhci, dev->out_ctx, ep_index);
+	state->new_cycle_state = 0x1 & ep_ctx->deq;
 
 	state->new_deq_ptr = cur_td->last_trb;
+	xhci_dbg(xhci, "Finding segment containing last TRB in TD.\n");
 	state->new_deq_seg = find_trb_seg(state->new_deq_seg,
 			state->new_deq_ptr,
 			&state->new_cycle_state);
@@ -367,6 +386,12 @@ static void find_new_dequeue_state(struct xhci_hcd *xhci,
 	next_trb(xhci, ep_ring, &state->new_deq_seg, &state->new_deq_ptr);
 
 	/* Don't update the ring cycle state for the producer (us). */
+	xhci_dbg(xhci, "New dequeue segment = %p (virtual)\n",
+			state->new_deq_seg);
+	addr = xhci_trb_virt_to_dma(state->new_deq_seg, state->new_deq_ptr);
+	xhci_dbg(xhci, "New dequeue pointer = 0x%llx (DMA)\n",
+			(unsigned long long) addr);
+	xhci_dbg(xhci, "Setting dequeue pointer in internal ring state.\n");
 	ep_ring->dequeue = state->new_deq_ptr;
 	ep_ring->deq_seg = state->new_deq_seg;
 }
@@ -416,6 +441,30 @@ static int queue_set_tr_deq(struct xhci_hcd *xhci, int slot_id,
 		unsigned int ep_index, struct xhci_segment *deq_seg,
 		union xhci_trb *deq_ptr, u32 cycle_state);
 
+void xhci_queue_new_dequeue_state(struct xhci_hcd *xhci,
+		struct xhci_ring *ep_ring, unsigned int slot_id,
+		unsigned int ep_index, struct xhci_dequeue_state *deq_state)
+{
+	xhci_dbg(xhci, "Set TR Deq Ptr cmd, new deq seg = %p (0x%llx dma), "
+			"new deq ptr = %p (0x%llx dma), new cycle = %u\n",
+			deq_state->new_deq_seg,
+			(unsigned long long)deq_state->new_deq_seg->dma,
+			deq_state->new_deq_ptr,
+			(unsigned long long)xhci_trb_virt_to_dma(deq_state->new_deq_seg, deq_state->new_deq_ptr),
+			deq_state->new_cycle_state);
+	queue_set_tr_deq(xhci, slot_id, ep_index,
+			deq_state->new_deq_seg,
+			deq_state->new_deq_ptr,
+			(u32) deq_state->new_cycle_state);
+	/* Stop the TD queueing code from ringing the doorbell until
+	 * this command completes.  The HC won't set the dequeue pointer
+	 * if the ring is running, and ringing the doorbell starts the
+	 * ring running.
+	 */
+	ep_ring->state |= SET_DEQ_PENDING;
+	xhci_ring_cmd_db(xhci);
+}
+
 /*
  * When we get a command completion for a Stop Endpoint Command, we need to
  * unlink any cancelled TDs from the ring.  There are two ways to do that:
@@ -436,7 +485,7 @@ static void handle_stopped_endpoint(struct xhci_hcd *xhci,
 	struct xhci_td *cur_td = 0;
 	struct xhci_td *last_unlinked_td;
 
-	struct dequeue_state deq_state;
+	struct xhci_dequeue_state deq_state;
 #ifdef CONFIG_USB_HCD_STAT
 	ktime_t stop_time = ktime_get();
 #endif
@@ -464,7 +513,7 @@ static void handle_stopped_endpoint(struct xhci_hcd *xhci,
 		 * move the xHC endpoint ring dequeue pointer past this TD.
 		 */
 		if (cur_td == ep_ring->stopped_td)
-			find_new_dequeue_state(xhci, slot_id, ep_index, cur_td,
+			xhci_find_new_dequeue_state(xhci, slot_id, ep_index, cur_td,
 					&deq_state);
 		else
 			td_to_noop(xhci, ep_ring, cur_td);
@@ -480,24 +529,8 @@ static void handle_stopped_endpoint(struct xhci_hcd *xhci,
 
 	/* If necessary, queue a Set Transfer Ring Dequeue Pointer command */
 	if (deq_state.new_deq_ptr && deq_state.new_deq_seg) {
-		xhci_dbg(xhci, "Set TR Deq Ptr cmd, new deq seg = %p (0x%llx dma), "
-				"new deq ptr = %p (0x%llx dma), new cycle = %u\n",
-				deq_state.new_deq_seg,
-				(unsigned long long)deq_state.new_deq_seg->dma,
-				deq_state.new_deq_ptr,
-				(unsigned long long)xhci_trb_virt_to_dma(deq_state.new_deq_seg, deq_state.new_deq_ptr),
-				deq_state.new_cycle_state);
-		queue_set_tr_deq(xhci, slot_id, ep_index,
-				deq_state.new_deq_seg,
-				deq_state.new_deq_ptr,
-				(u32) deq_state.new_cycle_state);
-		/* Stop the TD queueing code from ringing the doorbell until
-		 * this command completes.  The HC won't set the dequeue pointer
-		 * if the ring is running, and ringing the doorbell starts the
-		 * ring running.
-		 */
-		ep_ring->state |= SET_DEQ_PENDING;
-		xhci_ring_cmd_db(xhci);
+		xhci_queue_new_dequeue_state(xhci, ep_ring,
+				slot_id, ep_index, &deq_state);
 	} else {
 		/* Otherwise just ring the doorbell to restart the ring */
 		ring_ep_doorbell(xhci, slot_id, ep_index);
@@ -551,11 +584,15 @@ static void handle_set_deq_completion(struct xhci_hcd *xhci,
 	unsigned int ep_index;
 	struct xhci_ring *ep_ring;
 	struct xhci_virt_device *dev;
+	struct xhci_ep_ctx *ep_ctx;
+	struct xhci_slot_ctx *slot_ctx;
 
 	slot_id = TRB_TO_SLOT_ID(trb->generic.field[3]);
 	ep_index = TRB_TO_EP_INDEX(trb->generic.field[3]);
 	dev = xhci->devs[slot_id];
 	ep_ring = dev->ep_rings[ep_index];
+	ep_ctx = xhci_get_ep_ctx(xhci, dev->out_ctx, ep_index);
+	slot_ctx = xhci_get_slot_ctx(xhci, dev->out_ctx);
 
 	if (GET_COMP_CODE(event->status) != COMP_SUCCESS) {
 		unsigned int ep_state;
@@ -569,9 +606,9 @@ static void handle_set_deq_completion(struct xhci_hcd *xhci,
 		case COMP_CTX_STATE:
 			xhci_warn(xhci, "WARN Set TR Deq Ptr cmd failed due "
 					"to incorrect slot or ep state.\n");
-			ep_state = dev->out_ctx->ep[ep_index].ep_info;
+			ep_state = ep_ctx->ep_info;
 			ep_state &= EP_STATE_MASK;
-			slot_state = dev->out_ctx->slot.dev_state;
+			slot_state = slot_ctx->dev_state;
 			slot_state = GET_SLOT_STATE(slot_state);
 			xhci_dbg(xhci, "Slot state = %u, EP state = %u\n",
 					slot_state, ep_state);
@@ -593,16 +630,33 @@ static void handle_set_deq_completion(struct xhci_hcd *xhci,
 		 * cancelling URBs, which might not be an error...
 		 */
 	} else {
-		xhci_dbg(xhci, "Successful Set TR Deq Ptr cmd, deq[0] = 0x%x, "
-				"deq[1] = 0x%x.\n",
-				dev->out_ctx->ep[ep_index].deq[0],
-				dev->out_ctx->ep[ep_index].deq[1]);
+		xhci_dbg(xhci, "Successful Set TR Deq Ptr cmd, deq = @%08llx\n",
+				ep_ctx->deq);
 	}
 
 	ep_ring->state &= ~SET_DEQ_PENDING;
 	ring_ep_doorbell(xhci, slot_id, ep_index);
 }
 
+static void handle_reset_ep_completion(struct xhci_hcd *xhci,
+		struct xhci_event_cmd *event,
+		union xhci_trb *trb)
+{
+	int slot_id;
+	unsigned int ep_index;
+
+	slot_id = TRB_TO_SLOT_ID(trb->generic.field[3]);
+	ep_index = TRB_TO_EP_INDEX(trb->generic.field[3]);
+	/* This command will only fail if the endpoint wasn't halted,
+	 * but we don't care.
+	 */
+	xhci_dbg(xhci, "Ignoring reset ep completion code of %u\n",
+			(unsigned int) GET_COMP_CODE(event->status));
+
+	/* Clear our internal halted state and restart the ring */
+	xhci->devs[slot_id]->ep_rings[ep_index]->state &= ~EP_HALTED;
+	ring_ep_doorbell(xhci, slot_id, ep_index);
+}
 
 static void handle_cmd_completion(struct xhci_hcd *xhci,
 		struct xhci_event_cmd *event)
@@ -611,7 +665,7 @@ static void handle_cmd_completion(struct xhci_hcd *xhci,
 	u64 cmd_dma;
 	dma_addr_t cmd_dequeue_dma;
 
-	cmd_dma = (((u64) event->cmd_trb[1]) << 32) + event->cmd_trb[0];
+	cmd_dma = event->cmd_trb;
 	cmd_dequeue_dma = xhci_trb_virt_to_dma(xhci->cmd_ring->deq_seg,
 			xhci->cmd_ring->dequeue);
 	/* Is the command ring deq ptr out of sync with the deq seg ptr? */
@@ -653,6 +707,9 @@ static void handle_cmd_completion(struct xhci_hcd *xhci,
 	case TRB_TYPE(TRB_CMD_NOOP):
 		++xhci->noops_handled;
 		break;
+	case TRB_TYPE(TRB_RESET_EP):
+		handle_reset_ep_completion(xhci, event, xhci->cmd_ring->dequeue);
+		break;
 	default:
 		/* Skip over unknown commands on the event ring */
 		xhci->error_bitmask |= 1 << 6;
@@ -756,7 +813,9 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 	union xhci_trb *event_trb;
 	struct urb *urb = 0;
 	int status = -EINPROGRESS;
+	struct xhci_ep_ctx *ep_ctx;
 
+	xhci_dbg(xhci, "In %s\n", __func__);
 	xdev = xhci->devs[TRB_TO_SLOT_ID(event->flags)];
 	if (!xdev) {
 		xhci_err(xhci, "ERROR Transfer event pointed to bad slot\n");
@@ -765,17 +824,17 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 
 	/* Endpoint ID is 1 based, our index is zero based */
 	ep_index = TRB_TO_EP_ID(event->flags) - 1;
+	xhci_dbg(xhci, "%s - ep index = %d\n", __func__, ep_index);
 	ep_ring = xdev->ep_rings[ep_index];
-	if (!ep_ring || (xdev->out_ctx->ep[ep_index].ep_info & EP_STATE_MASK) == EP_STATE_DISABLED) {
+	ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
+	if (!ep_ring || (ep_ctx->ep_info & EP_STATE_MASK) == EP_STATE_DISABLED) {
 		xhci_err(xhci, "ERROR Transfer event pointed to disabled endpoint\n");
 		return -ENODEV;
 	}
 
-	event_dma = event->buffer[0];
-	if (event->buffer[1] != 0)
-		xhci_warn(xhci, "WARN ignoring upper 32-bits of 64-bit TRB dma address\n");
-
+	event_dma = event->buffer;
 	/* This TRB should be in the TD at the head of this ring's TD list */
+	xhci_dbg(xhci, "%s - checking for list empty\n", __func__);
 	if (list_empty(&ep_ring->td_list)) {
 		xhci_warn(xhci, "WARN Event TRB for slot %d ep %d with no TDs queued?\n",
 				TRB_TO_SLOT_ID(event->flags), ep_index);
@@ -785,11 +844,14 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 		urb = NULL;
 		goto cleanup;
 	}
+	xhci_dbg(xhci, "%s - getting list entry\n", __func__);
 	td = list_entry(ep_ring->td_list.next, struct xhci_td, td_list);
 
 	/* Is this a TRB in the currently executing TD? */
+	xhci_dbg(xhci, "%s - looking for TD\n", __func__);
 	event_seg = trb_in_td(ep_ring->deq_seg, ep_ring->dequeue,
 			td->last_trb, event_dma);
+	xhci_dbg(xhci, "%s - found event_seg = %p\n", __func__, event_seg);
 	if (!event_seg) {
 		/* HC is busted, give up! */
 		xhci_err(xhci, "ERROR Transfer event TRB DMA ptr not part of current TD\n");
@@ -798,10 +860,10 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 	event_trb = &event_seg->trbs[(event_dma - event_seg->dma) / sizeof(*event_trb)];
 	xhci_dbg(xhci, "Event TRB with TRB type ID %u\n",
 			(unsigned int) (event->flags & TRB_TYPE_BITMASK)>>10);
-	xhci_dbg(xhci, "Offset 0x00 (buffer[0]) = 0x%x\n",
-			(unsigned int) event->buffer[0]);
-	xhci_dbg(xhci, "Offset 0x04 (buffer[0]) = 0x%x\n",
-			(unsigned int) event->buffer[1]);
+	xhci_dbg(xhci, "Offset 0x00 (buffer lo) = 0x%x\n",
+			lower_32_bits(event->buffer));
+	xhci_dbg(xhci, "Offset 0x04 (buffer hi) = 0x%x\n",
+			upper_32_bits(event->buffer));
 	xhci_dbg(xhci, "Offset 0x08 (transfer length) = 0x%x\n",
 			(unsigned int) event->transfer_len);
 	xhci_dbg(xhci, "Offset 0x0C (flags) = 0x%x\n",
@@ -823,6 +885,7 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 		break;
 	case COMP_STALL:
 		xhci_warn(xhci, "WARN: Stalled endpoint\n");
+		ep_ring->state |= EP_HALTED;
 		status = -EPIPE;
 		break;
 	case COMP_TRB_ERR:
@@ -833,6 +896,10 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 		xhci_warn(xhci, "WARN: transfer error on endpoint\n");
 		status = -EPROTO;
 		break;
+	case COMP_BABBLE:
+		xhci_warn(xhci, "WARN: babble error on endpoint\n");
+		status = -EOVERFLOW;
+		break;
 	case COMP_DB_ERR:
 		xhci_warn(xhci, "WARN: HC couldn't access mem fast enough\n");
 		status = -ENOSR;
@@ -874,15 +941,26 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 		if (event_trb != ep_ring->dequeue) {
 			/* The event was for the status stage */
 			if (event_trb == td->last_trb) {
-				td->urb->actual_length =
-					td->urb->transfer_buffer_length;
+				if (td->urb->actual_length != 0) {
+					/* Don't overwrite a previously set error code */
+					if (status == -EINPROGRESS || status == 0)
+						/* Did we already see a short data stage? */
+						status = -EREMOTEIO;
+				} else {
+					td->urb->actual_length =
+						td->urb->transfer_buffer_length;
+				}
 			} else {
 			/* Maybe the event was for the data stage? */
-				if (GET_COMP_CODE(event->transfer_len) != COMP_STOP_INVAL)
+				if (GET_COMP_CODE(event->transfer_len) != COMP_STOP_INVAL) {
 					/* We didn't stop on a link TRB in the middle */
 					td->urb->actual_length =
 						td->urb->transfer_buffer_length -
 						TRB_LEN(event->transfer_len);
+					xhci_dbg(xhci, "Waiting for status stage event\n");
+					urb = NULL;
+					goto cleanup;
+				}
 			}
 		}
 	} else {
@@ -929,16 +1007,20 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 							TRB_LEN(event->transfer_len));
 					td->urb->actual_length = 0;
 				}
-				if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
-					status = -EREMOTEIO;
-				else
-					status = 0;
+				/* Don't overwrite a previously set error code */
+				if (status == -EINPROGRESS) {
+					if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
+						status = -EREMOTEIO;
+					else
+						status = 0;
+				}
 			} else {
 				td->urb->actual_length = td->urb->transfer_buffer_length;
 				/* Ignore a short packet completion if the
 				 * untransferred length was zero.
 				 */
-				status = 0;
+				if (status == -EREMOTEIO)
+					status = 0;
 			}
 		} else {
 			/* Slow path - walk the list, starting from the dequeue
@@ -965,19 +1047,30 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 					TRB_LEN(event->transfer_len);
 		}
 	}
-	/* The Endpoint Stop Command completion will take care of
-	 * any stopped TDs.  A stopped TD may be restarted, so don't update the
-	 * ring dequeue pointer or take this TD off any lists yet.
-	 */
 	if (GET_COMP_CODE(event->transfer_len) == COMP_STOP_INVAL ||
 			GET_COMP_CODE(event->transfer_len) == COMP_STOP) {
+		/* The Endpoint Stop Command completion will take care of any
+		 * stopped TDs.  A stopped TD may be restarted, so don't update
+		 * the ring dequeue pointer or take this TD off any lists yet.
+		 */
 		ep_ring->stopped_td = td;
 		ep_ring->stopped_trb = event_trb;
 	} else {
-		/* Update ring dequeue pointer */
-		while (ep_ring->dequeue != td->last_trb)
+		if (GET_COMP_CODE(event->transfer_len) == COMP_STALL) {
+			/* The transfer is completed from the driver's
+			 * perspective, but we need to issue a set dequeue
+			 * command for this stalled endpoint to move the dequeue
+			 * pointer past the TD.  We can't do that here because
+			 * the halt condition must be cleared first.
+			 */
+			ep_ring->stopped_td = td;
+			ep_ring->stopped_trb = event_trb;
+		} else {
+			/* Update ring dequeue pointer */
+			while (ep_ring->dequeue != td->last_trb)
+				inc_deq(xhci, ep_ring, false);
 			inc_deq(xhci, ep_ring, false);
-		inc_deq(xhci, ep_ring, false);
+		}
 
 		/* Clean up the endpoint's TD list */
 		urb = td->urb;
@@ -987,7 +1080,10 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 			list_del(&td->cancelled_td_list);
 			ep_ring->cancels_pending--;
 		}
-		kfree(td);
+		/* Leave the TD around for the reset endpoint function to use */
+		if (GET_COMP_CODE(event->transfer_len) != COMP_STALL) {
+			kfree(td);
+		}
 		urb->hcpriv = NULL;
 	}
 cleanup:
@@ -997,6 +1093,8 @@ cleanup:
 	/* FIXME for multi-TD URBs (who have buffers bigger than 64MB) */
 	if (urb) {
 		usb_hcd_unlink_urb_from_ep(xhci_to_hcd(xhci), urb);
+		xhci_dbg(xhci, "Giveback URB %p, len = %d, status = %d\n",
+				urb, td->urb->actual_length, status);
 		spin_unlock(&xhci->lock);
 		usb_hcd_giveback_urb(xhci_to_hcd(xhci), urb, status);
 		spin_lock(&xhci->lock);
@@ -1014,6 +1112,7 @@ void xhci_handle_event(struct xhci_hcd *xhci)
 	int update_ptrs = 1;
 	int ret;
 
+	xhci_dbg(xhci, "In %s\n", __func__);
 	if (!xhci->event_ring || !xhci->event_ring->dequeue) {
 		xhci->error_bitmask |= 1 << 1;
 		return;
@@ -1026,18 +1125,25 @@ void xhci_handle_event(struct xhci_hcd *xhci)
 		xhci->error_bitmask |= 1 << 2;
 		return;
 	}
+	xhci_dbg(xhci, "%s - OS owns TRB\n", __func__);
 
 	/* FIXME: Handle more event types. */
 	switch ((event->event_cmd.flags & TRB_TYPE_BITMASK)) {
 	case TRB_TYPE(TRB_COMPLETION):
+		xhci_dbg(xhci, "%s - calling handle_cmd_completion\n", __func__);
 		handle_cmd_completion(xhci, &event->event_cmd);
+		xhci_dbg(xhci, "%s - returned from handle_cmd_completion\n", __func__);
 		break;
 	case TRB_TYPE(TRB_PORT_STATUS):
+		xhci_dbg(xhci, "%s - calling handle_port_status\n", __func__);
 		handle_port_status(xhci, event);
+		xhci_dbg(xhci, "%s - returned from handle_port_status\n", __func__);
 		update_ptrs = 0;
 		break;
 	case TRB_TYPE(TRB_TRANSFER):
+		xhci_dbg(xhci, "%s - calling handle_tx_event\n", __func__);
 		ret = handle_tx_event(xhci, &event->trans_event);
+		xhci_dbg(xhci, "%s - returned from handle_tx_event\n", __func__);
 		if (ret < 0)
 			xhci->error_bitmask |= 1 << 9;
 		else
@@ -1093,13 +1199,13 @@ static int prepare_ring(struct xhci_hcd *xhci, struct xhci_ring *ep_ring,
 		 */
 		xhci_warn(xhci, "WARN urb submitted to disabled ep\n");
 		return -ENOENT;
-	case EP_STATE_HALTED:
 	case EP_STATE_ERROR:
-		xhci_warn(xhci, "WARN waiting for halt or error on ep "
-				"to be cleared\n");
+		xhci_warn(xhci, "WARN waiting for error on ep to be cleared\n");
 		/* FIXME event handling code for error needs to clear it */
 		/* XXX not sure if this should be -ENOENT or not */
 		return -EINVAL;
+	case EP_STATE_HALTED:
+		xhci_dbg(xhci, "WARN halted endpoint, queueing URB anyway.\n");
 	case EP_STATE_STOPPED:
 	case EP_STATE_RUNNING:
 		break;
@@ -1128,9 +1234,9 @@ static int prepare_transfer(struct xhci_hcd *xhci,
 		gfp_t mem_flags)
 {
 	int ret;
-
+	struct xhci_ep_ctx *ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
 	ret = prepare_ring(xhci, xdev->ep_rings[ep_index],
-			xdev->out_ctx->ep[ep_index].ep_info & EP_STATE_MASK,
+			ep_ctx->ep_info & EP_STATE_MASK,
 			num_trbs, mem_flags);
 	if (ret)
 		return ret;
@@ -1285,6 +1391,7 @@ static int queue_bulk_sg_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 	/* Queue the first TRB, even if it's zero-length */
 	do {
 		u32 field = 0;
+		u32 length_field = 0;
 
 		/* Don't change the cycle bit of the first TRB until later */
 		if (first_trb)
@@ -1314,10 +1421,13 @@ static int queue_bulk_sg_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 					(unsigned int) (addr + TRB_MAX_BUFF_SIZE) & ~(TRB_MAX_BUFF_SIZE - 1),
 					(unsigned int) addr + trb_buff_len);
 		}
+		length_field = TRB_LEN(trb_buff_len) |
+			TD_REMAINDER(urb->transfer_buffer_length - running_total) |
+			TRB_INTR_TARGET(0);
 		queue_trb(xhci, ep_ring, false,
-				(u32) addr,
-				(u32) ((u64) addr >> 32),
-				TRB_LEN(trb_buff_len) | TRB_INTR_TARGET(0),
+				lower_32_bits(addr),
+				upper_32_bits(addr),
+				length_field,
 				/* We always want to know if the TRB was short,
 				 * or we won't get an event when it completes.
 				 * (Unless we use event data TRBs, which are a
@@ -1365,7 +1475,7 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 	struct xhci_generic_trb *start_trb;
 	bool first_trb;
 	int start_cycle;
-	u32 field;
+	u32 field, length_field;
 
 	int running_total, trb_buff_len, ret;
 	u64 addr;
@@ -1443,10 +1553,13 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 			td->last_trb = ep_ring->enqueue;
 			field |= TRB_IOC;
 		}
+		length_field = TRB_LEN(trb_buff_len) |
+			TD_REMAINDER(urb->transfer_buffer_length - running_total) |
+			TRB_INTR_TARGET(0);
 		queue_trb(xhci, ep_ring, false,
-				(u32) addr,
-				(u32) ((u64) addr >> 32),
-				TRB_LEN(trb_buff_len) | TRB_INTR_TARGET(0),
+				lower_32_bits(addr),
+				upper_32_bits(addr),
+				length_field,
 				/* We always want to know if the TRB was short,
 				 * or we won't get an event when it completes.
 				 * (Unless we use event data TRBs, which are a
@@ -1478,7 +1591,7 @@ int xhci_queue_ctrl_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 	struct usb_ctrlrequest *setup;
 	struct xhci_generic_trb *start_trb;
 	int start_cycle;
-	u32 field;
+	u32 field, length_field;
 	struct xhci_td *td;
 
 	ep_ring = xhci->devs[slot_id]->ep_rings[ep_index];
@@ -1528,13 +1641,16 @@ int xhci_queue_ctrl_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 
 	/* If there's data, queue data TRBs */
 	field = 0;
+	length_field = TRB_LEN(urb->transfer_buffer_length) |
+		TD_REMAINDER(urb->transfer_buffer_length) |
+		TRB_INTR_TARGET(0);
 	if (urb->transfer_buffer_length > 0) {
 		if (setup->bRequestType & USB_DIR_IN)
 			field |= TRB_DIR_IN;
 		queue_trb(xhci, ep_ring, false,
 				lower_32_bits(urb->transfer_dma),
 				upper_32_bits(urb->transfer_dma),
-				TRB_LEN(urb->transfer_buffer_length) | TRB_INTR_TARGET(0),
+				length_field,
 				/* Event on short tx */
 				field | TRB_ISP | TRB_TYPE(TRB_DATA) | ep_ring->cycle_state);
 	}
@@ -1603,7 +1719,8 @@ int xhci_queue_slot_control(struct xhci_hcd *xhci, u32 trb_type, u32 slot_id)
 int xhci_queue_address_device(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
 		u32 slot_id)
 {
-	return queue_command(xhci, in_ctx_ptr, 0, 0,
+	return queue_command(xhci, lower_32_bits(in_ctx_ptr),
+			upper_32_bits(in_ctx_ptr), 0,
 			TRB_TYPE(TRB_ADDR_DEV) | SLOT_ID_FOR_TRB(slot_id));
 }
 
@@ -1611,7 +1728,8 @@ int xhci_queue_address_device(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
 int xhci_queue_configure_endpoint(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
 		u32 slot_id)
 {
-	return queue_command(xhci, in_ctx_ptr, 0, 0,
+	return queue_command(xhci, lower_32_bits(in_ctx_ptr),
+			upper_32_bits(in_ctx_ptr), 0,
 			TRB_TYPE(TRB_CONFIG_EP) | SLOT_ID_FOR_TRB(slot_id));
 }
 
@@ -1639,10 +1757,23 @@ static int queue_set_tr_deq(struct xhci_hcd *xhci, int slot_id,
 	u32 type = TRB_TYPE(TRB_SET_DEQ);
 
 	addr = xhci_trb_virt_to_dma(deq_seg, deq_ptr);
-	if (addr == 0)
+	if (addr == 0) {
 		xhci_warn(xhci, "WARN Cannot submit Set TR Deq Ptr\n");
 		xhci_warn(xhci, "WARN deq seg = %p, deq pt = %p\n",
 				deq_seg, deq_ptr);
-	return queue_command(xhci, (u32) addr | cycle_state, 0, 0,
+		return 0;
+	}
+	return queue_command(xhci, lower_32_bits(addr) | cycle_state,
+			upper_32_bits(addr), 0,
 			trb_slot_id | trb_ep_index | type);
 }
+
+int xhci_queue_reset_ep(struct xhci_hcd *xhci, int slot_id,
+		unsigned int ep_index)
+{
+	u32 trb_slot_id = SLOT_ID_FOR_TRB(slot_id);
+	u32 trb_ep_index = EP_ID_FOR_TRB(ep_index);
+	u32 type = TRB_TYPE(TRB_RESET_EP);
+
+	return queue_command(xhci, 0, 0, 0, trb_slot_id | trb_ep_index | type);
+}
diff --git a/drivers/usb/host/xhci.h b/drivers/usb/host/xhci.h
index 8936eeb5588..d31d32206ba 100644
--- a/drivers/usb/host/xhci.h
+++ b/drivers/usb/host/xhci.h
@@ -25,6 +25,7 @@
 
 #include <linux/usb.h>
 #include <linux/timer.h>
+#include <linux/kernel.h>
 
 #include "../core/hcd.h"
 /* Code sharing between pci-quirks and xhci hcd */
@@ -42,14 +43,6 @@
  * xHCI register interface.
  * This corresponds to the eXtensible Host Controller Interface (xHCI)
  * Revision 0.95 specification
- *
- * Registers should always be accessed with double word or quad word accesses.
- *
- * Some xHCI implementations may support 64-bit address pointers.  Registers
- * with 64-bit address pointers should be written to with dword accesses by
- * writing the low dword first (ptr[0]), then the high dword (ptr[1]) second.
- * xHCI implementations that do not support 64-bit address pointers will ignore
- * the high dword, and write order is irrelevant.
  */
 
 /**
@@ -96,6 +89,7 @@ struct xhci_cap_regs {
 #define HCS_ERST_MAX(p)		(((p) >> 4) & 0xf)
 /* bit 26 Scratchpad restore - for save/restore HW state - not used yet */
 /* bits 27:31 number of Scratchpad buffers SW must allocate for the HW */
+#define HCS_MAX_SCRATCHPAD(p)   (((p) >> 27) & 0x1f)
 
 /* HCSPARAMS3 - hcs_params3 - bitmasks */
 /* bits 0:7, Max U1 to U0 latency for the roothub ports */
@@ -166,10 +160,10 @@ struct xhci_op_regs {
 	u32	reserved1;
 	u32	reserved2;
 	u32	dev_notification;
-	u32	cmd_ring[2];
+	u64	cmd_ring;
 	/* rsvd: offset 0x20-2F */
 	u32	reserved3[4];
-	u32	dcbaa_ptr[2];
+	u64	dcbaa_ptr;
 	u32	config_reg;
 	/* rsvd: offset 0x3C-3FF */
 	u32	reserved4[241];
@@ -254,7 +248,7 @@ struct xhci_op_regs {
 #define CMD_RING_RUNNING	(1 << 3)
 /* bits 4:5 reserved and should be preserved */
 /* Command Ring pointer - bit mask for the lower 32 bits. */
-#define CMD_RING_ADDR_MASK	(0xffffffc0)
+#define CMD_RING_RSVD_BITS	(0x3f)
 
 /* CONFIG - Configure Register - config_reg bitmasks */
 /* bits 0:7 - maximum number of device slots enabled (NumSlotsEn) */
@@ -382,8 +376,8 @@ struct xhci_intr_reg {
 	u32	irq_control;
 	u32	erst_size;
 	u32	rsvd;
-	u32	erst_base[2];
-	u32	erst_dequeue[2];
+	u64	erst_base;
+	u64	erst_dequeue;
 };
 
 /* irq_pending bitmasks */
@@ -453,6 +447,27 @@ struct xhci_doorbell_array {
 
 
 /**
+ * struct xhci_container_ctx
+ * @type: Type of context.  Used to calculated offsets to contained contexts.
+ * @size: Size of the context data
+ * @bytes: The raw context data given to HW
+ * @dma: dma address of the bytes
+ *
+ * Represents either a Device or Input context.  Holds a pointer to the raw
+ * memory used for the context (bytes) and dma address of it (dma).
+ */
+struct xhci_container_ctx {
+	unsigned type;
+#define XHCI_CTX_TYPE_DEVICE  0x1
+#define XHCI_CTX_TYPE_INPUT   0x2
+
+	int size;
+
+	u8 *bytes;
+	dma_addr_t dma;
+};
+
+/**
  * struct xhci_slot_ctx
  * @dev_info:	Route string, device speed, hub info, and last valid endpoint
  * @dev_info2:	Max exit latency for device number, root hub port number
@@ -538,7 +553,7 @@ struct xhci_slot_ctx {
 struct xhci_ep_ctx {
 	u32	ep_info;
 	u32	ep_info2;
-	u32	deq[2];
+	u64	deq;
 	u32	tx_info;
 	/* offset 0x14 - 0x1f reserved for HC internal use */
 	u32	reserved[3];
@@ -589,18 +604,16 @@ struct xhci_ep_ctx {
 
 
 /**
- * struct xhci_device_control
- * Input/Output context; see section 6.2.5.
+ * struct xhci_input_control_context
+ * Input control context; see section 6.2.5.
  *
  * @drop_context:	set the bit of the endpoint context you want to disable
  * @add_context:	set the bit of the endpoint context you want to enable
  */
-struct xhci_device_control {
+struct xhci_input_control_ctx {
 	u32	drop_flags;
 	u32	add_flags;
-	u32	rsvd[6];
-	struct xhci_slot_ctx	slot;
-	struct xhci_ep_ctx	ep[31];
+	u32	rsvd2[6];
 };
 
 /* drop context bitmasks */
@@ -608,7 +621,6 @@ struct xhci_device_control {
 /* add context bitmasks */
 #define	ADD_EP(x)	(0x1 << x)
 
-
 struct xhci_virt_device {
 	/*
 	 * Commands to the hardware are passed an "input context" that
@@ -618,11 +630,10 @@ struct xhci_virt_device {
 	 * track of input and output contexts separately because
 	 * these commands might fail and we don't trust the hardware.
 	 */
-	struct xhci_device_control	*out_ctx;
-	dma_addr_t			out_ctx_dma;
+	struct xhci_container_ctx       *out_ctx;
 	/* Used for addressing devices and configuration changes */
-	struct xhci_device_control	*in_ctx;
-	dma_addr_t			in_ctx_dma;
+	struct xhci_container_ctx       *in_ctx;
+
 	/* FIXME when stream support is added */
 	struct xhci_ring		*ep_rings[31];
 	/* Temporary storage in case the configure endpoint command fails and we
@@ -641,7 +652,7 @@ struct xhci_virt_device {
  */
 struct xhci_device_context_array {
 	/* 64-bit device addresses; we only write 32-bit addresses */
-	u32			dev_context_ptrs[2*MAX_HC_SLOTS];
+	u64			dev_context_ptrs[MAX_HC_SLOTS];
 	/* private xHCD pointers */
 	dma_addr_t	dma;
 };
@@ -654,7 +665,7 @@ struct xhci_device_context_array {
 
 struct xhci_stream_ctx {
 	/* 64-bit stream ring address, cycle state, and stream type */
-	u32	stream_ring[2];
+	u64	stream_ring;
 	/* offset 0x14 - 0x1f reserved for HC internal use */
 	u32	reserved[2];
 };
@@ -662,7 +673,7 @@ struct xhci_stream_ctx {
 
 struct xhci_transfer_event {
 	/* 64-bit buffer address, or immediate data */
-	u32	buffer[2];
+	u64	buffer;
 	u32	transfer_len;
 	/* This field is interpreted differently based on the type of TRB */
 	u32	flags;
@@ -744,7 +755,7 @@ struct xhci_transfer_event {
 
 struct xhci_link_trb {
 	/* 64-bit segment pointer*/
-	u32 segment_ptr[2];
+	u64 segment_ptr;
 	u32 intr_target;
 	u32 control;
 };
@@ -755,7 +766,7 @@ struct xhci_link_trb {
 /* Command completion event TRB */
 struct xhci_event_cmd {
 	/* Pointer to command TRB, or the value passed by the event data trb */
-	u32 cmd_trb[2];
+	u64 cmd_trb;
 	u32 status;
 	u32 flags;
 };
@@ -848,8 +859,8 @@ union xhci_trb {
 #define TRB_CONFIG_EP		12
 /* Evaluate Context Command */
 #define TRB_EVAL_CONTEXT	13
-/* Reset Transfer Ring Command */
-#define TRB_RESET_RING		14
+/* Reset Endpoint Command */
+#define TRB_RESET_EP		14
 /* Stop Transfer Ring Command */
 #define TRB_STOP_RING		15
 /* Set Transfer Ring Dequeue Pointer Command */
@@ -929,6 +940,7 @@ struct xhci_ring {
 	unsigned int		cancels_pending;
 	unsigned int		state;
 #define SET_DEQ_PENDING		(1 << 0)
+#define EP_HALTED		(1 << 1)
 	/* The TRB that was last reported in a stopped endpoint ring */
 	union xhci_trb		*stopped_trb;
 	struct xhci_td		*stopped_td;
@@ -940,9 +952,15 @@ struct xhci_ring {
 	u32			cycle_state;
 };
 
+struct xhci_dequeue_state {
+	struct xhci_segment *new_deq_seg;
+	union xhci_trb *new_deq_ptr;
+	int new_cycle_state;
+};
+
 struct xhci_erst_entry {
 	/* 64-bit event ring segment address */
-	u32	seg_addr[2];
+	u64	seg_addr;
 	u32	seg_size;
 	/* Set to zero */
 	u32	rsvd;
@@ -957,6 +975,13 @@ struct xhci_erst {
 	unsigned int		erst_size;
 };
 
+struct xhci_scratchpad {
+	u64 *sp_array;
+	dma_addr_t sp_dma;
+	void **sp_buffers;
+	dma_addr_t *sp_dma_buffers;
+};
+
 /*
  * Each segment table entry is 4*32bits long.  1K seems like an ok size:
  * (1K bytes * 8bytes/bit) / (4*32 bits) = 64 segment entries in the table,
@@ -1011,6 +1036,9 @@ struct xhci_hcd {
 	struct xhci_ring	*cmd_ring;
 	struct xhci_ring	*event_ring;
 	struct xhci_erst	erst;
+	/* Scratchpad */
+	struct xhci_scratchpad  *scratchpad;
+
 	/* slot enabling and address device helpers */
 	struct completion	addr_dev;
 	int slot_id;
@@ -1071,13 +1099,43 @@ static inline unsigned int xhci_readl(const struct xhci_hcd *xhci,
 static inline void xhci_writel(struct xhci_hcd *xhci,
 		const unsigned int val, __u32 __iomem *regs)
 {
-	if (!in_interrupt())
-		xhci_dbg(xhci,
-			 "`MEM_WRITE_DWORD(3'b000, 32'h%p, 32'h%0x, 4'hf);\n",
-			 regs, val);
+	xhci_dbg(xhci,
+			"`MEM_WRITE_DWORD(3'b000, 32'h%p, 32'h%0x, 4'hf);\n",
+			regs, val);
 	writel(val, regs);
 }
 
+/*
+ * Registers should always be accessed with double word or quad word accesses.
+ *
+ * Some xHCI implementations may support 64-bit address pointers.  Registers
+ * with 64-bit address pointers should be written to with dword accesses by
+ * writing the low dword first (ptr[0]), then the high dword (ptr[1]) second.
+ * xHCI implementations that do not support 64-bit address pointers will ignore
+ * the high dword, and write order is irrelevant.
+ */
+static inline u64 xhci_read_64(const struct xhci_hcd *xhci,
+		__u64 __iomem *regs)
+{
+	__u32 __iomem *ptr = (__u32 __iomem *) regs;
+	u64 val_lo = readl(ptr);
+	u64 val_hi = readl(ptr + 1);
+	return val_lo + (val_hi << 32);
+}
+static inline void xhci_write_64(struct xhci_hcd *xhci,
+		const u64 val, __u64 __iomem *regs)
+{
+	__u32 __iomem *ptr = (__u32 __iomem *) regs;
+	u32 val_lo = lower_32_bits(val);
+	u32 val_hi = upper_32_bits(val);
+
+	xhci_dbg(xhci,
+			"`MEM_WRITE_DWORD(3'b000, 64'h%p, 64'h%0lx, 4'hf);\n",
+			regs, (long unsigned int) val);
+	writel(val_lo, ptr);
+	writel(val_hi, ptr + 1);
+}
+
 /* xHCI debugging */
 void xhci_print_ir_set(struct xhci_hcd *xhci, struct xhci_intr_reg *ir_set, int set_num);
 void xhci_print_registers(struct xhci_hcd *xhci);
@@ -1090,7 +1148,7 @@ void xhci_debug_ring(struct xhci_hcd *xhci, struct xhci_ring *ring);
 void xhci_dbg_erst(struct xhci_hcd *xhci, struct xhci_erst *erst);
 void xhci_dbg_cmd_ptrs(struct xhci_hcd *xhci);
 void xhci_dbg_ring_ptrs(struct xhci_hcd *xhci, struct xhci_ring *ring);
-void xhci_dbg_ctx(struct xhci_hcd *xhci, struct xhci_device_control *ctx, dma_addr_t dma, unsigned int last_ep);
+void xhci_dbg_ctx(struct xhci_hcd *xhci, struct xhci_container_ctx *ctx, unsigned int last_ep);
 
 /* xHCI memory managment */
 void xhci_mem_cleanup(struct xhci_hcd *xhci);
@@ -1128,6 +1186,7 @@ int xhci_urb_enqueue(struct usb_hcd *hcd, struct urb *urb, gfp_t mem_flags);
 int xhci_urb_dequeue(struct usb_hcd *hcd, struct urb *urb, int status);
 int xhci_add_endpoint(struct usb_hcd *hcd, struct usb_device *udev, struct usb_host_endpoint *ep);
 int xhci_drop_endpoint(struct usb_hcd *hcd, struct usb_device *udev, struct usb_host_endpoint *ep);
+void xhci_endpoint_reset(struct usb_hcd *hcd, struct usb_host_endpoint *ep);
 int xhci_check_bandwidth(struct usb_hcd *hcd, struct usb_device *udev);
 void xhci_reset_bandwidth(struct usb_hcd *hcd, struct usb_device *udev);
 
@@ -1148,10 +1207,23 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags, struct urb *urb,
 		int slot_id, unsigned int ep_index);
 int xhci_queue_configure_endpoint(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
 		u32 slot_id);
+int xhci_queue_reset_ep(struct xhci_hcd *xhci, int slot_id,
+		unsigned int ep_index);
+void xhci_find_new_dequeue_state(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		struct xhci_td *cur_td, struct xhci_dequeue_state *state);
+void xhci_queue_new_dequeue_state(struct xhci_hcd *xhci,
+		struct xhci_ring *ep_ring, unsigned int slot_id,
+		unsigned int ep_index, struct xhci_dequeue_state *deq_state);
 
 /* xHCI roothub code */
 int xhci_hub_control(struct usb_hcd *hcd, u16 typeReq, u16 wValue, u16 wIndex,
 		char *buf, u16 wLength);
 int xhci_hub_status_data(struct usb_hcd *hcd, char *buf);
 
+/* xHCI contexts */
+struct xhci_input_control_ctx *xhci_get_input_control_ctx(struct xhci_hcd *xhci, struct xhci_container_ctx *ctx);
+struct xhci_slot_ctx *xhci_get_slot_ctx(struct xhci_hcd *xhci, struct xhci_container_ctx *ctx);
+struct xhci_ep_ctx *xhci_get_ep_ctx(struct xhci_hcd *xhci, struct xhci_container_ctx *ctx, unsigned int ep_index);
+
 #endif /* __LINUX_XHCI_HCD_H */