[U-Boot] [RFC] [UBOOT] [PATCH v3 6/7] USB: Add xhci linux kernel host driver

Dan Murphy dmurphy at ti.com
Tue Jul 2 17:15:12 CEST 2013


Add xhci linux kernel host driver

Kernel base commit ID:aa4f608478acb7ed69dfcff4f3c404100b78ac49

Signed-off-by: Dan Murphy <dmurphy at ti.com>
---
 drivers/usb/host/xhci-ext-caps.h |  155 ++
 drivers/usb/host/xhci-hub.c      | 1216 ++++++++++
 drivers/usb/host/xhci-mem.c      | 2467 ++++++++++++++++++++
 drivers/usb/host/xhci-pci.c      |  356 +++
 drivers/usb/host/xhci-plat.c     |  205 ++
 drivers/usb/host/xhci-ring.c     | 4011 ++++++++++++++++++++++++++++++++
 drivers/usb/host/xhci.c          | 4769 ++++++++++++++++++++++++++++++++++++++
 drivers/usb/host/xhci.h          | 1856 +++++++++++++++
 include/linux/usb/ch11.h         |  266 +++
 include/linux/usb/hcd.h          |  672 ++++++
 10 files changed, 15973 insertions(+)
 create mode 100644 drivers/usb/host/xhci-ext-caps.h
 create mode 100644 drivers/usb/host/xhci-hub.c
 create mode 100644 drivers/usb/host/xhci-mem.c
 create mode 100644 drivers/usb/host/xhci-pci.c
 create mode 100644 drivers/usb/host/xhci-plat.c
 create mode 100644 drivers/usb/host/xhci-ring.c
 create mode 100644 drivers/usb/host/xhci.c
 create mode 100644 drivers/usb/host/xhci.h
 create mode 100644 include/linux/usb/ch11.h
 create mode 100644 include/linux/usb/hcd.h

diff --git a/drivers/usb/host/xhci-ext-caps.h b/drivers/usb/host/xhci-ext-caps.h
new file mode 100644
index 0000000..377f424
--- /dev/null
+++ b/drivers/usb/host/xhci-ext-caps.h
@@ -0,0 +1,155 @@
+/*
+ * xHCI host controller driver
+ *
+ * Copyright (C) 2008 Intel Corp.
+ *
+ * Author: Sarah Sharp
+ * Some code borrowed from the Linux EHCI driver.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+/* Up to 16 ms to halt an HC */
+#define XHCI_MAX_HALT_USEC	(16*1000)
+/* HC not running - set to 1 when run/stop bit is cleared. */
+#define XHCI_STS_HALT		(1<<0)
+
+/* HCCPARAMS offset from PCI base address */
+#define XHCI_HCC_PARAMS_OFFSET	0x10
+/* HCCPARAMS contains the first extended capability pointer */
+#define XHCI_HCC_EXT_CAPS(p)	(((p)>>16)&0xffff)
+
+/* Command and Status registers offset from the Operational Registers address */
+#define XHCI_CMD_OFFSET		0x00
+#define XHCI_STS_OFFSET		0x04
+
+#define XHCI_MAX_EXT_CAPS		50
+
+/* Capability Register */
+/* bits 7:0 - how long is the Capabilities register */
+#define XHCI_HC_LENGTH(p)	(((p)>>00)&0x00ff)
+
+/* Extended capability register fields */
+#define XHCI_EXT_CAPS_ID(p)	(((p)>>0)&0xff)
+#define XHCI_EXT_CAPS_NEXT(p)	(((p)>>8)&0xff)
+#define	XHCI_EXT_CAPS_VAL(p)	((p)>>16)
+/* Extended capability IDs - ID 0 reserved */
+#define XHCI_EXT_CAPS_LEGACY	1
+#define XHCI_EXT_CAPS_PROTOCOL	2
+#define XHCI_EXT_CAPS_PM	3
+#define XHCI_EXT_CAPS_VIRT	4
+#define XHCI_EXT_CAPS_ROUTE	5
+/* IDs 6-9 reserved */
+#define XHCI_EXT_CAPS_DEBUG	10
+/* USB Legacy Support Capability - section 7.1.1 */
+#define XHCI_HC_BIOS_OWNED	(1 << 16)
+#define XHCI_HC_OS_OWNED	(1 << 24)
+
+/* USB Legacy Support Capability - section 7.1.1 */
+/* Add this offset, plus the value of xECP in HCCPARAMS to the base address */
+#define XHCI_LEGACY_SUPPORT_OFFSET	(0x00)
+
+/* USB Legacy Support Control and Status Register  - section 7.1.2 */
+/* Add this offset, plus the value of xECP in HCCPARAMS to the base address */
+#define XHCI_LEGACY_CONTROL_OFFSET	(0x04)
+/* bits 1:3, 5:12, and 17:19 need to be preserved; bits 21:28 should be zero */
+#define	XHCI_LEGACY_DISABLE_SMI		((0x7 << 1) + (0xff << 5) + (0x7 << 17))
+#define XHCI_LEGACY_SMI_EVENTS		(0x7 << 29)
+
+/* USB 2.0 xHCI 0.96 L1C capability - section 7.2.2.1.3.2 */
+#define XHCI_L1C               (1 << 16)
+
+/* USB 2.0 xHCI 1.0 hardware LMP capability - section 7.2.2.1.3.2 */
+#define XHCI_HLC               (1 << 19)
+
+/* command register values to disable interrupts and halt the HC */
+/* start/stop HC execution - do not write unless HC is halted*/
+#define XHCI_CMD_RUN		(1 << 0)
+/* Event Interrupt Enable - get irq when EINT bit is set in USBSTS register */
+#define XHCI_CMD_EIE		(1 << 2)
+/* Host System Error Interrupt Enable - get irq when HSEIE bit set in USBSTS */
+#define XHCI_CMD_HSEIE		(1 << 3)
+/* Enable Wrap Event - '1' means xHC generates an event when MFINDEX wraps. */
+#define XHCI_CMD_EWE		(1 << 10)
+
+#define XHCI_IRQS		(XHCI_CMD_EIE | XHCI_CMD_HSEIE | XHCI_CMD_EWE)
+
+/* true: Controller Not Ready to accept doorbell or op reg writes after reset */
+#define XHCI_STS_CNR		(1 << 11)
+
+#include <linux/io.h>
+
+/**
+ * Return the next extended capability pointer register.
+ *
+ * @base	PCI register base address.
+ *
+ * @ext_offset	Offset of the 32-bit register that contains the extended
+ * capabilites pointer.  If searching for the first extended capability, pass
+ * in XHCI_HCC_PARAMS_OFFSET.  If searching for the next extended capability,
+ * pass in the offset of the current extended capability register.
+ *
+ * Returns 0 if there is no next extended capability register or returns the register offset
+ * from the PCI registers base address.
+ */
+static inline int xhci_find_next_cap_offset(void __iomem *base, int ext_offset)
+{
+	u32 next;
+
+	next = readl(base + ext_offset);
+
+	if (ext_offset == XHCI_HCC_PARAMS_OFFSET) {
+		/* Find the first extended capability */
+		next = XHCI_HCC_EXT_CAPS(next);
+		ext_offset = 0;
+	} else {
+		/* Find the next extended capability */
+		next = XHCI_EXT_CAPS_NEXT(next);
+	}
+
+	if (!next)
+		return 0;
+	/*
+	 * Address calculation from offset of extended capabilities
+	 * (or HCCPARAMS) register - see section 5.3.6 and section 7.
+	 */
+	return ext_offset + (next << 2);
+}
+
+/**
+ * Find the offset of the extended capabilities with capability ID id.
+ *
+ * @base PCI MMIO registers base address.
+ * @ext_offset Offset from base of the first extended capability to look at,
+ * 		or the address of HCCPARAMS.
+ * @id Extended capability ID to search for.
+ *
+ * This uses an arbitrary limit of XHCI_MAX_EXT_CAPS extended capabilities
+ * to make sure that the list doesn't contain a loop.
+ */
+static inline int xhci_find_ext_cap_by_id(void __iomem *base, int ext_offset, int id)
+{
+	u32 val;
+	int limit = XHCI_MAX_EXT_CAPS;
+
+	while (ext_offset && limit > 0) {
+		val = readl(base + ext_offset);
+		if (XHCI_EXT_CAPS_ID(val) == id)
+			break;
+		ext_offset = xhci_find_next_cap_offset(base, ext_offset);
+		limit--;
+	}
+	if (limit > 0)
+		return ext_offset;
+	return 0;
+}
diff --git a/drivers/usb/host/xhci-hub.c b/drivers/usb/host/xhci-hub.c
new file mode 100644
index 0000000..187a3ec
--- /dev/null
+++ b/drivers/usb/host/xhci-hub.c
@@ -0,0 +1,1216 @@
+/*
+ * xHCI host controller driver
+ *
+ * Copyright (C) 2008 Intel Corp.
+ *
+ * Author: Sarah Sharp
+ * Some code borrowed from the Linux EHCI driver.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#include <linux/gfp.h>
+#include <asm/unaligned.h>
+
+#include "xhci.h"
+
+#define	PORT_WAKE_BITS	(PORT_WKOC_E | PORT_WKDISC_E | PORT_WKCONN_E)
+#define	PORT_RWC_BITS	(PORT_CSC | PORT_PEC | PORT_WRC | PORT_OCC | \
+			 PORT_RC | PORT_PLC | PORT_PE)
+
+/* USB 3.0 BOS descriptor and a capability descriptor, combined */
+static u8 usb_bos_descriptor [] = {
+	USB_DT_BOS_SIZE,		/*  __u8 bLength, 5 bytes */
+	USB_DT_BOS,			/*  __u8 bDescriptorType */
+	0x0F, 0x00,			/*  __le16 wTotalLength, 15 bytes */
+	0x1,				/*  __u8 bNumDeviceCaps */
+	/* First device capability */
+	USB_DT_USB_SS_CAP_SIZE,		/*  __u8 bLength, 10 bytes */
+	USB_DT_DEVICE_CAPABILITY,	/* Device Capability */
+	USB_SS_CAP_TYPE,		/* bDevCapabilityType, SUPERSPEED_USB */
+	0x00,				/* bmAttributes, LTM off by default */
+	USB_5GBPS_OPERATION, 0x00,	/* wSpeedsSupported, 5Gbps only */
+	0x03,				/* bFunctionalitySupport,
+					   USB 3.0 speed only */
+	0x00,				/* bU1DevExitLat, set later. */
+	0x00, 0x00			/* __le16 bU2DevExitLat, set later. */
+};
+
+
+static void xhci_common_hub_descriptor(struct xhci_hcd *xhci,
+		struct usb_hub_descriptor *desc, int ports)
+{
+	u16 temp;
+
+	desc->bPwrOn2PwrGood = 10;	/* xhci section 5.4.9 says 20ms max */
+	desc->bHubContrCurrent = 0;
+
+	desc->bNbrPorts = ports;
+	temp = 0;
+	/* Bits 1:0 - support per-port power switching, or power always on */
+	if (HCC_PPC(xhci->hcc_params))
+		temp |= HUB_CHAR_INDV_PORT_LPSM;
+	else
+		temp |= HUB_CHAR_NO_LPSM;
+	/* Bit  2 - root hubs are not part of a compound device */
+	/* Bits 4:3 - individual port over current protection */
+	temp |= HUB_CHAR_INDV_PORT_OCPM;
+	/* Bits 6:5 - no TTs in root ports */
+	/* Bit  7 - no port indicators */
+	desc->wHubCharacteristics = cpu_to_le16(temp);
+}
+
+/* Fill in the USB 2.0 roothub descriptor */
+static void xhci_usb2_hub_descriptor(struct usb_hcd *hcd, struct xhci_hcd *xhci,
+		struct usb_hub_descriptor *desc)
+{
+	int ports;
+	u16 temp;
+	__u8 port_removable[(USB_MAXCHILDREN + 1 + 7) / 8];
+	u32 portsc;
+	unsigned int i;
+
+	ports = xhci->num_usb2_ports;
+
+	xhci_common_hub_descriptor(xhci, desc, ports);
+	desc->bDescriptorType = USB_DT_HUB;
+	temp = 1 + (ports / 8);
+	desc->bDescLength = USB_DT_HUB_NONVAR_SIZE + 2 * temp;
+
+	/* The Device Removable bits are reported on a byte granularity.
+	 * If the port doesn't exist within that byte, the bit is set to 0.
+	 */
+	memset(port_removable, 0, sizeof(port_removable));
+	for (i = 0; i < ports; i++) {
+		portsc = xhci_readl(xhci, xhci->usb2_ports[i]);
+		/* If a device is removable, PORTSC reports a 0, same as in the
+		 * hub descriptor DeviceRemovable bits.
+		 */
+		if (portsc & PORT_DEV_REMOVE)
+			/* This math is hairy because bit 0 of DeviceRemovable
+			 * is reserved, and bit 1 is for port 1, etc.
+			 */
+			port_removable[(i + 1) / 8] |= 1 << ((i + 1) % 8);
+	}
+
+	/* ch11.h defines a hub descriptor that has room for USB_MAXCHILDREN
+	 * ports on it.  The USB 2.0 specification says that there are two
+	 * variable length fields at the end of the hub descriptor:
+	 * DeviceRemovable and PortPwrCtrlMask.  But since we can have less than
+	 * USB_MAXCHILDREN ports, we may need to use the DeviceRemovable array
+	 * to set PortPwrCtrlMask bits.  PortPwrCtrlMask must always be set to
+	 * 0xFF, so we initialize the both arrays (DeviceRemovable and
+	 * PortPwrCtrlMask) to 0xFF.  Then we set the DeviceRemovable for each
+	 * set of ports that actually exist.
+	 */
+	memset(desc->u.hs.DeviceRemovable, 0xff,
+			sizeof(desc->u.hs.DeviceRemovable));
+	memset(desc->u.hs.PortPwrCtrlMask, 0xff,
+			sizeof(desc->u.hs.PortPwrCtrlMask));
+
+	for (i = 0; i < (ports + 1 + 7) / 8; i++)
+		memset(&desc->u.hs.DeviceRemovable[i], port_removable[i],
+				sizeof(__u8));
+}
+
+/* Fill in the USB 3.0 roothub descriptor */
+static void xhci_usb3_hub_descriptor(struct usb_hcd *hcd, struct xhci_hcd *xhci,
+		struct usb_hub_descriptor *desc)
+{
+	int ports;
+	u16 port_removable;
+	u32 portsc;
+	unsigned int i;
+
+	ports = xhci->num_usb3_ports;
+	xhci_common_hub_descriptor(xhci, desc, ports);
+	desc->bDescriptorType = USB_DT_SS_HUB;
+	desc->bDescLength = USB_DT_SS_HUB_SIZE;
+
+	/* header decode latency should be zero for roothubs,
+	 * see section 4.23.5.2.
+	 */
+	desc->u.ss.bHubHdrDecLat = 0;
+	desc->u.ss.wHubDelay = 0;
+
+	port_removable = 0;
+	/* bit 0 is reserved, bit 1 is for port 1, etc. */
+	for (i = 0; i < ports; i++) {
+		portsc = xhci_readl(xhci, xhci->usb3_ports[i]);
+		if (portsc & PORT_DEV_REMOVE)
+			port_removable |= 1 << (i + 1);
+	}
+
+	desc->u.ss.DeviceRemovable = cpu_to_le16(port_removable);
+}
+
+static void xhci_hub_descriptor(struct usb_hcd *hcd, struct xhci_hcd *xhci,
+		struct usb_hub_descriptor *desc)
+{
+
+	if (hcd->speed == HCD_USB3)
+		xhci_usb3_hub_descriptor(hcd, xhci, desc);
+	else
+		xhci_usb2_hub_descriptor(hcd, xhci, desc);
+
+}
+
+static unsigned int xhci_port_speed(unsigned int port_status)
+{
+	if (DEV_LOWSPEED(port_status))
+		return USB_PORT_STAT_LOW_SPEED;
+	if (DEV_HIGHSPEED(port_status))
+		return USB_PORT_STAT_HIGH_SPEED;
+	/*
+	 * FIXME: Yes, we should check for full speed, but the core uses that as
+	 * a default in portspeed() in usb/core/hub.c (which is the only place
+	 * USB_PORT_STAT_*_SPEED is used).
+	 */
+	return 0;
+}
+
+/*
+ * These bits are Read Only (RO) and should be saved and written to the
+ * registers: 0, 3, 10:13, 30
+ * connect status, over-current status, port speed, and device removable.
+ * connect status and port speed are also sticky - meaning they're in
+ * the AUX well and they aren't changed by a hot, warm, or cold reset.
+ */
+#define	XHCI_PORT_RO	((1<<0) | (1<<3) | (0xf<<10) | (1<<30))
+/*
+ * These bits are RW; writing a 0 clears the bit, writing a 1 sets the bit:
+ * bits 5:8, 9, 14:15, 25:27
+ * link state, port power, port indicator state, "wake on" enable state
+ */
+#define XHCI_PORT_RWS	((0xf<<5) | (1<<9) | (0x3<<14) | (0x7<<25))
+/*
+ * These bits are RW; writing a 1 sets the bit, writing a 0 has no effect:
+ * bit 4 (port reset)
+ */
+#define	XHCI_PORT_RW1S	((1<<4))
+/*
+ * These bits are RW; writing a 1 clears the bit, writing a 0 has no effect:
+ * bits 1, 17, 18, 19, 20, 21, 22, 23
+ * port enable/disable, and
+ * change bits: connect, PED, warm port reset changed (reserved zero for USB 2.0 ports),
+ * over-current, reset, link state, and L1 change
+ */
+#define XHCI_PORT_RW1CS	((1<<1) | (0x7f<<17))
+/*
+ * Bit 16 is RW, and writing a '1' to it causes the link state control to be
+ * latched in
+ */
+#define	XHCI_PORT_RW	((1<<16))
+/*
+ * These bits are Reserved Zero (RsvdZ) and zero should be written to them:
+ * bits 2, 24, 28:31
+ */
+#define	XHCI_PORT_RZ	((1<<2) | (1<<24) | (0xf<<28))
+
+/*
+ * Given a port state, this function returns a value that would result in the
+ * port being in the same state, if the value was written to the port status
+ * control register.
+ * Save Read Only (RO) bits and save read/write bits where
+ * writing a 0 clears the bit and writing a 1 sets the bit (RWS).
+ * For all other types (RW1S, RW1CS, RW, and RZ), writing a '0' has no effect.
+ */
+u32 xhci_port_state_to_neutral(u32 state)
+{
+	/* Save read-only status and port state */
+	return (state & XHCI_PORT_RO) | (state & XHCI_PORT_RWS);
+}
+
+/*
+ * find slot id based on port number.
+ * @port: The one-based port number from one of the two split roothubs.
+ */
+int xhci_find_slot_id_by_port(struct usb_hcd *hcd, struct xhci_hcd *xhci,
+		u16 port)
+{
+	int slot_id;
+	int i;
+	enum usb_device_speed speed;
+
+	slot_id = 0;
+	for (i = 0; i < MAX_HC_SLOTS; i++) {
+		if (!xhci->devs[i])
+			continue;
+		speed = xhci->devs[i]->udev->speed;
+		if (((speed == USB_SPEED_SUPER) == (hcd->speed == HCD_USB3))
+				&& xhci->devs[i]->fake_port == port) {
+			slot_id = i;
+			break;
+		}
+	}
+
+	return slot_id;
+}
+
+/*
+ * Stop device
+ * It issues stop endpoint command for EP 0 to 30. And wait the last command
+ * to complete.
+ * suspend will set to 1, if suspend bit need to set in command.
+ */
+static int xhci_stop_device(struct xhci_hcd *xhci, int slot_id, int suspend)
+{
+	struct xhci_virt_device *virt_dev;
+	struct xhci_command *cmd;
+	unsigned long flags;
+	int timeleft;
+	int ret;
+	int i;
+
+	ret = 0;
+	virt_dev = xhci->devs[slot_id];
+	cmd = xhci_alloc_command(xhci, false, true, GFP_NOIO);
+	if (!cmd) {
+		xhci_dbg(xhci, "Couldn't allocate command structure.\n");
+		return -ENOMEM;
+	}
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	for (i = LAST_EP_INDEX; i > 0; i--) {
+		if (virt_dev->eps[i].ring && virt_dev->eps[i].ring->dequeue)
+			xhci_queue_stop_endpoint(xhci, slot_id, i, suspend);
+	}
+	cmd->command_trb = xhci->cmd_ring->enqueue;
+	list_add_tail(&cmd->cmd_list, &virt_dev->cmd_list);
+	xhci_queue_stop_endpoint(xhci, slot_id, 0, suspend);
+	xhci_ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* Wait for last stop endpoint command to finish */
+	timeleft = wait_for_completion_interruptible_timeout(
+			cmd->completion,
+			USB_CTRL_SET_TIMEOUT);
+	if (timeleft <= 0) {
+		xhci_warn(xhci, "%s while waiting for stop endpoint command\n",
+				timeleft == 0 ? "Timeout" : "Signal");
+		spin_lock_irqsave(&xhci->lock, flags);
+		/* The timeout might have raced with the event ring handler, so
+		 * only delete from the list if the item isn't poisoned.
+		 */
+		if (cmd->cmd_list.next != LIST_POISON1)
+			list_del(&cmd->cmd_list);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		ret = -ETIME;
+		goto command_cleanup;
+	}
+
+command_cleanup:
+	xhci_free_command(xhci, cmd);
+	return ret;
+}
+
+/*
+ * Ring device, it rings the all doorbells unconditionally.
+ */
+void xhci_ring_device(struct xhci_hcd *xhci, int slot_id)
+{
+	int i;
+
+	for (i = 0; i < LAST_EP_INDEX + 1; i++)
+		if (xhci->devs[slot_id]->eps[i].ring &&
+		    xhci->devs[slot_id]->eps[i].ring->dequeue)
+			xhci_ring_ep_doorbell(xhci, slot_id, i, 0);
+
+	return;
+}
+
+static void xhci_disable_port(struct usb_hcd *hcd, struct xhci_hcd *xhci,
+		u16 wIndex, __le32 __iomem *addr, u32 port_status)
+{
+	/* Don't allow the USB core to disable SuperSpeed ports. */
+	if (hcd->speed == HCD_USB3) {
+		xhci_dbg(xhci, "Ignoring request to disable "
+				"SuperSpeed port.\n");
+		return;
+	}
+
+	/* Write 1 to disable the port */
+	xhci_writel(xhci, port_status | PORT_PE, addr);
+	port_status = xhci_readl(xhci, addr);
+	xhci_dbg(xhci, "disable port, actual port %d status  = 0x%x\n",
+			wIndex, port_status);
+}
+
+static void xhci_clear_port_change_bit(struct xhci_hcd *xhci, u16 wValue,
+		u16 wIndex, __le32 __iomem *addr, u32 port_status)
+{
+	char *port_change_bit;
+	u32 status;
+
+	switch (wValue) {
+	case USB_PORT_FEAT_C_RESET:
+		status = PORT_RC;
+		port_change_bit = "reset";
+		break;
+	case USB_PORT_FEAT_C_BH_PORT_RESET:
+		status = PORT_WRC;
+		port_change_bit = "warm(BH) reset";
+		break;
+	case USB_PORT_FEAT_C_CONNECTION:
+		status = PORT_CSC;
+		port_change_bit = "connect";
+		break;
+	case USB_PORT_FEAT_C_OVER_CURRENT:
+		status = PORT_OCC;
+		port_change_bit = "over-current";
+		break;
+	case USB_PORT_FEAT_C_ENABLE:
+		status = PORT_PEC;
+		port_change_bit = "enable/disable";
+		break;
+	case USB_PORT_FEAT_C_SUSPEND:
+		status = PORT_PLC;
+		port_change_bit = "suspend/resume";
+		break;
+	case USB_PORT_FEAT_C_PORT_LINK_STATE:
+		status = PORT_PLC;
+		port_change_bit = "link state";
+		break;
+	default:
+		/* Should never happen */
+		return;
+	}
+	/* Change bits are all write 1 to clear */
+	xhci_writel(xhci, port_status | status, addr);
+	port_status = xhci_readl(xhci, addr);
+	xhci_dbg(xhci, "clear port %s change, actual port %d status  = 0x%x\n",
+			port_change_bit, wIndex, port_status);
+}
+
+static int xhci_get_ports(struct usb_hcd *hcd, __le32 __iomem ***port_array)
+{
+	int max_ports;
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+
+	if (hcd->speed == HCD_USB3) {
+		max_ports = xhci->num_usb3_ports;
+		*port_array = xhci->usb3_ports;
+	} else {
+		max_ports = xhci->num_usb2_ports;
+		*port_array = xhci->usb2_ports;
+	}
+
+	return max_ports;
+}
+
+void xhci_set_link_state(struct xhci_hcd *xhci, __le32 __iomem **port_array,
+				int port_id, u32 link_state)
+{
+	u32 temp;
+
+	temp = xhci_readl(xhci, port_array[port_id]);
+	temp = xhci_port_state_to_neutral(temp);
+	temp &= ~PORT_PLS_MASK;
+	temp |= PORT_LINK_STROBE | link_state;
+	xhci_writel(xhci, temp, port_array[port_id]);
+}
+
+static void xhci_set_remote_wake_mask(struct xhci_hcd *xhci,
+		__le32 __iomem **port_array, int port_id, u16 wake_mask)
+{
+	u32 temp;
+
+	temp = xhci_readl(xhci, port_array[port_id]);
+	temp = xhci_port_state_to_neutral(temp);
+
+	if (wake_mask & USB_PORT_FEAT_REMOTE_WAKE_CONNECT)
+		temp |= PORT_WKCONN_E;
+	else
+		temp &= ~PORT_WKCONN_E;
+
+	if (wake_mask & USB_PORT_FEAT_REMOTE_WAKE_DISCONNECT)
+		temp |= PORT_WKDISC_E;
+	else
+		temp &= ~PORT_WKDISC_E;
+
+	if (wake_mask & USB_PORT_FEAT_REMOTE_WAKE_OVER_CURRENT)
+		temp |= PORT_WKOC_E;
+	else
+		temp &= ~PORT_WKOC_E;
+
+	xhci_writel(xhci, temp, port_array[port_id]);
+}
+
+/* Test and clear port RWC bit */
+void xhci_test_and_clear_bit(struct xhci_hcd *xhci, __le32 __iomem **port_array,
+				int port_id, u32 port_bit)
+{
+	u32 temp;
+
+	temp = xhci_readl(xhci, port_array[port_id]);
+	if (temp & port_bit) {
+		temp = xhci_port_state_to_neutral(temp);
+		temp |= port_bit;
+		xhci_writel(xhci, temp, port_array[port_id]);
+	}
+}
+
+/* Updates Link Status for super Speed port */
+static void xhci_hub_report_link_state(u32 *status, u32 status_reg)
+{
+	u32 pls = status_reg & PORT_PLS_MASK;
+
+	/* resume state is a xHCI internal state.
+	 * Do not report it to usb core.
+	 */
+	if (pls == XDEV_RESUME)
+		return;
+
+	/* When the CAS bit is set then warm reset
+	 * should be performed on port
+	 */
+	if (status_reg & PORT_CAS) {
+		/* The CAS bit can be set while the port is
+		 * in any link state.
+		 * Only roothubs have CAS bit, so we
+		 * pretend to be in compliance mode
+		 * unless we're already in compliance
+		 * or the inactive state.
+		 */
+		if (pls != USB_SS_PORT_LS_COMP_MOD &&
+		    pls != USB_SS_PORT_LS_SS_INACTIVE) {
+			pls = USB_SS_PORT_LS_COMP_MOD;
+		}
+		/* Return also connection bit -
+		 * hub state machine resets port
+		 * when this bit is set.
+		 */
+		pls |= USB_PORT_STAT_CONNECTION;
+	} else {
+		/*
+		 * If CAS bit isn't set but the Port is already at
+		 * Compliance Mode, fake a connection so the USB core
+		 * notices the Compliance state and resets the port.
+		 * This resolves an issue generated by the SN65LVPE502CP
+		 * in which sometimes the port enters compliance mode
+		 * caused by a delay on the host-device negotiation.
+		 */
+		if (pls == USB_SS_PORT_LS_COMP_MOD)
+			pls |= USB_PORT_STAT_CONNECTION;
+	}
+
+	/* update status field */
+	*status |= pls;
+}
+
+/*
+ * Function for Compliance Mode Quirk.
+ *
+ * This Function verifies if all xhc USB3 ports have entered U0, if so,
+ * the compliance mode timer is deleted. A port won't enter
+ * compliance mode if it has previously entered U0.
+ */
+void xhci_del_comp_mod_timer(struct xhci_hcd *xhci, u32 status, u16 wIndex)
+{
+	u32 all_ports_seen_u0 = ((1 << xhci->num_usb3_ports)-1);
+	bool port_in_u0 = ((status & PORT_PLS_MASK) == XDEV_U0);
+
+	if (!(xhci->quirks & XHCI_COMP_MODE_QUIRK))
+		return;
+
+	if ((xhci->port_status_u0 != all_ports_seen_u0) && port_in_u0) {
+		xhci->port_status_u0 |= 1 << wIndex;
+		if (xhci->port_status_u0 == all_ports_seen_u0) {
+			del_timer_sync(&xhci->comp_mode_recovery_timer);
+			xhci_dbg(xhci, "All USB3 ports have entered U0 already!\n");
+			xhci_dbg(xhci, "Compliance Mode Recovery Timer Deleted.\n");
+		}
+	}
+}
+
+int xhci_hub_control(struct usb_hcd *hcd, u16 typeReq, u16 wValue,
+		u16 wIndex, char *buf, u16 wLength)
+{
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+	int max_ports;
+	unsigned long flags;
+	u32 temp, status;
+	int retval = 0;
+	__le32 __iomem **port_array;
+	int slot_id;
+	struct xhci_bus_state *bus_state;
+	u16 link_state = 0;
+	u16 wake_mask = 0;
+	u16 timeout = 0;
+
+	max_ports = xhci_get_ports(hcd, &port_array);
+	bus_state = &xhci->bus_state[hcd_index(hcd)];
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	switch (typeReq) {
+	case GetHubStatus:
+		/* No power source, over-current reported per port */
+		memset(buf, 0, 4);
+		break;
+	case GetHubDescriptor:
+		/* Check to make sure userspace is asking for the USB 3.0 hub
+		 * descriptor for the USB 3.0 roothub.  If not, we stall the
+		 * endpoint, like external hubs do.
+		 */
+		if (hcd->speed == HCD_USB3 &&
+				(wLength < USB_DT_SS_HUB_SIZE ||
+				 wValue != (USB_DT_SS_HUB << 8))) {
+			xhci_dbg(xhci, "Wrong hub descriptor type for "
+					"USB 3.0 roothub.\n");
+			goto error;
+		}
+		xhci_hub_descriptor(hcd, xhci,
+				(struct usb_hub_descriptor *) buf);
+		break;
+	case DeviceRequest | USB_REQ_GET_DESCRIPTOR:
+		if ((wValue & 0xff00) != (USB_DT_BOS << 8))
+			goto error;
+
+		if (hcd->speed != HCD_USB3)
+			goto error;
+
+		/* Set the U1 and U2 exit latencies. */
+		memcpy(buf, &usb_bos_descriptor,
+				USB_DT_BOS_SIZE + USB_DT_USB_SS_CAP_SIZE);
+		temp = xhci_readl(xhci, &xhci->cap_regs->hcs_params3);
+		buf[12] = HCS_U1_LATENCY(temp);
+		put_unaligned_le16(HCS_U2_LATENCY(temp), &buf[13]);
+
+		/* Indicate whether the host has LTM support. */
+		temp = xhci_readl(xhci, &xhci->cap_regs->hcc_params);
+		if (HCC_LTC(temp))
+			buf[8] |= USB_LTM_SUPPORT;
+
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return USB_DT_BOS_SIZE + USB_DT_USB_SS_CAP_SIZE;
+	case GetPortStatus:
+		if (!wIndex || wIndex > max_ports)
+			goto error;
+		wIndex--;
+		status = 0;
+		temp = xhci_readl(xhci, port_array[wIndex]);
+		if (temp == 0xffffffff) {
+			retval = -ENODEV;
+			break;
+		}
+		xhci_dbg(xhci, "get port status, actual port %d status  = 0x%x\n", wIndex, temp);
+
+		/* wPortChange bits */
+		if (temp & PORT_CSC)
+			status |= USB_PORT_STAT_C_CONNECTION << 16;
+		if (temp & PORT_PEC)
+			status |= USB_PORT_STAT_C_ENABLE << 16;
+		if ((temp & PORT_OCC))
+			status |= USB_PORT_STAT_C_OVERCURRENT << 16;
+		if ((temp & PORT_RC))
+			status |= USB_PORT_STAT_C_RESET << 16;
+		/* USB3.0 only */
+		if (hcd->speed == HCD_USB3) {
+			if ((temp & PORT_PLC))
+				status |= USB_PORT_STAT_C_LINK_STATE << 16;
+			if ((temp & PORT_WRC))
+				status |= USB_PORT_STAT_C_BH_RESET << 16;
+		}
+
+		if (hcd->speed != HCD_USB3) {
+			if ((temp & PORT_PLS_MASK) == XDEV_U3
+					&& (temp & PORT_POWER))
+				status |= USB_PORT_STAT_SUSPEND;
+		}
+		if ((temp & PORT_PLS_MASK) == XDEV_RESUME &&
+				!DEV_SUPERSPEED(temp)) {
+			if ((temp & PORT_RESET) || !(temp & PORT_PE))
+				goto error;
+			if (time_after_eq(jiffies,
+					bus_state->resume_done[wIndex])) {
+				xhci_dbg(xhci, "Resume USB2 port %d\n",
+					wIndex + 1);
+				bus_state->resume_done[wIndex] = 0;
+				clear_bit(wIndex, &bus_state->resuming_ports);
+				xhci_set_link_state(xhci, port_array, wIndex,
+							XDEV_U0);
+				xhci_dbg(xhci, "set port %d resume\n",
+					wIndex + 1);
+				slot_id = xhci_find_slot_id_by_port(hcd, xhci,
+								 wIndex + 1);
+				if (!slot_id) {
+					xhci_dbg(xhci, "slot_id is zero\n");
+					goto error;
+				}
+				xhci_ring_device(xhci, slot_id);
+				bus_state->port_c_suspend |= 1 << wIndex;
+				bus_state->suspended_ports &= ~(1 << wIndex);
+			} else {
+				/*
+				 * The resume has been signaling for less than
+				 * 20ms. Report the port status as SUSPEND,
+				 * let the usbcore check port status again
+				 * and clear resume signaling later.
+				 */
+				status |= USB_PORT_STAT_SUSPEND;
+			}
+		}
+		if ((temp & PORT_PLS_MASK) == XDEV_U0
+			&& (temp & PORT_POWER)
+			&& (bus_state->suspended_ports & (1 << wIndex))) {
+			bus_state->suspended_ports &= ~(1 << wIndex);
+			if (hcd->speed != HCD_USB3)
+				bus_state->port_c_suspend |= 1 << wIndex;
+		}
+		if (temp & PORT_CONNECT) {
+			status |= USB_PORT_STAT_CONNECTION;
+			status |= xhci_port_speed(temp);
+		}
+		if (temp & PORT_PE)
+			status |= USB_PORT_STAT_ENABLE;
+		if (temp & PORT_OC)
+			status |= USB_PORT_STAT_OVERCURRENT;
+		if (temp & PORT_RESET)
+			status |= USB_PORT_STAT_RESET;
+		if (temp & PORT_POWER) {
+			if (hcd->speed == HCD_USB3)
+				status |= USB_SS_PORT_STAT_POWER;
+			else
+				status |= USB_PORT_STAT_POWER;
+		}
+		/* Update Port Link State for super speed ports*/
+		if (hcd->speed == HCD_USB3) {
+			xhci_hub_report_link_state(&status, temp);
+			/*
+			 * Verify if all USB3 Ports Have entered U0 already.
+			 * Delete Compliance Mode Timer if so.
+			 */
+			xhci_del_comp_mod_timer(xhci, temp, wIndex);
+		}
+		if (bus_state->port_c_suspend & (1 << wIndex))
+			status |= 1 << USB_PORT_FEAT_C_SUSPEND;
+		xhci_dbg(xhci, "Get port status returned 0x%x\n", status);
+		put_unaligned(cpu_to_le32(status), (__le32 *) buf);
+		break;
+	case SetPortFeature:
+		if (wValue == USB_PORT_FEAT_LINK_STATE)
+			link_state = (wIndex & 0xff00) >> 3;
+		if (wValue == USB_PORT_FEAT_REMOTE_WAKE_MASK)
+			wake_mask = wIndex & 0xff00;
+		/* The MSB of wIndex is the U1/U2 timeout */
+		timeout = (wIndex & 0xff00) >> 8;
+		wIndex &= 0xff;
+		if (!wIndex || wIndex > max_ports)
+			goto error;
+		wIndex--;
+		temp = xhci_readl(xhci, port_array[wIndex]);
+		if (temp == 0xffffffff) {
+			retval = -ENODEV;
+			break;
+		}
+		temp = xhci_port_state_to_neutral(temp);
+		/* FIXME: What new port features do we need to support? */
+		switch (wValue) {
+		case USB_PORT_FEAT_SUSPEND:
+			temp = xhci_readl(xhci, port_array[wIndex]);
+			if ((temp & PORT_PLS_MASK) != XDEV_U0) {
+				/* Resume the port to U0 first */
+				xhci_set_link_state(xhci, port_array, wIndex,
+							XDEV_U0);
+				spin_unlock_irqrestore(&xhci->lock, flags);
+				msleep(10);
+				spin_lock_irqsave(&xhci->lock, flags);
+			}
+			/* In spec software should not attempt to suspend
+			 * a port unless the port reports that it is in the
+			 * enabled (PED = ‘1’,PLS < ‘3’) state.
+			 */
+			temp = xhci_readl(xhci, port_array[wIndex]);
+			if ((temp & PORT_PE) == 0 || (temp & PORT_RESET)
+				|| (temp & PORT_PLS_MASK) >= XDEV_U3) {
+				xhci_warn(xhci, "USB core suspending device "
+					  "not in U0/U1/U2.\n");
+				goto error;
+			}
+
+			slot_id = xhci_find_slot_id_by_port(hcd, xhci,
+					wIndex + 1);
+			if (!slot_id) {
+				xhci_warn(xhci, "slot_id is zero\n");
+				goto error;
+			}
+			/* unlock to execute stop endpoint commands */
+			spin_unlock_irqrestore(&xhci->lock, flags);
+			xhci_stop_device(xhci, slot_id, 1);
+			spin_lock_irqsave(&xhci->lock, flags);
+
+			xhci_set_link_state(xhci, port_array, wIndex, XDEV_U3);
+
+			spin_unlock_irqrestore(&xhci->lock, flags);
+			msleep(10); /* wait device to enter */
+			spin_lock_irqsave(&xhci->lock, flags);
+
+			temp = xhci_readl(xhci, port_array[wIndex]);
+			bus_state->suspended_ports |= 1 << wIndex;
+			break;
+		case USB_PORT_FEAT_LINK_STATE:
+			temp = xhci_readl(xhci, port_array[wIndex]);
+
+			/* Disable port */
+			if (link_state == USB_SS_PORT_LS_SS_DISABLED) {
+				xhci_dbg(xhci, "Disable port %d\n", wIndex);
+				temp = xhci_port_state_to_neutral(temp);
+				/*
+				 * Clear all change bits, so that we get a new
+				 * connection event.
+				 */
+				temp |= PORT_CSC | PORT_PEC | PORT_WRC |
+					PORT_OCC | PORT_RC | PORT_PLC |
+					PORT_CEC;
+				xhci_writel(xhci, temp | PORT_PE,
+					port_array[wIndex]);
+				temp = xhci_readl(xhci, port_array[wIndex]);
+				break;
+			}
+
+			/* Put link in RxDetect (enable port) */
+			if (link_state == USB_SS_PORT_LS_RX_DETECT) {
+				xhci_dbg(xhci, "Enable port %d\n", wIndex);
+				xhci_set_link_state(xhci, port_array, wIndex,
+						link_state);
+				temp = xhci_readl(xhci, port_array[wIndex]);
+				break;
+			}
+
+			/* Software should not attempt to set
+			 * port link state above '3' (U3) and the port
+			 * must be enabled.
+			 */
+			if ((temp & PORT_PE) == 0 ||
+				(link_state > USB_SS_PORT_LS_U3)) {
+				xhci_warn(xhci, "Cannot set link state.\n");
+				goto error;
+			}
+
+			if (link_state == USB_SS_PORT_LS_U3) {
+				slot_id = xhci_find_slot_id_by_port(hcd, xhci,
+						wIndex + 1);
+				if (slot_id) {
+					/* unlock to execute stop endpoint
+					 * commands */
+					spin_unlock_irqrestore(&xhci->lock,
+								flags);
+					xhci_stop_device(xhci, slot_id, 1);
+					spin_lock_irqsave(&xhci->lock, flags);
+				}
+			}
+
+			xhci_set_link_state(xhci, port_array, wIndex,
+						link_state);
+
+			spin_unlock_irqrestore(&xhci->lock, flags);
+			msleep(20); /* wait device to enter */
+			spin_lock_irqsave(&xhci->lock, flags);
+
+			temp = xhci_readl(xhci, port_array[wIndex]);
+			if (link_state == USB_SS_PORT_LS_U3)
+				bus_state->suspended_ports |= 1 << wIndex;
+			break;
+		case USB_PORT_FEAT_POWER:
+			/*
+			 * Turn on ports, even if there isn't per-port switching.
+			 * HC will report connect events even before this is set.
+			 * However, khubd will ignore the roothub events until
+			 * the roothub is registered.
+			 */
+			xhci_writel(xhci, temp | PORT_POWER,
+					port_array[wIndex]);
+
+			temp = xhci_readl(xhci, port_array[wIndex]);
+			xhci_dbg(xhci, "set port power, actual port %d status  = 0x%x\n", wIndex, temp);
+
+			spin_unlock_irqrestore(&xhci->lock, flags);
+			temp = usb_acpi_power_manageable(hcd->self.root_hub,
+					wIndex);
+			if (temp)
+				usb_acpi_set_power_state(hcd->self.root_hub,
+						wIndex, true);
+			spin_lock_irqsave(&xhci->lock, flags);
+			break;
+		case USB_PORT_FEAT_RESET:
+			temp = (temp | PORT_RESET);
+			xhci_writel(xhci, temp, port_array[wIndex]);
+
+			temp = xhci_readl(xhci, port_array[wIndex]);
+			xhci_dbg(xhci, "set port reset, actual port %d status  = 0x%x\n", wIndex, temp);
+			break;
+		case USB_PORT_FEAT_REMOTE_WAKE_MASK:
+			xhci_set_remote_wake_mask(xhci, port_array,
+					wIndex, wake_mask);
+			temp = xhci_readl(xhci, port_array[wIndex]);
+			xhci_dbg(xhci, "set port remote wake mask, "
+					"actual port %d status  = 0x%x\n",
+					wIndex, temp);
+			break;
+		case USB_PORT_FEAT_BH_PORT_RESET:
+			temp |= PORT_WR;
+			xhci_writel(xhci, temp, port_array[wIndex]);
+
+			temp = xhci_readl(xhci, port_array[wIndex]);
+			break;
+		case USB_PORT_FEAT_U1_TIMEOUT:
+			if (hcd->speed != HCD_USB3)
+				goto error;
+			temp = xhci_readl(xhci, port_array[wIndex] + 1);
+			temp &= ~PORT_U1_TIMEOUT_MASK;
+			temp |= PORT_U1_TIMEOUT(timeout);
+			xhci_writel(xhci, temp, port_array[wIndex] + 1);
+			break;
+		case USB_PORT_FEAT_U2_TIMEOUT:
+			if (hcd->speed != HCD_USB3)
+				goto error;
+			temp = xhci_readl(xhci, port_array[wIndex] + 1);
+			temp &= ~PORT_U2_TIMEOUT_MASK;
+			temp |= PORT_U2_TIMEOUT(timeout);
+			xhci_writel(xhci, temp, port_array[wIndex] + 1);
+			break;
+		default:
+			goto error;
+		}
+		/* unblock any posted writes */
+		temp = xhci_readl(xhci, port_array[wIndex]);
+		break;
+	case ClearPortFeature:
+		if (!wIndex || wIndex > max_ports)
+			goto error;
+		wIndex--;
+		temp = xhci_readl(xhci, port_array[wIndex]);
+		if (temp == 0xffffffff) {
+			retval = -ENODEV;
+			break;
+		}
+		/* FIXME: What new port features do we need to support? */
+		temp = xhci_port_state_to_neutral(temp);
+		switch (wValue) {
+		case USB_PORT_FEAT_SUSPEND:
+			temp = xhci_readl(xhci, port_array[wIndex]);
+			xhci_dbg(xhci, "clear USB_PORT_FEAT_SUSPEND\n");
+			xhci_dbg(xhci, "PORTSC %04x\n", temp);
+			if (temp & PORT_RESET)
+				goto error;
+			if ((temp & PORT_PLS_MASK) == XDEV_U3) {
+				if ((temp & PORT_PE) == 0)
+					goto error;
+
+				xhci_set_link_state(xhci, port_array, wIndex,
+							XDEV_RESUME);
+				spin_unlock_irqrestore(&xhci->lock, flags);
+				msleep(20);
+				spin_lock_irqsave(&xhci->lock, flags);
+				xhci_set_link_state(xhci, port_array, wIndex,
+							XDEV_U0);
+			}
+			bus_state->port_c_suspend |= 1 << wIndex;
+
+			slot_id = xhci_find_slot_id_by_port(hcd, xhci,
+					wIndex + 1);
+			if (!slot_id) {
+				xhci_dbg(xhci, "slot_id is zero\n");
+				goto error;
+			}
+			xhci_ring_device(xhci, slot_id);
+			break;
+		case USB_PORT_FEAT_C_SUSPEND:
+			bus_state->port_c_suspend &= ~(1 << wIndex);
+		case USB_PORT_FEAT_C_RESET:
+		case USB_PORT_FEAT_C_BH_PORT_RESET:
+		case USB_PORT_FEAT_C_CONNECTION:
+		case USB_PORT_FEAT_C_OVER_CURRENT:
+		case USB_PORT_FEAT_C_ENABLE:
+		case USB_PORT_FEAT_C_PORT_LINK_STATE:
+			xhci_clear_port_change_bit(xhci, wValue, wIndex,
+					port_array[wIndex], temp);
+			break;
+		case USB_PORT_FEAT_ENABLE:
+			xhci_disable_port(hcd, xhci, wIndex,
+					port_array[wIndex], temp);
+			break;
+		case USB_PORT_FEAT_POWER:
+			xhci_writel(xhci, temp & ~PORT_POWER,
+				port_array[wIndex]);
+
+			spin_unlock_irqrestore(&xhci->lock, flags);
+			temp = usb_acpi_power_manageable(hcd->self.root_hub,
+					wIndex);
+			if (temp)
+				usb_acpi_set_power_state(hcd->self.root_hub,
+						wIndex, false);
+			spin_lock_irqsave(&xhci->lock, flags);
+			break;
+		default:
+			goto error;
+		}
+		break;
+	default:
+error:
+		/* "stall" on error */
+		retval = -EPIPE;
+	}
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return retval;
+}
+
+/*
+ * Returns 0 if the status hasn't changed, or the number of bytes in buf.
+ * Ports are 0-indexed from the HCD point of view,
+ * and 1-indexed from the USB core pointer of view.
+ *
+ * Note that the status change bits will be cleared as soon as a port status
+ * change event is generated, so we use the saved status from that event.
+ */
+int xhci_hub_status_data(struct usb_hcd *hcd, char *buf)
+{
+	unsigned long flags;
+	u32 temp, status;
+	u32 mask;
+	int i, retval;
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+	int max_ports;
+	__le32 __iomem **port_array;
+	struct xhci_bus_state *bus_state;
+	bool reset_change = false;
+
+	max_ports = xhci_get_ports(hcd, &port_array);
+	bus_state = &xhci->bus_state[hcd_index(hcd)];
+
+	/* Initial status is no changes */
+	retval = (max_ports + 8) / 8;
+	memset(buf, 0, retval);
+
+	/*
+	 * Inform the usbcore about resume-in-progress by returning
+	 * a non-zero value even if there are no status changes.
+	 */
+	status = bus_state->resuming_ports;
+
+	mask = PORT_CSC | PORT_PEC | PORT_OCC | PORT_PLC | PORT_WRC;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	/* For each port, did anything change?  If so, set that bit in buf. */
+	for (i = 0; i < max_ports; i++) {
+		temp = xhci_readl(xhci, port_array[i]);
+		if (temp == 0xffffffff) {
+			retval = -ENODEV;
+			break;
+		}
+		if ((temp & mask) != 0 ||
+			(bus_state->port_c_suspend & 1 << i) ||
+			(bus_state->resume_done[i] && time_after_eq(
+			    jiffies, bus_state->resume_done[i]))) {
+			buf[(i + 1) / 8] |= 1 << (i + 1) % 8;
+			status = 1;
+		}
+		if ((temp & PORT_RC))
+			reset_change = true;
+	}
+	if (!status && !reset_change) {
+		xhci_dbg(xhci, "%s: stopping port polling.\n", __func__);
+		clear_bit(HCD_FLAG_POLL_RH, &hcd->flags);
+	}
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return status ? retval : 0;
+}
+
+#ifdef CONFIG_PM
+
+int xhci_bus_suspend(struct usb_hcd *hcd)
+{
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+	int max_ports, port_index;
+	__le32 __iomem **port_array;
+	struct xhci_bus_state *bus_state;
+	unsigned long flags;
+
+	max_ports = xhci_get_ports(hcd, &port_array);
+	bus_state = &xhci->bus_state[hcd_index(hcd)];
+
+	spin_lock_irqsave(&xhci->lock, flags);
+
+	if (hcd->self.root_hub->do_remote_wakeup) {
+		if (bus_state->resuming_ports) {
+			spin_unlock_irqrestore(&xhci->lock, flags);
+			xhci_dbg(xhci, "suspend failed because "
+						"a port is resuming\n");
+			return -EBUSY;
+		}
+	}
+
+	port_index = max_ports;
+	bus_state->bus_suspended = 0;
+	while (port_index--) {
+		/* suspend the port if the port is not suspended */
+		u32 t1, t2;
+		int slot_id;
+
+		t1 = xhci_readl(xhci, port_array[port_index]);
+		t2 = xhci_port_state_to_neutral(t1);
+
+		if ((t1 & PORT_PE) && !(t1 & PORT_PLS_MASK)) {
+			xhci_dbg(xhci, "port %d not suspended\n", port_index);
+			slot_id = xhci_find_slot_id_by_port(hcd, xhci,
+					port_index + 1);
+			if (slot_id) {
+				spin_unlock_irqrestore(&xhci->lock, flags);
+				xhci_stop_device(xhci, slot_id, 1);
+				spin_lock_irqsave(&xhci->lock, flags);
+			}
+			t2 &= ~PORT_PLS_MASK;
+			t2 |= PORT_LINK_STROBE | XDEV_U3;
+			set_bit(port_index, &bus_state->bus_suspended);
+		}
+		/* USB core sets remote wake mask for USB 3.0 hubs,
+		 * including the USB 3.0 roothub, but only if CONFIG_PM_RUNTIME
+		 * is enabled, so also enable remote wake here.
+		 */
+		if (hcd->self.root_hub->do_remote_wakeup) {
+			if (t1 & PORT_CONNECT) {
+				t2 |= PORT_WKOC_E | PORT_WKDISC_E;
+				t2 &= ~PORT_WKCONN_E;
+			} else {
+				t2 |= PORT_WKOC_E | PORT_WKCONN_E;
+				t2 &= ~PORT_WKDISC_E;
+			}
+		} else
+			t2 &= ~PORT_WAKE_BITS;
+
+		t1 = xhci_port_state_to_neutral(t1);
+		if (t1 != t2)
+			xhci_writel(xhci, t2, port_array[port_index]);
+
+		if (hcd->speed != HCD_USB3) {
+			/* enable remote wake up for USB 2.0 */
+			__le32 __iomem *addr;
+			u32 tmp;
+
+			/* Add one to the port status register address to get
+			 * the port power control register address.
+			 */
+			addr = port_array[port_index] + 1;
+			tmp = xhci_readl(xhci, addr);
+			tmp |= PORT_RWE;
+			xhci_writel(xhci, tmp, addr);
+		}
+	}
+	hcd->state = HC_STATE_SUSPENDED;
+	bus_state->next_statechange = jiffies + msecs_to_jiffies(10);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return 0;
+}
+
+int xhci_bus_resume(struct usb_hcd *hcd)
+{
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+	int max_ports, port_index;
+	__le32 __iomem **port_array;
+	struct xhci_bus_state *bus_state;
+	u32 temp;
+	unsigned long flags;
+
+	max_ports = xhci_get_ports(hcd, &port_array);
+	bus_state = &xhci->bus_state[hcd_index(hcd)];
+
+	if (time_before(jiffies, bus_state->next_statechange))
+		msleep(5);
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	if (!HCD_HW_ACCESSIBLE(hcd)) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return -ESHUTDOWN;
+	}
+
+	/* delay the irqs */
+	temp = xhci_readl(xhci, &xhci->op_regs->command);
+	temp &= ~CMD_EIE;
+	xhci_writel(xhci, temp, &xhci->op_regs->command);
+
+	port_index = max_ports;
+	while (port_index--) {
+		/* Check whether need resume ports. If needed
+		   resume port and disable remote wakeup */
+		u32 temp;
+		int slot_id;
+
+		temp = xhci_readl(xhci, port_array[port_index]);
+		if (DEV_SUPERSPEED(temp))
+			temp &= ~(PORT_RWC_BITS | PORT_CEC | PORT_WAKE_BITS);
+		else
+			temp &= ~(PORT_RWC_BITS | PORT_WAKE_BITS);
+		if (test_bit(port_index, &bus_state->bus_suspended) &&
+		    (temp & PORT_PLS_MASK)) {
+			if (DEV_SUPERSPEED(temp)) {
+				xhci_set_link_state(xhci, port_array,
+							port_index, XDEV_U0);
+			} else {
+				xhci_set_link_state(xhci, port_array,
+						port_index, XDEV_RESUME);
+
+				spin_unlock_irqrestore(&xhci->lock, flags);
+				msleep(20);
+				spin_lock_irqsave(&xhci->lock, flags);
+
+				xhci_set_link_state(xhci, port_array,
+							port_index, XDEV_U0);
+			}
+			/* wait for the port to enter U0 and report port link
+			 * state change.
+			 */
+			spin_unlock_irqrestore(&xhci->lock, flags);
+			msleep(20);
+			spin_lock_irqsave(&xhci->lock, flags);
+
+			/* Clear PLC */
+			xhci_test_and_clear_bit(xhci, port_array, port_index,
+						PORT_PLC);
+
+			slot_id = xhci_find_slot_id_by_port(hcd,
+					xhci, port_index + 1);
+			if (slot_id)
+				xhci_ring_device(xhci, slot_id);
+		} else
+			xhci_writel(xhci, temp, port_array[port_index]);
+
+		if (hcd->speed != HCD_USB3) {
+			/* disable remote wake up for USB 2.0 */
+			__le32 __iomem *addr;
+			u32 tmp;
+
+			/* Add one to the port status register address to get
+			 * the port power control register address.
+			 */
+			addr = port_array[port_index] + 1;
+			tmp = xhci_readl(xhci, addr);
+			tmp &= ~PORT_RWE;
+			xhci_writel(xhci, tmp, addr);
+		}
+	}
+
+	(void) xhci_readl(xhci, &xhci->op_regs->command);
+
+	bus_state->next_statechange = jiffies + msecs_to_jiffies(5);
+	/* re-enable irqs */
+	temp = xhci_readl(xhci, &xhci->op_regs->command);
+	temp |= CMD_EIE;
+	xhci_writel(xhci, temp, &xhci->op_regs->command);
+	temp = xhci_readl(xhci, &xhci->op_regs->command);
+
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return 0;
+}
+
+#endif	/* CONFIG_PM */
diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
new file mode 100644
index 0000000..2cfc465
--- /dev/null
+++ b/drivers/usb/host/xhci-mem.c
@@ -0,0 +1,2467 @@
+/*
+ * xHCI host controller driver
+ *
+ * Copyright (C) 2008 Intel Corp.
+ *
+ * Author: Sarah Sharp
+ * Some code borrowed from the Linux EHCI driver.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#include <linux/usb.h>
+#include <linux/pci.h>
+#include <linux/slab.h>
+#include <linux/dmapool.h>
+
+#include "xhci.h"
+
+/*
+ * Allocates a generic ring segment from the ring pool, sets the dma address,
+ * initializes the segment to zero, and sets the private next pointer to NULL.
+ *
+ * Section 4.11.1.1:
+ * "All components of all Command and Transfer TRBs shall be initialized to '0'"
+ */
+static struct xhci_segment *xhci_segment_alloc(struct xhci_hcd *xhci,
+					unsigned int cycle_state, gfp_t flags)
+{
+	struct xhci_segment *seg;
+	dma_addr_t	dma;
+	int		i;
+
+	seg = kzalloc(sizeof *seg, flags);
+	if (!seg)
+		return NULL;
+
+	seg->trbs = dma_pool_alloc(xhci->segment_pool, flags, &dma);
+	if (!seg->trbs) {
+		kfree(seg);
+		return NULL;
+	}
+
+	memset(seg->trbs, 0, TRB_SEGMENT_SIZE);
+	/* If the cycle state is 0, set the cycle bit to 1 for all the TRBs */
+	if (cycle_state == 0) {
+		for (i = 0; i < TRBS_PER_SEGMENT; i++)
+			seg->trbs[i].link.control |= TRB_CYCLE;
+	}
+	seg->dma = dma;
+	seg->next = NULL;
+
+	return seg;
+}
+
+static void xhci_segment_free(struct xhci_hcd *xhci, struct xhci_segment *seg)
+{
+	if (seg->trbs) {
+		dma_pool_free(xhci->segment_pool, seg->trbs, seg->dma);
+		seg->trbs = NULL;
+	}
+	kfree(seg);
+}
+
+static void xhci_free_segments_for_ring(struct xhci_hcd *xhci,
+				struct xhci_segment *first)
+{
+	struct xhci_segment *seg;
+
+	seg = first->next;
+	while (seg != first) {
+		struct xhci_segment *next = seg->next;
+		xhci_segment_free(xhci, seg);
+		seg = next;
+	}
+	xhci_segment_free(xhci, first);
+}
+
+/*
+ * Make the prev segment point to the next segment.
+ *
+ * Change the last TRB in the prev segment to be a Link TRB which points to the
+ * DMA address of the next segment.  The caller needs to set any Link TRB
+ * related flags, such as End TRB, Toggle Cycle, and no snoop.
+ */
+static void xhci_link_segments(struct xhci_hcd *xhci, struct xhci_segment *prev,
+		struct xhci_segment *next, enum xhci_ring_type type)
+{
+	u32 val;
+
+	if (!prev || !next)
+		return;
+	prev->next = next;
+	if (type != TYPE_EVENT) {
+		prev->trbs[TRBS_PER_SEGMENT-1].link.segment_ptr =
+			cpu_to_le64(next->dma);
+
+		/* Set the last TRB in the segment to have a TRB type ID of Link TRB */
+		val = le32_to_cpu(prev->trbs[TRBS_PER_SEGMENT-1].link.control);
+		val &= ~TRB_TYPE_BITMASK;
+		val |= TRB_TYPE(TRB_LINK);
+		/* Always set the chain bit with 0.95 hardware */
+		/* Set chain bit for isoc rings on AMD 0.96 host */
+		if (xhci_link_trb_quirk(xhci) ||
+				(type == TYPE_ISOC &&
+				 (xhci->quirks & XHCI_AMD_0x96_HOST)))
+			val |= TRB_CHAIN;
+		prev->trbs[TRBS_PER_SEGMENT-1].link.control = cpu_to_le32(val);
+	}
+}
+
+/*
+ * Link the ring to the new segments.
+ * Set Toggle Cycle for the new ring if needed.
+ */
+static void xhci_link_rings(struct xhci_hcd *xhci, struct xhci_ring *ring,
+		struct xhci_segment *first, struct xhci_segment *last,
+		unsigned int num_segs)
+{
+	struct xhci_segment *next;
+
+	if (!ring || !first || !last)
+		return;
+
+	next = ring->enq_seg->next;
+	xhci_link_segments(xhci, ring->enq_seg, first, ring->type);
+	xhci_link_segments(xhci, last, next, ring->type);
+	ring->num_segs += num_segs;
+	ring->num_trbs_free += (TRBS_PER_SEGMENT - 1) * num_segs;
+
+	if (ring->type != TYPE_EVENT && ring->enq_seg == ring->last_seg) {
+		ring->last_seg->trbs[TRBS_PER_SEGMENT-1].link.control
+			&= ~cpu_to_le32(LINK_TOGGLE);
+		last->trbs[TRBS_PER_SEGMENT-1].link.control
+			|= cpu_to_le32(LINK_TOGGLE);
+		ring->last_seg = last;
+	}
+}
+
+/* XXX: Do we need the hcd structure in all these functions? */
+void xhci_ring_free(struct xhci_hcd *xhci, struct xhci_ring *ring)
+{
+	if (!ring)
+		return;
+
+	if (ring->first_seg)
+		xhci_free_segments_for_ring(xhci, ring->first_seg);
+
+	kfree(ring);
+}
+
+static void xhci_initialize_ring_info(struct xhci_ring *ring,
+					unsigned int cycle_state)
+{
+	/* The ring is empty, so the enqueue pointer == dequeue pointer */
+	ring->enqueue = ring->first_seg->trbs;
+	ring->enq_seg = ring->first_seg;
+	ring->dequeue = ring->enqueue;
+	ring->deq_seg = ring->first_seg;
+	/* The ring is initialized to 0. The producer must write 1 to the cycle
+	 * bit to handover ownership of the TRB, so PCS = 1.  The consumer must
+	 * compare CCS to the cycle bit to check ownership, so CCS = 1.
+	 *
+	 * New rings are initialized with cycle state equal to 1; if we are
+	 * handling ring expansion, set the cycle state equal to the old ring.
+	 */
+	ring->cycle_state = cycle_state;
+	/* Not necessary for new rings, but needed for re-initialized rings */
+	ring->enq_updates = 0;
+	ring->deq_updates = 0;
+
+	/*
+	 * Each segment has a link TRB, and leave an extra TRB for SW
+	 * accounting purpose
+	 */
+	ring->num_trbs_free = ring->num_segs * (TRBS_PER_SEGMENT - 1) - 1;
+}
+
+/* Allocate segments and link them for a ring */
+static int xhci_alloc_segments_for_ring(struct xhci_hcd *xhci,
+		struct xhci_segment **first, struct xhci_segment **last,
+		unsigned int num_segs, unsigned int cycle_state,
+		enum xhci_ring_type type, gfp_t flags)
+{
+	struct xhci_segment *prev;
+
+	prev = xhci_segment_alloc(xhci, cycle_state, flags);
+	if (!prev)
+		return -ENOMEM;
+	num_segs--;
+
+	*first = prev;
+	while (num_segs > 0) {
+		struct xhci_segment	*next;
+
+		next = xhci_segment_alloc(xhci, cycle_state, flags);
+		if (!next) {
+			prev = *first;
+			while (prev) {
+				next = prev->next;
+				xhci_segment_free(xhci, prev);
+				prev = next;
+			}
+			return -ENOMEM;
+		}
+		xhci_link_segments(xhci, prev, next, type);
+
+		prev = next;
+		num_segs--;
+	}
+	xhci_link_segments(xhci, prev, *first, type);
+	*last = prev;
+
+	return 0;
+}
+
+/**
+ * Create a new ring with zero or more segments.
+ *
+ * Link each segment together into a ring.
+ * Set the end flag and the cycle toggle bit on the last segment.
+ * See section 4.9.1 and figures 15 and 16.
+ */
+static struct xhci_ring *xhci_ring_alloc(struct xhci_hcd *xhci,
+		unsigned int num_segs, unsigned int cycle_state,
+		enum xhci_ring_type type, gfp_t flags)
+{
+	struct xhci_ring	*ring;
+	int ret;
+
+	ring = kzalloc(sizeof *(ring), flags);
+	if (!ring)
+		return NULL;
+
+	ring->num_segs = num_segs;
+	INIT_LIST_HEAD(&ring->td_list);
+	ring->type = type;
+	if (num_segs == 0)
+		return ring;
+
+	ret = xhci_alloc_segments_for_ring(xhci, &ring->first_seg,
+			&ring->last_seg, num_segs, cycle_state, type, flags);
+	if (ret)
+		goto fail;
+
+	/* Only event ring does not use link TRB */
+	if (type != TYPE_EVENT) {
+		/* See section 4.9.2.1 and 6.4.4.1 */
+		ring->last_seg->trbs[TRBS_PER_SEGMENT - 1].link.control |=
+			cpu_to_le32(LINK_TOGGLE);
+	}
+	xhci_initialize_ring_info(ring, cycle_state);
+	return ring;
+
+fail:
+	kfree(ring);
+	return NULL;
+}
+
+void xhci_free_or_cache_endpoint_ring(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		unsigned int ep_index)
+{
+	int rings_cached;
+
+	rings_cached = virt_dev->num_rings_cached;
+	if (rings_cached < XHCI_MAX_RINGS_CACHED) {
+		virt_dev->ring_cache[rings_cached] =
+			virt_dev->eps[ep_index].ring;
+		virt_dev->num_rings_cached++;
+		xhci_dbg(xhci, "Cached old ring, "
+				"%d ring%s cached\n",
+				virt_dev->num_rings_cached,
+				(virt_dev->num_rings_cached > 1) ? "s" : "");
+	} else {
+		xhci_ring_free(xhci, virt_dev->eps[ep_index].ring);
+		xhci_dbg(xhci, "Ring cache full (%d rings), "
+				"freeing ring\n",
+				virt_dev->num_rings_cached);
+	}
+	virt_dev->eps[ep_index].ring = NULL;
+}
+
+/* Zero an endpoint ring (except for link TRBs) and move the enqueue and dequeue
+ * pointers to the beginning of the ring.
+ */
+static void xhci_reinit_cached_ring(struct xhci_hcd *xhci,
+			struct xhci_ring *ring, unsigned int cycle_state,
+			enum xhci_ring_type type)
+{
+	struct xhci_segment	*seg = ring->first_seg;
+	int i;
+
+	do {
+		memset(seg->trbs, 0,
+				sizeof(union xhci_trb)*TRBS_PER_SEGMENT);
+		if (cycle_state == 0) {
+			for (i = 0; i < TRBS_PER_SEGMENT; i++)
+				seg->trbs[i].link.control |= TRB_CYCLE;
+		}
+		/* All endpoint rings have link TRBs */
+		xhci_link_segments(xhci, seg, seg->next, type);
+		seg = seg->next;
+	} while (seg != ring->first_seg);
+	ring->type = type;
+	xhci_initialize_ring_info(ring, cycle_state);
+	/* td list should be empty since all URBs have been cancelled,
+	 * but just in case...
+	 */
+	INIT_LIST_HEAD(&ring->td_list);
+}
+
+/*
+ * Expand an existing ring.
+ * Look for a cached ring or allocate a new ring which has same segment numbers
+ * and link the two rings.
+ */
+int xhci_ring_expansion(struct xhci_hcd *xhci, struct xhci_ring *ring,
+				unsigned int num_trbs, gfp_t flags)
+{
+	struct xhci_segment	*first;
+	struct xhci_segment	*last;
+	unsigned int		num_segs;
+	unsigned int		num_segs_needed;
+	int			ret;
+
+	num_segs_needed = (num_trbs + (TRBS_PER_SEGMENT - 1) - 1) /
+				(TRBS_PER_SEGMENT - 1);
+
+	/* Allocate number of segments we needed, or double the ring size */
+	num_segs = ring->num_segs > num_segs_needed ?
+			ring->num_segs : num_segs_needed;
+
+	ret = xhci_alloc_segments_for_ring(xhci, &first, &last,
+			num_segs, ring->cycle_state, ring->type, flags);
+	if (ret)
+		return -ENOMEM;
+
+	xhci_link_rings(xhci, ring, first, last, num_segs);
+	xhci_dbg(xhci, "ring expansion succeed, now has %d segments\n",
+			ring->num_segs);
+
+	return 0;
+}
+
+#define CTX_SIZE(_hcc) (HCC_64BYTE_CONTEXT(_hcc) ? 64 : 32)
+
+static 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;
+}
+
+static void xhci_free_container_ctx(struct xhci_hcd *xhci,
+			     struct xhci_container_ctx *ctx)
+{
+	if (!ctx)
+		return;
+	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)));
+}
+
+
+/***************** Streams structures manipulation *************************/
+
+static void xhci_free_stream_ctx(struct xhci_hcd *xhci,
+		unsigned int num_stream_ctxs,
+		struct xhci_stream_ctx *stream_ctx, dma_addr_t dma)
+{
+	struct pci_dev *pdev = to_pci_dev(xhci_to_hcd(xhci)->self.controller);
+
+	if (num_stream_ctxs > MEDIUM_STREAM_ARRAY_SIZE)
+		dma_free_coherent(&pdev->dev,
+				sizeof(struct xhci_stream_ctx)*num_stream_ctxs,
+				stream_ctx, dma);
+	else if (num_stream_ctxs <= SMALL_STREAM_ARRAY_SIZE)
+		return dma_pool_free(xhci->small_streams_pool,
+				stream_ctx, dma);
+	else
+		return dma_pool_free(xhci->medium_streams_pool,
+				stream_ctx, dma);
+}
+
+/*
+ * The stream context array for each endpoint with bulk streams enabled can
+ * vary in size, based on:
+ *  - how many streams the endpoint supports,
+ *  - the maximum primary stream array size the host controller supports,
+ *  - and how many streams the device driver asks for.
+ *
+ * The stream context array must be a power of 2, and can be as small as
+ * 64 bytes or as large as 1MB.
+ */
+static struct xhci_stream_ctx *xhci_alloc_stream_ctx(struct xhci_hcd *xhci,
+		unsigned int num_stream_ctxs, dma_addr_t *dma,
+		gfp_t mem_flags)
+{
+	struct pci_dev *pdev = to_pci_dev(xhci_to_hcd(xhci)->self.controller);
+
+	if (num_stream_ctxs > MEDIUM_STREAM_ARRAY_SIZE)
+		return dma_alloc_coherent(&pdev->dev,
+				sizeof(struct xhci_stream_ctx)*num_stream_ctxs,
+				dma, mem_flags);
+	else if (num_stream_ctxs <= SMALL_STREAM_ARRAY_SIZE)
+		return dma_pool_alloc(xhci->small_streams_pool,
+				mem_flags, dma);
+	else
+		return dma_pool_alloc(xhci->medium_streams_pool,
+				mem_flags, dma);
+}
+
+struct xhci_ring *xhci_dma_to_transfer_ring(
+		struct xhci_virt_ep *ep,
+		u64 address)
+{
+	if (ep->ep_state & EP_HAS_STREAMS)
+		return radix_tree_lookup(&ep->stream_info->trb_address_map,
+				address >> TRB_SEGMENT_SHIFT);
+	return ep->ring;
+}
+
+/* Only use this when you know stream_info is valid */
+#ifdef CONFIG_USB_XHCI_HCD_DEBUGGING
+static struct xhci_ring *dma_to_stream_ring(
+		struct xhci_stream_info *stream_info,
+		u64 address)
+{
+	return radix_tree_lookup(&stream_info->trb_address_map,
+			address >> TRB_SEGMENT_SHIFT);
+}
+#endif	/* CONFIG_USB_XHCI_HCD_DEBUGGING */
+
+struct xhci_ring *xhci_stream_id_to_ring(
+		struct xhci_virt_device *dev,
+		unsigned int ep_index,
+		unsigned int stream_id)
+{
+	struct xhci_virt_ep *ep = &dev->eps[ep_index];
+
+	if (stream_id == 0)
+		return ep->ring;
+	if (!ep->stream_info)
+		return NULL;
+
+	if (stream_id > ep->stream_info->num_streams)
+		return NULL;
+	return ep->stream_info->stream_rings[stream_id];
+}
+
+#ifdef CONFIG_USB_XHCI_HCD_DEBUGGING
+static int xhci_test_radix_tree(struct xhci_hcd *xhci,
+		unsigned int num_streams,
+		struct xhci_stream_info *stream_info)
+{
+	u32 cur_stream;
+	struct xhci_ring *cur_ring;
+	u64 addr;
+
+	for (cur_stream = 1; cur_stream < num_streams; cur_stream++) {
+		struct xhci_ring *mapped_ring;
+		int trb_size = sizeof(union xhci_trb);
+
+		cur_ring = stream_info->stream_rings[cur_stream];
+		for (addr = cur_ring->first_seg->dma;
+				addr < cur_ring->first_seg->dma + TRB_SEGMENT_SIZE;
+				addr += trb_size) {
+			mapped_ring = dma_to_stream_ring(stream_info, addr);
+			if (cur_ring != mapped_ring) {
+				xhci_warn(xhci, "WARN: DMA address 0x%08llx "
+						"didn't map to stream ID %u; "
+						"mapped to ring %p\n",
+						(unsigned long long) addr,
+						cur_stream,
+						mapped_ring);
+				return -EINVAL;
+			}
+		}
+		/* One TRB after the end of the ring segment shouldn't return a
+		 * pointer to the current ring (although it may be a part of a
+		 * different ring).
+		 */
+		mapped_ring = dma_to_stream_ring(stream_info, addr);
+		if (mapped_ring != cur_ring) {
+			/* One TRB before should also fail */
+			addr = cur_ring->first_seg->dma - trb_size;
+			mapped_ring = dma_to_stream_ring(stream_info, addr);
+		}
+		if (mapped_ring == cur_ring) {
+			xhci_warn(xhci, "WARN: Bad DMA address 0x%08llx "
+					"mapped to valid stream ID %u; "
+					"mapped ring = %p\n",
+					(unsigned long long) addr,
+					cur_stream,
+					mapped_ring);
+			return -EINVAL;
+		}
+	}
+	return 0;
+}
+#endif	/* CONFIG_USB_XHCI_HCD_DEBUGGING */
+
+/*
+ * Change an endpoint's internal structure so it supports stream IDs.  The
+ * number of requested streams includes stream 0, which cannot be used by device
+ * drivers.
+ *
+ * The number of stream contexts in the stream context array may be bigger than
+ * the number of streams the driver wants to use.  This is because the number of
+ * stream context array entries must be a power of two.
+ *
+ * We need a radix tree for mapping physical addresses of TRBs to which stream
+ * ID they belong to.  We need to do this because the host controller won't tell
+ * us which stream ring the TRB came from.  We could store the stream ID in an
+ * event data TRB, but that doesn't help us for the cancellation case, since the
+ * endpoint may stop before it reaches that event data TRB.
+ *
+ * The radix tree maps the upper portion of the TRB DMA address to a ring
+ * segment that has the same upper portion of DMA addresses.  For example, say I
+ * have segments of size 1KB, that are always 64-byte aligned.  A segment may
+ * start at 0x10c91000 and end at 0x10c913f0.  If I use the upper 10 bits, the
+ * key to the stream ID is 0x43244.  I can use the DMA address of the TRB to
+ * pass the radix tree a key to get the right stream ID:
+ *
+ * 	0x10c90fff >> 10 = 0x43243
+ * 	0x10c912c0 >> 10 = 0x43244
+ * 	0x10c91400 >> 10 = 0x43245
+ *
+ * Obviously, only those TRBs with DMA addresses that are within the segment
+ * will make the radix tree return the stream ID for that ring.
+ *
+ * Caveats for the radix tree:
+ *
+ * The radix tree uses an unsigned long as a key pair.  On 32-bit systems, an
+ * unsigned long will be 32-bits; on a 64-bit system an unsigned long will be
+ * 64-bits.  Since we only request 32-bit DMA addresses, we can use that as the
+ * key on 32-bit or 64-bit systems (it would also be fine if we asked for 64-bit
+ * PCI DMA addresses on a 64-bit system).  There might be a problem on 32-bit
+ * extended systems (where the DMA address can be bigger than 32-bits),
+ * if we allow the PCI dma mask to be bigger than 32-bits.  So don't do that.
+ */
+struct xhci_stream_info *xhci_alloc_stream_info(struct xhci_hcd *xhci,
+		unsigned int num_stream_ctxs,
+		unsigned int num_streams, gfp_t mem_flags)
+{
+	struct xhci_stream_info *stream_info;
+	u32 cur_stream;
+	struct xhci_ring *cur_ring;
+	unsigned long key;
+	u64 addr;
+	int ret;
+
+	xhci_dbg(xhci, "Allocating %u streams and %u "
+			"stream context array entries.\n",
+			num_streams, num_stream_ctxs);
+	if (xhci->cmd_ring_reserved_trbs == MAX_RSVD_CMD_TRBS) {
+		xhci_dbg(xhci, "Command ring has no reserved TRBs available\n");
+		return NULL;
+	}
+	xhci->cmd_ring_reserved_trbs++;
+
+	stream_info = kzalloc(sizeof(struct xhci_stream_info), mem_flags);
+	if (!stream_info)
+		goto cleanup_trbs;
+
+	stream_info->num_streams = num_streams;
+	stream_info->num_stream_ctxs = num_stream_ctxs;
+
+	/* Initialize the array of virtual pointers to stream rings. */
+	stream_info->stream_rings = kzalloc(
+			sizeof(struct xhci_ring *)*num_streams,
+			mem_flags);
+	if (!stream_info->stream_rings)
+		goto cleanup_info;
+
+	/* Initialize the array of DMA addresses for stream rings for the HW. */
+	stream_info->stream_ctx_array = xhci_alloc_stream_ctx(xhci,
+			num_stream_ctxs, &stream_info->ctx_array_dma,
+			mem_flags);
+	if (!stream_info->stream_ctx_array)
+		goto cleanup_ctx;
+	memset(stream_info->stream_ctx_array, 0,
+			sizeof(struct xhci_stream_ctx)*num_stream_ctxs);
+
+	/* Allocate everything needed to free the stream rings later */
+	stream_info->free_streams_command =
+		xhci_alloc_command(xhci, true, true, mem_flags);
+	if (!stream_info->free_streams_command)
+		goto cleanup_ctx;
+
+	INIT_RADIX_TREE(&stream_info->trb_address_map, GFP_ATOMIC);
+
+	/* Allocate rings for all the streams that the driver will use,
+	 * and add their segment DMA addresses to the radix tree.
+	 * Stream 0 is reserved.
+	 */
+	for (cur_stream = 1; cur_stream < num_streams; cur_stream++) {
+		stream_info->stream_rings[cur_stream] =
+			xhci_ring_alloc(xhci, 2, 1, TYPE_STREAM, mem_flags);
+		cur_ring = stream_info->stream_rings[cur_stream];
+		if (!cur_ring)
+			goto cleanup_rings;
+		cur_ring->stream_id = cur_stream;
+		/* Set deq ptr, cycle bit, and stream context type */
+		addr = cur_ring->first_seg->dma |
+			SCT_FOR_CTX(SCT_PRI_TR) |
+			cur_ring->cycle_state;
+		stream_info->stream_ctx_array[cur_stream].stream_ring =
+			cpu_to_le64(addr);
+		xhci_dbg(xhci, "Setting stream %d ring ptr to 0x%08llx\n",
+				cur_stream, (unsigned long long) addr);
+
+		key = (unsigned long)
+			(cur_ring->first_seg->dma >> TRB_SEGMENT_SHIFT);
+		ret = radix_tree_insert(&stream_info->trb_address_map,
+				key, cur_ring);
+		if (ret) {
+			xhci_ring_free(xhci, cur_ring);
+			stream_info->stream_rings[cur_stream] = NULL;
+			goto cleanup_rings;
+		}
+	}
+	/* Leave the other unused stream ring pointers in the stream context
+	 * array initialized to zero.  This will cause the xHC to give us an
+	 * error if the device asks for a stream ID we don't have setup (if it
+	 * was any other way, the host controller would assume the ring is
+	 * "empty" and wait forever for data to be queued to that stream ID).
+	 */
+#if XHCI_DEBUG
+	/* Do a little test on the radix tree to make sure it returns the
+	 * correct values.
+	 */
+	if (xhci_test_radix_tree(xhci, num_streams, stream_info))
+		goto cleanup_rings;
+#endif
+
+	return stream_info;
+
+cleanup_rings:
+	for (cur_stream = 1; cur_stream < num_streams; cur_stream++) {
+		cur_ring = stream_info->stream_rings[cur_stream];
+		if (cur_ring) {
+			addr = cur_ring->first_seg->dma;
+			radix_tree_delete(&stream_info->trb_address_map,
+					addr >> TRB_SEGMENT_SHIFT);
+			xhci_ring_free(xhci, cur_ring);
+			stream_info->stream_rings[cur_stream] = NULL;
+		}
+	}
+	xhci_free_command(xhci, stream_info->free_streams_command);
+cleanup_ctx:
+	kfree(stream_info->stream_rings);
+cleanup_info:
+	kfree(stream_info);
+cleanup_trbs:
+	xhci->cmd_ring_reserved_trbs--;
+	return NULL;
+}
+/*
+ * Sets the MaxPStreams field and the Linear Stream Array field.
+ * Sets the dequeue pointer to the stream context array.
+ */
+void xhci_setup_streams_ep_input_ctx(struct xhci_hcd *xhci,
+		struct xhci_ep_ctx *ep_ctx,
+		struct xhci_stream_info *stream_info)
+{
+	u32 max_primary_streams;
+	/* MaxPStreams is the number of stream context array entries, not the
+	 * number we're actually using.  Must be in 2^(MaxPstreams + 1) format.
+	 * fls(0) = 0, fls(0x1) = 1, fls(0x10) = 2, fls(0x100) = 3, etc.
+	 */
+	max_primary_streams = fls(stream_info->num_stream_ctxs) - 2;
+	xhci_dbg(xhci, "Setting number of stream ctx array entries to %u\n",
+			1 << (max_primary_streams + 1));
+	ep_ctx->ep_info &= cpu_to_le32(~EP_MAXPSTREAMS_MASK);
+	ep_ctx->ep_info |= cpu_to_le32(EP_MAXPSTREAMS(max_primary_streams)
+				       | EP_HAS_LSA);
+	ep_ctx->deq  = cpu_to_le64(stream_info->ctx_array_dma);
+}
+
+/*
+ * Sets the MaxPStreams field and the Linear Stream Array field to 0.
+ * Reinstalls the "normal" endpoint ring (at its previous dequeue mark,
+ * not at the beginning of the ring).
+ */
+void xhci_setup_no_streams_ep_input_ctx(struct xhci_hcd *xhci,
+		struct xhci_ep_ctx *ep_ctx,
+		struct xhci_virt_ep *ep)
+{
+	dma_addr_t addr;
+	ep_ctx->ep_info &= cpu_to_le32(~(EP_MAXPSTREAMS_MASK | EP_HAS_LSA));
+	addr = xhci_trb_virt_to_dma(ep->ring->deq_seg, ep->ring->dequeue);
+	ep_ctx->deq  = cpu_to_le64(addr | ep->ring->cycle_state);
+}
+
+/* Frees all stream contexts associated with the endpoint,
+ *
+ * Caller should fix the endpoint context streams fields.
+ */
+void xhci_free_stream_info(struct xhci_hcd *xhci,
+		struct xhci_stream_info *stream_info)
+{
+	int cur_stream;
+	struct xhci_ring *cur_ring;
+	dma_addr_t addr;
+
+	if (!stream_info)
+		return;
+
+	for (cur_stream = 1; cur_stream < stream_info->num_streams;
+			cur_stream++) {
+		cur_ring = stream_info->stream_rings[cur_stream];
+		if (cur_ring) {
+			addr = cur_ring->first_seg->dma;
+			radix_tree_delete(&stream_info->trb_address_map,
+					addr >> TRB_SEGMENT_SHIFT);
+			xhci_ring_free(xhci, cur_ring);
+			stream_info->stream_rings[cur_stream] = NULL;
+		}
+	}
+	xhci_free_command(xhci, stream_info->free_streams_command);
+	xhci->cmd_ring_reserved_trbs--;
+	if (stream_info->stream_ctx_array)
+		xhci_free_stream_ctx(xhci,
+				stream_info->num_stream_ctxs,
+				stream_info->stream_ctx_array,
+				stream_info->ctx_array_dma);
+
+	if (stream_info)
+		kfree(stream_info->stream_rings);
+	kfree(stream_info);
+}
+
+
+/***************** Device context manipulation *************************/
+
+static void xhci_init_endpoint_timer(struct xhci_hcd *xhci,
+		struct xhci_virt_ep *ep)
+{
+	init_timer(&ep->stop_cmd_timer);
+	ep->stop_cmd_timer.data = (unsigned long) ep;
+	ep->stop_cmd_timer.function = xhci_stop_endpoint_command_watchdog;
+	ep->xhci = xhci;
+}
+
+static void xhci_free_tt_info(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		int slot_id)
+{
+	struct list_head *tt_list_head;
+	struct xhci_tt_bw_info *tt_info, *next;
+	bool slot_found = false;
+
+	/* If the device never made it past the Set Address stage,
+	 * it may not have the real_port set correctly.
+	 */
+	if (virt_dev->real_port == 0 ||
+			virt_dev->real_port > HCS_MAX_PORTS(xhci->hcs_params1)) {
+		xhci_dbg(xhci, "Bad real port.\n");
+		return;
+	}
+
+	tt_list_head = &(xhci->rh_bw[virt_dev->real_port - 1].tts);
+	list_for_each_entry_safe(tt_info, next, tt_list_head, tt_list) {
+		/* Multi-TT hubs will have more than one entry */
+		if (tt_info->slot_id == slot_id) {
+			slot_found = true;
+			list_del(&tt_info->tt_list);
+			kfree(tt_info);
+		} else if (slot_found) {
+			break;
+		}
+	}
+}
+
+int xhci_alloc_tt_info(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		struct usb_device *hdev,
+		struct usb_tt *tt, gfp_t mem_flags)
+{
+	struct xhci_tt_bw_info		*tt_info;
+	unsigned int			num_ports;
+	int				i, j;
+
+	if (!tt->multi)
+		num_ports = 1;
+	else
+		num_ports = hdev->maxchild;
+
+	for (i = 0; i < num_ports; i++, tt_info++) {
+		struct xhci_interval_bw_table *bw_table;
+
+		tt_info = kzalloc(sizeof(*tt_info), mem_flags);
+		if (!tt_info)
+			goto free_tts;
+		INIT_LIST_HEAD(&tt_info->tt_list);
+		list_add(&tt_info->tt_list,
+				&xhci->rh_bw[virt_dev->real_port - 1].tts);
+		tt_info->slot_id = virt_dev->udev->slot_id;
+		if (tt->multi)
+			tt_info->ttport = i+1;
+		bw_table = &tt_info->bw_table;
+		for (j = 0; j < XHCI_MAX_INTERVAL; j++)
+			INIT_LIST_HEAD(&bw_table->interval_bw[j].endpoints);
+	}
+	return 0;
+
+free_tts:
+	xhci_free_tt_info(xhci, virt_dev, virt_dev->udev->slot_id);
+	return -ENOMEM;
+}
+
+
+/* All the xhci_tds in the ring's TD list should be freed at this point.
+ * Should be called with xhci->lock held if there is any chance the TT lists
+ * will be manipulated by the configure endpoint, allocate device, or update
+ * hub functions while this function is removing the TT entries from the list.
+ */
+void xhci_free_virt_device(struct xhci_hcd *xhci, int slot_id)
+{
+	struct xhci_virt_device *dev;
+	int i;
+	int old_active_eps = 0;
+
+	/* Slot ID 0 is reserved */
+	if (slot_id == 0 || !xhci->devs[slot_id])
+		return;
+
+	dev = xhci->devs[slot_id];
+	xhci->dcbaa->dev_context_ptrs[slot_id] = 0;
+	if (!dev)
+		return;
+
+	if (dev->tt_info)
+		old_active_eps = dev->tt_info->active_eps;
+
+	for (i = 0; i < 31; ++i) {
+		if (dev->eps[i].ring)
+			xhci_ring_free(xhci, dev->eps[i].ring);
+		if (dev->eps[i].stream_info)
+			xhci_free_stream_info(xhci,
+					dev->eps[i].stream_info);
+		/* Endpoints on the TT/root port lists should have been removed
+		 * when usb_disable_device() was called for the device.
+		 * We can't drop them anyway, because the udev might have gone
+		 * away by this point, and we can't tell what speed it was.
+		 */
+		if (!list_empty(&dev->eps[i].bw_endpoint_list))
+			xhci_warn(xhci, "Slot %u endpoint %u "
+					"not removed from BW list!\n",
+					slot_id, i);
+	}
+	/* If this is a hub, free the TT(s) from the TT list */
+	xhci_free_tt_info(xhci, dev, slot_id);
+	/* If necessary, update the number of active TTs on this root port */
+	xhci_update_tt_active_eps(xhci, dev, old_active_eps);
+
+	if (dev->ring_cache) {
+		for (i = 0; i < dev->num_rings_cached; i++)
+			xhci_ring_free(xhci, dev->ring_cache[i]);
+		kfree(dev->ring_cache);
+	}
+
+	if (dev->in_ctx)
+		xhci_free_container_ctx(xhci, dev->in_ctx);
+	if (dev->out_ctx)
+		xhci_free_container_ctx(xhci, dev->out_ctx);
+
+	kfree(xhci->devs[slot_id]);
+	xhci->devs[slot_id] = NULL;
+}
+
+int xhci_alloc_virt_device(struct xhci_hcd *xhci, int slot_id,
+		struct usb_device *udev, gfp_t flags)
+{
+	struct xhci_virt_device *dev;
+	int i;
+
+	/* Slot ID 0 is reserved */
+	if (slot_id == 0 || xhci->devs[slot_id]) {
+		xhci_warn(xhci, "Bad Slot ID %d\n", slot_id);
+		return 0;
+	}
+
+	xhci->devs[slot_id] = kzalloc(sizeof(*xhci->devs[slot_id]), flags);
+	if (!xhci->devs[slot_id])
+		return 0;
+	dev = xhci->devs[slot_id];
+
+	/* 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;
+
+	xhci_dbg(xhci, "Slot %d output ctx = 0x%llx (dma)\n", slot_id,
+			(unsigned long long)dev->out_ctx->dma);
+
+	/* Allocate the (input) device context for address device command */
+	dev->in_ctx = xhci_alloc_container_ctx(xhci, XHCI_CTX_TYPE_INPUT, flags);
+	if (!dev->in_ctx)
+		goto fail;
+
+	xhci_dbg(xhci, "Slot %d input ctx = 0x%llx (dma)\n", slot_id,
+			(unsigned long long)dev->in_ctx->dma);
+
+	/* Initialize the cancellation list and watchdog timers for each ep */
+	for (i = 0; i < 31; i++) {
+		xhci_init_endpoint_timer(xhci, &dev->eps[i]);
+		INIT_LIST_HEAD(&dev->eps[i].cancelled_td_list);
+		INIT_LIST_HEAD(&dev->eps[i].bw_endpoint_list);
+	}
+
+	/* Allocate endpoint 0 ring */
+	dev->eps[0].ring = xhci_ring_alloc(xhci, 2, 1, TYPE_CTRL, flags);
+	if (!dev->eps[0].ring)
+		goto fail;
+
+	/* Allocate pointers to the ring cache */
+	dev->ring_cache = kzalloc(
+			sizeof(struct xhci_ring *)*XHCI_MAX_RINGS_CACHED,
+			flags);
+	if (!dev->ring_cache)
+		goto fail;
+	dev->num_rings_cached = 0;
+
+	init_completion(&dev->cmd_completion);
+	INIT_LIST_HEAD(&dev->cmd_list);
+	dev->udev = udev;
+
+	/* Point to output device context in dcbaa. */
+	xhci->dcbaa->dev_context_ptrs[slot_id] = cpu_to_le64(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[slot_id],
+		 le64_to_cpu(xhci->dcbaa->dev_context_ptrs[slot_id]));
+
+	return 1;
+fail:
+	xhci_free_virt_device(xhci, slot_id);
+	return 0;
+}
+
+void xhci_copy_ep0_dequeue_into_input_ctx(struct xhci_hcd *xhci,
+		struct usb_device *udev)
+{
+	struct xhci_virt_device *virt_dev;
+	struct xhci_ep_ctx	*ep0_ctx;
+	struct xhci_ring	*ep_ring;
+
+	virt_dev = xhci->devs[udev->slot_id];
+	ep0_ctx = xhci_get_ep_ctx(xhci, virt_dev->in_ctx, 0);
+	ep_ring = virt_dev->eps[0].ring;
+	/*
+	 * FIXME we don't keep track of the dequeue pointer very well after a
+	 * Set TR dequeue pointer, so we're setting the dequeue pointer of the
+	 * host to our enqueue pointer.  This should only be called after a
+	 * configured device has reset, so all control transfers should have
+	 * been completed or cancelled before the reset.
+	 */
+	ep0_ctx->deq = cpu_to_le64(xhci_trb_virt_to_dma(ep_ring->enq_seg,
+							ep_ring->enqueue)
+				   | ep_ring->cycle_state);
+}
+
+/*
+ * The xHCI roothub may have ports of differing speeds in any order in the port
+ * status registers.  xhci->port_array provides an array of the port speed for
+ * each offset into the port status registers.
+ *
+ * The xHCI hardware wants to know the roothub port number that the USB device
+ * is attached to (or the roothub port its ancestor hub is attached to).  All we
+ * know is the index of that port under either the USB 2.0 or the USB 3.0
+ * roothub, but that doesn't give us the real index into the HW port status
+ * registers. Call xhci_find_raw_port_number() to get real index.
+ */
+static u32 xhci_find_real_port_number(struct xhci_hcd *xhci,
+		struct usb_device *udev)
+{
+	struct usb_device *top_dev;
+	struct usb_hcd *hcd;
+
+	if (udev->speed == USB_SPEED_SUPER)
+		hcd = xhci->shared_hcd;
+	else
+		hcd = xhci->main_hcd;
+
+	for (top_dev = udev; top_dev->parent && top_dev->parent->parent;
+			top_dev = top_dev->parent)
+		/* Found device below root hub */;
+
+	return	xhci_find_raw_port_number(hcd, top_dev->portnum);
+}
+
+/* Setup an xHCI virtual device for a Set Address command */
+int xhci_setup_addressable_virt_dev(struct xhci_hcd *xhci, struct usb_device *udev)
+{
+	struct xhci_virt_device *dev;
+	struct xhci_ep_ctx	*ep0_ctx;
+	struct xhci_slot_ctx    *slot_ctx;
+	u32			port_num;
+	struct usb_device *top_dev;
+
+	dev = xhci->devs[udev->slot_id];
+	/* Slot ID 0 is reserved */
+	if (udev->slot_id == 0 || !dev) {
+		xhci_warn(xhci, "Slot ID %d is not assigned to this device\n",
+				udev->slot_id);
+		return -EINVAL;
+	}
+	ep0_ctx = xhci_get_ep_ctx(xhci, dev->in_ctx, 0);
+	slot_ctx = xhci_get_slot_ctx(xhci, dev->in_ctx);
+
+	/* 3) Only the control endpoint is valid - one endpoint context */
+	slot_ctx->dev_info |= cpu_to_le32(LAST_CTX(1) | udev->route);
+	switch (udev->speed) {
+	case USB_SPEED_SUPER:
+		slot_ctx->dev_info |= cpu_to_le32(SLOT_SPEED_SS);
+		break;
+	case USB_SPEED_HIGH:
+		slot_ctx->dev_info |= cpu_to_le32(SLOT_SPEED_HS);
+		break;
+	case USB_SPEED_FULL:
+		slot_ctx->dev_info |= cpu_to_le32(SLOT_SPEED_FS);
+		break;
+	case USB_SPEED_LOW:
+		slot_ctx->dev_info |= cpu_to_le32(SLOT_SPEED_LS);
+		break;
+	case USB_SPEED_WIRELESS:
+		xhci_dbg(xhci, "FIXME xHCI doesn't support wireless speeds\n");
+		return -EINVAL;
+		break;
+	default:
+		/* Speed was set earlier, this shouldn't happen. */
+		BUG();
+	}
+	/* Find the root hub port this device is under */
+	port_num = xhci_find_real_port_number(xhci, udev);
+	if (!port_num)
+		return -EINVAL;
+	slot_ctx->dev_info2 |= cpu_to_le32(ROOT_HUB_PORT(port_num));
+	/* Set the port number in the virtual_device to the faked port number */
+	for (top_dev = udev; top_dev->parent && top_dev->parent->parent;
+			top_dev = top_dev->parent)
+		/* Found device below root hub */;
+	dev->fake_port = top_dev->portnum;
+	dev->real_port = port_num;
+	xhci_dbg(xhci, "Set root hub portnum to %d\n", port_num);
+	xhci_dbg(xhci, "Set fake root hub portnum to %d\n", dev->fake_port);
+
+	/* Find the right bandwidth table that this device will be a part of.
+	 * If this is a full speed device attached directly to a root port (or a
+	 * decendent of one), it counts as a primary bandwidth domain, not a
+	 * secondary bandwidth domain under a TT.  An xhci_tt_info structure
+	 * will never be created for the HS root hub.
+	 */
+	if (!udev->tt || !udev->tt->hub->parent) {
+		dev->bw_table = &xhci->rh_bw[port_num - 1].bw_table;
+	} else {
+		struct xhci_root_port_bw_info *rh_bw;
+		struct xhci_tt_bw_info *tt_bw;
+
+		rh_bw = &xhci->rh_bw[port_num - 1];
+		/* Find the right TT. */
+		list_for_each_entry(tt_bw, &rh_bw->tts, tt_list) {
+			if (tt_bw->slot_id != udev->tt->hub->slot_id)
+				continue;
+
+			if (!dev->udev->tt->multi ||
+					(udev->tt->multi &&
+					 tt_bw->ttport == dev->udev->ttport)) {
+				dev->bw_table = &tt_bw->bw_table;
+				dev->tt_info = tt_bw;
+				break;
+			}
+		}
+		if (!dev->tt_info)
+			xhci_warn(xhci, "WARN: Didn't find a matching TT\n");
+	}
+
+	/* Is this a LS/FS device under an external HS hub? */
+	if (udev->tt && udev->tt->hub->parent) {
+		slot_ctx->tt_info = cpu_to_le32(udev->tt->hub->slot_id |
+						(udev->ttport << 8));
+		if (udev->tt->multi)
+			slot_ctx->dev_info |= cpu_to_le32(DEV_MTT);
+	}
+	xhci_dbg(xhci, "udev->tt = %p\n", udev->tt);
+	xhci_dbg(xhci, "udev->ttport = 0x%x\n", udev->ttport);
+
+	/* Step 4 - ring already allocated */
+	/* Step 5 */
+	ep0_ctx->ep_info2 = cpu_to_le32(EP_TYPE(CTRL_EP));
+	/*
+	 * XXX: Not sure about wireless USB devices.
+	 */
+	switch (udev->speed) {
+	case USB_SPEED_SUPER:
+		ep0_ctx->ep_info2 |= cpu_to_le32(MAX_PACKET(512));
+		break;
+	case USB_SPEED_HIGH:
+	/* USB core guesses at a 64-byte max packet first for FS devices */
+	case USB_SPEED_FULL:
+		ep0_ctx->ep_info2 |= cpu_to_le32(MAX_PACKET(64));
+		break;
+	case USB_SPEED_LOW:
+		ep0_ctx->ep_info2 |= cpu_to_le32(MAX_PACKET(8));
+		break;
+	case USB_SPEED_WIRELESS:
+		xhci_dbg(xhci, "FIXME xHCI doesn't support wireless speeds\n");
+		return -EINVAL;
+		break;
+	default:
+		/* New speed? */
+		BUG();
+	}
+	/* EP 0 can handle "burst" sizes of 1, so Max Burst Size field is 0 */
+	ep0_ctx->ep_info2 |= cpu_to_le32(MAX_BURST(0) | ERROR_COUNT(3));
+
+	ep0_ctx->deq = cpu_to_le64(dev->eps[0].ring->first_seg->dma |
+				   dev->eps[0].ring->cycle_state);
+
+	/* Steps 7 and 8 were done in xhci_alloc_virt_device() */
+
+	return 0;
+}
+
+/*
+ * Convert interval expressed as 2^(bInterval - 1) == interval into
+ * straight exponent value 2^n == interval.
+ *
+ */
+static unsigned int xhci_parse_exponent_interval(struct usb_device *udev,
+		struct usb_host_endpoint *ep)
+{
+	unsigned int interval;
+
+	interval = clamp_val(ep->desc.bInterval, 1, 16) - 1;
+	if (interval != ep->desc.bInterval - 1)
+		dev_warn(&udev->dev,
+			 "ep %#x - rounding interval to %d %sframes\n",
+			 ep->desc.bEndpointAddress,
+			 1 << interval,
+			 udev->speed == USB_SPEED_FULL ? "" : "micro");
+
+	if (udev->speed == USB_SPEED_FULL) {
+		/*
+		 * Full speed isoc endpoints specify interval in frames,
+		 * not microframes. We are using microframes everywhere,
+		 * so adjust accordingly.
+		 */
+		interval += 3;	/* 1 frame = 2^3 uframes */
+	}
+
+	return interval;
+}
+
+/*
+ * Convert bInterval expressed in microframes (in 1-255 range) to exponent of
+ * microframes, rounded down to nearest power of 2.
+ */
+static unsigned int xhci_microframes_to_exponent(struct usb_device *udev,
+		struct usb_host_endpoint *ep, unsigned int desc_interval,
+		unsigned int min_exponent, unsigned int max_exponent)
+{
+	unsigned int interval;
+
+	interval = fls(desc_interval) - 1;
+	interval = clamp_val(interval, min_exponent, max_exponent);
+	if ((1 << interval) != desc_interval)
+		dev_warn(&udev->dev,
+			 "ep %#x - rounding interval to %d microframes, ep desc says %d microframes\n",
+			 ep->desc.bEndpointAddress,
+			 1 << interval,
+			 desc_interval);
+
+	return interval;
+}
+
+static unsigned int xhci_parse_microframe_interval(struct usb_device *udev,
+		struct usb_host_endpoint *ep)
+{
+	if (ep->desc.bInterval == 0)
+		return 0;
+	return xhci_microframes_to_exponent(udev, ep,
+			ep->desc.bInterval, 0, 15);
+}
+
+
+static unsigned int xhci_parse_frame_interval(struct usb_device *udev,
+		struct usb_host_endpoint *ep)
+{
+	return xhci_microframes_to_exponent(udev, ep,
+			ep->desc.bInterval * 8, 3, 10);
+}
+
+/* Return the polling or NAK interval.
+ *
+ * The polling interval is expressed in "microframes".  If xHCI's Interval field
+ * is set to N, it will service the endpoint every 2^(Interval)*125us.
+ *
+ * The NAK interval is one NAK per 1 to 255 microframes, or no NAKs if interval
+ * is set to 0.
+ */
+static unsigned int xhci_get_endpoint_interval(struct usb_device *udev,
+		struct usb_host_endpoint *ep)
+{
+	unsigned int interval = 0;
+
+	switch (udev->speed) {
+	case USB_SPEED_HIGH:
+		/* Max NAK rate */
+		if (usb_endpoint_xfer_control(&ep->desc) ||
+		    usb_endpoint_xfer_bulk(&ep->desc)) {
+			interval = xhci_parse_microframe_interval(udev, ep);
+			break;
+		}
+		/* Fall through - SS and HS isoc/int have same decoding */
+
+	case USB_SPEED_SUPER:
+		if (usb_endpoint_xfer_int(&ep->desc) ||
+		    usb_endpoint_xfer_isoc(&ep->desc)) {
+			interval = xhci_parse_exponent_interval(udev, ep);
+		}
+		break;
+
+	case USB_SPEED_FULL:
+		if (usb_endpoint_xfer_isoc(&ep->desc)) {
+			interval = xhci_parse_exponent_interval(udev, ep);
+			break;
+		}
+		/*
+		 * Fall through for interrupt endpoint interval decoding
+		 * since it uses the same rules as low speed interrupt
+		 * endpoints.
+		 */
+
+	case USB_SPEED_LOW:
+		if (usb_endpoint_xfer_int(&ep->desc) ||
+		    usb_endpoint_xfer_isoc(&ep->desc)) {
+
+			interval = xhci_parse_frame_interval(udev, ep);
+		}
+		break;
+
+	default:
+		BUG();
+	}
+	return EP_INTERVAL(interval);
+}
+
+/* The "Mult" field in the endpoint context is only set for SuperSpeed isoc eps.
+ * High speed endpoint descriptors can define "the number of additional
+ * transaction opportunities per microframe", but that goes in the Max Burst
+ * endpoint context field.
+ */
+static u32 xhci_get_endpoint_mult(struct usb_device *udev,
+		struct usb_host_endpoint *ep)
+{
+	if (udev->speed != USB_SPEED_SUPER ||
+			!usb_endpoint_xfer_isoc(&ep->desc))
+		return 0;
+	return ep->ss_ep_comp.bmAttributes;
+}
+
+static u32 xhci_get_endpoint_type(struct usb_device *udev,
+		struct usb_host_endpoint *ep)
+{
+	int in;
+	u32 type;
+
+	in = usb_endpoint_dir_in(&ep->desc);
+	if (usb_endpoint_xfer_control(&ep->desc)) {
+		type = EP_TYPE(CTRL_EP);
+	} else if (usb_endpoint_xfer_bulk(&ep->desc)) {
+		if (in)
+			type = EP_TYPE(BULK_IN_EP);
+		else
+			type = EP_TYPE(BULK_OUT_EP);
+	} else if (usb_endpoint_xfer_isoc(&ep->desc)) {
+		if (in)
+			type = EP_TYPE(ISOC_IN_EP);
+		else
+			type = EP_TYPE(ISOC_OUT_EP);
+	} else if (usb_endpoint_xfer_int(&ep->desc)) {
+		if (in)
+			type = EP_TYPE(INT_IN_EP);
+		else
+			type = EP_TYPE(INT_OUT_EP);
+	} else {
+		BUG();
+	}
+	return type;
+}
+
+/* Return the maximum endpoint service interval time (ESIT) payload.
+ * Basically, this is the maxpacket size, multiplied by the burst size
+ * and mult size.
+ */
+static u32 xhci_get_max_esit_payload(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		struct usb_host_endpoint *ep)
+{
+	int max_burst;
+	int max_packet;
+
+	/* Only applies for interrupt or isochronous endpoints */
+	if (usb_endpoint_xfer_control(&ep->desc) ||
+			usb_endpoint_xfer_bulk(&ep->desc))
+		return 0;
+
+	if (udev->speed == USB_SPEED_SUPER)
+		return le16_to_cpu(ep->ss_ep_comp.wBytesPerInterval);
+
+	max_packet = GET_MAX_PACKET(usb_endpoint_maxp(&ep->desc));
+	max_burst = (usb_endpoint_maxp(&ep->desc) & 0x1800) >> 11;
+	/* A 0 in max burst means 1 transfer per ESIT */
+	return max_packet * (max_burst + 1);
+}
+
+/* Set up an endpoint with one ring segment.  Do not allocate stream rings.
+ * Drivers will have to call usb_alloc_streams() to do that.
+ */
+int xhci_endpoint_init(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		struct usb_device *udev,
+		struct usb_host_endpoint *ep,
+		gfp_t mem_flags)
+{
+	unsigned int ep_index;
+	struct xhci_ep_ctx *ep_ctx;
+	struct xhci_ring *ep_ring;
+	unsigned int max_packet;
+	unsigned int max_burst;
+	enum xhci_ring_type type;
+	u32 max_esit_payload;
+
+	ep_index = xhci_get_endpoint_index(&ep->desc);
+	ep_ctx = xhci_get_ep_ctx(xhci, virt_dev->in_ctx, ep_index);
+
+	type = usb_endpoint_type(&ep->desc);
+	/* Set up the endpoint ring */
+	virt_dev->eps[ep_index].new_ring =
+		xhci_ring_alloc(xhci, 2, 1, type, mem_flags);
+	if (!virt_dev->eps[ep_index].new_ring) {
+		/* Attempt to use the ring cache */
+		if (virt_dev->num_rings_cached == 0)
+			return -ENOMEM;
+		virt_dev->eps[ep_index].new_ring =
+			virt_dev->ring_cache[virt_dev->num_rings_cached];
+		virt_dev->ring_cache[virt_dev->num_rings_cached] = NULL;
+		virt_dev->num_rings_cached--;
+		xhci_reinit_cached_ring(xhci, virt_dev->eps[ep_index].new_ring,
+					1, type);
+	}
+	virt_dev->eps[ep_index].skip = false;
+	ep_ring = virt_dev->eps[ep_index].new_ring;
+	ep_ctx->deq = cpu_to_le64(ep_ring->first_seg->dma | ep_ring->cycle_state);
+
+	ep_ctx->ep_info = cpu_to_le32(xhci_get_endpoint_interval(udev, ep)
+				      | EP_MULT(xhci_get_endpoint_mult(udev, ep)));
+
+	/* FIXME dig Mult and streams info out of ep companion desc */
+
+	/* Allow 3 retries for everything but isoc;
+	 * CErr shall be set to 0 for Isoch endpoints.
+	 */
+	if (!usb_endpoint_xfer_isoc(&ep->desc))
+		ep_ctx->ep_info2 = cpu_to_le32(ERROR_COUNT(3));
+	else
+		ep_ctx->ep_info2 = cpu_to_le32(ERROR_COUNT(0));
+
+	ep_ctx->ep_info2 |= cpu_to_le32(xhci_get_endpoint_type(udev, ep));
+
+	/* Set the max packet size and max burst */
+	max_packet = GET_MAX_PACKET(usb_endpoint_maxp(&ep->desc));
+	max_burst = 0;
+	switch (udev->speed) {
+	case USB_SPEED_SUPER:
+		/* dig out max burst from ep companion desc */
+		max_burst = ep->ss_ep_comp.bMaxBurst;
+		break;
+	case USB_SPEED_HIGH:
+		/* Some devices get this wrong */
+		if (usb_endpoint_xfer_bulk(&ep->desc))
+			max_packet = 512;
+		/* bits 11:12 specify the number of additional transaction
+		 * opportunities per microframe (USB 2.0, section 9.6.6)
+		 */
+		if (usb_endpoint_xfer_isoc(&ep->desc) ||
+				usb_endpoint_xfer_int(&ep->desc)) {
+			max_burst = (usb_endpoint_maxp(&ep->desc)
+				     & 0x1800) >> 11;
+		}
+		break;
+	case USB_SPEED_FULL:
+	case USB_SPEED_LOW:
+		break;
+	default:
+		BUG();
+	}
+	ep_ctx->ep_info2 |= cpu_to_le32(MAX_PACKET(max_packet) |
+			MAX_BURST(max_burst));
+	max_esit_payload = xhci_get_max_esit_payload(xhci, udev, ep);
+	ep_ctx->tx_info = cpu_to_le32(MAX_ESIT_PAYLOAD_FOR_EP(max_esit_payload));
+
+	/*
+	 * XXX no idea how to calculate the average TRB buffer length for bulk
+	 * endpoints, as the driver gives us no clue how big each scatter gather
+	 * list entry (or buffer) is going to be.
+	 *
+	 * For isochronous and interrupt endpoints, we set it to the max
+	 * available, until we have new API in the USB core to allow drivers to
+	 * declare how much bandwidth they actually need.
+	 *
+	 * Normally, it would be calculated by taking the total of the buffer
+	 * lengths in the TD and then dividing by the number of TRBs in a TD,
+	 * including link TRBs, No-op TRBs, and Event data TRBs.  Since we don't
+	 * use Event Data TRBs, and we don't chain in a link TRB on short
+	 * transfers, we're basically dividing by 1.
+	 *
+	 * xHCI 1.0 specification indicates that the Average TRB Length should
+	 * be set to 8 for control endpoints.
+	 */
+	if (usb_endpoint_xfer_control(&ep->desc) && xhci->hci_version == 0x100)
+		ep_ctx->tx_info |= cpu_to_le32(AVG_TRB_LENGTH_FOR_EP(8));
+	else
+		ep_ctx->tx_info |=
+			 cpu_to_le32(AVG_TRB_LENGTH_FOR_EP(max_esit_payload));
+
+	/* FIXME Debug endpoint context */
+	return 0;
+}
+
+void xhci_endpoint_zero(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		struct usb_host_endpoint *ep)
+{
+	unsigned int ep_index;
+	struct xhci_ep_ctx *ep_ctx;
+
+	ep_index = xhci_get_endpoint_index(&ep->desc);
+	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;
+	ep_ctx->tx_info = 0;
+	/* Don't free the endpoint ring until the set interface or configuration
+	 * request succeeds.
+	 */
+}
+
+void xhci_clear_endpoint_bw_info(struct xhci_bw_info *bw_info)
+{
+	bw_info->ep_interval = 0;
+	bw_info->mult = 0;
+	bw_info->num_packets = 0;
+	bw_info->max_packet_size = 0;
+	bw_info->type = 0;
+	bw_info->max_esit_payload = 0;
+}
+
+void xhci_update_bw_info(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx,
+		struct xhci_input_control_ctx *ctrl_ctx,
+		struct xhci_virt_device *virt_dev)
+{
+	struct xhci_bw_info *bw_info;
+	struct xhci_ep_ctx *ep_ctx;
+	unsigned int ep_type;
+	int i;
+
+	for (i = 1; i < 31; ++i) {
+		bw_info = &virt_dev->eps[i].bw_info;
+
+		/* We can't tell what endpoint type is being dropped, but
+		 * unconditionally clearing the bandwidth info for non-periodic
+		 * endpoints should be harmless because the info will never be
+		 * set in the first place.
+		 */
+		if (!EP_IS_ADDED(ctrl_ctx, i) && EP_IS_DROPPED(ctrl_ctx, i)) {
+			/* Dropped endpoint */
+			xhci_clear_endpoint_bw_info(bw_info);
+			continue;
+		}
+
+		if (EP_IS_ADDED(ctrl_ctx, i)) {
+			ep_ctx = xhci_get_ep_ctx(xhci, in_ctx, i);
+			ep_type = CTX_TO_EP_TYPE(le32_to_cpu(ep_ctx->ep_info2));
+
+			/* Ignore non-periodic endpoints */
+			if (ep_type != ISOC_OUT_EP && ep_type != INT_OUT_EP &&
+					ep_type != ISOC_IN_EP &&
+					ep_type != INT_IN_EP)
+				continue;
+
+			/* Added or changed endpoint */
+			bw_info->ep_interval = CTX_TO_EP_INTERVAL(
+					le32_to_cpu(ep_ctx->ep_info));
+			/* Number of packets and mult are zero-based in the
+			 * input context, but we want one-based for the
+			 * interval table.
+			 */
+			bw_info->mult = CTX_TO_EP_MULT(
+					le32_to_cpu(ep_ctx->ep_info)) + 1;
+			bw_info->num_packets = CTX_TO_MAX_BURST(
+					le32_to_cpu(ep_ctx->ep_info2)) + 1;
+			bw_info->max_packet_size = MAX_PACKET_DECODED(
+					le32_to_cpu(ep_ctx->ep_info2));
+			bw_info->type = ep_type;
+			bw_info->max_esit_payload = CTX_TO_MAX_ESIT_PAYLOAD(
+					le32_to_cpu(ep_ctx->tx_info));
+		}
+	}
+}
+
+/* Copy output xhci_ep_ctx to the input xhci_ep_ctx copy.
+ * Useful when you want to change one particular aspect of the endpoint and then
+ * issue a configure endpoint command.
+ */
+void xhci_endpoint_copy(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx,
+		struct xhci_container_ctx *out_ctx,
+		unsigned int ep_index)
+{
+	struct xhci_ep_ctx *out_ep_ctx;
+	struct xhci_ep_ctx *in_ep_ctx;
+
+	out_ep_ctx = xhci_get_ep_ctx(xhci, out_ctx, ep_index);
+	in_ep_ctx = xhci_get_ep_ctx(xhci, in_ctx, ep_index);
+
+	in_ep_ctx->ep_info = out_ep_ctx->ep_info;
+	in_ep_ctx->ep_info2 = out_ep_ctx->ep_info2;
+	in_ep_ctx->deq = out_ep_ctx->deq;
+	in_ep_ctx->tx_info = out_ep_ctx->tx_info;
+}
+
+/* Copy output xhci_slot_ctx to the input xhci_slot_ctx.
+ * Useful when you want to change one particular aspect of the endpoint and then
+ * issue a configure endpoint command.  Only the context entries field matters,
+ * but we'll copy the whole thing anyway.
+ */
+void xhci_slot_copy(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx,
+		struct xhci_container_ctx *out_ctx)
+{
+	struct xhci_slot_ctx *in_slot_ctx;
+	struct xhci_slot_ctx *out_slot_ctx;
+
+	in_slot_ctx = xhci_get_slot_ctx(xhci, in_ctx);
+	out_slot_ctx = xhci_get_slot_ctx(xhci, out_ctx);
+
+	in_slot_ctx->dev_info = out_slot_ctx->dev_info;
+	in_slot_ctx->dev_info2 = out_slot_ctx->dev_info2;
+	in_slot_ctx->tt_info = out_slot_ctx->tt_info;
+	in_slot_ctx->dev_state = out_slot_ctx->dev_state;
+}
+
+/* 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 = dma_alloc_coherent(dev,
+				     num_sp * sizeof(u64),
+				     &xhci->scratchpad->sp_dma, flags);
+	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] = cpu_to_le64(xhci->scratchpad->sp_dma);
+	for (i = 0; i < num_sp; i++) {
+		dma_addr_t dma;
+		void *buf = dma_alloc_coherent(dev, xhci->page_size, &dma,
+				flags);
+		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--) {
+		dma_free_coherent(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:
+	dma_free_coherent(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++) {
+		dma_free_coherent(&pdev->dev, 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);
+	dma_free_coherent(&pdev->dev, num_sp * sizeof(u64),
+			    xhci->scratchpad->sp_array,
+			    xhci->scratchpad->sp_dma);
+	kfree(xhci->scratchpad);
+	xhci->scratchpad = NULL;
+}
+
+struct xhci_command *xhci_alloc_command(struct xhci_hcd *xhci,
+		bool allocate_in_ctx, bool allocate_completion,
+		gfp_t mem_flags)
+{
+	struct xhci_command *command;
+
+	command = kzalloc(sizeof(*command), mem_flags);
+	if (!command)
+		return NULL;
+
+	if (allocate_in_ctx) {
+		command->in_ctx =
+			xhci_alloc_container_ctx(xhci, XHCI_CTX_TYPE_INPUT,
+					mem_flags);
+		if (!command->in_ctx) {
+			kfree(command);
+			return NULL;
+		}
+	}
+
+	if (allocate_completion) {
+		command->completion =
+			kzalloc(sizeof(struct completion), mem_flags);
+		if (!command->completion) {
+			xhci_free_container_ctx(xhci, command->in_ctx);
+			kfree(command);
+			return NULL;
+		}
+		init_completion(command->completion);
+	}
+
+	command->status = 0;
+	INIT_LIST_HEAD(&command->cmd_list);
+	return command;
+}
+
+void xhci_urb_free_priv(struct xhci_hcd *xhci, struct urb_priv *urb_priv)
+{
+	if (urb_priv) {
+		kfree(urb_priv->td[0]);
+		kfree(urb_priv);
+	}
+}
+
+void xhci_free_command(struct xhci_hcd *xhci,
+		struct xhci_command *command)
+{
+	xhci_free_container_ctx(xhci,
+			command->in_ctx);
+	kfree(command->completion);
+	kfree(command);
+}
+
+void xhci_mem_cleanup(struct xhci_hcd *xhci)
+{
+	struct pci_dev	*pdev = to_pci_dev(xhci_to_hcd(xhci)->self.controller);
+	struct dev_info	*dev_info, *next;
+	struct xhci_cd  *cur_cd, *next_cd;
+	unsigned long	flags;
+	int size;
+	int i, j, num_ports;
+
+	/* Free the Event Ring Segment Table and the actual Event Ring */
+	size = sizeof(struct xhci_erst_entry)*(xhci->erst.num_entries);
+	if (xhci->erst.entries)
+		dma_free_coherent(&pdev->dev, size,
+				xhci->erst.entries, xhci->erst.erst_dma_addr);
+	xhci->erst.entries = NULL;
+	xhci_dbg(xhci, "Freed ERST\n");
+	if (xhci->event_ring)
+		xhci_ring_free(xhci, xhci->event_ring);
+	xhci->event_ring = NULL;
+	xhci_dbg(xhci, "Freed event ring\n");
+
+	if (xhci->lpm_command)
+		xhci_free_command(xhci, xhci->lpm_command);
+	xhci->cmd_ring_reserved_trbs = 0;
+	if (xhci->cmd_ring)
+		xhci_ring_free(xhci, xhci->cmd_ring);
+	xhci->cmd_ring = NULL;
+	xhci_dbg(xhci, "Freed command ring\n");
+	list_for_each_entry_safe(cur_cd, next_cd,
+			&xhci->cancel_cmd_list, cancel_cmd_list) {
+		list_del(&cur_cd->cancel_cmd_list);
+		kfree(cur_cd);
+	}
+
+	for (i = 1; i < MAX_HC_SLOTS; ++i)
+		xhci_free_virt_device(xhci, i);
+
+	if (xhci->segment_pool)
+		dma_pool_destroy(xhci->segment_pool);
+	xhci->segment_pool = NULL;
+	xhci_dbg(xhci, "Freed segment pool\n");
+
+	if (xhci->device_pool)
+		dma_pool_destroy(xhci->device_pool);
+	xhci->device_pool = NULL;
+	xhci_dbg(xhci, "Freed device context pool\n");
+
+	if (xhci->small_streams_pool)
+		dma_pool_destroy(xhci->small_streams_pool);
+	xhci->small_streams_pool = NULL;
+	xhci_dbg(xhci, "Freed small stream array pool\n");
+
+	if (xhci->medium_streams_pool)
+		dma_pool_destroy(xhci->medium_streams_pool);
+	xhci->medium_streams_pool = NULL;
+	xhci_dbg(xhci, "Freed medium stream array pool\n");
+
+	if (xhci->dcbaa)
+		dma_free_coherent(&pdev->dev, sizeof(*xhci->dcbaa),
+				xhci->dcbaa, xhci->dcbaa->dma);
+	xhci->dcbaa = NULL;
+
+	scratchpad_free(xhci);
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	list_for_each_entry_safe(dev_info, next, &xhci->lpm_failed_devs, list) {
+		list_del(&dev_info->list);
+		kfree(dev_info);
+	}
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	num_ports = HCS_MAX_PORTS(xhci->hcs_params1);
+	for (i = 0; i < num_ports; i++) {
+		struct xhci_interval_bw_table *bwt = &xhci->rh_bw[i].bw_table;
+		for (j = 0; j < XHCI_MAX_INTERVAL; j++) {
+			struct list_head *ep = &bwt->interval_bw[j].endpoints;
+			while (!list_empty(ep))
+				list_del_init(ep->next);
+		}
+	}
+
+	for (i = 0; i < num_ports; i++) {
+		struct xhci_tt_bw_info *tt, *n;
+		list_for_each_entry_safe(tt, n, &xhci->rh_bw[i].tts, tt_list) {
+			list_del(&tt->tt_list);
+			kfree(tt);
+		}
+	}
+
+	xhci->num_usb2_ports = 0;
+	xhci->num_usb3_ports = 0;
+	xhci->num_active_eps = 0;
+	kfree(xhci->usb2_ports);
+	kfree(xhci->usb3_ports);
+	kfree(xhci->port_array);
+	kfree(xhci->rh_bw);
+
+	xhci->page_size = 0;
+	xhci->page_shift = 0;
+	xhci->bus_state[0].bus_suspended = 0;
+	xhci->bus_state[1].bus_suspended = 0;
+}
+
+static int xhci_test_trb_in_td(struct xhci_hcd *xhci,
+		struct xhci_segment *input_seg,
+		union xhci_trb *start_trb,
+		union xhci_trb *end_trb,
+		dma_addr_t input_dma,
+		struct xhci_segment *result_seg,
+		char *test_name, int test_number)
+{
+	unsigned long long start_dma;
+	unsigned long long end_dma;
+	struct xhci_segment *seg;
+
+	start_dma = xhci_trb_virt_to_dma(input_seg, start_trb);
+	end_dma = xhci_trb_virt_to_dma(input_seg, end_trb);
+
+	seg = trb_in_td(input_seg, start_trb, end_trb, input_dma);
+	if (seg != result_seg) {
+		xhci_warn(xhci, "WARN: %s TRB math test %d failed!\n",
+				test_name, test_number);
+		xhci_warn(xhci, "Tested TRB math w/ seg %p and "
+				"input DMA 0x%llx\n",
+				input_seg,
+				(unsigned long long) input_dma);
+		xhci_warn(xhci, "starting TRB %p (0x%llx DMA), "
+				"ending TRB %p (0x%llx DMA)\n",
+				start_trb, start_dma,
+				end_trb, end_dma);
+		xhci_warn(xhci, "Expected seg %p, got seg %p\n",
+				result_seg, seg);
+		return -1;
+	}
+	return 0;
+}
+
+/* TRB math checks for xhci_trb_in_td(), using the command and event rings. */
+static int xhci_check_trb_in_td_math(struct xhci_hcd *xhci, gfp_t mem_flags)
+{
+	struct {
+		dma_addr_t		input_dma;
+		struct xhci_segment	*result_seg;
+	} simple_test_vector [] = {
+		/* A zeroed DMA field should fail */
+		{ 0, NULL },
+		/* One TRB before the ring start should fail */
+		{ xhci->event_ring->first_seg->dma - 16, NULL },
+		/* One byte before the ring start should fail */
+		{ xhci->event_ring->first_seg->dma - 1, NULL },
+		/* Starting TRB should succeed */
+		{ xhci->event_ring->first_seg->dma, xhci->event_ring->first_seg },
+		/* Ending TRB should succeed */
+		{ xhci->event_ring->first_seg->dma + (TRBS_PER_SEGMENT - 1)*16,
+			xhci->event_ring->first_seg },
+		/* One byte after the ring end should fail */
+		{ xhci->event_ring->first_seg->dma + (TRBS_PER_SEGMENT - 1)*16 + 1, NULL },
+		/* One TRB after the ring end should fail */
+		{ xhci->event_ring->first_seg->dma + (TRBS_PER_SEGMENT)*16, NULL },
+		/* An address of all ones should fail */
+		{ (dma_addr_t) (~0), NULL },
+	};
+	struct {
+		struct xhci_segment	*input_seg;
+		union xhci_trb		*start_trb;
+		union xhci_trb		*end_trb;
+		dma_addr_t		input_dma;
+		struct xhci_segment	*result_seg;
+	} complex_test_vector [] = {
+		/* Test feeding a valid DMA address from a different ring */
+		{	.input_seg = xhci->event_ring->first_seg,
+			.start_trb = xhci->event_ring->first_seg->trbs,
+			.end_trb = &xhci->event_ring->first_seg->trbs[TRBS_PER_SEGMENT - 1],
+			.input_dma = xhci->cmd_ring->first_seg->dma,
+			.result_seg = NULL,
+		},
+		/* Test feeding a valid end TRB from a different ring */
+		{	.input_seg = xhci->event_ring->first_seg,
+			.start_trb = xhci->event_ring->first_seg->trbs,
+			.end_trb = &xhci->cmd_ring->first_seg->trbs[TRBS_PER_SEGMENT - 1],
+			.input_dma = xhci->cmd_ring->first_seg->dma,
+			.result_seg = NULL,
+		},
+		/* Test feeding a valid start and end TRB from a different ring */
+		{	.input_seg = xhci->event_ring->first_seg,
+			.start_trb = xhci->cmd_ring->first_seg->trbs,
+			.end_trb = &xhci->cmd_ring->first_seg->trbs[TRBS_PER_SEGMENT - 1],
+			.input_dma = xhci->cmd_ring->first_seg->dma,
+			.result_seg = NULL,
+		},
+		/* TRB in this ring, but after this TD */
+		{	.input_seg = xhci->event_ring->first_seg,
+			.start_trb = &xhci->event_ring->first_seg->trbs[0],
+			.end_trb = &xhci->event_ring->first_seg->trbs[3],
+			.input_dma = xhci->event_ring->first_seg->dma + 4*16,
+			.result_seg = NULL,
+		},
+		/* TRB in this ring, but before this TD */
+		{	.input_seg = xhci->event_ring->first_seg,
+			.start_trb = &xhci->event_ring->first_seg->trbs[3],
+			.end_trb = &xhci->event_ring->first_seg->trbs[6],
+			.input_dma = xhci->event_ring->first_seg->dma + 2*16,
+			.result_seg = NULL,
+		},
+		/* TRB in this ring, but after this wrapped TD */
+		{	.input_seg = xhci->event_ring->first_seg,
+			.start_trb = &xhci->event_ring->first_seg->trbs[TRBS_PER_SEGMENT - 3],
+			.end_trb = &xhci->event_ring->first_seg->trbs[1],
+			.input_dma = xhci->event_ring->first_seg->dma + 2*16,
+			.result_seg = NULL,
+		},
+		/* TRB in this ring, but before this wrapped TD */
+		{	.input_seg = xhci->event_ring->first_seg,
+			.start_trb = &xhci->event_ring->first_seg->trbs[TRBS_PER_SEGMENT - 3],
+			.end_trb = &xhci->event_ring->first_seg->trbs[1],
+			.input_dma = xhci->event_ring->first_seg->dma + (TRBS_PER_SEGMENT - 4)*16,
+			.result_seg = NULL,
+		},
+		/* TRB not in this ring, and we have a wrapped TD */
+		{	.input_seg = xhci->event_ring->first_seg,
+			.start_trb = &xhci->event_ring->first_seg->trbs[TRBS_PER_SEGMENT - 3],
+			.end_trb = &xhci->event_ring->first_seg->trbs[1],
+			.input_dma = xhci->cmd_ring->first_seg->dma + 2*16,
+			.result_seg = NULL,
+		},
+	};
+
+	unsigned int num_tests;
+	int i, ret;
+
+	num_tests = ARRAY_SIZE(simple_test_vector);
+	for (i = 0; i < num_tests; i++) {
+		ret = xhci_test_trb_in_td(xhci,
+				xhci->event_ring->first_seg,
+				xhci->event_ring->first_seg->trbs,
+				&xhci->event_ring->first_seg->trbs[TRBS_PER_SEGMENT - 1],
+				simple_test_vector[i].input_dma,
+				simple_test_vector[i].result_seg,
+				"Simple", i);
+		if (ret < 0)
+			return ret;
+	}
+
+	num_tests = ARRAY_SIZE(complex_test_vector);
+	for (i = 0; i < num_tests; i++) {
+		ret = xhci_test_trb_in_td(xhci,
+				complex_test_vector[i].input_seg,
+				complex_test_vector[i].start_trb,
+				complex_test_vector[i].end_trb,
+				complex_test_vector[i].input_dma,
+				complex_test_vector[i].result_seg,
+				"Complex", i);
+		if (ret < 0)
+			return ret;
+	}
+	xhci_dbg(xhci, "TRB math tests passed.\n");
+	return 0;
+}
+
+static void xhci_set_hc_event_deq(struct xhci_hcd *xhci)
+{
+	u64 temp;
+	dma_addr_t deq;
+
+	deq = xhci_trb_virt_to_dma(xhci->event_ring->deq_seg,
+			xhci->event_ring->dequeue);
+	if (deq == 0 && !in_interrupt())
+		xhci_warn(xhci, "WARN something wrong with SW event ring "
+				"dequeue ptr.\n");
+	/* Update HC event ring dequeue pointer */
+	temp = xhci_read_64(xhci, &xhci->ir_set->erst_dequeue);
+	temp &= ERST_PTR_MASK;
+	/* 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);
+}
+
+static void xhci_add_in_port(struct xhci_hcd *xhci, unsigned int num_ports,
+		__le32 __iomem *addr, u8 major_revision)
+{
+	u32 temp, port_offset, port_count;
+	int i;
+
+	if (major_revision > 0x03) {
+		xhci_warn(xhci, "Ignoring unknown port speed, "
+				"Ext Cap %p, revision = 0x%x\n",
+				addr, major_revision);
+		/* Ignoring port protocol we can't understand. FIXME */
+		return;
+	}
+
+	/* Port offset and count in the third dword, see section 7.2 */
+	temp = xhci_readl(xhci, addr + 2);
+	port_offset = XHCI_EXT_PORT_OFF(temp);
+	port_count = XHCI_EXT_PORT_COUNT(temp);
+	xhci_dbg(xhci, "Ext Cap %p, port offset = %u, "
+			"count = %u, revision = 0x%x\n",
+			addr, port_offset, port_count, major_revision);
+	/* Port count includes the current port offset */
+	if (port_offset == 0 || (port_offset + port_count - 1) > num_ports)
+		/* WTF? "Valid values are ‘1’ to MaxPorts" */
+		return;
+
+	/* Check the host's USB2 LPM capability */
+	if ((xhci->hci_version == 0x96) && (major_revision != 0x03) &&
+			(temp & XHCI_L1C)) {
+		xhci_dbg(xhci, "xHCI 0.96: support USB2 software lpm\n");
+		xhci->sw_lpm_support = 1;
+	}
+
+	if ((xhci->hci_version >= 0x100) && (major_revision != 0x03)) {
+		xhci_dbg(xhci, "xHCI 1.0: support USB2 software lpm\n");
+		xhci->sw_lpm_support = 1;
+		if (temp & XHCI_HLC) {
+			xhci_dbg(xhci, "xHCI 1.0: support USB2 hardware lpm\n");
+			xhci->hw_lpm_support = 1;
+		}
+	}
+
+	port_offset--;
+	for (i = port_offset; i < (port_offset + port_count); i++) {
+		/* Duplicate entry.  Ignore the port if the revisions differ. */
+		if (xhci->port_array[i] != 0) {
+			xhci_warn(xhci, "Duplicate port entry, Ext Cap %p,"
+					" port %u\n", addr, i);
+			xhci_warn(xhci, "Port was marked as USB %u, "
+					"duplicated as USB %u\n",
+					xhci->port_array[i], major_revision);
+			/* Only adjust the roothub port counts if we haven't
+			 * found a similar duplicate.
+			 */
+			if (xhci->port_array[i] != major_revision &&
+				xhci->port_array[i] != DUPLICATE_ENTRY) {
+				if (xhci->port_array[i] == 0x03)
+					xhci->num_usb3_ports--;
+				else
+					xhci->num_usb2_ports--;
+				xhci->port_array[i] = DUPLICATE_ENTRY;
+			}
+			/* FIXME: Should we disable the port? */
+			continue;
+		}
+		xhci->port_array[i] = major_revision;
+		if (major_revision == 0x03)
+			xhci->num_usb3_ports++;
+		else
+			xhci->num_usb2_ports++;
+	}
+	/* FIXME: Should we disable ports not in the Extended Capabilities? */
+}
+
+/*
+ * Scan the Extended Capabilities for the "Supported Protocol Capabilities" that
+ * specify what speeds each port is supposed to be.  We can't count on the port
+ * speed bits in the PORTSC register being correct until a device is connected,
+ * but we need to set up the two fake roothubs with the correct number of USB
+ * 3.0 and USB 2.0 ports at host controller initialization time.
+ */
+static int xhci_setup_port_arrays(struct xhci_hcd *xhci, gfp_t flags)
+{
+	__le32 __iomem *addr;
+	u32 offset;
+	unsigned int num_ports;
+	int i, j, port_index;
+
+	addr = &xhci->cap_regs->hcc_params;
+	offset = XHCI_HCC_EXT_CAPS(xhci_readl(xhci, addr));
+	if (offset == 0) {
+		xhci_err(xhci, "No Extended Capability registers, "
+				"unable to set up roothub.\n");
+		return -ENODEV;
+	}
+
+	num_ports = HCS_MAX_PORTS(xhci->hcs_params1);
+	xhci->port_array = kzalloc(sizeof(*xhci->port_array)*num_ports, flags);
+	if (!xhci->port_array)
+		return -ENOMEM;
+
+	xhci->rh_bw = kzalloc(sizeof(*xhci->rh_bw)*num_ports, flags);
+	if (!xhci->rh_bw)
+		return -ENOMEM;
+	for (i = 0; i < num_ports; i++) {
+		struct xhci_interval_bw_table *bw_table;
+
+		INIT_LIST_HEAD(&xhci->rh_bw[i].tts);
+		bw_table = &xhci->rh_bw[i].bw_table;
+		for (j = 0; j < XHCI_MAX_INTERVAL; j++)
+			INIT_LIST_HEAD(&bw_table->interval_bw[j].endpoints);
+	}
+
+	/*
+	 * For whatever reason, the first capability offset is from the
+	 * capability register base, not from the HCCPARAMS register.
+	 * See section 5.3.6 for offset calculation.
+	 */
+	addr = &xhci->cap_regs->hc_capbase + offset;
+	while (1) {
+		u32 cap_id;
+
+		cap_id = xhci_readl(xhci, addr);
+		if (XHCI_EXT_CAPS_ID(cap_id) == XHCI_EXT_CAPS_PROTOCOL)
+			xhci_add_in_port(xhci, num_ports, addr,
+					(u8) XHCI_EXT_PORT_MAJOR(cap_id));
+		offset = XHCI_EXT_CAPS_NEXT(cap_id);
+		if (!offset || (xhci->num_usb2_ports + xhci->num_usb3_ports)
+				== num_ports)
+			break;
+		/*
+		 * Once you're into the Extended Capabilities, the offset is
+		 * always relative to the register holding the offset.
+		 */
+		addr += offset;
+	}
+
+	if (xhci->num_usb2_ports == 0 && xhci->num_usb3_ports == 0) {
+		xhci_warn(xhci, "No ports on the roothubs?\n");
+		return -ENODEV;
+	}
+	xhci_dbg(xhci, "Found %u USB 2.0 ports and %u USB 3.0 ports.\n",
+			xhci->num_usb2_ports, xhci->num_usb3_ports);
+
+	/* Place limits on the number of roothub ports so that the hub
+	 * descriptors aren't longer than the USB core will allocate.
+	 */
+	if (xhci->num_usb3_ports > 15) {
+		xhci_dbg(xhci, "Limiting USB 3.0 roothub ports to 15.\n");
+		xhci->num_usb3_ports = 15;
+	}
+	if (xhci->num_usb2_ports > USB_MAXCHILDREN) {
+		xhci_dbg(xhci, "Limiting USB 2.0 roothub ports to %u.\n",
+				USB_MAXCHILDREN);
+		xhci->num_usb2_ports = USB_MAXCHILDREN;
+	}
+
+	/*
+	 * Note we could have all USB 3.0 ports, or all USB 2.0 ports.
+	 * Not sure how the USB core will handle a hub with no ports...
+	 */
+	if (xhci->num_usb2_ports) {
+		xhci->usb2_ports = kmalloc(sizeof(*xhci->usb2_ports)*
+				xhci->num_usb2_ports, flags);
+		if (!xhci->usb2_ports)
+			return -ENOMEM;
+
+		port_index = 0;
+		for (i = 0; i < num_ports; i++) {
+			if (xhci->port_array[i] == 0x03 ||
+					xhci->port_array[i] == 0 ||
+					xhci->port_array[i] == DUPLICATE_ENTRY)
+				continue;
+
+			xhci->usb2_ports[port_index] =
+				&xhci->op_regs->port_status_base +
+				NUM_PORT_REGS*i;
+			xhci_dbg(xhci, "USB 2.0 port at index %u, "
+					"addr = %p\n", i,
+					xhci->usb2_ports[port_index]);
+			port_index++;
+			if (port_index == xhci->num_usb2_ports)
+				break;
+		}
+	}
+	if (xhci->num_usb3_ports) {
+		xhci->usb3_ports = kmalloc(sizeof(*xhci->usb3_ports)*
+				xhci->num_usb3_ports, flags);
+		if (!xhci->usb3_ports)
+			return -ENOMEM;
+
+		port_index = 0;
+		for (i = 0; i < num_ports; i++)
+			if (xhci->port_array[i] == 0x03) {
+				xhci->usb3_ports[port_index] =
+					&xhci->op_regs->port_status_base +
+					NUM_PORT_REGS*i;
+				xhci_dbg(xhci, "USB 3.0 port at index %u, "
+						"addr = %p\n", i,
+						xhci->usb3_ports[port_index]);
+				port_index++;
+				if (port_index == xhci->num_usb3_ports)
+					break;
+			}
+	}
+	return 0;
+}
+
+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, temp;
+	int i;
+
+	page_size = xhci_readl(xhci, &xhci->op_regs->page_size);
+	xhci_dbg(xhci, "Supported page size register = 0x%x\n", page_size);
+	for (i = 0; i < 16; i++) {
+		if ((0x1 & page_size) != 0)
+			break;
+		page_size = page_size >> 1;
+	}
+	if (i < 16)
+		xhci_dbg(xhci, "Supported page size of %iK\n", (1 << (i+12)) / 1024);
+	else
+		xhci_warn(xhci, "WARN: no supported page size\n");
+	/* Use 4K pages, since that's common and the minimum the HC supports */
+	xhci->page_shift = 12;
+	xhci->page_size = 1 << xhci->page_shift;
+	xhci_dbg(xhci, "HCD page size set to %iK\n", xhci->page_size / 1024);
+
+	/*
+	 * Program the Number of Device Slots Enabled field in the CONFIG
+	 * register with the max value of slots the HC can handle.
+	 */
+	val = HCS_MAX_SLOTS(xhci_readl(xhci, &xhci->cap_regs->hcs_params1));
+	xhci_dbg(xhci, "// xHC can handle at most %d device slots.\n",
+			(unsigned int) val);
+	val2 = xhci_readl(xhci, &xhci->op_regs->config_reg);
+	val |= (val2 & ~HCS_SLOTS_MASK);
+	xhci_dbg(xhci, "// Setting Max device slots reg = 0x%x.\n",
+			(unsigned int) val);
+	xhci_writel(xhci, val, &xhci->op_regs->config_reg);
+
+	/*
+	 * Section 5.4.8 - doorbell array must be
+	 * "physically contiguous and 64-byte (cache line) aligned".
+	 */
+	xhci->dcbaa = dma_alloc_coherent(dev, sizeof(*xhci->dcbaa), &dma,
+			GFP_KERNEL);
+	if (!xhci->dcbaa)
+		goto fail;
+	memset(xhci->dcbaa, 0, sizeof *(xhci->dcbaa));
+	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_write_64(xhci, dma, &xhci->op_regs->dcbaa_ptr);
+
+	/*
+	 * Initialize the ring segment pool.  The ring must be a contiguous
+	 * structure comprised of TRBs.  The TRBs must be 16 byte aligned,
+	 * however, the command ring segment needs 64-byte aligned segments,
+	 * so we pick the greater alignment need.
+	 */
+	xhci->segment_pool = dma_pool_create("xHCI ring segments", dev,
+			TRB_SEGMENT_SIZE, 64, xhci->page_size);
+
+	/* See Table 46 and Note on Figure 55 */
+	xhci->device_pool = dma_pool_create("xHCI input/output contexts", dev,
+			2112, 64, xhci->page_size);
+	if (!xhci->segment_pool || !xhci->device_pool)
+		goto fail;
+
+	/* Linear stream context arrays don't have any boundary restrictions,
+	 * and only need to be 16-byte aligned.
+	 */
+	xhci->small_streams_pool =
+		dma_pool_create("xHCI 256 byte stream ctx arrays",
+			dev, SMALL_STREAM_ARRAY_SIZE, 16, 0);
+	xhci->medium_streams_pool =
+		dma_pool_create("xHCI 1KB stream ctx arrays",
+			dev, MEDIUM_STREAM_ARRAY_SIZE, 16, 0);
+	/* Any stream context array bigger than MEDIUM_STREAM_ARRAY_SIZE
+	 * will be allocated with dma_alloc_coherent()
+	 */
+
+	if (!xhci->small_streams_pool || !xhci->medium_streams_pool)
+		goto fail;
+
+	/* Set up the command ring to have one segments for now. */
+	xhci->cmd_ring = xhci_ring_alloc(xhci, 1, 1, TYPE_COMMAND, flags);
+	if (!xhci->cmd_ring)
+		goto fail;
+	INIT_LIST_HEAD(&xhci->cancel_cmd_list);
+	xhci_dbg(xhci, "Allocated command ring at %p\n", xhci->cmd_ring);
+	xhci_dbg(xhci, "First segment DMA is 0x%llx\n",
+			(unsigned long long)xhci->cmd_ring->first_seg->dma);
+
+	/* Set the address in the Command Ring Control register */
+	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 to 0x%x\n", val);
+	xhci_write_64(xhci, val_64, &xhci->op_regs->cmd_ring);
+	xhci_dbg_cmd_ptrs(xhci);
+
+	xhci->lpm_command = xhci_alloc_command(xhci, true, true, flags);
+	if (!xhci->lpm_command)
+		goto fail;
+
+	/* Reserve one command ring TRB for disabling LPM.
+	 * Since the USB core grabs the shared usb_bus bandwidth mutex before
+	 * disabling LPM, we only need to reserve one TRB for all devices.
+	 */
+	xhci->cmd_ring_reserved_trbs++;
+
+	val = xhci_readl(xhci, &xhci->cap_regs->db_off);
+	val &= DBOFF_MASK;
+	xhci_dbg(xhci, "// Doorbell array is located at offset 0x%x"
+			" from cap regs base addr\n", val);
+	xhci->dba = (void __iomem *) xhci->cap_regs + val;
+	xhci_dbg_regs(xhci);
+	xhci_print_run_regs(xhci);
+	/* Set ir_set to interrupt register set 0 */
+	xhci->ir_set = &xhci->run_regs->ir_set[0];
+
+	/*
+	 * Event ring setup: Allocate a normal ring, but also setup
+	 * the event ring segment table (ERST).  Section 4.9.3.
+	 */
+	xhci_dbg(xhci, "// Allocating event ring\n");
+	xhci->event_ring = xhci_ring_alloc(xhci, ERST_NUM_SEGS, 1, TYPE_EVENT,
+						flags);
+	if (!xhci->event_ring)
+		goto fail;
+	if (xhci_check_trb_in_td_math(xhci, flags) < 0)
+		goto fail;
+
+	xhci->erst.entries = dma_alloc_coherent(dev,
+			sizeof(struct xhci_erst_entry) * ERST_NUM_SEGS, &dma,
+			GFP_KERNEL);
+	if (!xhci->erst.entries)
+		goto fail;
+	xhci_dbg(xhci, "// Allocated event ring segment table at 0x%llx\n",
+			(unsigned long long)dma);
+
+	memset(xhci->erst.entries, 0, sizeof(struct xhci_erst_entry)*ERST_NUM_SEGS);
+	xhci->erst.num_entries = ERST_NUM_SEGS;
+	xhci->erst.erst_dma_addr = dma;
+	xhci_dbg(xhci, "Set ERST to 0; private num segs = %i, virt addr = %p, dma addr = 0x%llx\n",
+			xhci->erst.num_entries,
+			xhci->erst.entries,
+			(unsigned long long)xhci->erst.erst_dma_addr);
+
+	/* 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 = cpu_to_le64(seg->dma);
+		entry->seg_size = cpu_to_le32(TRBS_PER_SEGMENT);
+		entry->rsvd = 0;
+		seg = seg->next;
+	}
+
+	/* set ERST count with the number of entries in the segment table */
+	val = xhci_readl(xhci, &xhci->ir_set->erst_size);
+	val &= ERST_SIZE_MASK;
+	val |= ERST_NUM_SEGS;
+	xhci_dbg(xhci, "// Write ERST size = %i to ir_set 0 (some bits preserved)\n",
+			val);
+	xhci_writel(xhci, val, &xhci->ir_set->erst_size);
+
+	xhci_dbg(xhci, "// Set ERST entries to point to event ring.\n");
+	/* 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_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);
+	xhci_dbg(xhci, "Wrote ERST address to ir_set 0.\n");
+	xhci_print_ir_set(xhci, 0);
+
+	/*
+	 * XXX: Might need to set the Interrupter Moderation Register to
+	 * something other than the default (~1ms minimum between interrupts).
+	 * See section 5.5.1.2.
+	 */
+	init_completion(&xhci->addr_dev);
+	for (i = 0; i < MAX_HC_SLOTS; ++i)
+		xhci->devs[i] = NULL;
+	for (i = 0; i < USB_MAXCHILDREN; ++i) {
+		xhci->bus_state[0].resume_done[i] = 0;
+		xhci->bus_state[1].resume_done[i] = 0;
+	}
+
+	if (scratchpad_alloc(xhci, flags))
+		goto fail;
+	if (xhci_setup_port_arrays(xhci, flags))
+		goto fail;
+
+	INIT_LIST_HEAD(&xhci->lpm_failed_devs);
+
+	/* Enable USB 3.0 device notifications for function remote wake, which
+	 * is necessary for allowing USB 3.0 devices to do remote wakeup from
+	 * U3 (device suspend).
+	 */
+	temp = xhci_readl(xhci, &xhci->op_regs->dev_notification);
+	temp &= ~DEV_NOTE_MASK;
+	temp |= DEV_NOTE_FWAKE;
+	xhci_writel(xhci, temp, &xhci->op_regs->dev_notification);
+
+	return 0;
+
+fail:
+	xhci_warn(xhci, "Couldn't initialize memory\n");
+	xhci_halt(xhci);
+	xhci_reset(xhci);
+	xhci_mem_cleanup(xhci);
+	return -ENOMEM;
+}
diff --git a/drivers/usb/host/xhci-pci.c b/drivers/usb/host/xhci-pci.c
new file mode 100644
index 0000000..1a30c38
--- /dev/null
+++ b/drivers/usb/host/xhci-pci.c
@@ -0,0 +1,356 @@
+/*
+ * xHCI host controller driver PCI Bus Glue.
+ *
+ * Copyright (C) 2008 Intel Corp.
+ *
+ * Author: Sarah Sharp
+ * Some code borrowed from the Linux EHCI driver.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#include <linux/pci.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+
+#include "xhci.h"
+
+/* Device for a quirk */
+#define PCI_VENDOR_ID_FRESCO_LOGIC	0x1b73
+#define PCI_DEVICE_ID_FRESCO_LOGIC_PDK	0x1000
+#define PCI_DEVICE_ID_FRESCO_LOGIC_FL1400	0x1400
+
+#define PCI_VENDOR_ID_ETRON		0x1b6f
+#define PCI_DEVICE_ID_ASROCK_P67	0x7023
+
+static const char hcd_name[] = "xhci_hcd";
+
+/* called after powerup, by probe or system-pm "wakeup" */
+static int xhci_pci_reinit(struct xhci_hcd *xhci, struct pci_dev *pdev)
+{
+	/*
+	 * TODO: Implement finding debug ports later.
+	 * TODO: see if there are any quirks that need to be added to handle
+	 * new extended capabilities.
+	 */
+
+	/* PCI Memory-Write-Invalidate cycle support is optional (uncommon) */
+	if (!pci_set_mwi(pdev))
+		xhci_dbg(xhci, "MWI active\n");
+
+	xhci_dbg(xhci, "Finished xhci_pci_reinit\n");
+	return 0;
+}
+
+static void xhci_pci_quirks(struct device *dev, struct xhci_hcd *xhci)
+{
+	struct pci_dev		*pdev = to_pci_dev(dev);
+
+	/* Look for vendor-specific quirks */
+	if (pdev->vendor == PCI_VENDOR_ID_FRESCO_LOGIC &&
+			(pdev->device == PCI_DEVICE_ID_FRESCO_LOGIC_PDK ||
+			 pdev->device == PCI_DEVICE_ID_FRESCO_LOGIC_FL1400)) {
+		if (pdev->device == PCI_DEVICE_ID_FRESCO_LOGIC_PDK &&
+				pdev->revision == 0x0) {
+			xhci->quirks |= XHCI_RESET_EP_QUIRK;
+			xhci_dbg(xhci, "QUIRK: Fresco Logic xHC needs configure"
+					" endpoint cmd after reset endpoint\n");
+		}
+		/* Fresco Logic confirms: all revisions of this chip do not
+		 * support MSI, even though some of them claim to in their PCI
+		 * capabilities.
+		 */
+		xhci->quirks |= XHCI_BROKEN_MSI;
+		xhci_dbg(xhci, "QUIRK: Fresco Logic revision %u "
+				"has broken MSI implementation\n",
+				pdev->revision);
+		xhci->quirks |= XHCI_TRUST_TX_LENGTH;
+	}
+
+	if (pdev->vendor == PCI_VENDOR_ID_NEC)
+		xhci->quirks |= XHCI_NEC_HOST;
+
+	if (pdev->vendor == PCI_VENDOR_ID_AMD && xhci->hci_version == 0x96)
+		xhci->quirks |= XHCI_AMD_0x96_HOST;
+
+	/* AMD PLL quirk */
+	if (pdev->vendor == PCI_VENDOR_ID_AMD && usb_amd_find_chipset_info())
+		xhci->quirks |= XHCI_AMD_PLL_FIX;
+	if (pdev->vendor == PCI_VENDOR_ID_INTEL) {
+		xhci->quirks |= XHCI_LPM_SUPPORT;
+		xhci->quirks |= XHCI_INTEL_HOST;
+	}
+	if (pdev->vendor == PCI_VENDOR_ID_INTEL &&
+			pdev->device == PCI_DEVICE_ID_INTEL_PANTHERPOINT_XHCI) {
+		xhci->quirks |= XHCI_SPURIOUS_SUCCESS;
+		xhci->quirks |= XHCI_EP_LIMIT_QUIRK;
+		xhci->limit_active_eps = 64;
+		xhci->quirks |= XHCI_SW_BW_CHECKING;
+		/*
+		 * PPT desktop boards DH77EB and DH77DF will power back on after
+		 * a few seconds of being shutdown.  The fix for this is to
+		 * switch the ports from xHCI to EHCI on shutdown.  We can't use
+		 * DMI information to find those particular boards (since each
+		 * vendor will change the board name), so we have to key off all
+		 * PPT chipsets.
+		 */
+		xhci->quirks |= XHCI_SPURIOUS_REBOOT;
+		xhci->quirks |= XHCI_AVOID_BEI;
+	}
+	if (pdev->vendor == PCI_VENDOR_ID_ETRON &&
+			pdev->device == PCI_DEVICE_ID_ASROCK_P67) {
+		xhci->quirks |= XHCI_RESET_ON_RESUME;
+		xhci_dbg(xhci, "QUIRK: Resetting on resume\n");
+		xhci->quirks |= XHCI_TRUST_TX_LENGTH;
+	}
+	if (pdev->vendor == PCI_VENDOR_ID_VIA)
+		xhci->quirks |= XHCI_RESET_ON_RESUME;
+}
+
+/* called during probe() after chip reset completes */
+static int xhci_pci_setup(struct usb_hcd *hcd)
+{
+	struct xhci_hcd		*xhci;
+	struct pci_dev		*pdev = to_pci_dev(hcd->self.controller);
+	int			retval;
+
+	retval = xhci_gen_setup(hcd, xhci_pci_quirks);
+	if (retval)
+		return retval;
+
+	xhci = hcd_to_xhci(hcd);
+	if (!usb_hcd_is_primary_hcd(hcd))
+		return 0;
+
+	pci_read_config_byte(pdev, XHCI_SBRN_OFFSET, &xhci->sbrn);
+	xhci_dbg(xhci, "Got SBRN %u\n", (unsigned int) xhci->sbrn);
+
+	/* Find any debug ports */
+	retval = xhci_pci_reinit(xhci, pdev);
+	if (!retval)
+		return retval;
+
+	kfree(xhci);
+	return retval;
+}
+
+/*
+ * We need to register our own PCI probe function (instead of the USB core's
+ * function) in order to create a second roothub under xHCI.
+ */
+static int xhci_pci_probe(struct pci_dev *dev, const struct pci_device_id *id)
+{
+	int retval;
+	struct xhci_hcd *xhci;
+	struct hc_driver *driver;
+	struct usb_hcd *hcd;
+
+	driver = (struct hc_driver *)id->driver_data;
+	/* Register the USB 2.0 roothub.
+	 * FIXME: USB core must know to register the USB 2.0 roothub first.
+	 * This is sort of silly, because we could just set the HCD driver flags
+	 * to say USB 2.0, but I'm not sure what the implications would be in
+	 * the other parts of the HCD code.
+	 */
+	retval = usb_hcd_pci_probe(dev, id);
+
+	if (retval)
+		return retval;
+
+	/* USB 2.0 roothub is stored in the PCI device now. */
+	hcd = dev_get_drvdata(&dev->dev);
+	xhci = hcd_to_xhci(hcd);
+	xhci->shared_hcd = usb_create_shared_hcd(driver, &dev->dev,
+				pci_name(dev), hcd);
+	if (!xhci->shared_hcd) {
+		retval = -ENOMEM;
+		goto dealloc_usb2_hcd;
+	}
+
+	/* Set the xHCI pointer before xhci_pci_setup() (aka hcd_driver.reset)
+	 * is called by usb_add_hcd().
+	 */
+	*((struct xhci_hcd **) xhci->shared_hcd->hcd_priv) = xhci;
+
+	retval = usb_add_hcd(xhci->shared_hcd, dev->irq,
+			IRQF_SHARED);
+	if (retval)
+		goto put_usb3_hcd;
+	/* Roothub already marked as USB 3.0 speed */
+
+	/* We know the LPM timeout algorithms for this host, let the USB core
+	 * enable and disable LPM for devices under the USB 3.0 roothub.
+	 */
+	if (xhci->quirks & XHCI_LPM_SUPPORT)
+		hcd_to_bus(xhci->shared_hcd)->root_hub->lpm_capable = 1;
+
+	return 0;
+
+put_usb3_hcd:
+	usb_put_hcd(xhci->shared_hcd);
+dealloc_usb2_hcd:
+	usb_hcd_pci_remove(dev);
+	return retval;
+}
+
+static void xhci_pci_remove(struct pci_dev *dev)
+{
+	struct xhci_hcd *xhci;
+
+	xhci = hcd_to_xhci(pci_get_drvdata(dev));
+	if (xhci->shared_hcd) {
+		usb_remove_hcd(xhci->shared_hcd);
+		usb_put_hcd(xhci->shared_hcd);
+	}
+	usb_hcd_pci_remove(dev);
+	kfree(xhci);
+}
+
+#ifdef CONFIG_PM
+static int xhci_pci_suspend(struct usb_hcd *hcd, bool do_wakeup)
+{
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+
+	return xhci_suspend(xhci);
+}
+
+static int xhci_pci_resume(struct usb_hcd *hcd, bool hibernated)
+{
+	struct xhci_hcd		*xhci = hcd_to_xhci(hcd);
+	struct pci_dev		*pdev = to_pci_dev(hcd->self.controller);
+	int			retval = 0;
+
+	/* The BIOS on systems with the Intel Panther Point chipset may or may
+	 * not support xHCI natively.  That means that during system resume, it
+	 * may switch the ports back to EHCI so that users can use their
+	 * keyboard to select a kernel from GRUB after resume from hibernate.
+	 *
+	 * The BIOS is supposed to remember whether the OS had xHCI ports
+	 * enabled before resume, and switch the ports back to xHCI when the
+	 * BIOS/OS semaphore is written, but we all know we can't trust BIOS
+	 * writers.
+	 *
+	 * Unconditionally switch the ports back to xHCI after a system resume.
+	 * We can't tell whether the EHCI or xHCI controller will be resumed
+	 * first, so we have to do the port switchover in both drivers.  Writing
+	 * a '1' to the port switchover registers should have no effect if the
+	 * port was already switched over.
+	 */
+	if (usb_is_intel_switchable_xhci(pdev))
+		usb_enable_xhci_ports(pdev);
+
+	retval = xhci_resume(xhci, hibernated);
+	return retval;
+}
+#endif /* CONFIG_PM */
+
+static const struct hc_driver xhci_pci_hc_driver = {
+	.description =		hcd_name,
+	.product_desc =		"xHCI Host Controller",
+	.hcd_priv_size =	sizeof(struct xhci_hcd *),
+
+	/*
+	 * generic hardware linkage
+	 */
+	.irq =			xhci_irq,
+	.flags =		HCD_MEMORY | HCD_USB3 | HCD_SHARED,
+
+	/*
+	 * basic lifecycle operations
+	 */
+	.reset =		xhci_pci_setup,
+	.start =		xhci_run,
+#ifdef CONFIG_PM
+	.pci_suspend =          xhci_pci_suspend,
+	.pci_resume =           xhci_pci_resume,
+#endif
+	.stop =			xhci_stop,
+	.shutdown =		xhci_shutdown,
+
+	/*
+	 * managing i/o requests and associated device resources
+	 */
+	.urb_enqueue =		xhci_urb_enqueue,
+	.urb_dequeue =		xhci_urb_dequeue,
+	.alloc_dev =		xhci_alloc_dev,
+	.free_dev =		xhci_free_dev,
+	.alloc_streams =	xhci_alloc_streams,
+	.free_streams =		xhci_free_streams,
+	.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,
+	.update_hub_device =	xhci_update_hub_device,
+	.reset_device =		xhci_discover_or_reset_device,
+
+	/*
+	 * scheduling support
+	 */
+	.get_frame_number =	xhci_get_frame,
+
+	/* Root hub support */
+	.hub_control =		xhci_hub_control,
+	.hub_status_data =	xhci_hub_status_data,
+	.bus_suspend =		xhci_bus_suspend,
+	.bus_resume =		xhci_bus_resume,
+	/*
+	 * call back when device connected and addressed
+	 */
+	.update_device =        xhci_update_device,
+	.set_usb2_hw_lpm =	xhci_set_usb2_hardware_lpm,
+	.enable_usb3_lpm_timeout =	xhci_enable_usb3_lpm_timeout,
+	.disable_usb3_lpm_timeout =	xhci_disable_usb3_lpm_timeout,
+	.find_raw_port_number =	xhci_find_raw_port_number,
+};
+
+/*-------------------------------------------------------------------------*/
+
+/* PCI driver selection metadata; PCI hotplugging uses this */
+static const struct pci_device_id pci_ids[] = { {
+	/* handle any USB 3.0 xHCI controller */
+	PCI_DEVICE_CLASS(PCI_CLASS_SERIAL_USB_XHCI, ~0),
+	.driver_data =	(unsigned long) &xhci_pci_hc_driver,
+	},
+	{ /* end: all zeroes */ }
+};
+MODULE_DEVICE_TABLE(pci, pci_ids);
+
+/* pci driver glue; this is a "new style" PCI driver module */
+static struct pci_driver xhci_pci_driver = {
+	.name =		(char *) hcd_name,
+	.id_table =	pci_ids,
+
+	.probe =	xhci_pci_probe,
+	.remove =	xhci_pci_remove,
+	/* suspend and resume implemented later */
+
+	.shutdown = 	usb_hcd_pci_shutdown,
+#ifdef CONFIG_PM_SLEEP
+	.driver = {
+		.pm = &usb_hcd_pci_pm_ops
+	},
+#endif
+};
+
+int __init xhci_register_pci(void)
+{
+	return pci_register_driver(&xhci_pci_driver);
+}
+
+void xhci_unregister_pci(void)
+{
+	pci_unregister_driver(&xhci_pci_driver);
+}
diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c
new file mode 100644
index 0000000..df90fe5
--- /dev/null
+++ b/drivers/usb/host/xhci-plat.c
@@ -0,0 +1,205 @@
+/*
+ * xhci-plat.c - xHCI host controller driver platform Bus Glue.
+ *
+ * Copyright (C) 2012 Texas Instruments Incorporated - http://www.ti.com
+ * Author: Sebastian Andrzej Siewior <bigeasy at linutronix.de>
+ *
+ * A lot of code borrowed from the Linux xHCI driver.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * version 2 as published by the Free Software Foundation.
+ */
+
+#include <linux/platform_device.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+
+#include "xhci.h"
+
+static void xhci_plat_quirks(struct device *dev, struct xhci_hcd *xhci)
+{
+	/*
+	 * As of now platform drivers don't provide MSI support so we ensure
+	 * here that the generic code does not try to make a pci_dev from our
+	 * dev struct in order to setup MSI
+	 */
+	xhci->quirks |= XHCI_BROKEN_MSI;
+}
+
+/* called during probe() after chip reset completes */
+static int xhci_plat_setup(struct usb_hcd *hcd)
+{
+	return xhci_gen_setup(hcd, xhci_plat_quirks);
+}
+
+static const struct hc_driver xhci_plat_xhci_driver = {
+	.description =		"xhci-hcd",
+	.product_desc =		"xHCI Host Controller",
+	.hcd_priv_size =	sizeof(struct xhci_hcd *),
+
+	/*
+	 * generic hardware linkage
+	 */
+	.irq =			xhci_irq,
+	.flags =		HCD_MEMORY | HCD_USB3 | HCD_SHARED,
+
+	/*
+	 * basic lifecycle operations
+	 */
+	.reset =		xhci_plat_setup,
+	.start =		xhci_run,
+	.stop =			xhci_stop,
+	.shutdown =		xhci_shutdown,
+
+	/*
+	 * managing i/o requests and associated device resources
+	 */
+	.urb_enqueue =		xhci_urb_enqueue,
+	.urb_dequeue =		xhci_urb_dequeue,
+	.alloc_dev =		xhci_alloc_dev,
+	.free_dev =		xhci_free_dev,
+	.alloc_streams =	xhci_alloc_streams,
+	.free_streams =		xhci_free_streams,
+	.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,
+	.update_hub_device =	xhci_update_hub_device,
+	.reset_device =		xhci_discover_or_reset_device,
+
+	/*
+	 * scheduling support
+	 */
+	.get_frame_number =	xhci_get_frame,
+
+	/* Root hub support */
+	.hub_control =		xhci_hub_control,
+	.hub_status_data =	xhci_hub_status_data,
+	.bus_suspend =		xhci_bus_suspend,
+	.bus_resume =		xhci_bus_resume,
+};
+
+static int xhci_plat_probe(struct platform_device *pdev)
+{
+	const struct hc_driver	*driver;
+	struct xhci_hcd		*xhci;
+	struct resource         *res;
+	struct usb_hcd		*hcd;
+	int			ret;
+	int			irq;
+
+	if (usb_disabled())
+		return -ENODEV;
+
+	driver = &xhci_plat_xhci_driver;
+
+	irq = platform_get_irq(pdev, 0);
+	if (irq < 0)
+		return -ENODEV;
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!res)
+		return -ENODEV;
+
+	hcd = usb_create_hcd(driver, &pdev->dev, dev_name(&pdev->dev));
+	if (!hcd)
+		return -ENOMEM;
+
+	hcd->rsrc_start = res->start;
+	hcd->rsrc_len = resource_size(res);
+
+	if (!request_mem_region(hcd->rsrc_start, hcd->rsrc_len,
+				driver->description)) {
+		dev_dbg(&pdev->dev, "controller already in use\n");
+		ret = -EBUSY;
+		goto put_hcd;
+	}
+
+	hcd->regs = ioremap_nocache(hcd->rsrc_start, hcd->rsrc_len);
+	if (!hcd->regs) {
+		dev_dbg(&pdev->dev, "error mapping memory\n");
+		ret = -EFAULT;
+		goto release_mem_region;
+	}
+
+	ret = usb_add_hcd(hcd, irq, IRQF_SHARED);
+	if (ret)
+		goto unmap_registers;
+
+	/* USB 2.0 roothub is stored in the platform_device now. */
+	hcd = dev_get_drvdata(&pdev->dev);
+	xhci = hcd_to_xhci(hcd);
+	xhci->shared_hcd = usb_create_shared_hcd(driver, &pdev->dev,
+			dev_name(&pdev->dev), hcd);
+	if (!xhci->shared_hcd) {
+		ret = -ENOMEM;
+		goto dealloc_usb2_hcd;
+	}
+
+	/*
+	 * Set the xHCI pointer before xhci_plat_setup() (aka hcd_driver.reset)
+	 * is called by usb_add_hcd().
+	 */
+	*((struct xhci_hcd **) xhci->shared_hcd->hcd_priv) = xhci;
+
+	ret = usb_add_hcd(xhci->shared_hcd, irq, IRQF_SHARED);
+	if (ret)
+		goto put_usb3_hcd;
+
+	return 0;
+
+put_usb3_hcd:
+	usb_put_hcd(xhci->shared_hcd);
+
+dealloc_usb2_hcd:
+	usb_remove_hcd(hcd);
+
+unmap_registers:
+	iounmap(hcd->regs);
+
+release_mem_region:
+	release_mem_region(hcd->rsrc_start, hcd->rsrc_len);
+
+put_hcd:
+	usb_put_hcd(hcd);
+
+	return ret;
+}
+
+static int xhci_plat_remove(struct platform_device *dev)
+{
+	struct usb_hcd	*hcd = platform_get_drvdata(dev);
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+
+	usb_remove_hcd(xhci->shared_hcd);
+	usb_put_hcd(xhci->shared_hcd);
+
+	usb_remove_hcd(hcd);
+	iounmap(hcd->regs);
+	usb_put_hcd(hcd);
+	kfree(xhci);
+
+	return 0;
+}
+
+static struct platform_driver usb_xhci_driver = {
+	.probe	= xhci_plat_probe,
+	.remove	= xhci_plat_remove,
+	.driver	= {
+		.name = "xhci-hcd",
+	},
+};
+MODULE_ALIAS("platform:xhci-hcd");
+
+int xhci_register_plat(void)
+{
+	return platform_driver_register(&usb_xhci_driver);
+}
+
+void xhci_unregister_plat(void)
+{
+	platform_driver_unregister(&usb_xhci_driver);
+}
diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
new file mode 100644
index 0000000..1969c00
--- /dev/null
+++ b/drivers/usb/host/xhci-ring.c
@@ -0,0 +1,4011 @@
+/*
+ * xHCI host controller driver
+ *
+ * Copyright (C) 2008 Intel Corp.
+ *
+ * Author: Sarah Sharp
+ * Some code borrowed from the Linux EHCI driver.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+/*
+ * Ring initialization rules:
+ * 1. Each segment is initialized to zero, except for link TRBs.
+ * 2. Ring cycle state = 0.  This represents Producer Cycle State (PCS) or
+ *    Consumer Cycle State (CCS), depending on ring function.
+ * 3. Enqueue pointer = dequeue pointer = address of first TRB in the segment.
+ *
+ * Ring behavior rules:
+ * 1. A ring is empty if enqueue == dequeue.  This means there will always be at
+ *    least one free TRB in the ring.  This is useful if you want to turn that
+ *    into a link TRB and expand the ring.
+ * 2. When incrementing an enqueue or dequeue pointer, if the next TRB is a
+ *    link TRB, then load the pointer with the address in the link TRB.  If the
+ *    link TRB had its toggle bit set, you may need to update the ring cycle
+ *    state (see cycle bit rules).  You may have to do this multiple times
+ *    until you reach a non-link TRB.
+ * 3. A ring is full if enqueue++ (for the definition of increment above)
+ *    equals the dequeue pointer.
+ *
+ * Cycle bit rules:
+ * 1. When a consumer increments a dequeue pointer and encounters a toggle bit
+ *    in a link TRB, it must toggle the ring cycle state.
+ * 2. When a producer increments an enqueue pointer and encounters a toggle bit
+ *    in a link TRB, it must toggle the ring cycle state.
+ *
+ * Producer rules:
+ * 1. Check if ring is full before you enqueue.
+ * 2. Write the ring cycle state to the cycle bit in the TRB you're enqueuing.
+ *    Update enqueue pointer between each write (which may update the ring
+ *    cycle state).
+ * 3. Notify consumer.  If SW is producer, it rings the doorbell for command
+ *    and endpoint rings.  If HC is the producer for the event ring,
+ *    and it generates an interrupt according to interrupt modulation rules.
+ *
+ * Consumer rules:
+ * 1. Check if TRB belongs to you.  If the cycle bit == your ring cycle state,
+ *    the TRB is owned by the consumer.
+ * 2. Update dequeue pointer (which may update the ring cycle state) and
+ *    continue processing TRBs until you reach a TRB which is not owned by you.
+ * 3. Notify the producer.  SW is the consumer for the event ring, and it
+ *   updates event ring dequeue pointer.  HC is the consumer for the command and
+ *   endpoint rings; it generates events on the event ring for these.
+ */
+
+#include <linux/scatterlist.h>
+#include <linux/slab.h>
+#include "xhci.h"
+
+static int handle_cmd_in_cmd_wait_list(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		struct xhci_event_cmd *event);
+
+/*
+ * Returns zero if the TRB isn't in this segment, otherwise it returns the DMA
+ * address of the TRB.
+ */
+dma_addr_t xhci_trb_virt_to_dma(struct xhci_segment *seg,
+		union xhci_trb *trb)
+{
+	unsigned long segment_offset;
+
+	if (!seg || !trb || trb < seg->trbs)
+		return 0;
+	/* offset in TRBs */
+	segment_offset = trb - seg->trbs;
+	if (segment_offset > TRBS_PER_SEGMENT)
+		return 0;
+	return seg->dma + (segment_offset * sizeof(*trb));
+}
+
+/* Does this link TRB point to the first segment in a ring,
+ * or was the previous TRB the last TRB on the last segment in the ERST?
+ */
+static bool last_trb_on_last_seg(struct xhci_hcd *xhci, struct xhci_ring *ring,
+		struct xhci_segment *seg, union xhci_trb *trb)
+{
+	if (ring == xhci->event_ring)
+		return (trb == &seg->trbs[TRBS_PER_SEGMENT]) &&
+			(seg->next == xhci->event_ring->first_seg);
+	else
+		return le32_to_cpu(trb->link.control) & LINK_TOGGLE;
+}
+
+/* Is this TRB a link TRB or was the last TRB the last TRB in this event ring
+ * segment?  I.e. would the updated event TRB pointer step off the end of the
+ * event seg?
+ */
+static int last_trb(struct xhci_hcd *xhci, struct xhci_ring *ring,
+		struct xhci_segment *seg, union xhci_trb *trb)
+{
+	if (ring == xhci->event_ring)
+		return trb == &seg->trbs[TRBS_PER_SEGMENT];
+	else
+		return TRB_TYPE_LINK_LE32(trb->link.control);
+}
+
+static int enqueue_is_link_trb(struct xhci_ring *ring)
+{
+	struct xhci_link_trb *link = &ring->enqueue->link;
+	return TRB_TYPE_LINK_LE32(link->control);
+}
+
+/* Updates trb to point to the next TRB in the ring, and updates seg if the next
+ * TRB is in a new segment.  This does not skip over link TRBs, and it does not
+ * effect the ring dequeue or enqueue pointers.
+ */
+static void next_trb(struct xhci_hcd *xhci,
+		struct xhci_ring *ring,
+		struct xhci_segment **seg,
+		union xhci_trb **trb)
+{
+	if (last_trb(xhci, ring, *seg, *trb)) {
+		*seg = (*seg)->next;
+		*trb = ((*seg)->trbs);
+	} else {
+		(*trb)++;
+	}
+}
+
+/*
+ * See Cycle bit rules. SW is the consumer for the event ring only.
+ * Don't make a ring full of link TRBs.  That would be dumb and this would loop.
+ */
+static void inc_deq(struct xhci_hcd *xhci, struct xhci_ring *ring)
+{
+	unsigned long long addr;
+
+	ring->deq_updates++;
+
+	/*
+	 * If this is not event ring, and the dequeue pointer
+	 * is not on a link TRB, there is one more usable TRB
+	 */
+	if (ring->type != TYPE_EVENT &&
+			!last_trb(xhci, ring, ring->deq_seg, ring->dequeue))
+		ring->num_trbs_free++;
+
+	do {
+		/*
+		 * Update the dequeue pointer further if that was a link TRB or
+		 * we're at the end of an event ring segment (which doesn't have
+		 * link TRBS)
+		 */
+		if (last_trb(xhci, ring, ring->deq_seg, ring->dequeue)) {
+			if (ring->type == TYPE_EVENT &&
+					last_trb_on_last_seg(xhci, ring,
+						ring->deq_seg, ring->dequeue)) {
+				ring->cycle_state = (ring->cycle_state ? 0 : 1);
+			}
+			ring->deq_seg = ring->deq_seg->next;
+			ring->dequeue = ring->deq_seg->trbs;
+		} else {
+			ring->dequeue++;
+		}
+	} while (last_trb(xhci, ring, ring->deq_seg, ring->dequeue));
+
+	addr = (unsigned long long) xhci_trb_virt_to_dma(ring->deq_seg, ring->dequeue);
+}
+
+/*
+ * See Cycle bit rules. SW is the consumer for the event ring only.
+ * Don't make a ring full of link TRBs.  That would be dumb and this would loop.
+ *
+ * If we've just enqueued a TRB that is in the middle of a TD (meaning the
+ * chain bit is set), then set the chain bit in all the following link TRBs.
+ * If we've enqueued the last TRB in a TD, make sure the following link TRBs
+ * have their chain bit cleared (so that each Link TRB is a separate TD).
+ *
+ * Section 6.4.4.1 of the 0.95 spec says link TRBs cannot have the chain bit
+ * set, but other sections talk about dealing with the chain bit set.  This was
+ * fixed in the 0.96 specification errata, but we have to assume that all 0.95
+ * xHCI hardware can't handle the chain bit being cleared on a link TRB.
+ *
+ * @more_trbs_coming:	Will you enqueue more TRBs before calling
+ *			prepare_transfer()?
+ */
+static void inc_enq(struct xhci_hcd *xhci, struct xhci_ring *ring,
+			bool more_trbs_coming)
+{
+	u32 chain;
+	union xhci_trb *next;
+	unsigned long long addr;
+
+	chain = le32_to_cpu(ring->enqueue->generic.field[3]) & TRB_CHAIN;
+	/* If this is not event ring, there is one less usable TRB */
+	if (ring->type != TYPE_EVENT &&
+			!last_trb(xhci, ring, ring->enq_seg, ring->enqueue))
+		ring->num_trbs_free--;
+	next = ++(ring->enqueue);
+
+	ring->enq_updates++;
+	/* Update the dequeue pointer further if that was a link TRB or we're at
+	 * the end of an event ring segment (which doesn't have link TRBS)
+	 */
+	while (last_trb(xhci, ring, ring->enq_seg, next)) {
+		if (ring->type != TYPE_EVENT) {
+			/*
+			 * If the caller doesn't plan on enqueueing more
+			 * TDs before ringing the doorbell, then we
+			 * don't want to give the link TRB to the
+			 * hardware just yet.  We'll give the link TRB
+			 * back in prepare_ring() just before we enqueue
+			 * the TD at the top of the ring.
+			 */
+			if (!chain && !more_trbs_coming)
+				break;
+
+			/* If we're not dealing with 0.95 hardware or
+			 * isoc rings on AMD 0.96 host,
+			 * carry over the chain bit of the previous TRB
+			 * (which may mean the chain bit is cleared).
+			 */
+			if (!(ring->type == TYPE_ISOC &&
+					(xhci->quirks & XHCI_AMD_0x96_HOST))
+						&& !xhci_link_trb_quirk(xhci)) {
+				next->link.control &=
+					cpu_to_le32(~TRB_CHAIN);
+				next->link.control |=
+					cpu_to_le32(chain);
+			}
+			/* Give this link TRB to the hardware */
+			wmb();
+			next->link.control ^= cpu_to_le32(TRB_CYCLE);
+
+			/* Toggle the cycle bit after the last ring segment. */
+			if (last_trb_on_last_seg(xhci, ring, ring->enq_seg, next)) {
+				ring->cycle_state = (ring->cycle_state ? 0 : 1);
+			}
+		}
+		ring->enq_seg = ring->enq_seg->next;
+		ring->enqueue = ring->enq_seg->trbs;
+		next = ring->enqueue;
+	}
+	addr = (unsigned long long) xhci_trb_virt_to_dma(ring->enq_seg, ring->enqueue);
+}
+
+/*
+ * Check to see if there's room to enqueue num_trbs on the ring and make sure
+ * enqueue pointer will not advance into dequeue segment. See rules above.
+ */
+static inline int room_on_ring(struct xhci_hcd *xhci, struct xhci_ring *ring,
+		unsigned int num_trbs)
+{
+	int num_trbs_in_deq_seg;
+
+	if (ring->num_trbs_free < num_trbs)
+		return 0;
+
+	if (ring->type != TYPE_COMMAND && ring->type != TYPE_EVENT) {
+		num_trbs_in_deq_seg = ring->dequeue - ring->deq_seg->trbs;
+		if (ring->num_trbs_free < num_trbs + num_trbs_in_deq_seg)
+			return 0;
+	}
+
+	return 1;
+}
+
+/* Ring the host controller doorbell after placing a command on the ring */
+void xhci_ring_cmd_db(struct xhci_hcd *xhci)
+{
+	if (!(xhci->cmd_ring_state & CMD_RING_STATE_RUNNING))
+		return;
+
+	xhci_dbg(xhci, "// Ding dong!\n");
+	xhci_writel(xhci, DB_VALUE_HOST, &xhci->dba->doorbell[0]);
+	/* Flush PCI posted writes */
+	xhci_readl(xhci, &xhci->dba->doorbell[0]);
+}
+
+static int xhci_abort_cmd_ring(struct xhci_hcd *xhci)
+{
+	u64 temp_64;
+	int ret;
+
+	xhci_dbg(xhci, "Abort command ring\n");
+
+	if (!(xhci->cmd_ring_state & CMD_RING_STATE_RUNNING)) {
+		xhci_dbg(xhci, "The command ring isn't running, "
+				"Have the command ring been stopped?\n");
+		return 0;
+	}
+
+	temp_64 = xhci_read_64(xhci, &xhci->op_regs->cmd_ring);
+	if (!(temp_64 & CMD_RING_RUNNING)) {
+		xhci_dbg(xhci, "Command ring had been stopped\n");
+		return 0;
+	}
+	xhci->cmd_ring_state = CMD_RING_STATE_ABORTED;
+	xhci_write_64(xhci, temp_64 | CMD_RING_ABORT,
+			&xhci->op_regs->cmd_ring);
+
+	/* Section 4.6.1.2 of xHCI 1.0 spec says software should
+	 * time the completion od all xHCI commands, including
+	 * the Command Abort operation. If software doesn't see
+	 * CRR negated in a timely manner (e.g. longer than 5
+	 * seconds), then it should assume that the there are
+	 * larger problems with the xHC and assert HCRST.
+	 */
+	ret = xhci_handshake(xhci, &xhci->op_regs->cmd_ring,
+			CMD_RING_RUNNING, 0, 5 * 1000 * 1000);
+	if (ret < 0) {
+		xhci_err(xhci, "Stopped the command ring failed, "
+				"maybe the host is dead\n");
+		xhci->xhc_state |= XHCI_STATE_DYING;
+		xhci_quiesce(xhci);
+		xhci_halt(xhci);
+		return -ESHUTDOWN;
+	}
+
+	return 0;
+}
+
+static int xhci_queue_cd(struct xhci_hcd *xhci,
+		struct xhci_command *command,
+		union xhci_trb *cmd_trb)
+{
+	struct xhci_cd *cd;
+	cd = kzalloc(sizeof(struct xhci_cd), GFP_ATOMIC);
+	if (!cd)
+		return -ENOMEM;
+	INIT_LIST_HEAD(&cd->cancel_cmd_list);
+
+	cd->command = command;
+	cd->cmd_trb = cmd_trb;
+	list_add_tail(&cd->cancel_cmd_list, &xhci->cancel_cmd_list);
+
+	return 0;
+}
+
+/*
+ * Cancel the command which has issue.
+ *
+ * Some commands may hang due to waiting for acknowledgement from
+ * usb device. It is outside of the xHC's ability to control and
+ * will cause the command ring is blocked. When it occurs software
+ * should intervene to recover the command ring.
+ * See Section 4.6.1.1 and 4.6.1.2
+ */
+int xhci_cancel_cmd(struct xhci_hcd *xhci, struct xhci_command *command,
+		union xhci_trb *cmd_trb)
+{
+	int retval = 0;
+	unsigned long flags;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+
+	if (xhci->xhc_state & XHCI_STATE_DYING) {
+		xhci_warn(xhci, "Abort the command ring,"
+				" but the xHCI is dead.\n");
+		retval = -ESHUTDOWN;
+		goto fail;
+	}
+
+	/* queue the cmd desriptor to cancel_cmd_list */
+	retval = xhci_queue_cd(xhci, command, cmd_trb);
+	if (retval) {
+		xhci_warn(xhci, "Queuing command descriptor failed.\n");
+		goto fail;
+	}
+
+	/* abort command ring */
+	retval = xhci_abort_cmd_ring(xhci);
+	if (retval) {
+		xhci_err(xhci, "Abort command ring failed\n");
+		if (unlikely(retval == -ESHUTDOWN)) {
+			spin_unlock_irqrestore(&xhci->lock, flags);
+			usb_hc_died(xhci_to_hcd(xhci)->primary_hcd);
+			xhci_dbg(xhci, "xHCI host controller is dead.\n");
+			return retval;
+		}
+	}
+
+fail:
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return retval;
+}
+
+void xhci_ring_ep_doorbell(struct xhci_hcd *xhci,
+		unsigned int slot_id,
+		unsigned int ep_index,
+		unsigned int stream_id)
+{
+	__le32 __iomem *db_addr = &xhci->dba->doorbell[slot_id];
+	struct xhci_virt_ep *ep = &xhci->devs[slot_id]->eps[ep_index];
+	unsigned int ep_state = ep->ep_state;
+
+	/* Don't ring the doorbell for this endpoint if there are pending
+	 * cancellations because we don't want to interrupt processing.
+	 * We don't want to restart any stream rings if there's a set dequeue
+	 * pointer command pending because the device can choose to start any
+	 * stream once the endpoint is on the HW schedule.
+	 * FIXME - check all the stream rings for pending cancellations.
+	 */
+	if ((ep_state & EP_HALT_PENDING) || (ep_state & SET_DEQ_PENDING) ||
+	    (ep_state & EP_HALTED))
+		return;
+	xhci_writel(xhci, DB_VALUE(ep_index, stream_id), db_addr);
+	/* The CPU has better things to do at this point than wait for a
+	 * write-posting flush.  It'll get there soon enough.
+	 */
+}
+
+/* Ring the doorbell for any rings with pending URBs */
+static void ring_doorbell_for_active_rings(struct xhci_hcd *xhci,
+		unsigned int slot_id,
+		unsigned int ep_index)
+{
+	unsigned int stream_id;
+	struct xhci_virt_ep *ep;
+
+	ep = &xhci->devs[slot_id]->eps[ep_index];
+
+	/* A ring has pending URBs if its TD list is not empty */
+	if (!(ep->ep_state & EP_HAS_STREAMS)) {
+		if (!(list_empty(&ep->ring->td_list)))
+			xhci_ring_ep_doorbell(xhci, slot_id, ep_index, 0);
+		return;
+	}
+
+	for (stream_id = 1; stream_id < ep->stream_info->num_streams;
+			stream_id++) {
+		struct xhci_stream_info *stream_info = ep->stream_info;
+		if (!list_empty(&stream_info->stream_rings[stream_id]->td_list))
+			xhci_ring_ep_doorbell(xhci, slot_id, ep_index,
+						stream_id);
+	}
+}
+
+/*
+ * Find the segment that trb is in.  Start searching in start_seg.
+ * If we must move past a segment that has a link TRB with a toggle cycle state
+ * bit set, then we will toggle the value pointed at by cycle_state.
+ */
+static struct xhci_segment *find_trb_seg(
+		struct xhci_segment *start_seg,
+		union xhci_trb	*trb, int *cycle_state)
+{
+	struct xhci_segment *cur_seg = start_seg;
+	struct xhci_generic_trb *generic_trb;
+
+	while (cur_seg->trbs > trb ||
+			&cur_seg->trbs[TRBS_PER_SEGMENT - 1] < trb) {
+		generic_trb = &cur_seg->trbs[TRBS_PER_SEGMENT - 1].generic;
+		if (generic_trb->field[3] & cpu_to_le32(LINK_TOGGLE))
+			*cycle_state ^= 0x1;
+		cur_seg = cur_seg->next;
+		if (cur_seg == start_seg)
+			/* Looped over the entire list.  Oops! */
+			return NULL;
+	}
+	return cur_seg;
+}
+
+
+static struct xhci_ring *xhci_triad_to_transfer_ring(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		unsigned int stream_id)
+{
+	struct xhci_virt_ep *ep;
+
+	ep = &xhci->devs[slot_id]->eps[ep_index];
+	/* Common case: no streams */
+	if (!(ep->ep_state & EP_HAS_STREAMS))
+		return ep->ring;
+
+	if (stream_id == 0) {
+		xhci_warn(xhci,
+				"WARN: Slot ID %u, ep index %u has streams, "
+				"but URB has no stream ID.\n",
+				slot_id, ep_index);
+		return NULL;
+	}
+
+	if (stream_id < ep->stream_info->num_streams)
+		return ep->stream_info->stream_rings[stream_id];
+
+	xhci_warn(xhci,
+			"WARN: Slot ID %u, ep index %u has "
+			"stream IDs 1 to %u allocated, "
+			"but stream ID %u is requested.\n",
+			slot_id, ep_index,
+			ep->stream_info->num_streams - 1,
+			stream_id);
+	return NULL;
+}
+
+/* Get the right ring for the given URB.
+ * If the endpoint supports streams, boundary check the URB's stream ID.
+ * If the endpoint doesn't support streams, return the singular endpoint ring.
+ */
+static struct xhci_ring *xhci_urb_to_transfer_ring(struct xhci_hcd *xhci,
+		struct urb *urb)
+{
+	return xhci_triad_to_transfer_ring(xhci, urb->dev->slot_id,
+		xhci_get_endpoint_index(&urb->ep->desc), urb->stream_id);
+}
+
+/*
+ * Move the xHC's endpoint ring dequeue pointer past cur_td.
+ * Record the new state of the xHC's endpoint ring dequeue segment,
+ * dequeue pointer, and new consumer cycle state in state.
+ * Update our internal representation of the ring's dequeue pointer.
+ *
+ * We do this in three jumps:
+ *  - First we update our new ring state to be the same as when the xHC stopped.
+ *  - Then we traverse the ring to find the segment that contains
+ *    the last TRB in the TD.  We toggle the xHC's new cycle state when we pass
+ *    any link TRBs with the toggle cycle bit set.
+ *  - 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.
+ *
+ * Some of the uses of xhci_generic_trb are grotty, but if they're done
+ * with correct __le32 accesses they should work fine.  Only users of this are
+ * in here.
+ */
+void xhci_find_new_dequeue_state(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		unsigned int stream_id, struct xhci_td *cur_td,
+		struct xhci_dequeue_state *state)
+{
+	struct xhci_virt_device *dev = xhci->devs[slot_id];
+	struct xhci_ring *ep_ring;
+	struct xhci_generic_trb *trb;
+	struct xhci_ep_ctx *ep_ctx;
+	dma_addr_t addr;
+
+	ep_ring = xhci_triad_to_transfer_ring(xhci, slot_id,
+			ep_index, stream_id);
+	if (!ep_ring) {
+		xhci_warn(xhci, "WARN can't find new dequeue state "
+				"for invalid stream ID %u.\n",
+				stream_id);
+		return;
+	}
+	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,
+			dev->eps[ep_index].stopped_trb,
+			&state->new_cycle_state);
+	if (!state->new_deq_seg) {
+		WARN_ON(1);
+		return;
+	}
+
+	/* Dig out the cycle state saved by the xHC during the stop ep cmd */
+	xhci_dbg(xhci, "Finding endpoint context\n");
+	ep_ctx = xhci_get_ep_ctx(xhci, dev->out_ctx, ep_index);
+	state->new_cycle_state = 0x1 & le64_to_cpu(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);
+	if (!state->new_deq_seg) {
+		WARN_ON(1);
+		return;
+	}
+
+	trb = &state->new_deq_ptr->generic;
+	if (TRB_TYPE_LINK_LE32(trb->field[3]) &&
+	    (trb->field[3] & cpu_to_le32(LINK_TOGGLE)))
+		state->new_cycle_state ^= 0x1;
+	next_trb(xhci, ep_ring, &state->new_deq_seg, &state->new_deq_ptr);
+
+	/*
+	 * If there is only one segment in a ring, find_trb_seg()'s while loop
+	 * will not run, and it will return before it has a chance to see if it
+	 * needs to toggle the cycle bit.  It can't tell if the stalled transfer
+	 * ended just before the link TRB on a one-segment ring, or if the TD
+	 * wrapped around the top of the ring, because it doesn't have the TD in
+	 * question.  Look for the one-segment case where stalled TRB's address
+	 * is greater than the new dequeue pointer address.
+	 */
+	if (ep_ring->first_seg == ep_ring->first_seg->next &&
+			state->new_deq_ptr < dev->eps[ep_index].stopped_trb)
+		state->new_cycle_state ^= 0x1;
+	xhci_dbg(xhci, "Cycle state = 0x%x\n", state->new_cycle_state);
+
+	/* 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);
+}
+
+/* flip_cycle means flip the cycle bit of all but the first and last TRB.
+ * (The last TRB actually points to the ring enqueue pointer, which is not part
+ * of this TD.)  This is used to remove partially enqueued isoc TDs from a ring.
+ */
+static void td_to_noop(struct xhci_hcd *xhci, struct xhci_ring *ep_ring,
+		struct xhci_td *cur_td, bool flip_cycle)
+{
+	struct xhci_segment *cur_seg;
+	union xhci_trb *cur_trb;
+
+	for (cur_seg = cur_td->start_seg, cur_trb = cur_td->first_trb;
+			true;
+			next_trb(xhci, ep_ring, &cur_seg, &cur_trb)) {
+		if (TRB_TYPE_LINK_LE32(cur_trb->generic.field[3])) {
+			/* Unchain any chained Link TRBs, but
+			 * leave the pointers intact.
+			 */
+			cur_trb->generic.field[3] &= cpu_to_le32(~TRB_CHAIN);
+			/* Flip the cycle bit (link TRBs can't be the first
+			 * or last TRB).
+			 */
+			if (flip_cycle)
+				cur_trb->generic.field[3] ^=
+					cpu_to_le32(TRB_CYCLE);
+			xhci_dbg(xhci, "Cancel (unchain) link TRB\n");
+			xhci_dbg(xhci, "Address = %p (0x%llx dma); "
+					"in seg %p (0x%llx dma)\n",
+					cur_trb,
+					(unsigned long long)xhci_trb_virt_to_dma(cur_seg, cur_trb),
+					cur_seg,
+					(unsigned long long)cur_seg->dma);
+		} else {
+			cur_trb->generic.field[0] = 0;
+			cur_trb->generic.field[1] = 0;
+			cur_trb->generic.field[2] = 0;
+			/* Preserve only the cycle bit of this TRB */
+			cur_trb->generic.field[3] &= cpu_to_le32(TRB_CYCLE);
+			/* Flip the cycle bit except on the first or last TRB */
+			if (flip_cycle && cur_trb != cur_td->first_trb &&
+					cur_trb != cur_td->last_trb)
+				cur_trb->generic.field[3] ^=
+					cpu_to_le32(TRB_CYCLE);
+			cur_trb->generic.field[3] |= cpu_to_le32(
+				TRB_TYPE(TRB_TR_NOOP));
+			xhci_dbg(xhci, "TRB to noop at offset 0x%llx\n",
+					(unsigned long long)
+					xhci_trb_virt_to_dma(cur_seg, cur_trb));
+		}
+		if (cur_trb == cur_td->last_trb)
+			break;
+	}
+}
+
+static int queue_set_tr_deq(struct xhci_hcd *xhci, int slot_id,
+		unsigned int ep_index, unsigned int stream_id,
+		struct xhci_segment *deq_seg,
+		union xhci_trb *deq_ptr, u32 cycle_state);
+
+void xhci_queue_new_dequeue_state(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		unsigned int stream_id,
+		struct xhci_dequeue_state *deq_state)
+{
+	struct xhci_virt_ep *ep = &xhci->devs[slot_id]->eps[ep_index];
+
+	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, stream_id,
+			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->ep_state |= SET_DEQ_PENDING;
+}
+
+static void xhci_stop_watchdog_timer_in_irq(struct xhci_hcd *xhci,
+		struct xhci_virt_ep *ep)
+{
+	ep->ep_state &= ~EP_HALT_PENDING;
+	/* Can't del_timer_sync in interrupt, so we attempt to cancel.  If the
+	 * timer is running on another CPU, we don't decrement stop_cmds_pending
+	 * (since we didn't successfully stop the watchdog timer).
+	 */
+	if (del_timer(&ep->stop_cmd_timer))
+		ep->stop_cmds_pending--;
+}
+
+/* Must be called with xhci->lock held in interrupt context */
+static void xhci_giveback_urb_in_irq(struct xhci_hcd *xhci,
+		struct xhci_td *cur_td, int status, char *adjective)
+{
+	struct usb_hcd *hcd;
+	struct urb	*urb;
+	struct urb_priv	*urb_priv;
+
+	urb = cur_td->urb;
+	urb_priv = urb->hcpriv;
+	urb_priv->td_cnt++;
+	hcd = bus_to_hcd(urb->dev->bus);
+
+	/* Only giveback urb when this is the last td in urb */
+	if (urb_priv->td_cnt == urb_priv->length) {
+		if (usb_pipetype(urb->pipe) == PIPE_ISOCHRONOUS) {
+			xhci_to_hcd(xhci)->self.bandwidth_isoc_reqs--;
+			if (xhci_to_hcd(xhci)->self.bandwidth_isoc_reqs	== 0) {
+				if (xhci->quirks & XHCI_AMD_PLL_FIX)
+					usb_amd_quirk_pll_enable();
+			}
+		}
+		usb_hcd_unlink_urb_from_ep(hcd, urb);
+
+		spin_unlock(&xhci->lock);
+		usb_hcd_giveback_urb(hcd, urb, status);
+		xhci_urb_free_priv(xhci, urb_priv);
+		spin_lock(&xhci->lock);
+	}
+}
+
+/*
+ * 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:
+ *
+ *  1. If the HW was in the middle of processing the TD that needs to be
+ *     cancelled, then we must move the ring's dequeue pointer past the last TRB
+ *     in the TD with a Set Dequeue Pointer Command.
+ *  2. Otherwise, we turn all the TRBs in the TD into No-op TRBs (with the chain
+ *     bit cleared) so that the HW will skip over them.
+ */
+static void handle_stopped_endpoint(struct xhci_hcd *xhci,
+		union xhci_trb *trb, struct xhci_event_cmd *event)
+{
+	unsigned int slot_id;
+	unsigned int ep_index;
+	struct xhci_virt_device *virt_dev;
+	struct xhci_ring *ep_ring;
+	struct xhci_virt_ep *ep;
+	struct list_head *entry;
+	struct xhci_td *cur_td = NULL;
+	struct xhci_td *last_unlinked_td;
+
+	struct xhci_dequeue_state deq_state;
+
+	if (unlikely(TRB_TO_SUSPEND_PORT(
+			     le32_to_cpu(xhci->cmd_ring->dequeue->generic.field[3])))) {
+		slot_id = TRB_TO_SLOT_ID(
+			le32_to_cpu(xhci->cmd_ring->dequeue->generic.field[3]));
+		virt_dev = xhci->devs[slot_id];
+		if (virt_dev)
+			handle_cmd_in_cmd_wait_list(xhci, virt_dev,
+				event);
+		else
+			xhci_warn(xhci, "Stop endpoint command "
+				"completion for disabled slot %u\n",
+				slot_id);
+		return;
+	}
+
+	memset(&deq_state, 0, sizeof(deq_state));
+	slot_id = TRB_TO_SLOT_ID(le32_to_cpu(trb->generic.field[3]));
+	ep_index = TRB_TO_EP_INDEX(le32_to_cpu(trb->generic.field[3]));
+	ep = &xhci->devs[slot_id]->eps[ep_index];
+
+	if (list_empty(&ep->cancelled_td_list)) {
+		xhci_stop_watchdog_timer_in_irq(xhci, ep);
+		ep->stopped_td = NULL;
+		ep->stopped_trb = NULL;
+		ring_doorbell_for_active_rings(xhci, slot_id, ep_index);
+		return;
+	}
+
+	/* Fix up the ep ring first, so HW stops executing cancelled TDs.
+	 * We have the xHCI lock, so nothing can modify this list until we drop
+	 * it.  We're also in the event handler, so we can't get re-interrupted
+	 * if another Stop Endpoint command completes
+	 */
+	list_for_each(entry, &ep->cancelled_td_list) {
+		cur_td = list_entry(entry, struct xhci_td, cancelled_td_list);
+		xhci_dbg(xhci, "Removing canceled TD starting at 0x%llx (dma).\n",
+				(unsigned long long)xhci_trb_virt_to_dma(
+					cur_td->start_seg, cur_td->first_trb));
+		ep_ring = xhci_urb_to_transfer_ring(xhci, cur_td->urb);
+		if (!ep_ring) {
+			/* This shouldn't happen unless a driver is mucking
+			 * with the stream ID after submission.  This will
+			 * leave the TD on the hardware ring, and the hardware
+			 * will try to execute it, and may access a buffer
+			 * that has already been freed.  In the best case, the
+			 * hardware will execute it, and the event handler will
+			 * ignore the completion event for that TD, since it was
+			 * removed from the td_list for that endpoint.  In
+			 * short, don't muck with the stream ID after
+			 * submission.
+			 */
+			xhci_warn(xhci, "WARN Cancelled URB %p "
+					"has invalid stream ID %u.\n",
+					cur_td->urb,
+					cur_td->urb->stream_id);
+			goto remove_finished_td;
+		}
+		/*
+		 * If we stopped on the TD we need to cancel, then we have to
+		 * move the xHC endpoint ring dequeue pointer past this TD.
+		 */
+		if (cur_td == ep->stopped_td)
+			xhci_find_new_dequeue_state(xhci, slot_id, ep_index,
+					cur_td->urb->stream_id,
+					cur_td, &deq_state);
+		else
+			td_to_noop(xhci, ep_ring, cur_td, false);
+remove_finished_td:
+		/*
+		 * The event handler won't see a completion for this TD anymore,
+		 * so remove it from the endpoint ring's TD list.  Keep it in
+		 * the cancelled TD list for URB completion later.
+		 */
+		list_del_init(&cur_td->td_list);
+	}
+	last_unlinked_td = cur_td;
+	xhci_stop_watchdog_timer_in_irq(xhci, ep);
+
+	/* If necessary, queue a Set Transfer Ring Dequeue Pointer command */
+	if (deq_state.new_deq_ptr && deq_state.new_deq_seg) {
+		xhci_queue_new_dequeue_state(xhci,
+				slot_id, ep_index,
+				ep->stopped_td->urb->stream_id,
+				&deq_state);
+		xhci_ring_cmd_db(xhci);
+	} else {
+		/* Otherwise ring the doorbell(s) to restart queued transfers */
+		ring_doorbell_for_active_rings(xhci, slot_id, ep_index);
+	}
+	ep->stopped_td = NULL;
+	ep->stopped_trb = NULL;
+
+	/*
+	 * Drop the lock and complete the URBs in the cancelled TD list.
+	 * New TDs to be cancelled might be added to the end of the list before
+	 * we can complete all the URBs for the TDs we already unlinked.
+	 * So stop when we've completed the URB for the last TD we unlinked.
+	 */
+	do {
+		cur_td = list_entry(ep->cancelled_td_list.next,
+				struct xhci_td, cancelled_td_list);
+		list_del_init(&cur_td->cancelled_td_list);
+
+		/* Clean up the cancelled URB */
+		/* Doesn't matter what we pass for status, since the core will
+		 * just overwrite it (because the URB has been unlinked).
+		 */
+		xhci_giveback_urb_in_irq(xhci, cur_td, 0, "cancelled");
+
+		/* Stop processing the cancelled list if the watchdog timer is
+		 * running.
+		 */
+		if (xhci->xhc_state & XHCI_STATE_DYING)
+			return;
+	} while (cur_td != last_unlinked_td);
+
+	/* Return to the event handler with xhci->lock re-acquired */
+}
+
+/* Watchdog timer function for when a stop endpoint command fails to complete.
+ * In this case, we assume the host controller is broken or dying or dead.  The
+ * host may still be completing some other events, so we have to be careful to
+ * let the event ring handler and the URB dequeueing/enqueueing functions know
+ * through xhci->state.
+ *
+ * The timer may also fire if the host takes a very long time to respond to the
+ * command, and the stop endpoint command completion handler cannot delete the
+ * timer before the timer function is called.  Another endpoint cancellation may
+ * sneak in before the timer function can grab the lock, and that may queue
+ * another stop endpoint command and add the timer back.  So we cannot use a
+ * simple flag to say whether there is a pending stop endpoint command for a
+ * particular endpoint.
+ *
+ * Instead we use a combination of that flag and a counter for the number of
+ * pending stop endpoint commands.  If the timer is the tail end of the last
+ * stop endpoint command, and the endpoint's command is still pending, we assume
+ * the host is dying.
+ */
+void xhci_stop_endpoint_command_watchdog(unsigned long arg)
+{
+	struct xhci_hcd *xhci;
+	struct xhci_virt_ep *ep;
+	struct xhci_virt_ep *temp_ep;
+	struct xhci_ring *ring;
+	struct xhci_td *cur_td;
+	int ret, i, j;
+	unsigned long flags;
+
+	ep = (struct xhci_virt_ep *) arg;
+	xhci = ep->xhci;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+
+	ep->stop_cmds_pending--;
+	if (xhci->xhc_state & XHCI_STATE_DYING) {
+		xhci_dbg(xhci, "Stop EP timer ran, but another timer marked "
+				"xHCI as DYING, exiting.\n");
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return;
+	}
+	if (!(ep->stop_cmds_pending == 0 && (ep->ep_state & EP_HALT_PENDING))) {
+		xhci_dbg(xhci, "Stop EP timer ran, but no command pending, "
+				"exiting.\n");
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return;
+	}
+
+	xhci_warn(xhci, "xHCI host not responding to stop endpoint command.\n");
+	xhci_warn(xhci, "Assuming host is dying, halting host.\n");
+	/* Oops, HC is dead or dying or at least not responding to the stop
+	 * endpoint command.
+	 */
+	xhci->xhc_state |= XHCI_STATE_DYING;
+	/* Disable interrupts from the host controller and start halting it */
+	xhci_quiesce(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	ret = xhci_halt(xhci);
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	if (ret < 0) {
+		/* This is bad; the host is not responding to commands and it's
+		 * not allowing itself to be halted.  At least interrupts are
+		 * disabled. If we call usb_hc_died(), it will attempt to
+		 * disconnect all device drivers under this host.  Those
+		 * disconnect() methods will wait for all URBs to be unlinked,
+		 * so we must complete them.
+		 */
+		xhci_warn(xhci, "Non-responsive xHCI host is not halting.\n");
+		xhci_warn(xhci, "Completing active URBs anyway.\n");
+		/* We could turn all TDs on the rings to no-ops.  This won't
+		 * help if the host has cached part of the ring, and is slow if
+		 * we want to preserve the cycle bit.  Skip it and hope the host
+		 * doesn't touch the memory.
+		 */
+	}
+	for (i = 0; i < MAX_HC_SLOTS; i++) {
+		if (!xhci->devs[i])
+			continue;
+		for (j = 0; j < 31; j++) {
+			temp_ep = &xhci->devs[i]->eps[j];
+			ring = temp_ep->ring;
+			if (!ring)
+				continue;
+			xhci_dbg(xhci, "Killing URBs for slot ID %u, "
+					"ep index %u\n", i, j);
+			while (!list_empty(&ring->td_list)) {
+				cur_td = list_first_entry(&ring->td_list,
+						struct xhci_td,
+						td_list);
+				list_del_init(&cur_td->td_list);
+				if (!list_empty(&cur_td->cancelled_td_list))
+					list_del_init(&cur_td->cancelled_td_list);
+				xhci_giveback_urb_in_irq(xhci, cur_td,
+						-ESHUTDOWN, "killed");
+			}
+			while (!list_empty(&temp_ep->cancelled_td_list)) {
+				cur_td = list_first_entry(
+						&temp_ep->cancelled_td_list,
+						struct xhci_td,
+						cancelled_td_list);
+				list_del_init(&cur_td->cancelled_td_list);
+				xhci_giveback_urb_in_irq(xhci, cur_td,
+						-ESHUTDOWN, "killed");
+			}
+		}
+	}
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	xhci_dbg(xhci, "Calling usb_hc_died()\n");
+	usb_hc_died(xhci_to_hcd(xhci)->primary_hcd);
+	xhci_dbg(xhci, "xHCI host controller is dead.\n");
+}
+
+
+static void update_ring_for_set_deq_completion(struct xhci_hcd *xhci,
+		struct xhci_virt_device *dev,
+		struct xhci_ring *ep_ring,
+		unsigned int ep_index)
+{
+	union xhci_trb *dequeue_temp;
+	int num_trbs_free_temp;
+	bool revert = false;
+
+	num_trbs_free_temp = ep_ring->num_trbs_free;
+	dequeue_temp = ep_ring->dequeue;
+
+	/* If we get two back-to-back stalls, and the first stalled transfer
+	 * ends just before a link TRB, the dequeue pointer will be left on
+	 * the link TRB by the code in the while loop.  So we have to update
+	 * the dequeue pointer one segment further, or we'll jump off
+	 * the segment into la-la-land.
+	 */
+	if (last_trb(xhci, ep_ring, ep_ring->deq_seg, ep_ring->dequeue)) {
+		ep_ring->deq_seg = ep_ring->deq_seg->next;
+		ep_ring->dequeue = ep_ring->deq_seg->trbs;
+	}
+
+	while (ep_ring->dequeue != dev->eps[ep_index].queued_deq_ptr) {
+		/* We have more usable TRBs */
+		ep_ring->num_trbs_free++;
+		ep_ring->dequeue++;
+		if (last_trb(xhci, ep_ring, ep_ring->deq_seg,
+				ep_ring->dequeue)) {
+			if (ep_ring->dequeue ==
+					dev->eps[ep_index].queued_deq_ptr)
+				break;
+			ep_ring->deq_seg = ep_ring->deq_seg->next;
+			ep_ring->dequeue = ep_ring->deq_seg->trbs;
+		}
+		if (ep_ring->dequeue == dequeue_temp) {
+			revert = true;
+			break;
+		}
+	}
+
+	if (revert) {
+		xhci_dbg(xhci, "Unable to find new dequeue pointer\n");
+		ep_ring->num_trbs_free = num_trbs_free_temp;
+	}
+}
+
+/*
+ * When we get a completion for a Set Transfer Ring Dequeue Pointer command,
+ * we need to clear the set deq pending flag in the endpoint ring state, so that
+ * the TD queueing code can ring the doorbell again.  We also need to ring the
+ * endpoint doorbell to restart the ring, but only if there aren't more
+ * cancellations pending.
+ */
+static void handle_set_deq_completion(struct xhci_hcd *xhci,
+		struct xhci_event_cmd *event,
+		union xhci_trb *trb)
+{
+	unsigned int slot_id;
+	unsigned int ep_index;
+	unsigned int stream_id;
+	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(le32_to_cpu(trb->generic.field[3]));
+	ep_index = TRB_TO_EP_INDEX(le32_to_cpu(trb->generic.field[3]));
+	stream_id = TRB_TO_STREAM_ID(le32_to_cpu(trb->generic.field[2]));
+	dev = xhci->devs[slot_id];
+
+	ep_ring = xhci_stream_id_to_ring(dev, ep_index, stream_id);
+	if (!ep_ring) {
+		xhci_warn(xhci, "WARN Set TR deq ptr command for "
+				"freed stream ID %u\n",
+				stream_id);
+		/* XXX: Harmless??? */
+		dev->eps[ep_index].ep_state &= ~SET_DEQ_PENDING;
+		return;
+	}
+
+	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(le32_to_cpu(event->status)) != COMP_SUCCESS) {
+		unsigned int ep_state;
+		unsigned int slot_state;
+
+		switch (GET_COMP_CODE(le32_to_cpu(event->status))) {
+		case COMP_TRB_ERR:
+			xhci_warn(xhci, "WARN Set TR Deq Ptr cmd invalid because "
+					"of stream ID configuration\n");
+			break;
+		case COMP_CTX_STATE:
+			xhci_warn(xhci, "WARN Set TR Deq Ptr cmd failed due "
+					"to incorrect slot or ep state.\n");
+			ep_state = le32_to_cpu(ep_ctx->ep_info);
+			ep_state &= EP_STATE_MASK;
+			slot_state = le32_to_cpu(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);
+			break;
+		case COMP_EBADSLT:
+			xhci_warn(xhci, "WARN Set TR Deq Ptr cmd failed because "
+					"slot %u was not enabled.\n", slot_id);
+			break;
+		default:
+			xhci_warn(xhci, "WARN Set TR Deq Ptr cmd with unknown "
+					"completion code of %u.\n",
+				  GET_COMP_CODE(le32_to_cpu(event->status)));
+			break;
+		}
+		/* OK what do we do now?  The endpoint state is hosed, and we
+		 * should never get to this point if the synchronization between
+		 * queueing, and endpoint state are correct.  This might happen
+		 * if the device gets disconnected after we've finished
+		 * cancelling URBs, which might not be an error...
+		 */
+	} else {
+		xhci_dbg(xhci, "Successful Set TR Deq Ptr cmd, deq = @%08llx\n",
+			 le64_to_cpu(ep_ctx->deq));
+		if (xhci_trb_virt_to_dma(dev->eps[ep_index].queued_deq_seg,
+					 dev->eps[ep_index].queued_deq_ptr) ==
+		    (le64_to_cpu(ep_ctx->deq) & ~(EP_CTX_CYCLE_MASK))) {
+			/* Update the ring's dequeue segment and dequeue pointer
+			 * to reflect the new position.
+			 */
+			update_ring_for_set_deq_completion(xhci, dev,
+				ep_ring, ep_index);
+		} else {
+			xhci_warn(xhci, "Mismatch between completed Set TR Deq "
+					"Ptr command & xHCI internal state.\n");
+			xhci_warn(xhci, "ep deq seg = %p, deq ptr = %p\n",
+					dev->eps[ep_index].queued_deq_seg,
+					dev->eps[ep_index].queued_deq_ptr);
+		}
+	}
+
+	dev->eps[ep_index].ep_state &= ~SET_DEQ_PENDING;
+	dev->eps[ep_index].queued_deq_seg = NULL;
+	dev->eps[ep_index].queued_deq_ptr = NULL;
+	/* Restart any rings with pending URBs */
+	ring_doorbell_for_active_rings(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(le32_to_cpu(trb->generic.field[3]));
+	ep_index = TRB_TO_EP_INDEX(le32_to_cpu(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",
+		 GET_COMP_CODE(le32_to_cpu(event->status)));
+
+	/* HW with the reset endpoint quirk needs to have a configure endpoint
+	 * command complete before the endpoint can be used.  Queue that here
+	 * because the HW can't handle two commands being queued in a row.
+	 */
+	if (xhci->quirks & XHCI_RESET_EP_QUIRK) {
+		xhci_dbg(xhci, "Queueing configure endpoint command\n");
+		xhci_queue_configure_endpoint(xhci,
+				xhci->devs[slot_id]->in_ctx->dma, slot_id,
+				false);
+		xhci_ring_cmd_db(xhci);
+	} else {
+		/* Clear our internal halted state and restart the ring(s) */
+		xhci->devs[slot_id]->eps[ep_index].ep_state &= ~EP_HALTED;
+		ring_doorbell_for_active_rings(xhci, slot_id, ep_index);
+	}
+}
+
+/* Complete the command and detele it from the devcie's command queue.
+ */
+static void xhci_complete_cmd_in_cmd_wait_list(struct xhci_hcd *xhci,
+		struct xhci_command *command, u32 status)
+{
+	command->status = status;
+	list_del(&command->cmd_list);
+	if (command->completion)
+		complete(command->completion);
+	else
+		xhci_free_command(xhci, command);
+}
+
+
+/* Check to see if a command in the device's command queue matches this one.
+ * Signal the completion or free the command, and return 1.  Return 0 if the
+ * completed command isn't at the head of the command list.
+ */
+static int handle_cmd_in_cmd_wait_list(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		struct xhci_event_cmd *event)
+{
+	struct xhci_command *command;
+
+	if (list_empty(&virt_dev->cmd_list))
+		return 0;
+
+	command = list_entry(virt_dev->cmd_list.next,
+			struct xhci_command, cmd_list);
+	if (xhci->cmd_ring->dequeue != command->command_trb)
+		return 0;
+
+	xhci_complete_cmd_in_cmd_wait_list(xhci, command,
+			GET_COMP_CODE(le32_to_cpu(event->status)));
+	return 1;
+}
+
+/*
+ * Finding the command trb need to be cancelled and modifying it to
+ * NO OP command. And if the command is in device's command wait
+ * list, finishing and freeing it.
+ *
+ * If we can't find the command trb, we think it had already been
+ * executed.
+ */
+static void xhci_cmd_to_noop(struct xhci_hcd *xhci, struct xhci_cd *cur_cd)
+{
+	struct xhci_segment *cur_seg;
+	union xhci_trb *cmd_trb;
+	u32 cycle_state;
+
+	if (xhci->cmd_ring->dequeue == xhci->cmd_ring->enqueue)
+		return;
+
+	/* find the current segment of command ring */
+	cur_seg = find_trb_seg(xhci->cmd_ring->first_seg,
+			xhci->cmd_ring->dequeue, &cycle_state);
+
+	if (!cur_seg) {
+		xhci_warn(xhci, "Command ring mismatch, dequeue = %p %llx (dma)\n",
+				xhci->cmd_ring->dequeue,
+				(unsigned long long)
+				xhci_trb_virt_to_dma(xhci->cmd_ring->deq_seg,
+					xhci->cmd_ring->dequeue));
+		xhci_debug_ring(xhci, xhci->cmd_ring);
+		xhci_dbg_ring_ptrs(xhci, xhci->cmd_ring);
+		return;
+	}
+
+	/* find the command trb matched by cd from command ring */
+	for (cmd_trb = xhci->cmd_ring->dequeue;
+			cmd_trb != xhci->cmd_ring->enqueue;
+			next_trb(xhci, xhci->cmd_ring, &cur_seg, &cmd_trb)) {
+		/* If the trb is link trb, continue */
+		if (TRB_TYPE_LINK_LE32(cmd_trb->generic.field[3]))
+			continue;
+
+		if (cur_cd->cmd_trb == cmd_trb) {
+
+			/* If the command in device's command list, we should
+			 * finish it and free the command structure.
+			 */
+			if (cur_cd->command)
+				xhci_complete_cmd_in_cmd_wait_list(xhci,
+					cur_cd->command, COMP_CMD_STOP);
+
+			/* get cycle state from the origin command trb */
+			cycle_state = le32_to_cpu(cmd_trb->generic.field[3])
+				& TRB_CYCLE;
+
+			/* modify the command trb to NO OP command */
+			cmd_trb->generic.field[0] = 0;
+			cmd_trb->generic.field[1] = 0;
+			cmd_trb->generic.field[2] = 0;
+			cmd_trb->generic.field[3] = cpu_to_le32(
+					TRB_TYPE(TRB_CMD_NOOP) | cycle_state);
+			break;
+		}
+	}
+}
+
+static void xhci_cancel_cmd_in_cd_list(struct xhci_hcd *xhci)
+{
+	struct xhci_cd *cur_cd, *next_cd;
+
+	if (list_empty(&xhci->cancel_cmd_list))
+		return;
+
+	list_for_each_entry_safe(cur_cd, next_cd,
+			&xhci->cancel_cmd_list, cancel_cmd_list) {
+		xhci_cmd_to_noop(xhci, cur_cd);
+		list_del(&cur_cd->cancel_cmd_list);
+		kfree(cur_cd);
+	}
+}
+
+/*
+ * traversing the cancel_cmd_list. If the command descriptor according
+ * to cmd_trb is found, the function free it and return 1, otherwise
+ * return 0.
+ */
+static int xhci_search_cmd_trb_in_cd_list(struct xhci_hcd *xhci,
+		union xhci_trb *cmd_trb)
+{
+	struct xhci_cd *cur_cd, *next_cd;
+
+	if (list_empty(&xhci->cancel_cmd_list))
+		return 0;
+
+	list_for_each_entry_safe(cur_cd, next_cd,
+			&xhci->cancel_cmd_list, cancel_cmd_list) {
+		if (cur_cd->cmd_trb == cmd_trb) {
+			if (cur_cd->command)
+				xhci_complete_cmd_in_cmd_wait_list(xhci,
+					cur_cd->command, COMP_CMD_STOP);
+			list_del(&cur_cd->cancel_cmd_list);
+			kfree(cur_cd);
+			return 1;
+		}
+	}
+
+	return 0;
+}
+
+/*
+ * If the cmd_trb_comp_code is COMP_CMD_ABORT, we just check whether the
+ * trb pointed by the command ring dequeue pointer is the trb we want to
+ * cancel or not. And if the cmd_trb_comp_code is COMP_CMD_STOP, we will
+ * traverse the cancel_cmd_list to trun the all of the commands according
+ * to command descriptor to NO-OP trb.
+ */
+static int handle_stopped_cmd_ring(struct xhci_hcd *xhci,
+		int cmd_trb_comp_code)
+{
+	int cur_trb_is_good = 0;
+
+	/* Searching the cmd trb pointed by the command ring dequeue
+	 * pointer in command descriptor list. If it is found, free it.
+	 */
+	cur_trb_is_good = xhci_search_cmd_trb_in_cd_list(xhci,
+			xhci->cmd_ring->dequeue);
+
+	if (cmd_trb_comp_code == COMP_CMD_ABORT)
+		xhci->cmd_ring_state = CMD_RING_STATE_STOPPED;
+	else if (cmd_trb_comp_code == COMP_CMD_STOP) {
+		/* traversing the cancel_cmd_list and canceling
+		 * the command according to command descriptor
+		 */
+		xhci_cancel_cmd_in_cd_list(xhci);
+
+		xhci->cmd_ring_state = CMD_RING_STATE_RUNNING;
+		/*
+		 * ring command ring doorbell again to restart the
+		 * command ring
+		 */
+		if (xhci->cmd_ring->dequeue != xhci->cmd_ring->enqueue)
+			xhci_ring_cmd_db(xhci);
+	}
+	return cur_trb_is_good;
+}
+
+static void handle_cmd_completion(struct xhci_hcd *xhci,
+		struct xhci_event_cmd *event)
+{
+	int slot_id = TRB_TO_SLOT_ID(le32_to_cpu(event->flags));
+	u64 cmd_dma;
+	dma_addr_t cmd_dequeue_dma;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	struct xhci_virt_device *virt_dev;
+	unsigned int ep_index;
+	struct xhci_ring *ep_ring;
+	unsigned int ep_state;
+
+	cmd_dma = le64_to_cpu(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? */
+	if (cmd_dequeue_dma == 0) {
+		xhci->error_bitmask |= 1 << 4;
+		return;
+	}
+	/* Does the DMA address match our internal dequeue pointer address? */
+	if (cmd_dma != (u64) cmd_dequeue_dma) {
+		xhci->error_bitmask |= 1 << 5;
+		return;
+	}
+
+	if ((GET_COMP_CODE(le32_to_cpu(event->status)) == COMP_CMD_ABORT) ||
+		(GET_COMP_CODE(le32_to_cpu(event->status)) == COMP_CMD_STOP)) {
+		/* If the return value is 0, we think the trb pointed by
+		 * command ring dequeue pointer is a good trb. The good
+		 * trb means we don't want to cancel the trb, but it have
+		 * been stopped by host. So we should handle it normally.
+		 * Otherwise, driver should invoke inc_deq() and return.
+		 */
+		if (handle_stopped_cmd_ring(xhci,
+				GET_COMP_CODE(le32_to_cpu(event->status)))) {
+			inc_deq(xhci, xhci->cmd_ring);
+			return;
+		}
+	}
+
+	switch (le32_to_cpu(xhci->cmd_ring->dequeue->generic.field[3])
+		& TRB_TYPE_BITMASK) {
+	case TRB_TYPE(TRB_ENABLE_SLOT):
+		if (GET_COMP_CODE(le32_to_cpu(event->status)) == COMP_SUCCESS)
+			xhci->slot_id = slot_id;
+		else
+			xhci->slot_id = 0;
+		complete(&xhci->addr_dev);
+		break;
+	case TRB_TYPE(TRB_DISABLE_SLOT):
+		if (xhci->devs[slot_id]) {
+			if (xhci->quirks & XHCI_EP_LIMIT_QUIRK)
+				/* Delete default control endpoint resources */
+				xhci_free_device_endpoint_resources(xhci,
+						xhci->devs[slot_id], true);
+			xhci_free_virt_device(xhci, slot_id);
+		}
+		break;
+	case TRB_TYPE(TRB_CONFIG_EP):
+		virt_dev = xhci->devs[slot_id];
+		if (handle_cmd_in_cmd_wait_list(xhci, virt_dev, event))
+			break;
+		/*
+		 * Configure endpoint commands can come from the USB core
+		 * configuration or alt setting changes, or because the HW
+		 * needed an extra configure endpoint command after a reset
+		 * endpoint command or streams were being configured.
+		 * If the command was for a halted endpoint, the xHCI driver
+		 * is not waiting on the configure endpoint command.
+		 */
+		ctrl_ctx = xhci_get_input_control_ctx(xhci,
+				virt_dev->in_ctx);
+		/* Input ctx add_flags are the endpoint index plus one */
+		ep_index = xhci_last_valid_endpoint(le32_to_cpu(ctrl_ctx->add_flags)) - 1;
+		/* A usb_set_interface() call directly after clearing a halted
+		 * condition may race on this quirky hardware.  Not worth
+		 * worrying about, since this is prototype hardware.  Not sure
+		 * if this will work for streams, but streams support was
+		 * untested on this prototype.
+		 */
+		if (xhci->quirks & XHCI_RESET_EP_QUIRK &&
+				ep_index != (unsigned int) -1 &&
+		    le32_to_cpu(ctrl_ctx->add_flags) - SLOT_FLAG ==
+		    le32_to_cpu(ctrl_ctx->drop_flags)) {
+			ep_ring = xhci->devs[slot_id]->eps[ep_index].ring;
+			ep_state = xhci->devs[slot_id]->eps[ep_index].ep_state;
+			if (!(ep_state & EP_HALTED))
+				goto bandwidth_change;
+			xhci_dbg(xhci, "Completed config ep cmd - "
+					"last ep index = %d, state = %d\n",
+					ep_index, ep_state);
+			/* Clear internal halted state and restart ring(s) */
+			xhci->devs[slot_id]->eps[ep_index].ep_state &=
+				~EP_HALTED;
+			ring_doorbell_for_active_rings(xhci, slot_id, ep_index);
+			break;
+		}
+bandwidth_change:
+		xhci_dbg(xhci, "Completed config ep cmd\n");
+		xhci->devs[slot_id]->cmd_status =
+			GET_COMP_CODE(le32_to_cpu(event->status));
+		complete(&xhci->devs[slot_id]->cmd_completion);
+		break;
+	case TRB_TYPE(TRB_EVAL_CONTEXT):
+		virt_dev = xhci->devs[slot_id];
+		if (handle_cmd_in_cmd_wait_list(xhci, virt_dev, event))
+			break;
+		xhci->devs[slot_id]->cmd_status = GET_COMP_CODE(le32_to_cpu(event->status));
+		complete(&xhci->devs[slot_id]->cmd_completion);
+		break;
+	case TRB_TYPE(TRB_ADDR_DEV):
+		xhci->devs[slot_id]->cmd_status = GET_COMP_CODE(le32_to_cpu(event->status));
+		complete(&xhci->addr_dev);
+		break;
+	case TRB_TYPE(TRB_STOP_RING):
+		handle_stopped_endpoint(xhci, xhci->cmd_ring->dequeue, event);
+		break;
+	case TRB_TYPE(TRB_SET_DEQ):
+		handle_set_deq_completion(xhci, event, xhci->cmd_ring->dequeue);
+		break;
+	case TRB_TYPE(TRB_CMD_NOOP):
+		break;
+	case TRB_TYPE(TRB_RESET_EP):
+		handle_reset_ep_completion(xhci, event, xhci->cmd_ring->dequeue);
+		break;
+	case TRB_TYPE(TRB_RESET_DEV):
+		xhci_dbg(xhci, "Completed reset device command.\n");
+		slot_id = TRB_TO_SLOT_ID(
+			le32_to_cpu(xhci->cmd_ring->dequeue->generic.field[3]));
+		virt_dev = xhci->devs[slot_id];
+		if (virt_dev)
+			handle_cmd_in_cmd_wait_list(xhci, virt_dev, event);
+		else
+			xhci_warn(xhci, "Reset device command completion "
+					"for disabled slot %u\n", slot_id);
+		break;
+	case TRB_TYPE(TRB_NEC_GET_FW):
+		if (!(xhci->quirks & XHCI_NEC_HOST)) {
+			xhci->error_bitmask |= 1 << 6;
+			break;
+		}
+		xhci_dbg(xhci, "NEC firmware version %2x.%02x\n",
+			 NEC_FW_MAJOR(le32_to_cpu(event->status)),
+			 NEC_FW_MINOR(le32_to_cpu(event->status)));
+		break;
+	default:
+		/* Skip over unknown commands on the event ring */
+		xhci->error_bitmask |= 1 << 6;
+		break;
+	}
+	inc_deq(xhci, xhci->cmd_ring);
+}
+
+static void handle_vendor_event(struct xhci_hcd *xhci,
+		union xhci_trb *event)
+{
+	u32 trb_type;
+
+	trb_type = TRB_FIELD_TO_TYPE(le32_to_cpu(event->generic.field[3]));
+	xhci_dbg(xhci, "Vendor specific event TRB type = %u\n", trb_type);
+	if (trb_type == TRB_NEC_CMD_COMP && (xhci->quirks & XHCI_NEC_HOST))
+		handle_cmd_completion(xhci, &event->event_cmd);
+}
+
+/* @port_id: the one-based port ID from the hardware (indexed from array of all
+ * port registers -- USB 3.0 and USB 2.0).
+ *
+ * Returns a zero-based port number, which is suitable for indexing into each of
+ * the split roothubs' port arrays and bus state arrays.
+ * Add one to it in order to call xhci_find_slot_id_by_port.
+ */
+static unsigned int find_faked_portnum_from_hw_portnum(struct usb_hcd *hcd,
+		struct xhci_hcd *xhci, u32 port_id)
+{
+	unsigned int i;
+	unsigned int num_similar_speed_ports = 0;
+
+	/* port_id from the hardware is 1-based, but port_array[], usb3_ports[],
+	 * and usb2_ports are 0-based indexes.  Count the number of similar
+	 * speed ports, up to 1 port before this port.
+	 */
+	for (i = 0; i < (port_id - 1); i++) {
+		u8 port_speed = xhci->port_array[i];
+
+		/*
+		 * Skip ports that don't have known speeds, or have duplicate
+		 * Extended Capabilities port speed entries.
+		 */
+		if (port_speed == 0 || port_speed == DUPLICATE_ENTRY)
+			continue;
+
+		/*
+		 * USB 3.0 ports are always under a USB 3.0 hub.  USB 2.0 and
+		 * 1.1 ports are under the USB 2.0 hub.  If the port speed
+		 * matches the device speed, it's a similar speed port.
+		 */
+		if ((port_speed == 0x03) == (hcd->speed == HCD_USB3))
+			num_similar_speed_ports++;
+	}
+	return num_similar_speed_ports;
+}
+
+static void handle_device_notification(struct xhci_hcd *xhci,
+		union xhci_trb *event)
+{
+	u32 slot_id;
+	struct usb_device *udev;
+
+	slot_id = TRB_TO_SLOT_ID(event->generic.field[3]);
+	if (!xhci->devs[slot_id]) {
+		xhci_warn(xhci, "Device Notification event for "
+				"unused slot %u\n", slot_id);
+		return;
+	}
+
+	xhci_dbg(xhci, "Device Wake Notification event for slot ID %u\n",
+			slot_id);
+	udev = xhci->devs[slot_id]->udev;
+	if (udev && udev->parent)
+		usb_wakeup_notification(udev->parent, udev->portnum);
+}
+
+static void handle_port_status(struct xhci_hcd *xhci,
+		union xhci_trb *event)
+{
+	struct usb_hcd *hcd;
+	u32 port_id;
+	u32 temp, temp1;
+	int max_ports;
+	int slot_id;
+	unsigned int faked_port_index;
+	u8 major_revision;
+	struct xhci_bus_state *bus_state;
+	__le32 __iomem **port_array;
+	bool bogus_port_status = false;
+
+	/* Port status change events always have a successful completion code */
+	if (GET_COMP_CODE(le32_to_cpu(event->generic.field[2])) != COMP_SUCCESS) {
+		xhci_warn(xhci, "WARN: xHC returned failed port status event\n");
+		xhci->error_bitmask |= 1 << 8;
+	}
+	port_id = GET_PORT_ID(le32_to_cpu(event->generic.field[0]));
+	xhci_dbg(xhci, "Port Status Change Event for port %d\n", port_id);
+
+	max_ports = HCS_MAX_PORTS(xhci->hcs_params1);
+	if ((port_id <= 0) || (port_id > max_ports)) {
+		xhci_warn(xhci, "Invalid port id %d\n", port_id);
+		inc_deq(xhci, xhci->event_ring);
+		return;
+	}
+
+	/* Figure out which usb_hcd this port is attached to:
+	 * is it a USB 3.0 port or a USB 2.0/1.1 port?
+	 */
+	major_revision = xhci->port_array[port_id - 1];
+
+	/* Find the right roothub. */
+	hcd = xhci_to_hcd(xhci);
+	if ((major_revision == 0x03) != (hcd->speed == HCD_USB3))
+		hcd = xhci->shared_hcd;
+
+	if (major_revision == 0) {
+		xhci_warn(xhci, "Event for port %u not in "
+				"Extended Capabilities, ignoring.\n",
+				port_id);
+		bogus_port_status = true;
+		goto cleanup;
+	}
+	if (major_revision == DUPLICATE_ENTRY) {
+		xhci_warn(xhci, "Event for port %u duplicated in"
+				"Extended Capabilities, ignoring.\n",
+				port_id);
+		bogus_port_status = true;
+		goto cleanup;
+	}
+
+	/*
+	 * Hardware port IDs reported by a Port Status Change Event include USB
+	 * 3.0 and USB 2.0 ports.  We want to check if the port has reported a
+	 * resume event, but we first need to translate the hardware port ID
+	 * into the index into the ports on the correct split roothub, and the
+	 * correct bus_state structure.
+	 */
+	bus_state = &xhci->bus_state[hcd_index(hcd)];
+	if (hcd->speed == HCD_USB3)
+		port_array = xhci->usb3_ports;
+	else
+		port_array = xhci->usb2_ports;
+	/* Find the faked port hub number */
+	faked_port_index = find_faked_portnum_from_hw_portnum(hcd, xhci,
+			port_id);
+
+	temp = xhci_readl(xhci, port_array[faked_port_index]);
+	if (hcd->state == HC_STATE_SUSPENDED) {
+		xhci_dbg(xhci, "resume root hub\n");
+		usb_hcd_resume_root_hub(hcd);
+	}
+
+	if ((temp & PORT_PLC) && (temp & PORT_PLS_MASK) == XDEV_RESUME) {
+		xhci_dbg(xhci, "port resume event for port %d\n", port_id);
+
+		temp1 = xhci_readl(xhci, &xhci->op_regs->command);
+		if (!(temp1 & CMD_RUN)) {
+			xhci_warn(xhci, "xHC is not running.\n");
+			goto cleanup;
+		}
+
+		if (DEV_SUPERSPEED(temp)) {
+			xhci_dbg(xhci, "remote wake SS port %d\n", port_id);
+			/* Set a flag to say the port signaled remote wakeup,
+			 * so we can tell the difference between the end of
+			 * device and host initiated resume.
+			 */
+			bus_state->port_remote_wakeup |= 1 << faked_port_index;
+			xhci_test_and_clear_bit(xhci, port_array,
+					faked_port_index, PORT_PLC);
+			xhci_set_link_state(xhci, port_array, faked_port_index,
+						XDEV_U0);
+			/* Need to wait until the next link state change
+			 * indicates the device is actually in U0.
+			 */
+			bogus_port_status = true;
+			goto cleanup;
+		} else {
+			xhci_dbg(xhci, "resume HS port %d\n", port_id);
+			bus_state->resume_done[faked_port_index] = jiffies +
+				msecs_to_jiffies(20);
+			set_bit(faked_port_index, &bus_state->resuming_ports);
+			mod_timer(&hcd->rh_timer,
+				  bus_state->resume_done[faked_port_index]);
+			/* Do the rest in GetPortStatus */
+		}
+	}
+
+	if ((temp & PORT_PLC) && (temp & PORT_PLS_MASK) == XDEV_U0 &&
+			DEV_SUPERSPEED(temp)) {
+		xhci_dbg(xhci, "resume SS port %d finished\n", port_id);
+		/* We've just brought the device into U0 through either the
+		 * Resume state after a device remote wakeup, or through the
+		 * U3Exit state after a host-initiated resume.  If it's a device
+		 * initiated remote wake, don't pass up the link state change,
+		 * so the roothub behavior is consistent with external
+		 * USB 3.0 hub behavior.
+		 */
+		slot_id = xhci_find_slot_id_by_port(hcd, xhci,
+				faked_port_index + 1);
+		if (slot_id && xhci->devs[slot_id])
+			xhci_ring_device(xhci, slot_id);
+		if (bus_state->port_remote_wakeup & (1 << faked_port_index)) {
+			bus_state->port_remote_wakeup &=
+				~(1 << faked_port_index);
+			xhci_test_and_clear_bit(xhci, port_array,
+					faked_port_index, PORT_PLC);
+			usb_wakeup_notification(hcd->self.root_hub,
+					faked_port_index + 1);
+			bogus_port_status = true;
+			goto cleanup;
+		}
+	}
+
+	if (hcd->speed != HCD_USB3)
+		xhci_test_and_clear_bit(xhci, port_array, faked_port_index,
+					PORT_PLC);
+
+cleanup:
+	/* Update event ring dequeue pointer before dropping the lock */
+	inc_deq(xhci, xhci->event_ring);
+
+	/* Don't make the USB core poll the roothub if we got a bad port status
+	 * change event.  Besides, at that point we can't tell which roothub
+	 * (USB 2.0 or USB 3.0) to kick.
+	 */
+	if (bogus_port_status)
+		return;
+
+	/*
+	 * xHCI port-status-change events occur when the "or" of all the
+	 * status-change bits in the portsc register changes from 0 to 1.
+	 * New status changes won't cause an event if any other change
+	 * bits are still set.  When an event occurs, switch over to
+	 * polling to avoid losing status changes.
+	 */
+	xhci_dbg(xhci, "%s: starting port polling.\n", __func__);
+	set_bit(HCD_FLAG_POLL_RH, &hcd->flags);
+	spin_unlock(&xhci->lock);
+	/* Pass this up to the core */
+	usb_hcd_poll_rh_status(hcd);
+	spin_lock(&xhci->lock);
+}
+
+/*
+ * This TD is defined by the TRBs starting at start_trb in start_seg and ending
+ * at end_trb, which may be in another segment.  If the suspect DMA address is a
+ * TRB in this TD, this function returns that TRB's segment.  Otherwise it
+ * returns 0.
+ */
+struct xhci_segment *trb_in_td(struct xhci_segment *start_seg,
+		union xhci_trb	*start_trb,
+		union xhci_trb	*end_trb,
+		dma_addr_t	suspect_dma)
+{
+	dma_addr_t start_dma;
+	dma_addr_t end_seg_dma;
+	dma_addr_t end_trb_dma;
+	struct xhci_segment *cur_seg;
+
+	start_dma = xhci_trb_virt_to_dma(start_seg, start_trb);
+	cur_seg = start_seg;
+
+	do {
+		if (start_dma == 0)
+			return NULL;
+		/* We may get an event for a Link TRB in the middle of a TD */
+		end_seg_dma = xhci_trb_virt_to_dma(cur_seg,
+				&cur_seg->trbs[TRBS_PER_SEGMENT - 1]);
+		/* If the end TRB isn't in this segment, this is set to 0 */
+		end_trb_dma = xhci_trb_virt_to_dma(cur_seg, end_trb);
+
+		if (end_trb_dma > 0) {
+			/* The end TRB is in this segment, so suspect should be here */
+			if (start_dma <= end_trb_dma) {
+				if (suspect_dma >= start_dma && suspect_dma <= end_trb_dma)
+					return cur_seg;
+			} else {
+				/* Case for one segment with
+				 * a TD wrapped around to the top
+				 */
+				if ((suspect_dma >= start_dma &&
+							suspect_dma <= end_seg_dma) ||
+						(suspect_dma >= cur_seg->dma &&
+						 suspect_dma <= end_trb_dma))
+					return cur_seg;
+			}
+			return NULL;
+		} else {
+			/* Might still be somewhere in this segment */
+			if (suspect_dma >= start_dma && suspect_dma <= end_seg_dma)
+				return cur_seg;
+		}
+		cur_seg = cur_seg->next;
+		start_dma = xhci_trb_virt_to_dma(cur_seg, &cur_seg->trbs[0]);
+	} while (cur_seg != start_seg);
+
+	return NULL;
+}
+
+static void xhci_cleanup_halted_endpoint(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		unsigned int stream_id,
+		struct xhci_td *td, union xhci_trb *event_trb)
+{
+	struct xhci_virt_ep *ep = &xhci->devs[slot_id]->eps[ep_index];
+	ep->ep_state |= EP_HALTED;
+	ep->stopped_td = td;
+	ep->stopped_trb = event_trb;
+	ep->stopped_stream = stream_id;
+
+	xhci_queue_reset_ep(xhci, slot_id, ep_index);
+	xhci_cleanup_stalled_ring(xhci, td->urb->dev, ep_index);
+
+	ep->stopped_td = NULL;
+	ep->stopped_trb = NULL;
+	ep->stopped_stream = 0;
+
+	xhci_ring_cmd_db(xhci);
+}
+
+/* Check if an error has halted the endpoint ring.  The class driver will
+ * cleanup the halt for a non-default control endpoint if we indicate a stall.
+ * However, a babble and other errors also halt the endpoint ring, and the class
+ * driver won't clear the halt in that case, so we need to issue a Set Transfer
+ * Ring Dequeue Pointer command manually.
+ */
+static int xhci_requires_manual_halt_cleanup(struct xhci_hcd *xhci,
+		struct xhci_ep_ctx *ep_ctx,
+		unsigned int trb_comp_code)
+{
+	/* TRB completion codes that may require a manual halt cleanup */
+	if (trb_comp_code == COMP_TX_ERR ||
+			trb_comp_code == COMP_BABBLE ||
+			trb_comp_code == COMP_SPLIT_ERR)
+		/* The 0.96 spec says a babbling control endpoint
+		 * is not halted. The 0.96 spec says it is.  Some HW
+		 * claims to be 0.95 compliant, but it halts the control
+		 * endpoint anyway.  Check if a babble halted the
+		 * endpoint.
+		 */
+		if ((ep_ctx->ep_info & cpu_to_le32(EP_STATE_MASK)) ==
+		    cpu_to_le32(EP_STATE_HALTED))
+			return 1;
+
+	return 0;
+}
+
+int xhci_is_vendor_info_code(struct xhci_hcd *xhci, unsigned int trb_comp_code)
+{
+	if (trb_comp_code >= 224 && trb_comp_code <= 255) {
+		/* Vendor defined "informational" completion code,
+		 * treat as not-an-error.
+		 */
+		xhci_dbg(xhci, "Vendor defined info completion code %u\n",
+				trb_comp_code);
+		xhci_dbg(xhci, "Treating code as success.\n");
+		return 1;
+	}
+	return 0;
+}
+
+/*
+ * Finish the td processing, remove the td from td list;
+ * Return 1 if the urb can be given back.
+ */
+static int finish_td(struct xhci_hcd *xhci, struct xhci_td *td,
+	union xhci_trb *event_trb, struct xhci_transfer_event *event,
+	struct xhci_virt_ep *ep, int *status, bool skip)
+{
+	struct xhci_virt_device *xdev;
+	struct xhci_ring *ep_ring;
+	unsigned int slot_id;
+	int ep_index;
+	struct urb *urb = NULL;
+	struct xhci_ep_ctx *ep_ctx;
+	int ret = 0;
+	struct urb_priv	*urb_priv;
+	u32 trb_comp_code;
+
+	slot_id = TRB_TO_SLOT_ID(le32_to_cpu(event->flags));
+	xdev = xhci->devs[slot_id];
+	ep_index = TRB_TO_EP_ID(le32_to_cpu(event->flags)) - 1;
+	ep_ring = xhci_dma_to_transfer_ring(ep, le64_to_cpu(event->buffer));
+	ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
+	trb_comp_code = GET_COMP_CODE(le32_to_cpu(event->transfer_len));
+
+	if (skip)
+		goto td_cleanup;
+
+	if (trb_comp_code == COMP_STOP_INVAL ||
+			trb_comp_code == 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->stopped_td = td;
+		ep->stopped_trb = event_trb;
+		return 0;
+	} else {
+		if (trb_comp_code == 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.  Let the
+			 * USB class driver clear the stall later.
+			 */
+			ep->stopped_td = td;
+			ep->stopped_trb = event_trb;
+			ep->stopped_stream = ep_ring->stream_id;
+		} else if (xhci_requires_manual_halt_cleanup(xhci,
+					ep_ctx, trb_comp_code)) {
+			/* Other types of errors halt the endpoint, but the
+			 * class driver doesn't call usb_reset_endpoint() unless
+			 * the error is -EPIPE.  Clear the halted status in the
+			 * xHCI hardware manually.
+			 */
+			xhci_cleanup_halted_endpoint(xhci,
+					slot_id, ep_index, ep_ring->stream_id,
+					td, event_trb);
+		} else {
+			/* Update ring dequeue pointer */
+			while (ep_ring->dequeue != td->last_trb)
+				inc_deq(xhci, ep_ring);
+			inc_deq(xhci, ep_ring);
+		}
+
+td_cleanup:
+		/* Clean up the endpoint's TD list */
+		urb = td->urb;
+		urb_priv = urb->hcpriv;
+
+		/* Do one last check of the actual transfer length.
+		 * If the host controller said we transferred more data than
+		 * the buffer length, urb->actual_length will be a very big
+		 * number (since it's unsigned).  Play it safe and say we didn't
+		 * transfer anything.
+		 */
+		if (urb->actual_length > urb->transfer_buffer_length) {
+			xhci_warn(xhci, "URB transfer length is wrong, "
+					"xHC issue? req. len = %u, "
+					"act. len = %u\n",
+					urb->transfer_buffer_length,
+					urb->actual_length);
+			urb->actual_length = 0;
+			if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
+				*status = -EREMOTEIO;
+			else
+				*status = 0;
+		}
+		list_del_init(&td->td_list);
+		/* Was this TD slated to be cancelled but completed anyway? */
+		if (!list_empty(&td->cancelled_td_list))
+			list_del_init(&td->cancelled_td_list);
+
+		urb_priv->td_cnt++;
+		/* Giveback the urb when all the tds are completed */
+		if (urb_priv->td_cnt == urb_priv->length) {
+			ret = 1;
+			if (usb_pipetype(urb->pipe) == PIPE_ISOCHRONOUS) {
+				xhci_to_hcd(xhci)->self.bandwidth_isoc_reqs--;
+				if (xhci_to_hcd(xhci)->self.bandwidth_isoc_reqs
+					== 0) {
+					if (xhci->quirks & XHCI_AMD_PLL_FIX)
+						usb_amd_quirk_pll_enable();
+				}
+			}
+		}
+	}
+
+	return ret;
+}
+
+/*
+ * Process control tds, update urb status and actual_length.
+ */
+static int process_ctrl_td(struct xhci_hcd *xhci, struct xhci_td *td,
+	union xhci_trb *event_trb, struct xhci_transfer_event *event,
+	struct xhci_virt_ep *ep, int *status)
+{
+	struct xhci_virt_device *xdev;
+	struct xhci_ring *ep_ring;
+	unsigned int slot_id;
+	int ep_index;
+	struct xhci_ep_ctx *ep_ctx;
+	u32 trb_comp_code;
+
+	slot_id = TRB_TO_SLOT_ID(le32_to_cpu(event->flags));
+	xdev = xhci->devs[slot_id];
+	ep_index = TRB_TO_EP_ID(le32_to_cpu(event->flags)) - 1;
+	ep_ring = xhci_dma_to_transfer_ring(ep, le64_to_cpu(event->buffer));
+	ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
+	trb_comp_code = GET_COMP_CODE(le32_to_cpu(event->transfer_len));
+
+	switch (trb_comp_code) {
+	case COMP_SUCCESS:
+		if (event_trb == ep_ring->dequeue) {
+			xhci_warn(xhci, "WARN: Success on ctrl setup TRB "
+					"without IOC set??\n");
+			*status = -ESHUTDOWN;
+		} else if (event_trb != td->last_trb) {
+			xhci_warn(xhci, "WARN: Success on ctrl data TRB "
+					"without IOC set??\n");
+			*status = -ESHUTDOWN;
+		} else {
+			*status = 0;
+		}
+		break;
+	case COMP_SHORT_TX:
+		if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
+			*status = -EREMOTEIO;
+		else
+			*status = 0;
+		break;
+	case COMP_STOP_INVAL:
+	case COMP_STOP:
+		return finish_td(xhci, td, event_trb, event, ep, status, false);
+	default:
+		if (!xhci_requires_manual_halt_cleanup(xhci,
+					ep_ctx, trb_comp_code))
+			break;
+		xhci_dbg(xhci, "TRB error code %u, "
+				"halted endpoint index = %u\n",
+				trb_comp_code, ep_index);
+		/* else fall through */
+	case COMP_STALL:
+		/* Did we transfer part of the data (middle) phase? */
+		if (event_trb != ep_ring->dequeue &&
+				event_trb != td->last_trb)
+			td->urb->actual_length =
+				td->urb->transfer_buffer_length -
+				EVENT_TRB_LEN(le32_to_cpu(event->transfer_len));
+		else
+			td->urb->actual_length = 0;
+
+		xhci_cleanup_halted_endpoint(xhci,
+			slot_id, ep_index, 0, td, event_trb);
+		return finish_td(xhci, td, event_trb, event, ep, status, true);
+	}
+	/*
+	 * Did we transfer any data, despite the errors that might have
+	 * happened?  I.e. did we get past the setup stage?
+	 */
+	if (event_trb != ep_ring->dequeue) {
+		/* The event was for the status stage */
+		if (event_trb == td->last_trb) {
+			if (td->urb->actual_length != 0) {
+				/* Don't overwrite a previously set error code
+				 */
+				if ((*status == -EINPROGRESS || *status == 0) &&
+						(td->urb->transfer_flags
+						 & URB_SHORT_NOT_OK))
+					/* 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? */
+			td->urb->actual_length =
+				td->urb->transfer_buffer_length -
+				EVENT_TRB_LEN(le32_to_cpu(event->transfer_len));
+			xhci_dbg(xhci, "Waiting for status "
+					"stage event\n");
+			return 0;
+		}
+	}
+
+	return finish_td(xhci, td, event_trb, event, ep, status, false);
+}
+
+/*
+ * Process isochronous tds, update urb packet status and actual_length.
+ */
+static int process_isoc_td(struct xhci_hcd *xhci, struct xhci_td *td,
+	union xhci_trb *event_trb, struct xhci_transfer_event *event,
+	struct xhci_virt_ep *ep, int *status)
+{
+	struct xhci_ring *ep_ring;
+	struct urb_priv *urb_priv;
+	int idx;
+	int len = 0;
+	union xhci_trb *cur_trb;
+	struct xhci_segment *cur_seg;
+	struct usb_iso_packet_descriptor *frame;
+	u32 trb_comp_code;
+	bool skip_td = false;
+
+	ep_ring = xhci_dma_to_transfer_ring(ep, le64_to_cpu(event->buffer));
+	trb_comp_code = GET_COMP_CODE(le32_to_cpu(event->transfer_len));
+	urb_priv = td->urb->hcpriv;
+	idx = urb_priv->td_cnt;
+	frame = &td->urb->iso_frame_desc[idx];
+
+	/* handle completion code */
+	switch (trb_comp_code) {
+	case COMP_SUCCESS:
+		if (EVENT_TRB_LEN(le32_to_cpu(event->transfer_len)) == 0) {
+			frame->status = 0;
+			break;
+		}
+		if ((xhci->quirks & XHCI_TRUST_TX_LENGTH))
+			trb_comp_code = COMP_SHORT_TX;
+	case COMP_SHORT_TX:
+		frame->status = td->urb->transfer_flags & URB_SHORT_NOT_OK ?
+				-EREMOTEIO : 0;
+		break;
+	case COMP_BW_OVER:
+		frame->status = -ECOMM;
+		skip_td = true;
+		break;
+	case COMP_BUFF_OVER:
+	case COMP_BABBLE:
+		frame->status = -EOVERFLOW;
+		skip_td = true;
+		break;
+	case COMP_DEV_ERR:
+	case COMP_STALL:
+	case COMP_TX_ERR:
+		frame->status = -EPROTO;
+		skip_td = true;
+		break;
+	case COMP_STOP:
+	case COMP_STOP_INVAL:
+		break;
+	default:
+		frame->status = -1;
+		break;
+	}
+
+	if (trb_comp_code == COMP_SUCCESS || skip_td) {
+		frame->actual_length = frame->length;
+		td->urb->actual_length += frame->length;
+	} else {
+		for (cur_trb = ep_ring->dequeue,
+		     cur_seg = ep_ring->deq_seg; cur_trb != event_trb;
+		     next_trb(xhci, ep_ring, &cur_seg, &cur_trb)) {
+			if (!TRB_TYPE_NOOP_LE32(cur_trb->generic.field[3]) &&
+			    !TRB_TYPE_LINK_LE32(cur_trb->generic.field[3]))
+				len += TRB_LEN(le32_to_cpu(cur_trb->generic.field[2]));
+		}
+		len += TRB_LEN(le32_to_cpu(cur_trb->generic.field[2])) -
+			EVENT_TRB_LEN(le32_to_cpu(event->transfer_len));
+
+		if (trb_comp_code != COMP_STOP_INVAL) {
+			frame->actual_length = len;
+			td->urb->actual_length += len;
+		}
+	}
+
+	return finish_td(xhci, td, event_trb, event, ep, status, false);
+}
+
+static int skip_isoc_td(struct xhci_hcd *xhci, struct xhci_td *td,
+			struct xhci_transfer_event *event,
+			struct xhci_virt_ep *ep, int *status)
+{
+	struct xhci_ring *ep_ring;
+	struct urb_priv *urb_priv;
+	struct usb_iso_packet_descriptor *frame;
+	int idx;
+
+	ep_ring = xhci_dma_to_transfer_ring(ep, le64_to_cpu(event->buffer));
+	urb_priv = td->urb->hcpriv;
+	idx = urb_priv->td_cnt;
+	frame = &td->urb->iso_frame_desc[idx];
+
+	/* The transfer is partly done. */
+	frame->status = -EXDEV;
+
+	/* calc actual length */
+	frame->actual_length = 0;
+
+	/* Update ring dequeue pointer */
+	while (ep_ring->dequeue != td->last_trb)
+		inc_deq(xhci, ep_ring);
+	inc_deq(xhci, ep_ring);
+
+	return finish_td(xhci, td, NULL, event, ep, status, true);
+}
+
+/*
+ * Process bulk and interrupt tds, update urb status and actual_length.
+ */
+static int process_bulk_intr_td(struct xhci_hcd *xhci, struct xhci_td *td,
+	union xhci_trb *event_trb, struct xhci_transfer_event *event,
+	struct xhci_virt_ep *ep, int *status)
+{
+	struct xhci_ring *ep_ring;
+	union xhci_trb *cur_trb;
+	struct xhci_segment *cur_seg;
+	u32 trb_comp_code;
+
+	ep_ring = xhci_dma_to_transfer_ring(ep, le64_to_cpu(event->buffer));
+	trb_comp_code = GET_COMP_CODE(le32_to_cpu(event->transfer_len));
+
+	switch (trb_comp_code) {
+	case COMP_SUCCESS:
+		/* Double check that the HW transferred everything. */
+		if (event_trb != td->last_trb ||
+		    EVENT_TRB_LEN(le32_to_cpu(event->transfer_len)) != 0) {
+			xhci_warn(xhci, "WARN Successful completion "
+					"on short TX\n");
+			if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
+				*status = -EREMOTEIO;
+			else
+				*status = 0;
+			if ((xhci->quirks & XHCI_TRUST_TX_LENGTH))
+				trb_comp_code = COMP_SHORT_TX;
+		} else {
+			*status = 0;
+		}
+		break;
+	case COMP_SHORT_TX:
+		if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
+			*status = -EREMOTEIO;
+		else
+			*status = 0;
+		break;
+	default:
+		/* Others already handled above */
+		break;
+	}
+	if (trb_comp_code == COMP_SHORT_TX)
+		xhci_dbg(xhci, "ep %#x - asked for %d bytes, "
+				"%d bytes untransferred\n",
+				td->urb->ep->desc.bEndpointAddress,
+				td->urb->transfer_buffer_length,
+				EVENT_TRB_LEN(le32_to_cpu(event->transfer_len)));
+	/* Fast path - was this the last TRB in the TD for this URB? */
+	if (event_trb == td->last_trb) {
+		if (EVENT_TRB_LEN(le32_to_cpu(event->transfer_len)) != 0) {
+			td->urb->actual_length =
+				td->urb->transfer_buffer_length -
+				EVENT_TRB_LEN(le32_to_cpu(event->transfer_len));
+			if (td->urb->transfer_buffer_length <
+					td->urb->actual_length) {
+				xhci_warn(xhci, "HC gave bad length "
+						"of %d bytes left\n",
+					  EVENT_TRB_LEN(le32_to_cpu(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.
+			 */
+			if (*status == -EREMOTEIO)
+				*status = 0;
+		}
+	} else {
+		/* Slow path - walk the list, starting from the dequeue
+		 * pointer, to get the actual length transferred.
+		 */
+		td->urb->actual_length = 0;
+		for (cur_trb = ep_ring->dequeue, cur_seg = ep_ring->deq_seg;
+				cur_trb != event_trb;
+				next_trb(xhci, ep_ring, &cur_seg, &cur_trb)) {
+			if (!TRB_TYPE_NOOP_LE32(cur_trb->generic.field[3]) &&
+			    !TRB_TYPE_LINK_LE32(cur_trb->generic.field[3]))
+				td->urb->actual_length +=
+					TRB_LEN(le32_to_cpu(cur_trb->generic.field[2]));
+		}
+		/* If the ring didn't stop on a Link or No-op TRB, add
+		 * in the actual bytes transferred from the Normal TRB
+		 */
+		if (trb_comp_code != COMP_STOP_INVAL)
+			td->urb->actual_length +=
+				TRB_LEN(le32_to_cpu(cur_trb->generic.field[2])) -
+				EVENT_TRB_LEN(le32_to_cpu(event->transfer_len));
+	}
+
+	return finish_td(xhci, td, event_trb, event, ep, status, false);
+}
+
+/*
+ * If this function returns an error condition, it means it got a Transfer
+ * event with a corrupted Slot ID, Endpoint ID, or TRB DMA address.
+ * At this point, the host controller is probably hosed and should be reset.
+ */
+static int handle_tx_event(struct xhci_hcd *xhci,
+		struct xhci_transfer_event *event)
+	__releases(&xhci->lock)
+	__acquires(&xhci->lock)
+{
+	struct xhci_virt_device *xdev;
+	struct xhci_virt_ep *ep;
+	struct xhci_ring *ep_ring;
+	unsigned int slot_id;
+	int ep_index;
+	struct xhci_td *td = NULL;
+	dma_addr_t event_dma;
+	struct xhci_segment *event_seg;
+	union xhci_trb *event_trb;
+	struct urb *urb = NULL;
+	int status = -EINPROGRESS;
+	struct urb_priv *urb_priv;
+	struct xhci_ep_ctx *ep_ctx;
+	struct list_head *tmp;
+	u32 trb_comp_code;
+	int ret = 0;
+	int td_num = 0;
+
+	slot_id = TRB_TO_SLOT_ID(le32_to_cpu(event->flags));
+	xdev = xhci->devs[slot_id];
+	if (!xdev) {
+		xhci_err(xhci, "ERROR Transfer event pointed to bad slot\n");
+		xhci_err(xhci, "@%016llx %08x %08x %08x %08x\n",
+			 (unsigned long long) xhci_trb_virt_to_dma(
+				 xhci->event_ring->deq_seg,
+				 xhci->event_ring->dequeue),
+			 lower_32_bits(le64_to_cpu(event->buffer)),
+			 upper_32_bits(le64_to_cpu(event->buffer)),
+			 le32_to_cpu(event->transfer_len),
+			 le32_to_cpu(event->flags));
+		xhci_dbg(xhci, "Event ring:\n");
+		xhci_debug_segment(xhci, xhci->event_ring->deq_seg);
+		return -ENODEV;
+	}
+
+	/* Endpoint ID is 1 based, our index is zero based */
+	ep_index = TRB_TO_EP_ID(le32_to_cpu(event->flags)) - 1;
+	ep = &xdev->eps[ep_index];
+	ep_ring = xhci_dma_to_transfer_ring(ep, le64_to_cpu(event->buffer));
+	ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
+	if (!ep_ring ||
+	    (le32_to_cpu(ep_ctx->ep_info) & EP_STATE_MASK) ==
+	    EP_STATE_DISABLED) {
+		xhci_err(xhci, "ERROR Transfer event for disabled endpoint "
+				"or incorrect stream ring\n");
+		xhci_err(xhci, "@%016llx %08x %08x %08x %08x\n",
+			 (unsigned long long) xhci_trb_virt_to_dma(
+				 xhci->event_ring->deq_seg,
+				 xhci->event_ring->dequeue),
+			 lower_32_bits(le64_to_cpu(event->buffer)),
+			 upper_32_bits(le64_to_cpu(event->buffer)),
+			 le32_to_cpu(event->transfer_len),
+			 le32_to_cpu(event->flags));
+		xhci_dbg(xhci, "Event ring:\n");
+		xhci_debug_segment(xhci, xhci->event_ring->deq_seg);
+		return -ENODEV;
+	}
+
+	/* Count current td numbers if ep->skip is set */
+	if (ep->skip) {
+		list_for_each(tmp, &ep_ring->td_list)
+			td_num++;
+	}
+
+	event_dma = le64_to_cpu(event->buffer);
+	trb_comp_code = GET_COMP_CODE(le32_to_cpu(event->transfer_len));
+	/* Look for common error cases */
+	switch (trb_comp_code) {
+	/* Skip codes that require special handling depending on
+	 * transfer type
+	 */
+	case COMP_SUCCESS:
+		if (EVENT_TRB_LEN(le32_to_cpu(event->transfer_len)) == 0)
+			break;
+		if (xhci->quirks & XHCI_TRUST_TX_LENGTH)
+			trb_comp_code = COMP_SHORT_TX;
+		else
+			xhci_warn_ratelimited(xhci,
+					"WARN Successful completion on short TX: needs XHCI_TRUST_TX_LENGTH quirk?\n");
+	case COMP_SHORT_TX:
+		break;
+	case COMP_STOP:
+		xhci_dbg(xhci, "Stopped on Transfer TRB\n");
+		break;
+	case COMP_STOP_INVAL:
+		xhci_dbg(xhci, "Stopped on No-op or Link TRB\n");
+		break;
+	case COMP_STALL:
+		xhci_dbg(xhci, "Stalled endpoint\n");
+		ep->ep_state |= EP_HALTED;
+		status = -EPIPE;
+		break;
+	case COMP_TRB_ERR:
+		xhci_warn(xhci, "WARN: TRB error on endpoint\n");
+		status = -EILSEQ;
+		break;
+	case COMP_SPLIT_ERR:
+	case COMP_TX_ERR:
+		xhci_dbg(xhci, "Transfer error on endpoint\n");
+		status = -EPROTO;
+		break;
+	case COMP_BABBLE:
+		xhci_dbg(xhci, "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;
+		break;
+	case COMP_BW_OVER:
+		xhci_warn(xhci, "WARN: bandwidth overrun event on endpoint\n");
+		break;
+	case COMP_BUFF_OVER:
+		xhci_warn(xhci, "WARN: buffer overrun event on endpoint\n");
+		break;
+	case COMP_UNDERRUN:
+		/*
+		 * When the Isoch ring is empty, the xHC will generate
+		 * a Ring Overrun Event for IN Isoch endpoint or Ring
+		 * Underrun Event for OUT Isoch endpoint.
+		 */
+		xhci_dbg(xhci, "underrun event on endpoint\n");
+		if (!list_empty(&ep_ring->td_list))
+			xhci_dbg(xhci, "Underrun Event for slot %d ep %d "
+					"still with TDs queued?\n",
+				 TRB_TO_SLOT_ID(le32_to_cpu(event->flags)),
+				 ep_index);
+		goto cleanup;
+	case COMP_OVERRUN:
+		xhci_dbg(xhci, "overrun event on endpoint\n");
+		if (!list_empty(&ep_ring->td_list))
+			xhci_dbg(xhci, "Overrun Event for slot %d ep %d "
+					"still with TDs queued?\n",
+				 TRB_TO_SLOT_ID(le32_to_cpu(event->flags)),
+				 ep_index);
+		goto cleanup;
+	case COMP_DEV_ERR:
+		xhci_warn(xhci, "WARN: detect an incompatible device");
+		status = -EPROTO;
+		break;
+	case COMP_MISSED_INT:
+		/*
+		 * When encounter missed service error, one or more isoc tds
+		 * may be missed by xHC.
+		 * Set skip flag of the ep_ring; Complete the missed tds as
+		 * short transfer when process the ep_ring next time.
+		 */
+		ep->skip = true;
+		xhci_dbg(xhci, "Miss service interval error, set skip flag\n");
+		goto cleanup;
+	default:
+		if (xhci_is_vendor_info_code(xhci, trb_comp_code)) {
+			status = 0;
+			break;
+		}
+		xhci_warn(xhci, "ERROR Unknown event condition, HC probably "
+				"busted\n");
+		goto cleanup;
+	}
+
+	do {
+		/* This TRB should be in the TD at the head of this ring's
+		 * TD list.
+		 */
+		if (list_empty(&ep_ring->td_list)) {
+			/*
+			 * A stopped endpoint may generate an extra completion
+			 * event if the device was suspended.  Don't print
+			 * warnings.
+			 */
+			if (!(trb_comp_code == COMP_STOP ||
+						trb_comp_code == COMP_STOP_INVAL)) {
+				xhci_warn(xhci, "WARN Event TRB for slot %d ep %d with no TDs queued?\n",
+						TRB_TO_SLOT_ID(le32_to_cpu(event->flags)),
+						ep_index);
+				xhci_dbg(xhci, "Event TRB with TRB type ID %u\n",
+						(le32_to_cpu(event->flags) &
+						 TRB_TYPE_BITMASK)>>10);
+				xhci_print_trb_offsets(xhci, (union xhci_trb *) event);
+			}
+			if (ep->skip) {
+				ep->skip = false;
+				xhci_dbg(xhci, "td_list is empty while skip "
+						"flag set. Clear skip flag.\n");
+			}
+			ret = 0;
+			goto cleanup;
+		}
+
+		/* We've skipped all the TDs on the ep ring when ep->skip set */
+		if (ep->skip && td_num == 0) {
+			ep->skip = false;
+			xhci_dbg(xhci, "All tds on the ep_ring skipped. "
+						"Clear skip flag.\n");
+			ret = 0;
+			goto cleanup;
+		}
+
+		td = list_entry(ep_ring->td_list.next, struct xhci_td, td_list);
+		if (ep->skip)
+			td_num--;
+
+		/* Is this a TRB in the currently executing TD? */
+		event_seg = trb_in_td(ep_ring->deq_seg, ep_ring->dequeue,
+				td->last_trb, event_dma);
+
+		/*
+		 * Skip the Force Stopped Event. The event_trb(event_dma) of FSE
+		 * is not in the current TD pointed by ep_ring->dequeue because
+		 * that the hardware dequeue pointer still at the previous TRB
+		 * of the current TD. The previous TRB maybe a Link TD or the
+		 * last TRB of the previous TD. The command completion handle
+		 * will take care the rest.
+		 */
+		if (!event_seg && trb_comp_code == COMP_STOP_INVAL) {
+			ret = 0;
+			goto cleanup;
+		}
+
+		if (!event_seg) {
+			if (!ep->skip ||
+			    !usb_endpoint_xfer_isoc(&td->urb->ep->desc)) {
+				/* Some host controllers give a spurious
+				 * successful event after a short transfer.
+				 * Ignore it.
+				 */
+				if ((xhci->quirks & XHCI_SPURIOUS_SUCCESS) && 
+						ep_ring->last_td_was_short) {
+					ep_ring->last_td_was_short = false;
+					ret = 0;
+					goto cleanup;
+				}
+				/* HC is busted, give up! */
+				xhci_err(xhci,
+					"ERROR Transfer event TRB DMA ptr not "
+					"part of current TD\n");
+				return -ESHUTDOWN;
+			}
+
+			ret = skip_isoc_td(xhci, td, event, ep, &status);
+			goto cleanup;
+		}
+		if (trb_comp_code == COMP_SHORT_TX)
+			ep_ring->last_td_was_short = true;
+		else
+			ep_ring->last_td_was_short = false;
+
+		if (ep->skip) {
+			xhci_dbg(xhci, "Found td. Clear skip flag.\n");
+			ep->skip = false;
+		}
+
+		event_trb = &event_seg->trbs[(event_dma - event_seg->dma) /
+						sizeof(*event_trb)];
+		/*
+		 * No-op TRB should not trigger interrupts.
+		 * If event_trb is a no-op TRB, it means the
+		 * corresponding TD has been cancelled. Just ignore
+		 * the TD.
+		 */
+		if (TRB_TYPE_NOOP_LE32(event_trb->generic.field[3])) {
+			xhci_dbg(xhci,
+				 "event_trb is a no-op TRB. Skip it\n");
+			goto cleanup;
+		}
+
+		/* Now update the urb's actual_length and give back to
+		 * the core
+		 */
+		if (usb_endpoint_xfer_control(&td->urb->ep->desc))
+			ret = process_ctrl_td(xhci, td, event_trb, event, ep,
+						 &status);
+		else if (usb_endpoint_xfer_isoc(&td->urb->ep->desc))
+			ret = process_isoc_td(xhci, td, event_trb, event, ep,
+						 &status);
+		else
+			ret = process_bulk_intr_td(xhci, td, event_trb, event,
+						 ep, &status);
+
+cleanup:
+		/*
+		 * Do not update event ring dequeue pointer if ep->skip is set.
+		 * Will roll back to continue process missed tds.
+		 */
+		if (trb_comp_code == COMP_MISSED_INT || !ep->skip) {
+			inc_deq(xhci, xhci->event_ring);
+		}
+
+		if (ret) {
+			urb = td->urb;
+			urb_priv = urb->hcpriv;
+			/* Leave the TD around for the reset endpoint function
+			 * to use(but only if it's not a control endpoint,
+			 * since we already queued the Set TR dequeue pointer
+			 * command for stalled control endpoints).
+			 */
+			if (usb_endpoint_xfer_control(&urb->ep->desc) ||
+				(trb_comp_code != COMP_STALL &&
+					trb_comp_code != COMP_BABBLE))
+				xhci_urb_free_priv(xhci, urb_priv);
+			else
+				kfree(urb_priv);
+
+			usb_hcd_unlink_urb_from_ep(bus_to_hcd(urb->dev->bus), urb);
+			if ((urb->actual_length != urb->transfer_buffer_length &&
+						(urb->transfer_flags &
+						 URB_SHORT_NOT_OK)) ||
+					(status != 0 &&
+					 !usb_endpoint_xfer_isoc(&urb->ep->desc)))
+				xhci_dbg(xhci, "Giveback URB %p, len = %d, "
+						"expected = %d, status = %d\n",
+						urb, urb->actual_length,
+						urb->transfer_buffer_length,
+						status);
+			spin_unlock(&xhci->lock);
+			/* EHCI, UHCI, and OHCI always unconditionally set the
+			 * urb->status of an isochronous endpoint to 0.
+			 */
+			if (usb_pipetype(urb->pipe) == PIPE_ISOCHRONOUS)
+				status = 0;
+			usb_hcd_giveback_urb(bus_to_hcd(urb->dev->bus), urb, status);
+			spin_lock(&xhci->lock);
+		}
+
+	/*
+	 * If ep->skip is set, it means there are missed tds on the
+	 * endpoint ring need to take care of.
+	 * Process them as short transfer until reach the td pointed by
+	 * the event.
+	 */
+	} while (ep->skip && trb_comp_code != COMP_MISSED_INT);
+
+	return 0;
+}
+
+/*
+ * This function handles all OS-owned events on the event ring.  It may drop
+ * xhci->lock between event processing (e.g. to pass up port status changes).
+ * Returns >0 for "possibly more events to process" (caller should call again),
+ * otherwise 0 if done.  In future, <0 returns should indicate error code.
+ */
+static int xhci_handle_event(struct xhci_hcd *xhci)
+{
+	union xhci_trb *event;
+	int update_ptrs = 1;
+	int ret;
+
+	if (!xhci->event_ring || !xhci->event_ring->dequeue) {
+		xhci->error_bitmask |= 1 << 1;
+		return 0;
+	}
+
+	event = xhci->event_ring->dequeue;
+	/* Does the HC or OS own the TRB? */
+	if ((le32_to_cpu(event->event_cmd.flags) & TRB_CYCLE) !=
+	    xhci->event_ring->cycle_state) {
+		xhci->error_bitmask |= 1 << 2;
+		return 0;
+	}
+
+	/*
+	 * Barrier between reading the TRB_CYCLE (valid) flag above and any
+	 * speculative reads of the event's flags/data below.
+	 */
+	rmb();
+	/* FIXME: Handle more event types. */
+	switch ((le32_to_cpu(event->event_cmd.flags) & TRB_TYPE_BITMASK)) {
+	case TRB_TYPE(TRB_COMPLETION):
+		handle_cmd_completion(xhci, &event->event_cmd);
+		break;
+	case TRB_TYPE(TRB_PORT_STATUS):
+		handle_port_status(xhci, event);
+		update_ptrs = 0;
+		break;
+	case TRB_TYPE(TRB_TRANSFER):
+		ret = handle_tx_event(xhci, &event->trans_event);
+		if (ret < 0)
+			xhci->error_bitmask |= 1 << 9;
+		else
+			update_ptrs = 0;
+		break;
+	case TRB_TYPE(TRB_DEV_NOTE):
+		handle_device_notification(xhci, event);
+		break;
+	default:
+		if ((le32_to_cpu(event->event_cmd.flags) & TRB_TYPE_BITMASK) >=
+		    TRB_TYPE(48))
+			handle_vendor_event(xhci, event);
+		else
+			xhci->error_bitmask |= 1 << 3;
+	}
+	/* Any of the above functions may drop and re-acquire the lock, so check
+	 * to make sure a watchdog timer didn't mark the host as non-responsive.
+	 */
+	if (xhci->xhc_state & XHCI_STATE_DYING) {
+		xhci_dbg(xhci, "xHCI host dying, returning from "
+				"event handler.\n");
+		return 0;
+	}
+
+	if (update_ptrs)
+		/* Update SW event ring dequeue pointer */
+		inc_deq(xhci, xhci->event_ring);
+
+	/* Are there more items on the event ring?  Caller will call us again to
+	 * check.
+	 */
+	return 1;
+}
+
+/*
+ * xHCI spec says we can get an interrupt, and if the HC has an error condition,
+ * we might get bad data out of the event ring.  Section 4.10.2.7 has a list of
+ * indicators of an event TRB error, but we check the status *first* to be safe.
+ */
+irqreturn_t xhci_irq(struct usb_hcd *hcd)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	u32 status;
+	u64 temp_64;
+	union xhci_trb *event_ring_deq;
+	dma_addr_t deq;
+
+	spin_lock(&xhci->lock);
+	/* Check if the xHC generated the interrupt, or the irq is shared */
+	status = xhci_readl(xhci, &xhci->op_regs->status);
+	if (status == 0xffffffff)
+		goto hw_died;
+
+	if (!(status & STS_EINT)) {
+		spin_unlock(&xhci->lock);
+		return IRQ_NONE;
+	}
+	if (status & STS_FATAL) {
+		xhci_warn(xhci, "WARNING: Host System Error\n");
+		xhci_halt(xhci);
+hw_died:
+		spin_unlock(&xhci->lock);
+		return -ESHUTDOWN;
+	}
+
+	/*
+	 * Clear the op reg interrupt status first,
+	 * so we can receive interrupts from other MSI-X interrupters.
+	 * Write 1 to clear the interrupt status.
+	 */
+	status |= STS_EINT;
+	xhci_writel(xhci, status, &xhci->op_regs->status);
+	/* FIXME when MSI-X is supported and there are multiple vectors */
+	/* Clear the MSI-X event interrupt status */
+
+	if (hcd->irq) {
+		u32 irq_pending;
+		/* Acknowledge the PCI interrupt */
+		irq_pending = xhci_readl(xhci, &xhci->ir_set->irq_pending);
+		irq_pending |= IMAN_IP;
+		xhci_writel(xhci, irq_pending, &xhci->ir_set->irq_pending);
+	}
+
+	if (xhci->xhc_state & XHCI_STATE_DYING) {
+		xhci_dbg(xhci, "xHCI dying, ignoring interrupt. "
+				"Shouldn't IRQs be disabled?\n");
+		/* 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);
+		spin_unlock(&xhci->lock);
+
+		return IRQ_HANDLED;
+	}
+
+	event_ring_deq = xhci->event_ring->dequeue;
+	/* FIXME this should be a delayed service routine
+	 * that clears the EHB.
+	 */
+	while (xhci_handle_event(xhci) > 0) {}
+
+	temp_64 = xhci_read_64(xhci, &xhci->ir_set->erst_dequeue);
+	/* If necessary, update the HW's version of the event ring deq ptr. */
+	if (event_ring_deq != xhci->event_ring->dequeue) {
+		deq = xhci_trb_virt_to_dma(xhci->event_ring->deq_seg,
+				xhci->event_ring->dequeue);
+		if (deq == 0)
+			xhci_warn(xhci, "WARN something wrong with SW event "
+					"ring dequeue ptr.\n");
+		/* Update HC event ring dequeue pointer */
+		temp_64 &= ERST_PTR_MASK;
+		temp_64 |= ((u64) deq & (u64) ~ERST_PTR_MASK);
+	}
+
+	/* Clear the event handler busy flag (RW1C); event ring is empty. */
+	temp_64 |= ERST_EHB;
+	xhci_write_64(xhci, temp_64, &xhci->ir_set->erst_dequeue);
+
+	spin_unlock(&xhci->lock);
+
+	return IRQ_HANDLED;
+}
+
+irqreturn_t xhci_msi_irq(int irq, struct usb_hcd *hcd)
+{
+	return xhci_irq(hcd);
+}
+
+/****		Endpoint Ring Operations	****/
+
+/*
+ * Generic function for queueing a TRB on a ring.
+ * The caller must have checked to make sure there's room on the ring.
+ *
+ * @more_trbs_coming:	Will you enqueue more TRBs before calling
+ *			prepare_transfer()?
+ */
+static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring,
+		bool more_trbs_coming,
+		u32 field1, u32 field2, u32 field3, u32 field4)
+{
+	struct xhci_generic_trb *trb;
+
+	trb = &ring->enqueue->generic;
+	trb->field[0] = cpu_to_le32(field1);
+	trb->field[1] = cpu_to_le32(field2);
+	trb->field[2] = cpu_to_le32(field3);
+	trb->field[3] = cpu_to_le32(field4);
+	inc_enq(xhci, ring, more_trbs_coming);
+}
+
+/*
+ * Does various checks on the endpoint ring, and makes it ready to queue num_trbs.
+ * FIXME allocate segments if the ring is full.
+ */
+static int prepare_ring(struct xhci_hcd *xhci, struct xhci_ring *ep_ring,
+		u32 ep_state, unsigned int num_trbs, gfp_t mem_flags)
+{
+	unsigned int num_trbs_needed;
+
+	/* Make sure the endpoint has been added to xHC schedule */
+	switch (ep_state) {
+	case EP_STATE_DISABLED:
+		/*
+		 * USB core changed config/interfaces without notifying us,
+		 * or hardware is reporting the wrong state.
+		 */
+		xhci_warn(xhci, "WARN urb submitted to disabled ep\n");
+		return -ENOENT;
+	case EP_STATE_ERROR:
+		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;
+	default:
+		xhci_err(xhci, "ERROR unknown endpoint state for ep\n");
+		/*
+		 * FIXME issue Configure Endpoint command to try to get the HC
+		 * back into a known state.
+		 */
+		return -EINVAL;
+	}
+
+	while (1) {
+		if (room_on_ring(xhci, ep_ring, num_trbs))
+			break;
+
+		if (ep_ring == xhci->cmd_ring) {
+			xhci_err(xhci, "Do not support expand command ring\n");
+			return -ENOMEM;
+		}
+
+		xhci_dbg(xhci, "ERROR no room on ep ring, "
+					"try ring expansion\n");
+		num_trbs_needed = num_trbs - ep_ring->num_trbs_free;
+		if (xhci_ring_expansion(xhci, ep_ring, num_trbs_needed,
+					mem_flags)) {
+			xhci_err(xhci, "Ring expansion failed\n");
+			return -ENOMEM;
+		}
+	}
+
+	if (enqueue_is_link_trb(ep_ring)) {
+		struct xhci_ring *ring = ep_ring;
+		union xhci_trb *next;
+
+		next = ring->enqueue;
+
+		while (last_trb(xhci, ring, ring->enq_seg, next)) {
+			/* If we're not dealing with 0.95 hardware or isoc rings
+			 * on AMD 0.96 host, clear the chain bit.
+			 */
+			if (!xhci_link_trb_quirk(xhci) &&
+					!(ring->type == TYPE_ISOC &&
+					 (xhci->quirks & XHCI_AMD_0x96_HOST)))
+				next->link.control &= cpu_to_le32(~TRB_CHAIN);
+			else
+				next->link.control |= cpu_to_le32(TRB_CHAIN);
+
+			wmb();
+			next->link.control ^= cpu_to_le32(TRB_CYCLE);
+
+			/* Toggle the cycle bit after the last ring segment. */
+			if (last_trb_on_last_seg(xhci, ring, ring->enq_seg, next)) {
+				ring->cycle_state = (ring->cycle_state ? 0 : 1);
+			}
+			ring->enq_seg = ring->enq_seg->next;
+			ring->enqueue = ring->enq_seg->trbs;
+			next = ring->enqueue;
+		}
+	}
+
+	return 0;
+}
+
+static int prepare_transfer(struct xhci_hcd *xhci,
+		struct xhci_virt_device *xdev,
+		unsigned int ep_index,
+		unsigned int stream_id,
+		unsigned int num_trbs,
+		struct urb *urb,
+		unsigned int td_index,
+		gfp_t mem_flags)
+{
+	int ret;
+	struct urb_priv *urb_priv;
+	struct xhci_td	*td;
+	struct xhci_ring *ep_ring;
+	struct xhci_ep_ctx *ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
+
+	ep_ring = xhci_stream_id_to_ring(xdev, ep_index, stream_id);
+	if (!ep_ring) {
+		xhci_dbg(xhci, "Can't prepare ring for bad stream ID %u\n",
+				stream_id);
+		return -EINVAL;
+	}
+
+	ret = prepare_ring(xhci, ep_ring,
+			   le32_to_cpu(ep_ctx->ep_info) & EP_STATE_MASK,
+			   num_trbs, mem_flags);
+	if (ret)
+		return ret;
+
+	urb_priv = urb->hcpriv;
+	td = urb_priv->td[td_index];
+
+	INIT_LIST_HEAD(&td->td_list);
+	INIT_LIST_HEAD(&td->cancelled_td_list);
+
+	if (td_index == 0) {
+		ret = usb_hcd_link_urb_to_ep(bus_to_hcd(urb->dev->bus), urb);
+		if (unlikely(ret))
+			return ret;
+	}
+
+	td->urb = urb;
+	/* Add this TD to the tail of the endpoint ring's TD list */
+	list_add_tail(&td->td_list, &ep_ring->td_list);
+	td->start_seg = ep_ring->enq_seg;
+	td->first_trb = ep_ring->enqueue;
+
+	urb_priv->td[td_index] = td;
+
+	return 0;
+}
+
+static unsigned int count_sg_trbs_needed(struct xhci_hcd *xhci, struct urb *urb)
+{
+	int num_sgs, num_trbs, running_total, temp, i;
+	struct scatterlist *sg;
+
+	sg = NULL;
+	num_sgs = urb->num_mapped_sgs;
+	temp = urb->transfer_buffer_length;
+
+	num_trbs = 0;
+	for_each_sg(urb->sg, sg, num_sgs, i) {
+		unsigned int len = sg_dma_len(sg);
+
+		/* Scatter gather list entries may cross 64KB boundaries */
+		running_total = TRB_MAX_BUFF_SIZE -
+			(sg_dma_address(sg) & (TRB_MAX_BUFF_SIZE - 1));
+		running_total &= TRB_MAX_BUFF_SIZE - 1;
+		if (running_total != 0)
+			num_trbs++;
+
+		/* How many more 64KB chunks to transfer, how many more TRBs? */
+		while (running_total < sg_dma_len(sg) && running_total < temp) {
+			num_trbs++;
+			running_total += TRB_MAX_BUFF_SIZE;
+		}
+		len = min_t(int, len, temp);
+		temp -= len;
+		if (temp == 0)
+			break;
+	}
+	return num_trbs;
+}
+
+static void check_trb_math(struct urb *urb, int num_trbs, int running_total)
+{
+	if (num_trbs != 0)
+		dev_err(&urb->dev->dev, "%s - ep %#x - Miscalculated number of "
+				"TRBs, %d left\n", __func__,
+				urb->ep->desc.bEndpointAddress, num_trbs);
+	if (running_total != urb->transfer_buffer_length)
+		dev_err(&urb->dev->dev, "%s - ep %#x - Miscalculated tx length, "
+				"queued %#x (%d), asked for %#x (%d)\n",
+				__func__,
+				urb->ep->desc.bEndpointAddress,
+				running_total, running_total,
+				urb->transfer_buffer_length,
+				urb->transfer_buffer_length);
+}
+
+static void giveback_first_trb(struct xhci_hcd *xhci, int slot_id,
+		unsigned int ep_index, unsigned int stream_id, int start_cycle,
+		struct xhci_generic_trb *start_trb)
+{
+	/*
+	 * Pass all the TRBs to the hardware at once and make sure this write
+	 * isn't reordered.
+	 */
+	wmb();
+	if (start_cycle)
+		start_trb->field[3] |= cpu_to_le32(start_cycle);
+	else
+		start_trb->field[3] &= cpu_to_le32(~TRB_CYCLE);
+	xhci_ring_ep_doorbell(xhci, slot_id, ep_index, stream_id);
+}
+
+/*
+ * xHCI uses normal TRBs for both bulk and interrupt.  When the interrupt
+ * endpoint is to be serviced, the xHC will consume (at most) one TD.  A TD
+ * (comprised of sg list entries) can take several service intervals to
+ * transmit.
+ */
+int xhci_queue_intr_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
+		struct urb *urb, int slot_id, unsigned int ep_index)
+{
+	struct xhci_ep_ctx *ep_ctx = xhci_get_ep_ctx(xhci,
+			xhci->devs[slot_id]->out_ctx, ep_index);
+	int xhci_interval;
+	int ep_interval;
+
+	xhci_interval = EP_INTERVAL_TO_UFRAMES(le32_to_cpu(ep_ctx->ep_info));
+	ep_interval = urb->interval;
+	/* Convert to microframes */
+	if (urb->dev->speed == USB_SPEED_LOW ||
+			urb->dev->speed == USB_SPEED_FULL)
+		ep_interval *= 8;
+	/* FIXME change this to a warning and a suggestion to use the new API
+	 * to set the polling interval (once the API is added).
+	 */
+	if (xhci_interval != ep_interval) {
+		if (printk_ratelimit())
+			dev_dbg(&urb->dev->dev, "Driver uses different interval"
+					" (%d microframe%s) than xHCI "
+					"(%d microframe%s)\n",
+					ep_interval,
+					ep_interval == 1 ? "" : "s",
+					xhci_interval,
+					xhci_interval == 1 ? "" : "s");
+		urb->interval = xhci_interval;
+		/* Convert back to frames for LS/FS devices */
+		if (urb->dev->speed == USB_SPEED_LOW ||
+				urb->dev->speed == USB_SPEED_FULL)
+			urb->interval /= 8;
+	}
+	return xhci_queue_bulk_tx(xhci, mem_flags, urb, slot_id, ep_index);
+}
+
+/*
+ * The TD size is the number of bytes remaining in the TD (including this TRB),
+ * right shifted by 10.
+ * It must fit in bits 21:17, so it can't be bigger than 31.
+ */
+static u32 xhci_td_remainder(unsigned int remainder)
+{
+	u32 max = (1 << (21 - 17 + 1)) - 1;
+
+	if ((remainder >> 10) >= max)
+		return max << 17;
+	else
+		return (remainder >> 10) << 17;
+}
+
+/*
+ * For xHCI 1.0 host controllers, TD size is the number of max packet sized
+ * packets remaining in the TD (*not* including this TRB).
+ *
+ * Total TD packet count = total_packet_count =
+ *     DIV_ROUND_UP(TD size in bytes / wMaxPacketSize)
+ *
+ * Packets transferred up to and including this TRB = packets_transferred =
+ *     rounddown(total bytes transferred including this TRB / wMaxPacketSize)
+ *
+ * TD size = total_packet_count - packets_transferred
+ *
+ * It must fit in bits 21:17, so it can't be bigger than 31.
+ * The last TRB in a TD must have the TD size set to zero.
+ */
+static u32 xhci_v1_0_td_remainder(int running_total, int trb_buff_len,
+		unsigned int total_packet_count, struct urb *urb,
+		unsigned int num_trbs_left)
+{
+	int packets_transferred;
+
+	/* One TRB with a zero-length data packet. */
+	if (num_trbs_left == 0 || (running_total == 0 && trb_buff_len == 0))
+		return 0;
+
+	/* All the TRB queueing functions don't count the current TRB in
+	 * running_total.
+	 */
+	packets_transferred = (running_total + trb_buff_len) /
+		GET_MAX_PACKET(usb_endpoint_maxp(&urb->ep->desc));
+
+	if ((total_packet_count - packets_transferred) > 31)
+		return 31 << 17;
+	return (total_packet_count - packets_transferred) << 17;
+}
+
+static int queue_bulk_sg_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
+		struct urb *urb, int slot_id, unsigned int ep_index)
+{
+	struct xhci_ring *ep_ring;
+	unsigned int num_trbs;
+	struct urb_priv *urb_priv;
+	struct xhci_td *td;
+	struct scatterlist *sg;
+	int num_sgs;
+	int trb_buff_len, this_sg_len, running_total;
+	unsigned int total_packet_count;
+	bool first_trb;
+	u64 addr;
+	bool more_trbs_coming;
+
+	struct xhci_generic_trb *start_trb;
+	int start_cycle;
+
+	ep_ring = xhci_urb_to_transfer_ring(xhci, urb);
+	if (!ep_ring)
+		return -EINVAL;
+
+	num_trbs = count_sg_trbs_needed(xhci, urb);
+	num_sgs = urb->num_mapped_sgs;
+	total_packet_count = DIV_ROUND_UP(urb->transfer_buffer_length,
+			usb_endpoint_maxp(&urb->ep->desc));
+
+	trb_buff_len = prepare_transfer(xhci, xhci->devs[slot_id],
+			ep_index, urb->stream_id,
+			num_trbs, urb, 0, mem_flags);
+	if (trb_buff_len < 0)
+		return trb_buff_len;
+
+	urb_priv = urb->hcpriv;
+	td = urb_priv->td[0];
+
+	/*
+	 * Don't give the first TRB to the hardware (by toggling the cycle bit)
+	 * until we've finished creating all the other TRBs.  The ring's cycle
+	 * state may change as we enqueue the other TRBs, so save it too.
+	 */
+	start_trb = &ep_ring->enqueue->generic;
+	start_cycle = ep_ring->cycle_state;
+
+	running_total = 0;
+	/*
+	 * How much data is in the first TRB?
+	 *
+	 * There are three forces at work for TRB buffer pointers and lengths:
+	 * 1. We don't want to walk off the end of this sg-list entry buffer.
+	 * 2. The transfer length that the driver requested may be smaller than
+	 *    the amount of memory allocated for this scatter-gather list.
+	 * 3. TRBs buffers can't cross 64KB boundaries.
+	 */
+	sg = urb->sg;
+	addr = (u64) sg_dma_address(sg);
+	this_sg_len = sg_dma_len(sg);
+	trb_buff_len = TRB_MAX_BUFF_SIZE - (addr & (TRB_MAX_BUFF_SIZE - 1));
+	trb_buff_len = min_t(int, trb_buff_len, this_sg_len);
+	if (trb_buff_len > urb->transfer_buffer_length)
+		trb_buff_len = urb->transfer_buffer_length;
+
+	first_trb = true;
+	/* Queue the first TRB, even if it's zero-length */
+	do {
+		u32 field = 0;
+		u32 length_field = 0;
+		u32 remainder = 0;
+
+		/* Don't change the cycle bit of the first TRB until later */
+		if (first_trb) {
+			first_trb = false;
+			if (start_cycle == 0)
+				field |= 0x1;
+		} else
+			field |= ep_ring->cycle_state;
+
+		/* Chain all the TRBs together; clear the chain bit in the last
+		 * TRB to indicate it's the last TRB in the chain.
+		 */
+		if (num_trbs > 1) {
+			field |= TRB_CHAIN;
+		} else {
+			/* FIXME - add check for ZERO_PACKET flag before this */
+			td->last_trb = ep_ring->enqueue;
+			field |= TRB_IOC;
+		}
+
+		/* Only set interrupt on short packet for IN endpoints */
+		if (usb_urb_dir_in(urb))
+			field |= TRB_ISP;
+
+		if (TRB_MAX_BUFF_SIZE -
+				(addr & (TRB_MAX_BUFF_SIZE - 1)) < trb_buff_len) {
+			xhci_warn(xhci, "WARN: sg dma xfer crosses 64KB boundaries!\n");
+			xhci_dbg(xhci, "Next boundary at %#x, end dma = %#x\n",
+					(unsigned int) (addr + TRB_MAX_BUFF_SIZE) & ~(TRB_MAX_BUFF_SIZE - 1),
+					(unsigned int) addr + trb_buff_len);
+		}
+
+		/* Set the TRB length, TD size, and interrupter fields. */
+		if (xhci->hci_version < 0x100) {
+			remainder = xhci_td_remainder(
+					urb->transfer_buffer_length -
+					running_total);
+		} else {
+			remainder = xhci_v1_0_td_remainder(running_total,
+					trb_buff_len, total_packet_count, urb,
+					num_trbs - 1);
+		}
+		length_field = TRB_LEN(trb_buff_len) |
+			remainder |
+			TRB_INTR_TARGET(0);
+
+		if (num_trbs > 1)
+			more_trbs_coming = true;
+		else
+			more_trbs_coming = false;
+		queue_trb(xhci, ep_ring, more_trbs_coming,
+				lower_32_bits(addr),
+				upper_32_bits(addr),
+				length_field,
+				field | TRB_TYPE(TRB_NORMAL));
+		--num_trbs;
+		running_total += trb_buff_len;
+
+		/* Calculate length for next transfer --
+		 * Are we done queueing all the TRBs for this sg entry?
+		 */
+		this_sg_len -= trb_buff_len;
+		if (this_sg_len == 0) {
+			--num_sgs;
+			if (num_sgs == 0)
+				break;
+			sg = sg_next(sg);
+			addr = (u64) sg_dma_address(sg);
+			this_sg_len = sg_dma_len(sg);
+		} else {
+			addr += trb_buff_len;
+		}
+
+		trb_buff_len = TRB_MAX_BUFF_SIZE -
+			(addr & (TRB_MAX_BUFF_SIZE - 1));
+		trb_buff_len = min_t(int, trb_buff_len, this_sg_len);
+		if (running_total + trb_buff_len > urb->transfer_buffer_length)
+			trb_buff_len =
+				urb->transfer_buffer_length - running_total;
+	} while (running_total < urb->transfer_buffer_length);
+
+	check_trb_math(urb, num_trbs, running_total);
+	giveback_first_trb(xhci, slot_id, ep_index, urb->stream_id,
+			start_cycle, start_trb);
+	return 0;
+}
+
+/* This is very similar to what ehci-q.c qtd_fill() does */
+int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
+		struct urb *urb, int slot_id, unsigned int ep_index)
+{
+	struct xhci_ring *ep_ring;
+	struct urb_priv *urb_priv;
+	struct xhci_td *td;
+	int num_trbs;
+	struct xhci_generic_trb *start_trb;
+	bool first_trb;
+	bool more_trbs_coming;
+	int start_cycle;
+	u32 field, length_field;
+
+	int running_total, trb_buff_len, ret;
+	unsigned int total_packet_count;
+	u64 addr;
+
+	if (urb->num_sgs)
+		return queue_bulk_sg_tx(xhci, mem_flags, urb, slot_id, ep_index);
+
+	ep_ring = xhci_urb_to_transfer_ring(xhci, urb);
+	if (!ep_ring)
+		return -EINVAL;
+
+	num_trbs = 0;
+	/* How much data is (potentially) left before the 64KB boundary? */
+	running_total = TRB_MAX_BUFF_SIZE -
+		(urb->transfer_dma & (TRB_MAX_BUFF_SIZE - 1));
+	running_total &= TRB_MAX_BUFF_SIZE - 1;
+
+	/* If there's some data on this 64KB chunk, or we have to send a
+	 * zero-length transfer, we need at least one TRB
+	 */
+	if (running_total != 0 || urb->transfer_buffer_length == 0)
+		num_trbs++;
+	/* How many more 64KB chunks to transfer, how many more TRBs? */
+	while (running_total < urb->transfer_buffer_length) {
+		num_trbs++;
+		running_total += TRB_MAX_BUFF_SIZE;
+	}
+	/* FIXME: this doesn't deal with URB_ZERO_PACKET - need one more */
+
+	ret = prepare_transfer(xhci, xhci->devs[slot_id],
+			ep_index, urb->stream_id,
+			num_trbs, urb, 0, mem_flags);
+	if (ret < 0)
+		return ret;
+
+	urb_priv = urb->hcpriv;
+	td = urb_priv->td[0];
+
+	/*
+	 * Don't give the first TRB to the hardware (by toggling the cycle bit)
+	 * until we've finished creating all the other TRBs.  The ring's cycle
+	 * state may change as we enqueue the other TRBs, so save it too.
+	 */
+	start_trb = &ep_ring->enqueue->generic;
+	start_cycle = ep_ring->cycle_state;
+
+	running_total = 0;
+	total_packet_count = DIV_ROUND_UP(urb->transfer_buffer_length,
+			usb_endpoint_maxp(&urb->ep->desc));
+	/* How much data is in the first TRB? */
+	addr = (u64) urb->transfer_dma;
+	trb_buff_len = TRB_MAX_BUFF_SIZE -
+		(urb->transfer_dma & (TRB_MAX_BUFF_SIZE - 1));
+	if (trb_buff_len > urb->transfer_buffer_length)
+		trb_buff_len = urb->transfer_buffer_length;
+
+	first_trb = true;
+
+	/* Queue the first TRB, even if it's zero-length */
+	do {
+		u32 remainder = 0;
+		field = 0;
+
+		/* Don't change the cycle bit of the first TRB until later */
+		if (first_trb) {
+			first_trb = false;
+			if (start_cycle == 0)
+				field |= 0x1;
+		} else
+			field |= ep_ring->cycle_state;
+
+		/* Chain all the TRBs together; clear the chain bit in the last
+		 * TRB to indicate it's the last TRB in the chain.
+		 */
+		if (num_trbs > 1) {
+			field |= TRB_CHAIN;
+		} else {
+			/* FIXME - add check for ZERO_PACKET flag before this */
+			td->last_trb = ep_ring->enqueue;
+			field |= TRB_IOC;
+		}
+
+		/* Only set interrupt on short packet for IN endpoints */
+		if (usb_urb_dir_in(urb))
+			field |= TRB_ISP;
+
+		/* Set the TRB length, TD size, and interrupter fields. */
+		if (xhci->hci_version < 0x100) {
+			remainder = xhci_td_remainder(
+					urb->transfer_buffer_length -
+					running_total);
+		} else {
+			remainder = xhci_v1_0_td_remainder(running_total,
+					trb_buff_len, total_packet_count, urb,
+					num_trbs - 1);
+		}
+		length_field = TRB_LEN(trb_buff_len) |
+			remainder |
+			TRB_INTR_TARGET(0);
+
+		if (num_trbs > 1)
+			more_trbs_coming = true;
+		else
+			more_trbs_coming = false;
+		queue_trb(xhci, ep_ring, more_trbs_coming,
+				lower_32_bits(addr),
+				upper_32_bits(addr),
+				length_field,
+				field | TRB_TYPE(TRB_NORMAL));
+		--num_trbs;
+		running_total += trb_buff_len;
+
+		/* Calculate length for next transfer */
+		addr += trb_buff_len;
+		trb_buff_len = urb->transfer_buffer_length - running_total;
+		if (trb_buff_len > TRB_MAX_BUFF_SIZE)
+			trb_buff_len = TRB_MAX_BUFF_SIZE;
+	} while (running_total < urb->transfer_buffer_length);
+
+	check_trb_math(urb, num_trbs, running_total);
+	giveback_first_trb(xhci, slot_id, ep_index, urb->stream_id,
+			start_cycle, start_trb);
+	return 0;
+}
+
+/* Caller must have locked xhci->lock */
+int xhci_queue_ctrl_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
+		struct urb *urb, int slot_id, unsigned int ep_index)
+{
+	struct xhci_ring *ep_ring;
+	int num_trbs;
+	int ret;
+	struct usb_ctrlrequest *setup;
+	struct xhci_generic_trb *start_trb;
+	int start_cycle;
+	u32 field, length_field;
+	struct urb_priv *urb_priv;
+	struct xhci_td *td;
+
+	ep_ring = xhci_urb_to_transfer_ring(xhci, urb);
+	if (!ep_ring)
+		return -EINVAL;
+
+	/*
+	 * Need to copy setup packet into setup TRB, so we can't use the setup
+	 * DMA address.
+	 */
+	if (!urb->setup_packet)
+		return -EINVAL;
+
+	/* 1 TRB for setup, 1 for status */
+	num_trbs = 2;
+	/*
+	 * Don't need to check if we need additional event data and normal TRBs,
+	 * since data in control transfers will never get bigger than 16MB
+	 * XXX: can we get a buffer that crosses 64KB boundaries?
+	 */
+	if (urb->transfer_buffer_length > 0)
+		num_trbs++;
+	ret = prepare_transfer(xhci, xhci->devs[slot_id],
+			ep_index, urb->stream_id,
+			num_trbs, urb, 0, mem_flags);
+	if (ret < 0)
+		return ret;
+
+	urb_priv = urb->hcpriv;
+	td = urb_priv->td[0];
+
+	/*
+	 * Don't give the first TRB to the hardware (by toggling the cycle bit)
+	 * until we've finished creating all the other TRBs.  The ring's cycle
+	 * state may change as we enqueue the other TRBs, so save it too.
+	 */
+	start_trb = &ep_ring->enqueue->generic;
+	start_cycle = ep_ring->cycle_state;
+
+	/* Queue setup TRB - see section 6.4.1.2.1 */
+	/* FIXME better way to translate setup_packet into two u32 fields? */
+	setup = (struct usb_ctrlrequest *) urb->setup_packet;
+	field = 0;
+	field |= TRB_IDT | TRB_TYPE(TRB_SETUP);
+	if (start_cycle == 0)
+		field |= 0x1;
+
+	/* xHCI 1.0 6.4.1.2.1: Transfer Type field */
+	if (xhci->hci_version == 0x100) {
+		if (urb->transfer_buffer_length > 0) {
+			if (setup->bRequestType & USB_DIR_IN)
+				field |= TRB_TX_TYPE(TRB_DATA_IN);
+			else
+				field |= TRB_TX_TYPE(TRB_DATA_OUT);
+		}
+	}
+
+	queue_trb(xhci, ep_ring, true,
+		  setup->bRequestType | setup->bRequest << 8 | le16_to_cpu(setup->wValue) << 16,
+		  le16_to_cpu(setup->wIndex) | le16_to_cpu(setup->wLength) << 16,
+		  TRB_LEN(8) | TRB_INTR_TARGET(0),
+		  /* Immediate data in pointer */
+		  field);
+
+	/* If there's data, queue data TRBs */
+	/* Only set interrupt on short packet for IN endpoints */
+	if (usb_urb_dir_in(urb))
+		field = TRB_ISP | TRB_TYPE(TRB_DATA);
+	else
+		field = TRB_TYPE(TRB_DATA);
+
+	length_field = TRB_LEN(urb->transfer_buffer_length) |
+		xhci_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, true,
+				lower_32_bits(urb->transfer_dma),
+				upper_32_bits(urb->transfer_dma),
+				length_field,
+				field | ep_ring->cycle_state);
+	}
+
+	/* Save the DMA address of the last TRB in the TD */
+	td->last_trb = ep_ring->enqueue;
+
+	/* Queue status TRB - see Table 7 and sections 4.11.2.2 and 6.4.1.2.3 */
+	/* If the device sent data, the status stage is an OUT transfer */
+	if (urb->transfer_buffer_length > 0 && setup->bRequestType & USB_DIR_IN)
+		field = 0;
+	else
+		field = TRB_DIR_IN;
+	queue_trb(xhci, ep_ring, false,
+			0,
+			0,
+			TRB_INTR_TARGET(0),
+			/* Event on completion */
+			field | TRB_IOC | TRB_TYPE(TRB_STATUS) | ep_ring->cycle_state);
+
+	giveback_first_trb(xhci, slot_id, ep_index, 0,
+			start_cycle, start_trb);
+	return 0;
+}
+
+static int count_isoc_trbs_needed(struct xhci_hcd *xhci,
+		struct urb *urb, int i)
+{
+	int num_trbs = 0;
+	u64 addr, td_len;
+
+	addr = (u64) (urb->transfer_dma + urb->iso_frame_desc[i].offset);
+	td_len = urb->iso_frame_desc[i].length;
+
+	num_trbs = DIV_ROUND_UP(td_len + (addr & (TRB_MAX_BUFF_SIZE - 1)),
+			TRB_MAX_BUFF_SIZE);
+	if (num_trbs == 0)
+		num_trbs++;
+
+	return num_trbs;
+}
+
+/*
+ * The transfer burst count field of the isochronous TRB defines the number of
+ * bursts that are required to move all packets in this TD.  Only SuperSpeed
+ * devices can burst up to bMaxBurst number of packets per service interval.
+ * This field is zero based, meaning a value of zero in the field means one
+ * burst.  Basically, for everything but SuperSpeed devices, this field will be
+ * zero.  Only xHCI 1.0 host controllers support this field.
+ */
+static unsigned int xhci_get_burst_count(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		struct urb *urb, unsigned int total_packet_count)
+{
+	unsigned int max_burst;
+
+	if (xhci->hci_version < 0x100 || udev->speed != USB_SPEED_SUPER)
+		return 0;
+
+	max_burst = urb->ep->ss_ep_comp.bMaxBurst;
+	return roundup(total_packet_count, max_burst + 1) - 1;
+}
+
+/*
+ * Returns the number of packets in the last "burst" of packets.  This field is
+ * valid for all speeds of devices.  USB 2.0 devices can only do one "burst", so
+ * the last burst packet count is equal to the total number of packets in the
+ * TD.  SuperSpeed endpoints can have up to 3 bursts.  All but the last burst
+ * must contain (bMaxBurst + 1) number of packets, but the last burst can
+ * contain 1 to (bMaxBurst + 1) packets.
+ */
+static unsigned int xhci_get_last_burst_packet_count(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		struct urb *urb, unsigned int total_packet_count)
+{
+	unsigned int max_burst;
+	unsigned int residue;
+
+	if (xhci->hci_version < 0x100)
+		return 0;
+
+	switch (udev->speed) {
+	case USB_SPEED_SUPER:
+		/* bMaxBurst is zero based: 0 means 1 packet per burst */
+		max_burst = urb->ep->ss_ep_comp.bMaxBurst;
+		residue = total_packet_count % (max_burst + 1);
+		/* If residue is zero, the last burst contains (max_burst + 1)
+		 * number of packets, but the TLBPC field is zero-based.
+		 */
+		if (residue == 0)
+			return max_burst;
+		return residue - 1;
+	default:
+		if (total_packet_count == 0)
+			return 0;
+		return total_packet_count - 1;
+	}
+}
+
+/* This is for isoc transfer */
+static int xhci_queue_isoc_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
+		struct urb *urb, int slot_id, unsigned int ep_index)
+{
+	struct xhci_ring *ep_ring;
+	struct urb_priv *urb_priv;
+	struct xhci_td *td;
+	int num_tds, trbs_per_td;
+	struct xhci_generic_trb *start_trb;
+	bool first_trb;
+	int start_cycle;
+	u32 field, length_field;
+	int running_total, trb_buff_len, td_len, td_remain_len, ret;
+	u64 start_addr, addr;
+	int i, j;
+	bool more_trbs_coming;
+
+	ep_ring = xhci->devs[slot_id]->eps[ep_index].ring;
+
+	num_tds = urb->number_of_packets;
+	if (num_tds < 1) {
+		xhci_dbg(xhci, "Isoc URB with zero packets?\n");
+		return -EINVAL;
+	}
+
+	start_addr = (u64) urb->transfer_dma;
+	start_trb = &ep_ring->enqueue->generic;
+	start_cycle = ep_ring->cycle_state;
+
+	urb_priv = urb->hcpriv;
+	/* Queue the first TRB, even if it's zero-length */
+	for (i = 0; i < num_tds; i++) {
+		unsigned int total_packet_count;
+		unsigned int burst_count;
+		unsigned int residue;
+
+		first_trb = true;
+		running_total = 0;
+		addr = start_addr + urb->iso_frame_desc[i].offset;
+		td_len = urb->iso_frame_desc[i].length;
+		td_remain_len = td_len;
+		total_packet_count = DIV_ROUND_UP(td_len,
+				GET_MAX_PACKET(
+					usb_endpoint_maxp(&urb->ep->desc)));
+		/* A zero-length transfer still involves at least one packet. */
+		if (total_packet_count == 0)
+			total_packet_count++;
+		burst_count = xhci_get_burst_count(xhci, urb->dev, urb,
+				total_packet_count);
+		residue = xhci_get_last_burst_packet_count(xhci,
+				urb->dev, urb, total_packet_count);
+
+		trbs_per_td = count_isoc_trbs_needed(xhci, urb, i);
+
+		ret = prepare_transfer(xhci, xhci->devs[slot_id], ep_index,
+				urb->stream_id, trbs_per_td, urb, i, mem_flags);
+		if (ret < 0) {
+			if (i == 0)
+				return ret;
+			goto cleanup;
+		}
+
+		td = urb_priv->td[i];
+		for (j = 0; j < trbs_per_td; j++) {
+			u32 remainder = 0;
+			field = 0;
+
+			if (first_trb) {
+				field = TRB_TBC(burst_count) |
+					TRB_TLBPC(residue);
+				/* Queue the isoc TRB */
+				field |= TRB_TYPE(TRB_ISOC);
+				/* Assume URB_ISO_ASAP is set */
+				field |= TRB_SIA;
+				if (i == 0) {
+					if (start_cycle == 0)
+						field |= 0x1;
+				} else
+					field |= ep_ring->cycle_state;
+				first_trb = false;
+			} else {
+				/* Queue other normal TRBs */
+				field |= TRB_TYPE(TRB_NORMAL);
+				field |= ep_ring->cycle_state;
+			}
+
+			/* Only set interrupt on short packet for IN EPs */
+			if (usb_urb_dir_in(urb))
+				field |= TRB_ISP;
+
+			/* Chain all the TRBs together; clear the chain bit in
+			 * the last TRB to indicate it's the last TRB in the
+			 * chain.
+			 */
+			if (j < trbs_per_td - 1) {
+				field |= TRB_CHAIN;
+				more_trbs_coming = true;
+			} else {
+				td->last_trb = ep_ring->enqueue;
+				field |= TRB_IOC;
+				if (xhci->hci_version == 0x100 &&
+						!(xhci->quirks &
+							XHCI_AVOID_BEI)) {
+					/* Set BEI bit except for the last td */
+					if (i < num_tds - 1)
+						field |= TRB_BEI;
+				}
+				more_trbs_coming = false;
+			}
+
+			/* Calculate TRB length */
+			trb_buff_len = TRB_MAX_BUFF_SIZE -
+				(addr & ((1 << TRB_MAX_BUFF_SHIFT) - 1));
+			if (trb_buff_len > td_remain_len)
+				trb_buff_len = td_remain_len;
+
+			/* Set the TRB length, TD size, & interrupter fields. */
+			if (xhci->hci_version < 0x100) {
+				remainder = xhci_td_remainder(
+						td_len - running_total);
+			} else {
+				remainder = xhci_v1_0_td_remainder(
+						running_total, trb_buff_len,
+						total_packet_count, urb,
+						(trbs_per_td - j - 1));
+			}
+			length_field = TRB_LEN(trb_buff_len) |
+				remainder |
+				TRB_INTR_TARGET(0);
+
+			queue_trb(xhci, ep_ring, more_trbs_coming,
+				lower_32_bits(addr),
+				upper_32_bits(addr),
+				length_field,
+				field);
+			running_total += trb_buff_len;
+
+			addr += trb_buff_len;
+			td_remain_len -= trb_buff_len;
+		}
+
+		/* Check TD length */
+		if (running_total != td_len) {
+			xhci_err(xhci, "ISOC TD length unmatch\n");
+			ret = -EINVAL;
+			goto cleanup;
+		}
+	}
+
+	if (xhci_to_hcd(xhci)->self.bandwidth_isoc_reqs == 0) {
+		if (xhci->quirks & XHCI_AMD_PLL_FIX)
+			usb_amd_quirk_pll_disable();
+	}
+	xhci_to_hcd(xhci)->self.bandwidth_isoc_reqs++;
+
+	giveback_first_trb(xhci, slot_id, ep_index, urb->stream_id,
+			start_cycle, start_trb);
+	return 0;
+cleanup:
+	/* Clean up a partially enqueued isoc transfer. */
+
+	for (i--; i >= 0; i--)
+		list_del_init(&urb_priv->td[i]->td_list);
+
+	/* Use the first TD as a temporary variable to turn the TDs we've queued
+	 * into No-ops with a software-owned cycle bit. That way the hardware
+	 * won't accidentally start executing bogus TDs when we partially
+	 * overwrite them.  td->first_trb and td->start_seg are already set.
+	 */
+	urb_priv->td[0]->last_trb = ep_ring->enqueue;
+	/* Every TRB except the first & last will have its cycle bit flipped. */
+	td_to_noop(xhci, ep_ring, urb_priv->td[0], true);
+
+	/* Reset the ring enqueue back to the first TRB and its cycle bit. */
+	ep_ring->enqueue = urb_priv->td[0]->first_trb;
+	ep_ring->enq_seg = urb_priv->td[0]->start_seg;
+	ep_ring->cycle_state = start_cycle;
+	ep_ring->num_trbs_free = ep_ring->num_trbs_free_temp;
+	usb_hcd_unlink_urb_from_ep(bus_to_hcd(urb->dev->bus), urb);
+	return ret;
+}
+
+/*
+ * Check transfer ring to guarantee there is enough room for the urb.
+ * Update ISO URB start_frame and interval.
+ * Update interval as xhci_queue_intr_tx does. Just use xhci frame_index to
+ * update the urb->start_frame by now.
+ * Always assume URB_ISO_ASAP set, and NEVER use urb->start_frame as input.
+ */
+int xhci_queue_isoc_tx_prepare(struct xhci_hcd *xhci, gfp_t mem_flags,
+		struct urb *urb, int slot_id, unsigned int ep_index)
+{
+	struct xhci_virt_device *xdev;
+	struct xhci_ring *ep_ring;
+	struct xhci_ep_ctx *ep_ctx;
+	int start_frame;
+	int xhci_interval;
+	int ep_interval;
+	int num_tds, num_trbs, i;
+	int ret;
+
+	xdev = xhci->devs[slot_id];
+	ep_ring = xdev->eps[ep_index].ring;
+	ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
+
+	num_trbs = 0;
+	num_tds = urb->number_of_packets;
+	for (i = 0; i < num_tds; i++)
+		num_trbs += count_isoc_trbs_needed(xhci, urb, i);
+
+	/* Check the ring to guarantee there is enough room for the whole urb.
+	 * Do not insert any td of the urb to the ring if the check failed.
+	 */
+	ret = prepare_ring(xhci, ep_ring, le32_to_cpu(ep_ctx->ep_info) & EP_STATE_MASK,
+			   num_trbs, mem_flags);
+	if (ret)
+		return ret;
+
+	start_frame = xhci_readl(xhci, &xhci->run_regs->microframe_index);
+	start_frame &= 0x3fff;
+
+	urb->start_frame = start_frame;
+	if (urb->dev->speed == USB_SPEED_LOW ||
+			urb->dev->speed == USB_SPEED_FULL)
+		urb->start_frame >>= 3;
+
+	xhci_interval = EP_INTERVAL_TO_UFRAMES(le32_to_cpu(ep_ctx->ep_info));
+	ep_interval = urb->interval;
+	/* Convert to microframes */
+	if (urb->dev->speed == USB_SPEED_LOW ||
+			urb->dev->speed == USB_SPEED_FULL)
+		ep_interval *= 8;
+	/* FIXME change this to a warning and a suggestion to use the new API
+	 * to set the polling interval (once the API is added).
+	 */
+	if (xhci_interval != ep_interval) {
+		if (printk_ratelimit())
+			dev_dbg(&urb->dev->dev, "Driver uses different interval"
+					" (%d microframe%s) than xHCI "
+					"(%d microframe%s)\n",
+					ep_interval,
+					ep_interval == 1 ? "" : "s",
+					xhci_interval,
+					xhci_interval == 1 ? "" : "s");
+		urb->interval = xhci_interval;
+		/* Convert back to frames for LS/FS devices */
+		if (urb->dev->speed == USB_SPEED_LOW ||
+				urb->dev->speed == USB_SPEED_FULL)
+			urb->interval /= 8;
+	}
+	ep_ring->num_trbs_free_temp = ep_ring->num_trbs_free;
+
+	return xhci_queue_isoc_tx(xhci, mem_flags, urb, slot_id, ep_index);
+}
+
+/****		Command Ring Operations		****/
+
+/* Generic function for queueing a command TRB on the command ring.
+ * Check to make sure there's room on the command ring for one command TRB.
+ * Also check that there's room reserved for commands that must not fail.
+ * If this is a command that must not fail, meaning command_must_succeed = TRUE,
+ * then only check for the number of reserved spots.
+ * Don't decrement xhci->cmd_ring_reserved_trbs after we've queued the TRB
+ * because the command event handler may want to resubmit a failed command.
+ */
+static int queue_command(struct xhci_hcd *xhci, u32 field1, u32 field2,
+		u32 field3, u32 field4, bool command_must_succeed)
+{
+	int reserved_trbs = xhci->cmd_ring_reserved_trbs;
+	int ret;
+
+	if (!command_must_succeed)
+		reserved_trbs++;
+
+	ret = prepare_ring(xhci, xhci->cmd_ring, EP_STATE_RUNNING,
+			reserved_trbs, GFP_ATOMIC);
+	if (ret < 0) {
+		xhci_err(xhci, "ERR: No room for command on command ring\n");
+		if (command_must_succeed)
+			xhci_err(xhci, "ERR: Reserved TRB counting for "
+					"unfailable commands failed.\n");
+		return ret;
+	}
+	queue_trb(xhci, xhci->cmd_ring, false, field1, field2, field3,
+			field4 | xhci->cmd_ring->cycle_state);
+	return 0;
+}
+
+/* Queue a slot enable or disable request on the command ring */
+int xhci_queue_slot_control(struct xhci_hcd *xhci, u32 trb_type, u32 slot_id)
+{
+	return queue_command(xhci, 0, 0, 0,
+			TRB_TYPE(trb_type) | SLOT_ID_FOR_TRB(slot_id), false);
+}
+
+/* Queue an address device command TRB */
+int xhci_queue_address_device(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
+		u32 slot_id)
+{
+	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),
+			false);
+}
+
+int xhci_queue_vendor_command(struct xhci_hcd *xhci,
+		u32 field1, u32 field2, u32 field3, u32 field4)
+{
+	return queue_command(xhci, field1, field2, field3, field4, false);
+}
+
+/* Queue a reset device command TRB */
+int xhci_queue_reset_device(struct xhci_hcd *xhci, u32 slot_id)
+{
+	return queue_command(xhci, 0, 0, 0,
+			TRB_TYPE(TRB_RESET_DEV) | SLOT_ID_FOR_TRB(slot_id),
+			false);
+}
+
+/* Queue a configure endpoint command TRB */
+int xhci_queue_configure_endpoint(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
+		u32 slot_id, bool command_must_succeed)
+{
+	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),
+			command_must_succeed);
+}
+
+/* Queue an evaluate context command TRB */
+int xhci_queue_evaluate_context(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
+		u32 slot_id, bool command_must_succeed)
+{
+	return queue_command(xhci, lower_32_bits(in_ctx_ptr),
+			upper_32_bits(in_ctx_ptr), 0,
+			TRB_TYPE(TRB_EVAL_CONTEXT) | SLOT_ID_FOR_TRB(slot_id),
+			command_must_succeed);
+}
+
+/*
+ * Suspend is set to indicate "Stop Endpoint Command" is being issued to stop
+ * activity on an endpoint that is about to be suspended.
+ */
+int xhci_queue_stop_endpoint(struct xhci_hcd *xhci, int slot_id,
+		unsigned int ep_index, int suspend)
+{
+	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_STOP_RING);
+	u32 trb_suspend = SUSPEND_PORT_FOR_TRB(suspend);
+
+	return queue_command(xhci, 0, 0, 0,
+			trb_slot_id | trb_ep_index | type | trb_suspend, false);
+}
+
+/* Set Transfer Ring Dequeue Pointer command.
+ * This should not be used for endpoints that have streams enabled.
+ */
+static int queue_set_tr_deq(struct xhci_hcd *xhci, int slot_id,
+		unsigned int ep_index, unsigned int stream_id,
+		struct xhci_segment *deq_seg,
+		union xhci_trb *deq_ptr, u32 cycle_state)
+{
+	dma_addr_t addr;
+	u32 trb_slot_id = SLOT_ID_FOR_TRB(slot_id);
+	u32 trb_ep_index = EP_ID_FOR_TRB(ep_index);
+	u32 trb_stream_id = STREAM_ID_FOR_TRB(stream_id);
+	u32 type = TRB_TYPE(TRB_SET_DEQ);
+	struct xhci_virt_ep *ep;
+
+	addr = xhci_trb_virt_to_dma(deq_seg, deq_ptr);
+	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 0;
+	}
+	ep = &xhci->devs[slot_id]->eps[ep_index];
+	if ((ep->ep_state & SET_DEQ_PENDING)) {
+		xhci_warn(xhci, "WARN Cannot submit Set TR Deq Ptr\n");
+		xhci_warn(xhci, "A Set TR Deq Ptr command is pending.\n");
+		return 0;
+	}
+	ep->queued_deq_seg = deq_seg;
+	ep->queued_deq_ptr = deq_ptr;
+	return queue_command(xhci, lower_32_bits(addr) | cycle_state,
+			upper_32_bits(addr), trb_stream_id,
+			trb_slot_id | trb_ep_index | type, false);
+}
+
+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,
+			false);
+}
diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
new file mode 100644
index 0000000..b4aa79d
--- /dev/null
+++ b/drivers/usb/host/xhci.c
@@ -0,0 +1,4769 @@
+/*
+ * xHCI host controller driver
+ *
+ * Copyright (C) 2008 Intel Corp.
+ *
+ * Author: Sarah Sharp
+ * Some code borrowed from the Linux EHCI driver.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#include <linux/pci.h>
+#include <linux/irq.h>
+#include <linux/log2.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/slab.h>
+#include <linux/dmi.h>
+
+#include "xhci.h"
+
+#define DRIVER_AUTHOR "Sarah Sharp"
+#define DRIVER_DESC "'eXtensible' Host Controller (xHC) Driver"
+
+/* Some 0.95 hardware can't handle the chain bit on a Link TRB being cleared */
+static int link_quirk;
+module_param(link_quirk, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(link_quirk, "Don't clear the chain bit on a link TRB");
+
+/* TODO: copied from ehci-hcd.c - can this be refactored? */
+/*
+ * xhci_handshake - spin reading hc until handshake completes or fails
+ * @ptr: address of hc register to be read
+ * @mask: bits to look at in result of read
+ * @done: value of those bits when handshake succeeds
+ * @usec: timeout in microseconds
+ *
+ * Returns negative errno, or zero on success
+ *
+ * Success happens when the "mask" bits have the specified value (hardware
+ * handshake done).  There are two failure modes:  "usec" have passed (major
+ * hardware flakeout), or the register reads as all-ones (hardware removed).
+ */
+int xhci_handshake(struct xhci_hcd *xhci, void __iomem *ptr,
+		      u32 mask, u32 done, int usec)
+{
+	u32	result;
+
+	do {
+		result = xhci_readl(xhci, ptr);
+		if (result == ~(u32)0)		/* card removed */
+			return -ENODEV;
+		result &= mask;
+		if (result == done)
+			return 0;
+		udelay(1);
+		usec--;
+	} while (usec > 0);
+	return -ETIMEDOUT;
+}
+
+/*
+ * Disable interrupts and begin the xHCI halting process.
+ */
+void xhci_quiesce(struct xhci_hcd *xhci)
+{
+	u32 halted;
+	u32 cmd;
+	u32 mask;
+
+	mask = ~(XHCI_IRQS);
+	halted = xhci_readl(xhci, &xhci->op_regs->status) & STS_HALT;
+	if (!halted)
+		mask &= ~CMD_RUN;
+
+	cmd = xhci_readl(xhci, &xhci->op_regs->command);
+	cmd &= mask;
+	xhci_writel(xhci, cmd, &xhci->op_regs->command);
+}
+
+/*
+ * Force HC into halt state.
+ *
+ * Disable any IRQs and clear the run/stop bit.
+ * HC will complete any current and actively pipelined transactions, and
+ * should halt within 16 ms of the run/stop bit being cleared.
+ * Read HC Halted bit in the status register to see when the HC is finished.
+ */
+int xhci_halt(struct xhci_hcd *xhci)
+{
+	int ret;
+	xhci_dbg(xhci, "// Halt the HC\n");
+	xhci_quiesce(xhci);
+
+	ret = xhci_handshake(xhci, &xhci->op_regs->status,
+			STS_HALT, STS_HALT, XHCI_MAX_HALT_USEC);
+	if (!ret) {
+		xhci->xhc_state |= XHCI_STATE_HALTED;
+		xhci->cmd_ring_state = CMD_RING_STATE_STOPPED;
+	} else
+		xhci_warn(xhci, "Host not halted after %u microseconds.\n",
+				XHCI_MAX_HALT_USEC);
+	return ret;
+}
+
+/*
+ * Set the run bit and wait for the host to be running.
+ */
+static int xhci_start(struct xhci_hcd *xhci)
+{
+	u32 temp;
+	int ret;
+
+	temp = xhci_readl(xhci, &xhci->op_regs->command);
+	temp |= (CMD_RUN);
+	xhci_dbg(xhci, "// Turn on HC, cmd = 0x%x.\n",
+			temp);
+	xhci_writel(xhci, temp, &xhci->op_regs->command);
+
+	/*
+	 * Wait for the HCHalted Status bit to be 0 to indicate the host is
+	 * running.
+	 */
+	ret = xhci_handshake(xhci, &xhci->op_regs->status,
+			STS_HALT, 0, XHCI_MAX_HALT_USEC);
+	if (ret == -ETIMEDOUT)
+		xhci_err(xhci, "Host took too long to start, "
+				"waited %u microseconds.\n",
+				XHCI_MAX_HALT_USEC);
+	if (!ret)
+		xhci->xhc_state &= ~XHCI_STATE_HALTED;
+	return ret;
+}
+
+/*
+ * Reset a halted HC.
+ *
+ * This resets pipelines, timers, counters, state machines, etc.
+ * Transactions will be terminated immediately, and operational registers
+ * will be set to their defaults.
+ */
+int xhci_reset(struct xhci_hcd *xhci)
+{
+	u32 command;
+	u32 state;
+	int ret, i;
+
+	state = xhci_readl(xhci, &xhci->op_regs->status);
+	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);
+	command |= CMD_RESET;
+	xhci_writel(xhci, command, &xhci->op_regs->command);
+
+	ret = xhci_handshake(xhci, &xhci->op_regs->command,
+			CMD_RESET, 0, 10 * 1000 * 1000);
+	if (ret)
+		return ret;
+
+	xhci_dbg(xhci, "Wait for controller to be ready for doorbell rings\n");
+	/*
+	 * xHCI cannot write to any doorbells or operational registers other
+	 * than status until the "Controller Not Ready" flag is cleared.
+	 */
+	ret = xhci_handshake(xhci, &xhci->op_regs->status,
+			STS_CNR, 0, 10 * 1000 * 1000);
+
+	for (i = 0; i < 2; ++i) {
+		xhci->bus_state[i].port_c_suspend = 0;
+		xhci->bus_state[i].suspended_ports = 0;
+		xhci->bus_state[i].resuming_ports = 0;
+	}
+
+	return ret;
+}
+
+#ifdef CONFIG_PCI
+static int xhci_free_msi(struct xhci_hcd *xhci)
+{
+	int i;
+
+	if (!xhci->msix_entries)
+		return -EINVAL;
+
+	for (i = 0; i < xhci->msix_count; i++)
+		if (xhci->msix_entries[i].vector)
+			free_irq(xhci->msix_entries[i].vector,
+					xhci_to_hcd(xhci));
+	return 0;
+}
+
+/*
+ * Set up MSI
+ */
+static int xhci_setup_msi(struct xhci_hcd *xhci)
+{
+	int ret;
+	struct pci_dev  *pdev = to_pci_dev(xhci_to_hcd(xhci)->self.controller);
+
+	ret = pci_enable_msi(pdev);
+	if (ret) {
+		xhci_dbg(xhci, "failed to allocate MSI entry\n");
+		return ret;
+	}
+
+	ret = request_irq(pdev->irq, (irq_handler_t)xhci_msi_irq,
+				0, "xhci_hcd", xhci_to_hcd(xhci));
+	if (ret) {
+		xhci_dbg(xhci, "disable MSI interrupt\n");
+		pci_disable_msi(pdev);
+	}
+
+	return ret;
+}
+
+/*
+ * Free IRQs
+ * free all IRQs request
+ */
+static void xhci_free_irq(struct xhci_hcd *xhci)
+{
+	struct pci_dev *pdev = to_pci_dev(xhci_to_hcd(xhci)->self.controller);
+	int ret;
+
+	/* return if using legacy interrupt */
+	if (xhci_to_hcd(xhci)->irq > 0)
+		return;
+
+	ret = xhci_free_msi(xhci);
+	if (!ret)
+		return;
+	if (pdev->irq > 0)
+		free_irq(pdev->irq, xhci_to_hcd(xhci));
+
+	return;
+}
+
+/*
+ * Set up MSI-X
+ */
+static int xhci_setup_msix(struct xhci_hcd *xhci)
+{
+	int i, ret = 0;
+	struct usb_hcd *hcd = xhci_to_hcd(xhci);
+	struct pci_dev *pdev = to_pci_dev(hcd->self.controller);
+
+	/*
+	 * calculate number of msi-x vectors supported.
+	 * - HCS_MAX_INTRS: the max number of interrupts the host can handle,
+	 *   with max number of interrupters based on the xhci HCSPARAMS1.
+	 * - num_online_cpus: maximum msi-x vectors per CPUs core.
+	 *   Add additional 1 vector to ensure always available interrupt.
+	 */
+	xhci->msix_count = min(num_online_cpus() + 1,
+				HCS_MAX_INTRS(xhci->hcs_params1));
+
+	xhci->msix_entries =
+		kmalloc((sizeof(struct msix_entry))*xhci->msix_count,
+				GFP_KERNEL);
+	if (!xhci->msix_entries) {
+		xhci_err(xhci, "Failed to allocate MSI-X entries\n");
+		return -ENOMEM;
+	}
+
+	for (i = 0; i < xhci->msix_count; i++) {
+		xhci->msix_entries[i].entry = i;
+		xhci->msix_entries[i].vector = 0;
+	}
+
+	ret = pci_enable_msix(pdev, xhci->msix_entries, xhci->msix_count);
+	if (ret) {
+		xhci_dbg(xhci, "Failed to enable MSI-X\n");
+		goto free_entries;
+	}
+
+	for (i = 0; i < xhci->msix_count; i++) {
+		ret = request_irq(xhci->msix_entries[i].vector,
+				(irq_handler_t)xhci_msi_irq,
+				0, "xhci_hcd", xhci_to_hcd(xhci));
+		if (ret)
+			goto disable_msix;
+	}
+
+	hcd->msix_enabled = 1;
+	return ret;
+
+disable_msix:
+	xhci_dbg(xhci, "disable MSI-X interrupt\n");
+	xhci_free_irq(xhci);
+	pci_disable_msix(pdev);
+free_entries:
+	kfree(xhci->msix_entries);
+	xhci->msix_entries = NULL;
+	return ret;
+}
+
+/* Free any IRQs and disable MSI-X */
+static void xhci_cleanup_msix(struct xhci_hcd *xhci)
+{
+	struct usb_hcd *hcd = xhci_to_hcd(xhci);
+	struct pci_dev *pdev = to_pci_dev(hcd->self.controller);
+
+	xhci_free_irq(xhci);
+
+	if (xhci->msix_entries) {
+		pci_disable_msix(pdev);
+		kfree(xhci->msix_entries);
+		xhci->msix_entries = NULL;
+	} else {
+		pci_disable_msi(pdev);
+	}
+
+	hcd->msix_enabled = 0;
+	return;
+}
+
+static void xhci_msix_sync_irqs(struct xhci_hcd *xhci)
+{
+	int i;
+
+	if (xhci->msix_entries) {
+		for (i = 0; i < xhci->msix_count; i++)
+			synchronize_irq(xhci->msix_entries[i].vector);
+	}
+}
+
+static int xhci_try_enable_msi(struct usb_hcd *hcd)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	struct pci_dev  *pdev = to_pci_dev(xhci_to_hcd(xhci)->self.controller);
+	int ret;
+
+	/*
+	 * Some Fresco Logic host controllers advertise MSI, but fail to
+	 * generate interrupts.  Don't even try to enable MSI.
+	 */
+	if (xhci->quirks & XHCI_BROKEN_MSI)
+		goto legacy_irq;
+
+	/* unregister the legacy interrupt */
+	if (hcd->irq)
+		free_irq(hcd->irq, hcd);
+	hcd->irq = 0;
+
+	ret = xhci_setup_msix(xhci);
+	if (ret)
+		/* fall back to msi*/
+		ret = xhci_setup_msi(xhci);
+
+	if (!ret)
+		/* hcd->irq is 0, we have MSI */
+		return 0;
+
+	if (!pdev->irq) {
+		xhci_err(xhci, "No msi-x/msi found and no IRQ in BIOS\n");
+		return -EINVAL;
+	}
+
+ legacy_irq:
+	/* fall back to legacy interrupt*/
+	ret = request_irq(pdev->irq, &usb_hcd_irq, IRQF_SHARED,
+			hcd->irq_descr, hcd);
+	if (ret) {
+		xhci_err(xhci, "request interrupt %d failed\n",
+				pdev->irq);
+		return ret;
+	}
+	hcd->irq = pdev->irq;
+	return 0;
+}
+
+#else
+
+static int xhci_try_enable_msi(struct usb_hcd *hcd)
+{
+	return 0;
+}
+
+static void xhci_cleanup_msix(struct xhci_hcd *xhci)
+{
+}
+
+static void xhci_msix_sync_irqs(struct xhci_hcd *xhci)
+{
+}
+
+#endif
+
+static void compliance_mode_recovery(unsigned long arg)
+{
+	struct xhci_hcd *xhci;
+	struct usb_hcd *hcd;
+	u32 temp;
+	int i;
+
+	xhci = (struct xhci_hcd *)arg;
+
+	for (i = 0; i < xhci->num_usb3_ports; i++) {
+		temp = xhci_readl(xhci, xhci->usb3_ports[i]);
+		if ((temp & PORT_PLS_MASK) == USB_SS_PORT_LS_COMP_MOD) {
+			/*
+			 * Compliance Mode Detected. Letting USB Core
+			 * handle the Warm Reset
+			 */
+			xhci_dbg(xhci, "Compliance mode detected->port %d\n",
+					i + 1);
+			xhci_dbg(xhci, "Attempting compliance mode recovery\n");
+			hcd = xhci->shared_hcd;
+
+			if (hcd->state == HC_STATE_SUSPENDED)
+				usb_hcd_resume_root_hub(hcd);
+
+			usb_hcd_poll_rh_status(hcd);
+		}
+	}
+
+	if (xhci->port_status_u0 != ((1 << xhci->num_usb3_ports)-1))
+		mod_timer(&xhci->comp_mode_recovery_timer,
+			jiffies + msecs_to_jiffies(COMP_MODE_RCVRY_MSECS));
+}
+
+/*
+ * Quirk to work around issue generated by the SN65LVPE502CP USB3.0 re-driver
+ * that causes ports behind that hardware to enter compliance mode sometimes.
+ * The quirk creates a timer that polls every 2 seconds the link state of
+ * each host controller's port and recovers it by issuing a Warm reset
+ * if Compliance mode is detected, otherwise the port will become "dead" (no
+ * device connections or disconnections will be detected anymore). Becasue no
+ * status event is generated when entering compliance mode (per xhci spec),
+ * this quirk is needed on systems that have the failing hardware installed.
+ */
+static void compliance_mode_recovery_timer_init(struct xhci_hcd *xhci)
+{
+	xhci->port_status_u0 = 0;
+	init_timer(&xhci->comp_mode_recovery_timer);
+
+	xhci->comp_mode_recovery_timer.data = (unsigned long) xhci;
+	xhci->comp_mode_recovery_timer.function = compliance_mode_recovery;
+	xhci->comp_mode_recovery_timer.expires = jiffies +
+			msecs_to_jiffies(COMP_MODE_RCVRY_MSECS);
+
+	set_timer_slack(&xhci->comp_mode_recovery_timer,
+			msecs_to_jiffies(COMP_MODE_RCVRY_MSECS));
+	add_timer(&xhci->comp_mode_recovery_timer);
+	xhci_dbg(xhci, "Compliance mode recovery timer initialized\n");
+}
+
+/*
+ * This function identifies the systems that have installed the SN65LVPE502CP
+ * USB3.0 re-driver and that need the Compliance Mode Quirk.
+ * Systems:
+ * Vendor: Hewlett-Packard -> System Models: Z420, Z620 and Z820
+ */
+static bool compliance_mode_recovery_timer_quirk_check(void)
+{
+	const char *dmi_product_name, *dmi_sys_vendor;
+
+	dmi_product_name = dmi_get_system_info(DMI_PRODUCT_NAME);
+	dmi_sys_vendor = dmi_get_system_info(DMI_SYS_VENDOR);
+	if (!dmi_product_name || !dmi_sys_vendor)
+		return false;
+
+	if (!(strstr(dmi_sys_vendor, "Hewlett-Packard")))
+		return false;
+
+	if (strstr(dmi_product_name, "Z420") ||
+			strstr(dmi_product_name, "Z620") ||
+			strstr(dmi_product_name, "Z820") ||
+			strstr(dmi_product_name, "Z1 Workstation"))
+		return true;
+
+	return false;
+}
+
+static int xhci_all_ports_seen_u0(struct xhci_hcd *xhci)
+{
+	return (xhci->port_status_u0 == ((1 << xhci->num_usb3_ports)-1));
+}
+
+
+/*
+ * Initialize memory for HCD and xHC (one-time init).
+ *
+ * Program the PAGESIZE register, initialize the device context array, create
+ * device contexts (?), set up a command ring segment (or two?), create event
+ * ring (one for now).
+ */
+int xhci_init(struct usb_hcd *hcd)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	int retval = 0;
+
+	xhci_dbg(xhci, "xhci_init\n");
+	spin_lock_init(&xhci->lock);
+	if (xhci->hci_version == 0x95 && link_quirk) {
+		xhci_dbg(xhci, "QUIRK: Not clearing Link TRB chain bits.\n");
+		xhci->quirks |= XHCI_LINK_TRB_QUIRK;
+	} else {
+		xhci_dbg(xhci, "xHCI doesn't need link TRB QUIRK\n");
+	}
+	retval = xhci_mem_init(xhci, GFP_KERNEL);
+	xhci_dbg(xhci, "Finished xhci_init\n");
+
+	/* Initializing Compliance Mode Recovery Data If Needed */
+	if (compliance_mode_recovery_timer_quirk_check()) {
+		xhci->quirks |= XHCI_COMP_MODE_QUIRK;
+		compliance_mode_recovery_timer_init(xhci);
+	}
+
+	return retval;
+}
+
+/*-------------------------------------------------------------------------*/
+
+
+#ifdef CONFIG_USB_XHCI_HCD_DEBUGGING
+static 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;
+
+	xhci_dbg(xhci, "Poll event ring: %lu\n", jiffies);
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	temp = xhci_readl(xhci, &xhci->op_regs->status);
+	xhci_dbg(xhci, "op reg status = 0x%x\n", temp);
+	if (temp == 0xffffffff || (xhci->xhc_state & XHCI_STATE_DYING) ||
+			(xhci->xhc_state & XHCI_STATE_HALTED)) {
+		xhci_dbg(xhci, "HW died, polling stopped.\n");
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return;
+	}
+
+	temp = xhci_readl(xhci, &xhci->ir_set->irq_pending);
+	xhci_dbg(xhci, "ir_set 0 pending = 0x%x\n", temp);
+	xhci_dbg(xhci, "HC error bitmask = 0x%x\n", xhci->error_bitmask);
+	xhci->error_bitmask = 0;
+	xhci_dbg(xhci, "Event ring:\n");
+	xhci_debug_segment(xhci, xhci->event_ring->deq_seg);
+	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, "Command ring:\n");
+	xhci_debug_segment(xhci, xhci->cmd_ring->deq_seg);
+	xhci_dbg_ring_ptrs(xhci, xhci->cmd_ring);
+	xhci_dbg_cmd_ptrs(xhci);
+	for (i = 0; i < MAX_HC_SLOTS; ++i) {
+		if (!xhci->devs[i])
+			continue;
+		for (j = 0; j < 31; ++j) {
+			xhci_dbg_ep_rings(xhci, i, j, &xhci->devs[i]->eps[j]);
+		}
+	}
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	if (!xhci->zombie)
+		mod_timer(&xhci->event_ring_timer, jiffies + POLL_TIMEOUT * HZ);
+	else
+		xhci_dbg(xhci, "Quit polling the event ring.\n");
+}
+#endif
+
+static int xhci_run_finished(struct xhci_hcd *xhci)
+{
+	if (xhci_start(xhci)) {
+		xhci_halt(xhci);
+		return -ENODEV;
+	}
+	xhci->shared_hcd->state = HC_STATE_RUNNING;
+	xhci->cmd_ring_state = CMD_RING_STATE_RUNNING;
+
+	if (xhci->quirks & XHCI_NEC_HOST)
+		xhci_ring_cmd_db(xhci);
+
+	xhci_dbg(xhci, "Finished xhci_run for USB3 roothub\n");
+	return 0;
+}
+
+/*
+ * Start the HC after it was halted.
+ *
+ * This function is called by the USB core when the HC driver is added.
+ * Its opposite is xhci_stop().
+ *
+ * xhci_init() must be called once before this function can be called.
+ * Reset the HC, enable device slot contexts, program DCBAAP, and
+ * set command ring pointer and event ring pointer.
+ *
+ * Setup MSI-X vectors and enable interrupts.
+ */
+int xhci_run(struct usb_hcd *hcd)
+{
+	u32 temp;
+	u64 temp_64;
+	int ret;
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+
+	/* Start the xHCI host controller running only after the USB 2.0 roothub
+	 * is setup.
+	 */
+
+	hcd->uses_new_polling = 1;
+	if (!usb_hcd_is_primary_hcd(hcd))
+		return xhci_run_finished(xhci);
+
+	xhci_dbg(xhci, "xhci_run\n");
+
+	ret = xhci_try_enable_msi(hcd);
+	if (ret)
+		return ret;
+
+#ifdef CONFIG_USB_XHCI_HCD_DEBUGGING
+	init_timer(&xhci->event_ring_timer);
+	xhci->event_ring_timer.data = (unsigned long) xhci;
+	xhci->event_ring_timer.function = xhci_event_ring_work;
+	/* Poll the event ring */
+	xhci->event_ring_timer.expires = jiffies + POLL_TIMEOUT * HZ;
+	xhci->zombie = 0;
+	xhci_dbg(xhci, "Setting event ring polling timer\n");
+	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;
+	temp |= (u32) 160;
+	xhci_writel(xhci, temp, &xhci->ir_set->irq_control);
+
+	/* Set the HCD state before we enable the irqs */
+	temp = xhci_readl(xhci, &xhci->op_regs->command);
+	temp |= (CMD_EIE);
+	xhci_dbg(xhci, "// Enable interrupts, cmd = 0x%x.\n",
+			temp);
+	xhci_writel(xhci, temp, &xhci->op_regs->command);
+
+	temp = xhci_readl(xhci, &xhci->ir_set->irq_pending);
+	xhci_dbg(xhci, "// Enabling event ring interrupter %p by writing 0x%x to irq_pending\n",
+			xhci->ir_set, (unsigned int) ER_IRQ_ENABLE(temp));
+	xhci_writel(xhci, ER_IRQ_ENABLE(temp),
+			&xhci->ir_set->irq_pending);
+	xhci_print_ir_set(xhci, 0);
+
+	if (xhci->quirks & XHCI_NEC_HOST)
+		xhci_queue_vendor_command(xhci, 0, 0, 0,
+				TRB_TYPE(TRB_NEC_GET_FW));
+
+	xhci_dbg(xhci, "Finished xhci_run for USB2 roothub\n");
+	return 0;
+}
+
+static void xhci_only_stop_hcd(struct usb_hcd *hcd)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+
+	spin_lock_irq(&xhci->lock);
+	xhci_halt(xhci);
+
+	/* The shared_hcd is going to be deallocated shortly (the USB core only
+	 * calls this function when allocation fails in usb_add_hcd(), or
+	 * usb_remove_hcd() is called).  So we need to unset xHCI's pointer.
+	 */
+	xhci->shared_hcd = NULL;
+	spin_unlock_irq(&xhci->lock);
+}
+
+/*
+ * Stop xHCI driver.
+ *
+ * This function is called by the USB core when the HC driver is removed.
+ * Its opposite is xhci_run().
+ *
+ * Disable device contexts, disable IRQs, and quiesce the HC.
+ * Reset the HC, finish any completed transactions, and cleanup memory.
+ */
+void xhci_stop(struct usb_hcd *hcd)
+{
+	u32 temp;
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+
+	if (!usb_hcd_is_primary_hcd(hcd)) {
+		xhci_only_stop_hcd(xhci->shared_hcd);
+		return;
+	}
+
+	spin_lock_irq(&xhci->lock);
+	/* Make sure the xHC is halted for a USB3 roothub
+	 * (xhci_stop() could be called as part of failed init).
+	 */
+	xhci_halt(xhci);
+	xhci_reset(xhci);
+	spin_unlock_irq(&xhci->lock);
+
+	xhci_cleanup_msix(xhci);
+
+#ifdef CONFIG_USB_XHCI_HCD_DEBUGGING
+	/* Tell the event ring poll function not to reschedule */
+	xhci->zombie = 1;
+	del_timer_sync(&xhci->event_ring_timer);
+#endif
+
+	/* Deleting Compliance Mode Recovery Timer */
+	if ((xhci->quirks & XHCI_COMP_MODE_QUIRK) &&
+			(!(xhci_all_ports_seen_u0(xhci)))) {
+		del_timer_sync(&xhci->comp_mode_recovery_timer);
+		xhci_dbg(xhci, "%s: compliance mode recovery timer deleted\n",
+				__func__);
+	}
+
+	if (xhci->quirks & XHCI_AMD_PLL_FIX)
+		usb_amd_dev_put();
+
+	xhci_dbg(xhci, "// Disabling event ring interrupts\n");
+	temp = xhci_readl(xhci, &xhci->op_regs->status);
+	xhci_writel(xhci, temp & ~STS_EINT, &xhci->op_regs->status);
+	temp = xhci_readl(xhci, &xhci->ir_set->irq_pending);
+	xhci_writel(xhci, ER_IRQ_DISABLE(temp),
+			&xhci->ir_set->irq_pending);
+	xhci_print_ir_set(xhci, 0);
+
+	xhci_dbg(xhci, "cleaning up memory\n");
+	xhci_mem_cleanup(xhci);
+	xhci_dbg(xhci, "xhci_stop completed - status = %x\n",
+		    xhci_readl(xhci, &xhci->op_regs->status));
+}
+
+/*
+ * Shutdown HC (not bus-specific)
+ *
+ * This is called when the machine is rebooting or halting.  We assume that the
+ * machine will be powered off, and the HC's internal state will be reset.
+ * Don't bother to free memory.
+ *
+ * This will only ever be called with the main usb_hcd (the USB3 roothub).
+ */
+void xhci_shutdown(struct usb_hcd *hcd)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+
+	if (xhci->quirks & XHCI_SPURIOUS_REBOOT)
+		usb_disable_xhci_ports(to_pci_dev(hcd->self.controller));
+
+	spin_lock_irq(&xhci->lock);
+	xhci_halt(xhci);
+	spin_unlock_irq(&xhci->lock);
+
+	xhci_cleanup_msix(xhci);
+
+	xhci_dbg(xhci, "xhci_shutdown completed - status = %x\n",
+		    xhci_readl(xhci, &xhci->op_regs->status));
+}
+
+#ifdef CONFIG_PM
+static void xhci_save_registers(struct xhci_hcd *xhci)
+{
+	xhci->s3.command = xhci_readl(xhci, &xhci->op_regs->command);
+	xhci->s3.dev_nt = xhci_readl(xhci, &xhci->op_regs->dev_notification);
+	xhci->s3.dcbaa_ptr = xhci_read_64(xhci, &xhci->op_regs->dcbaa_ptr);
+	xhci->s3.config_reg = xhci_readl(xhci, &xhci->op_regs->config_reg);
+	xhci->s3.erst_size = xhci_readl(xhci, &xhci->ir_set->erst_size);
+	xhci->s3.erst_base = xhci_read_64(xhci, &xhci->ir_set->erst_base);
+	xhci->s3.erst_dequeue = xhci_read_64(xhci, &xhci->ir_set->erst_dequeue);
+	xhci->s3.irq_pending = xhci_readl(xhci, &xhci->ir_set->irq_pending);
+	xhci->s3.irq_control = xhci_readl(xhci, &xhci->ir_set->irq_control);
+}
+
+static void xhci_restore_registers(struct xhci_hcd *xhci)
+{
+	xhci_writel(xhci, xhci->s3.command, &xhci->op_regs->command);
+	xhci_writel(xhci, xhci->s3.dev_nt, &xhci->op_regs->dev_notification);
+	xhci_write_64(xhci, xhci->s3.dcbaa_ptr, &xhci->op_regs->dcbaa_ptr);
+	xhci_writel(xhci, xhci->s3.config_reg, &xhci->op_regs->config_reg);
+	xhci_writel(xhci, xhci->s3.erst_size, &xhci->ir_set->erst_size);
+	xhci_write_64(xhci, xhci->s3.erst_base, &xhci->ir_set->erst_base);
+	xhci_write_64(xhci, xhci->s3.erst_dequeue, &xhci->ir_set->erst_dequeue);
+	xhci_writel(xhci, xhci->s3.irq_pending, &xhci->ir_set->irq_pending);
+	xhci_writel(xhci, xhci->s3.irq_control, &xhci->ir_set->irq_control);
+}
+
+static void xhci_set_cmd_ring_deq(struct xhci_hcd *xhci)
+{
+	u64	val_64;
+
+	/* step 2: initialize command ring buffer */
+	val_64 = xhci_read_64(xhci, &xhci->op_regs->cmd_ring);
+	val_64 = (val_64 & (u64) CMD_RING_RSVD_BITS) |
+		(xhci_trb_virt_to_dma(xhci->cmd_ring->deq_seg,
+				      xhci->cmd_ring->dequeue) &
+		 (u64) ~CMD_RING_RSVD_BITS) |
+		xhci->cmd_ring->cycle_state;
+	xhci_dbg(xhci, "// Setting command ring address to 0x%llx\n",
+			(long unsigned long) val_64);
+	xhci_write_64(xhci, val_64, &xhci->op_regs->cmd_ring);
+}
+
+/*
+ * The whole command ring must be cleared to zero when we suspend the host.
+ *
+ * The host doesn't save the command ring pointer in the suspend well, so we
+ * need to re-program it on resume.  Unfortunately, the pointer must be 64-byte
+ * aligned, because of the reserved bits in the command ring dequeue pointer
+ * register.  Therefore, we can't just set the dequeue pointer back in the
+ * middle of the ring (TRBs are 16-byte aligned).
+ */
+static void xhci_clear_command_ring(struct xhci_hcd *xhci)
+{
+	struct xhci_ring *ring;
+	struct xhci_segment *seg;
+
+	ring = xhci->cmd_ring;
+	seg = ring->deq_seg;
+	do {
+		memset(seg->trbs, 0,
+			sizeof(union xhci_trb) * (TRBS_PER_SEGMENT - 1));
+		seg->trbs[TRBS_PER_SEGMENT - 1].link.control &=
+			cpu_to_le32(~TRB_CYCLE);
+		seg = seg->next;
+	} while (seg != ring->deq_seg);
+
+	/* Reset the software enqueue and dequeue pointers */
+	ring->deq_seg = ring->first_seg;
+	ring->dequeue = ring->first_seg->trbs;
+	ring->enq_seg = ring->deq_seg;
+	ring->enqueue = ring->dequeue;
+
+	ring->num_trbs_free = ring->num_segs * (TRBS_PER_SEGMENT - 1) - 1;
+	/*
+	 * Ring is now zeroed, so the HW should look for change of ownership
+	 * when the cycle bit is set to 1.
+	 */
+	ring->cycle_state = 1;
+
+	/*
+	 * Reset the hardware dequeue pointer.
+	 * Yes, this will need to be re-written after resume, but we're paranoid
+	 * and want to make sure the hardware doesn't access bogus memory
+	 * because, say, the BIOS or an SMI started the host without changing
+	 * the command ring pointers.
+	 */
+	xhci_set_cmd_ring_deq(xhci);
+}
+
+/*
+ * Stop HC (not bus-specific)
+ *
+ * This is called when the machine transition into S3/S4 mode.
+ *
+ */
+int xhci_suspend(struct xhci_hcd *xhci)
+{
+	int			rc = 0;
+	struct usb_hcd		*hcd = xhci_to_hcd(xhci);
+	u32			command;
+
+	if (hcd->state != HC_STATE_SUSPENDED ||
+			xhci->shared_hcd->state != HC_STATE_SUSPENDED)
+		return -EINVAL;
+
+	/* Don't poll the roothubs on bus suspend. */
+	xhci_dbg(xhci, "%s: stopping port polling.\n", __func__);
+	clear_bit(HCD_FLAG_POLL_RH, &hcd->flags);
+	del_timer_sync(&hcd->rh_timer);
+
+	spin_lock_irq(&xhci->lock);
+	clear_bit(HCD_FLAG_HW_ACCESSIBLE, &hcd->flags);
+	clear_bit(HCD_FLAG_HW_ACCESSIBLE, &xhci->shared_hcd->flags);
+	/* step 1: stop endpoint */
+	/* skipped assuming that port suspend has done */
+
+	/* step 2: clear Run/Stop bit */
+	command = xhci_readl(xhci, &xhci->op_regs->command);
+	command &= ~CMD_RUN;
+	xhci_writel(xhci, command, &xhci->op_regs->command);
+	if (xhci_handshake(xhci, &xhci->op_regs->status,
+		      STS_HALT, STS_HALT, XHCI_MAX_HALT_USEC)) {
+		xhci_warn(xhci, "WARN: xHC CMD_RUN timeout\n");
+		spin_unlock_irq(&xhci->lock);
+		return -ETIMEDOUT;
+	}
+	xhci_clear_command_ring(xhci);
+
+	/* step 3: save registers */
+	xhci_save_registers(xhci);
+
+	/* step 4: set CSS flag */
+	command = xhci_readl(xhci, &xhci->op_regs->command);
+	command |= CMD_CSS;
+	xhci_writel(xhci, command, &xhci->op_regs->command);
+	if (xhci_handshake(xhci, &xhci->op_regs->status,
+				STS_SAVE, 0, 10 * 1000)) {
+		xhci_warn(xhci, "WARN: xHC save state timeout\n");
+		spin_unlock_irq(&xhci->lock);
+		return -ETIMEDOUT;
+	}
+	spin_unlock_irq(&xhci->lock);
+
+	/*
+	 * Deleting Compliance Mode Recovery Timer because the xHCI Host
+	 * is about to be suspended.
+	 */
+	if ((xhci->quirks & XHCI_COMP_MODE_QUIRK) &&
+			(!(xhci_all_ports_seen_u0(xhci)))) {
+		del_timer_sync(&xhci->comp_mode_recovery_timer);
+		xhci_dbg(xhci, "%s: compliance mode recovery timer deleted\n",
+				__func__);
+	}
+
+	/* step 5: remove core well power */
+	/* synchronize irq when using MSI-X */
+	xhci_msix_sync_irqs(xhci);
+
+	return rc;
+}
+
+/*
+ * start xHC (not bus-specific)
+ *
+ * This is called when the machine transition from S3/S4 mode.
+ *
+ */
+int xhci_resume(struct xhci_hcd *xhci, bool hibernated)
+{
+	u32			command, temp = 0;
+	struct usb_hcd		*hcd = xhci_to_hcd(xhci);
+	struct usb_hcd		*secondary_hcd;
+	int			retval = 0;
+
+	/* Wait a bit if either of the roothubs need to settle from the
+	 * transition into bus suspend.
+	 */
+	if (time_before(jiffies, xhci->bus_state[0].next_statechange) ||
+			time_before(jiffies,
+				xhci->bus_state[1].next_statechange))
+		msleep(100);
+
+	set_bit(HCD_FLAG_HW_ACCESSIBLE, &hcd->flags);
+	set_bit(HCD_FLAG_HW_ACCESSIBLE, &xhci->shared_hcd->flags);
+
+	spin_lock_irq(&xhci->lock);
+	if (xhci->quirks & XHCI_RESET_ON_RESUME)
+		hibernated = true;
+
+	if (!hibernated) {
+		/* step 1: restore register */
+		xhci_restore_registers(xhci);
+		/* step 2: initialize command ring buffer */
+		xhci_set_cmd_ring_deq(xhci);
+		/* step 3: restore state and start state*/
+		/* step 3: set CRS flag */
+		command = xhci_readl(xhci, &xhci->op_regs->command);
+		command |= CMD_CRS;
+		xhci_writel(xhci, command, &xhci->op_regs->command);
+		if (xhci_handshake(xhci, &xhci->op_regs->status,
+			      STS_RESTORE, 0, 10 * 1000)) {
+			xhci_warn(xhci, "WARN: xHC restore state timeout\n");
+			spin_unlock_irq(&xhci->lock);
+			return -ETIMEDOUT;
+		}
+		temp = xhci_readl(xhci, &xhci->op_regs->status);
+	}
+
+	/* If restore operation fails, re-initialize the HC during resume */
+	if ((temp & STS_SRE) || hibernated) {
+		/* Let the USB core know _both_ roothubs lost power. */
+		usb_root_hub_lost_power(xhci->main_hcd->self.root_hub);
+		usb_root_hub_lost_power(xhci->shared_hcd->self.root_hub);
+
+		xhci_dbg(xhci, "Stop HCD\n");
+		xhci_halt(xhci);
+		xhci_reset(xhci);
+		spin_unlock_irq(&xhci->lock);
+		xhci_cleanup_msix(xhci);
+
+#ifdef CONFIG_USB_XHCI_HCD_DEBUGGING
+		/* Tell the event ring poll function not to reschedule */
+		xhci->zombie = 1;
+		del_timer_sync(&xhci->event_ring_timer);
+#endif
+
+		xhci_dbg(xhci, "// Disabling event ring interrupts\n");
+		temp = xhci_readl(xhci, &xhci->op_regs->status);
+		xhci_writel(xhci, temp & ~STS_EINT, &xhci->op_regs->status);
+		temp = xhci_readl(xhci, &xhci->ir_set->irq_pending);
+		xhci_writel(xhci, ER_IRQ_DISABLE(temp),
+				&xhci->ir_set->irq_pending);
+		xhci_print_ir_set(xhci, 0);
+
+		xhci_dbg(xhci, "cleaning up memory\n");
+		xhci_mem_cleanup(xhci);
+		xhci_dbg(xhci, "xhci_stop completed - status = %x\n",
+			    xhci_readl(xhci, &xhci->op_regs->status));
+
+		/* USB core calls the PCI reinit and start functions twice:
+		 * first with the primary HCD, and then with the secondary HCD.
+		 * If we don't do the same, the host will never be started.
+		 */
+		if (!usb_hcd_is_primary_hcd(hcd))
+			secondary_hcd = hcd;
+		else
+			secondary_hcd = xhci->shared_hcd;
+
+		xhci_dbg(xhci, "Initialize the xhci_hcd\n");
+		retval = xhci_init(hcd->primary_hcd);
+		if (retval)
+			return retval;
+		xhci_dbg(xhci, "Start the primary HCD\n");
+		retval = xhci_run(hcd->primary_hcd);
+		if (!retval) {
+			xhci_dbg(xhci, "Start the secondary HCD\n");
+			retval = xhci_run(secondary_hcd);
+		}
+		hcd->state = HC_STATE_SUSPENDED;
+		xhci->shared_hcd->state = HC_STATE_SUSPENDED;
+		goto done;
+	}
+
+	/* step 4: set Run/Stop bit */
+	command = xhci_readl(xhci, &xhci->op_regs->command);
+	command |= CMD_RUN;
+	xhci_writel(xhci, command, &xhci->op_regs->command);
+	xhci_handshake(xhci, &xhci->op_regs->status, STS_HALT,
+		  0, 250 * 1000);
+
+	/* step 5: walk topology and initialize portsc,
+	 * portpmsc and portli
+	 */
+	/* this is done in bus_resume */
+
+	/* step 6: restart each of the previously
+	 * Running endpoints by ringing their doorbells
+	 */
+
+	spin_unlock_irq(&xhci->lock);
+
+ done:
+	if (retval == 0) {
+		usb_hcd_resume_root_hub(hcd);
+		usb_hcd_resume_root_hub(xhci->shared_hcd);
+	}
+
+	/*
+	 * If system is subject to the Quirk, Compliance Mode Timer needs to
+	 * be re-initialized Always after a system resume. Ports are subject
+	 * to suffer the Compliance Mode issue again. It doesn't matter if
+	 * ports have entered previously to U0 before system's suspension.
+	 */
+	if (xhci->quirks & XHCI_COMP_MODE_QUIRK)
+		compliance_mode_recovery_timer_init(xhci);
+
+	/* Re-enable port polling. */
+	xhci_dbg(xhci, "%s: starting port polling.\n", __func__);
+	set_bit(HCD_FLAG_POLL_RH, &hcd->flags);
+	usb_hcd_poll_rh_status(hcd);
+
+	return retval;
+}
+#endif	/* CONFIG_PM */
+
+/*-------------------------------------------------------------------------*/
+
+/**
+ * xhci_get_endpoint_index - Used for passing endpoint bitmasks between the core and
+ * HCDs.  Find the index for an endpoint given its descriptor.  Use the return
+ * value to right shift 1 for the bitmask.
+ *
+ * Index  = (epnum * 2) + direction - 1,
+ * where direction = 0 for OUT, 1 for IN.
+ * For control endpoints, the IN index is used (OUT index is unused), so
+ * index = (epnum * 2) + direction - 1 = (epnum * 2) + 1 - 1 = (epnum * 2)
+ */
+unsigned int xhci_get_endpoint_index(struct usb_endpoint_descriptor *desc)
+{
+	unsigned int index;
+	if (usb_endpoint_xfer_control(desc))
+		index = (unsigned int) (usb_endpoint_num(desc)*2);
+	else
+		index = (unsigned int) (usb_endpoint_num(desc)*2) +
+			(usb_endpoint_dir_in(desc) ? 1 : 0) - 1;
+	return index;
+}
+
+/* Find the flag for this endpoint (for use in the control context).  Use the
+ * endpoint index to create a bitmask.  The slot context is bit 0, endpoint 0 is
+ * bit 1, etc.
+ */
+unsigned int xhci_get_endpoint_flag(struct usb_endpoint_descriptor *desc)
+{
+	return 1 << (xhci_get_endpoint_index(desc) + 1);
+}
+
+/* Find the flag for this endpoint (for use in the control context).  Use the
+ * endpoint index to create a bitmask.  The slot context is bit 0, endpoint 0 is
+ * bit 1, etc.
+ */
+unsigned int xhci_get_endpoint_flag_from_index(unsigned int ep_index)
+{
+	return 1 << (ep_index + 1);
+}
+
+/* Compute the last valid endpoint context index.  Basically, this is the
+ * endpoint index plus one.  For slot contexts with more than valid endpoint,
+ * we find the most significant bit set in the added contexts flags.
+ * e.g. ep 1 IN (with epnum 0x81) => added_ctxs = 0b1000
+ * fls(0b1000) = 4, but the endpoint context index is 3, so subtract one.
+ */
+unsigned int xhci_last_valid_endpoint(u32 added_ctxs)
+{
+	return fls(added_ctxs) - 1;
+}
+
+/* Returns 1 if the arguments are OK;
+ * returns 0 this is a root hub; returns -EINVAL for NULL pointers.
+ */
+static int xhci_check_args(struct usb_hcd *hcd, struct usb_device *udev,
+		struct usb_host_endpoint *ep, int check_ep, bool check_virt_dev,
+		const char *func) {
+	struct xhci_hcd	*xhci;
+	struct xhci_virt_device	*virt_dev;
+
+	if (!hcd || (check_ep && !ep) || !udev) {
+		printk(KERN_DEBUG "xHCI %s called with invalid args\n",
+				func);
+		return -EINVAL;
+	}
+	if (!udev->parent) {
+		printk(KERN_DEBUG "xHCI %s called for root hub\n",
+				func);
+		return 0;
+	}
+
+	xhci = hcd_to_xhci(hcd);
+	if (xhci->xhc_state & XHCI_STATE_HALTED)
+		return -ENODEV;
+
+	if (check_virt_dev) {
+		if (!udev->slot_id || !xhci->devs[udev->slot_id]) {
+			printk(KERN_DEBUG "xHCI %s called with unaddressed "
+						"device\n", func);
+			return -EINVAL;
+		}
+
+		virt_dev = xhci->devs[udev->slot_id];
+		if (virt_dev->udev != udev) {
+			printk(KERN_DEBUG "xHCI %s called with udev and "
+					  "virt_dev does not match\n", func);
+			return -EINVAL;
+		}
+	}
+
+	return 1;
+}
+
+static int xhci_configure_endpoint(struct xhci_hcd *xhci,
+		struct usb_device *udev, struct xhci_command *command,
+		bool ctx_change, bool must_succeed);
+
+/*
+ * Full speed devices may have a max packet size greater than 8 bytes, but the
+ * USB core doesn't know that until it reads the first 8 bytes of the
+ * descriptor.  If the usb_device's max packet size changes after that point,
+ * we need to issue an evaluate context command and wait on it.
+ */
+static int xhci_check_maxpacket(struct xhci_hcd *xhci, unsigned int slot_id,
+		unsigned int ep_index, struct urb *urb)
+{
+	struct xhci_container_ctx *in_ctx;
+	struct xhci_container_ctx *out_ctx;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	struct xhci_ep_ctx *ep_ctx;
+	int max_packet_size;
+	int hw_max_packet_size;
+	int ret = 0;
+
+	out_ctx = xhci->devs[slot_id]->out_ctx;
+	ep_ctx = xhci_get_ep_ctx(xhci, out_ctx, ep_index);
+	hw_max_packet_size = MAX_PACKET_DECODED(le32_to_cpu(ep_ctx->ep_info2));
+	max_packet_size = usb_endpoint_maxp(&urb->dev->ep0.desc);
+	if (hw_max_packet_size != max_packet_size) {
+		xhci_dbg(xhci, "Max Packet Size for ep 0 changed.\n");
+		xhci_dbg(xhci, "Max packet size in usb_device = %d\n",
+				max_packet_size);
+		xhci_dbg(xhci, "Max packet size in xHCI HW = %d\n",
+				hw_max_packet_size);
+		xhci_dbg(xhci, "Issuing evaluate context command.\n");
+
+		/* Set up the modified control endpoint 0 */
+		xhci_endpoint_copy(xhci, xhci->devs[slot_id]->in_ctx,
+				xhci->devs[slot_id]->out_ctx, ep_index);
+		in_ctx = xhci->devs[slot_id]->in_ctx;
+		ep_ctx = xhci_get_ep_ctx(xhci, in_ctx, ep_index);
+		ep_ctx->ep_info2 &= cpu_to_le32(~MAX_PACKET_MASK);
+		ep_ctx->ep_info2 |= cpu_to_le32(MAX_PACKET(max_packet_size));
+
+		/* Set up the input context flags for the command */
+		/* FIXME: This won't work if a non-default control endpoint
+		 * changes max packet sizes.
+		 */
+		ctrl_ctx = xhci_get_input_control_ctx(xhci, in_ctx);
+		ctrl_ctx->add_flags = cpu_to_le32(EP0_FLAG);
+		ctrl_ctx->drop_flags = 0;
+
+		xhci_dbg(xhci, "Slot %d input context\n", slot_id);
+		xhci_dbg_ctx(xhci, in_ctx, ep_index);
+		xhci_dbg(xhci, "Slot %d output context\n", slot_id);
+		xhci_dbg_ctx(xhci, out_ctx, ep_index);
+
+		ret = xhci_configure_endpoint(xhci, urb->dev, NULL,
+				true, false);
+
+		/* Clean up the input context for later use by bandwidth
+		 * functions.
+		 */
+		ctrl_ctx->add_flags = cpu_to_le32(SLOT_FLAG);
+	}
+	return ret;
+}
+
+/*
+ * non-error returns are a promise to giveback() the urb later
+ * we drop ownership so next owner (or urb unlink) can get it
+ */
+int xhci_urb_enqueue(struct usb_hcd *hcd, struct urb *urb, gfp_t mem_flags)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	struct xhci_td *buffer;
+	unsigned long flags;
+	int ret = 0;
+	unsigned int slot_id, ep_index;
+	struct urb_priv	*urb_priv;
+	int size, i;
+
+	if (!urb || xhci_check_args(hcd, urb->dev, urb->ep,
+					true, true, __func__) <= 0)
+		return -EINVAL;
+
+	slot_id = urb->dev->slot_id;
+	ep_index = xhci_get_endpoint_index(&urb->ep->desc);
+
+	if (!HCD_HW_ACCESSIBLE(hcd)) {
+		if (!in_interrupt())
+			xhci_dbg(xhci, "urb submitted during PCI suspend\n");
+		ret = -ESHUTDOWN;
+		goto exit;
+	}
+
+	if (usb_endpoint_xfer_isoc(&urb->ep->desc))
+		size = urb->number_of_packets;
+	else
+		size = 1;
+
+	urb_priv = kzalloc(sizeof(struct urb_priv) +
+				  size * sizeof(struct xhci_td *), mem_flags);
+	if (!urb_priv)
+		return -ENOMEM;
+
+	buffer = kzalloc(size * sizeof(struct xhci_td), mem_flags);
+	if (!buffer) {
+		kfree(urb_priv);
+		return -ENOMEM;
+	}
+
+	for (i = 0; i < size; i++) {
+		urb_priv->td[i] = buffer;
+		buffer++;
+	}
+
+	urb_priv->length = size;
+	urb_priv->td_cnt = 0;
+	urb->hcpriv = urb_priv;
+
+	if (usb_endpoint_xfer_control(&urb->ep->desc)) {
+		/* Check to see if the max packet size for the default control
+		 * endpoint changed during FS device enumeration
+		 */
+		if (urb->dev->speed == USB_SPEED_FULL) {
+			ret = xhci_check_maxpacket(xhci, slot_id,
+					ep_index, urb);
+			if (ret < 0) {
+				xhci_urb_free_priv(xhci, urb_priv);
+				urb->hcpriv = NULL;
+				return ret;
+			}
+		}
+
+		/* We have a spinlock and interrupts disabled, so we must pass
+		 * atomic context to this function, which may allocate memory.
+		 */
+		spin_lock_irqsave(&xhci->lock, flags);
+		if (xhci->xhc_state & XHCI_STATE_DYING)
+			goto dying;
+		ret = xhci_queue_ctrl_tx(xhci, GFP_ATOMIC, urb,
+				slot_id, ep_index);
+		if (ret)
+			goto free_priv;
+		spin_unlock_irqrestore(&xhci->lock, flags);
+	} else if (usb_endpoint_xfer_bulk(&urb->ep->desc)) {
+		spin_lock_irqsave(&xhci->lock, flags);
+		if (xhci->xhc_state & XHCI_STATE_DYING)
+			goto dying;
+		if (xhci->devs[slot_id]->eps[ep_index].ep_state &
+				EP_GETTING_STREAMS) {
+			xhci_warn(xhci, "WARN: Can't enqueue URB while bulk ep "
+					"is transitioning to using streams.\n");
+			ret = -EINVAL;
+		} else if (xhci->devs[slot_id]->eps[ep_index].ep_state &
+				EP_GETTING_NO_STREAMS) {
+			xhci_warn(xhci, "WARN: Can't enqueue URB while bulk ep "
+					"is transitioning to "
+					"not having streams.\n");
+			ret = -EINVAL;
+		} else {
+			ret = xhci_queue_bulk_tx(xhci, GFP_ATOMIC, urb,
+					slot_id, ep_index);
+		}
+		if (ret)
+			goto free_priv;
+		spin_unlock_irqrestore(&xhci->lock, flags);
+	} else if (usb_endpoint_xfer_int(&urb->ep->desc)) {
+		spin_lock_irqsave(&xhci->lock, flags);
+		if (xhci->xhc_state & XHCI_STATE_DYING)
+			goto dying;
+		ret = xhci_queue_intr_tx(xhci, GFP_ATOMIC, urb,
+				slot_id, ep_index);
+		if (ret)
+			goto free_priv;
+		spin_unlock_irqrestore(&xhci->lock, flags);
+	} else {
+		spin_lock_irqsave(&xhci->lock, flags);
+		if (xhci->xhc_state & XHCI_STATE_DYING)
+			goto dying;
+		ret = xhci_queue_isoc_tx_prepare(xhci, GFP_ATOMIC, urb,
+				slot_id, ep_index);
+		if (ret)
+			goto free_priv;
+		spin_unlock_irqrestore(&xhci->lock, flags);
+	}
+exit:
+	return ret;
+dying:
+	xhci_dbg(xhci, "Ep 0x%x: URB %p submitted for "
+			"non-responsive xHCI host.\n",
+			urb->ep->desc.bEndpointAddress, urb);
+	ret = -ESHUTDOWN;
+free_priv:
+	xhci_urb_free_priv(xhci, urb_priv);
+	urb->hcpriv = NULL;
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return ret;
+}
+
+/* Get the right ring for the given URB.
+ * If the endpoint supports streams, boundary check the URB's stream ID.
+ * If the endpoint doesn't support streams, return the singular endpoint ring.
+ */
+static struct xhci_ring *xhci_urb_to_transfer_ring(struct xhci_hcd *xhci,
+		struct urb *urb)
+{
+	unsigned int slot_id;
+	unsigned int ep_index;
+	unsigned int stream_id;
+	struct xhci_virt_ep *ep;
+
+	slot_id = urb->dev->slot_id;
+	ep_index = xhci_get_endpoint_index(&urb->ep->desc);
+	stream_id = urb->stream_id;
+	ep = &xhci->devs[slot_id]->eps[ep_index];
+	/* Common case: no streams */
+	if (!(ep->ep_state & EP_HAS_STREAMS))
+		return ep->ring;
+
+	if (stream_id == 0) {
+		xhci_warn(xhci,
+				"WARN: Slot ID %u, ep index %u has streams, "
+				"but URB has no stream ID.\n",
+				slot_id, ep_index);
+		return NULL;
+	}
+
+	if (stream_id < ep->stream_info->num_streams)
+		return ep->stream_info->stream_rings[stream_id];
+
+	xhci_warn(xhci,
+			"WARN: Slot ID %u, ep index %u has "
+			"stream IDs 1 to %u allocated, "
+			"but stream ID %u is requested.\n",
+			slot_id, ep_index,
+			ep->stream_info->num_streams - 1,
+			stream_id);
+	return NULL;
+}
+
+/*
+ * Remove the URB's TD from the endpoint ring.  This may cause the HC to stop
+ * USB transfers, potentially stopping in the middle of a TRB buffer.  The HC
+ * should pick up where it left off in the TD, unless a Set Transfer Ring
+ * Dequeue Pointer is issued.
+ *
+ * The TRBs that make up the buffers for the canceled URB will be "removed" from
+ * the ring.  Since the ring is a contiguous structure, they can't be physically
+ * removed.  Instead, there are two options:
+ *
+ *  1) If the HC is in the middle of processing the URB to be canceled, we
+ *     simply move the ring's dequeue pointer past those TRBs using the Set
+ *     Transfer Ring Dequeue Pointer command.  This will be the common case,
+ *     when drivers timeout on the last submitted URB and attempt to cancel.
+ *
+ *  2) If the HC is in the middle of a different TD, we turn the TRBs into a
+ *     series of 1-TRB transfer no-op TDs.  (No-ops shouldn't be chained.)  The
+ *     HC will need to invalidate the any TRBs it has cached after the stop
+ *     endpoint command, as noted in the xHCI 0.95 errata.
+ *
+ *  3) The TD may have completed by the time the Stop Endpoint Command
+ *     completes, so software needs to handle that case too.
+ *
+ * This function should protect against the TD enqueueing code ringing the
+ * doorbell while this code is waiting for a Stop Endpoint command to complete.
+ * It also needs to account for multiple cancellations on happening at the same
+ * time for the same endpoint.
+ *
+ * Note that this function can be called in any context, or so says
+ * usb_hcd_unlink_urb()
+ */
+int xhci_urb_dequeue(struct usb_hcd *hcd, struct urb *urb, int status)
+{
+	unsigned long flags;
+	int ret, i;
+	u32 temp;
+	struct xhci_hcd *xhci;
+	struct urb_priv	*urb_priv;
+	struct xhci_td *td;
+	unsigned int ep_index;
+	struct xhci_ring *ep_ring;
+	struct xhci_virt_ep *ep;
+
+	xhci = hcd_to_xhci(hcd);
+	spin_lock_irqsave(&xhci->lock, flags);
+	/* Make sure the URB hasn't completed or been unlinked already */
+	ret = usb_hcd_check_unlink_urb(hcd, urb, status);
+	if (ret || !urb->hcpriv)
+		goto done;
+	temp = xhci_readl(xhci, &xhci->op_regs->status);
+	if (temp == 0xffffffff || (xhci->xhc_state & XHCI_STATE_HALTED)) {
+		xhci_dbg(xhci, "HW died, freeing TD.\n");
+		urb_priv = urb->hcpriv;
+		for (i = urb_priv->td_cnt; i < urb_priv->length; i++) {
+			td = urb_priv->td[i];
+			if (!list_empty(&td->td_list))
+				list_del_init(&td->td_list);
+			if (!list_empty(&td->cancelled_td_list))
+				list_del_init(&td->cancelled_td_list);
+		}
+
+		usb_hcd_unlink_urb_from_ep(hcd, urb);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		usb_hcd_giveback_urb(hcd, urb, -ESHUTDOWN);
+		xhci_urb_free_priv(xhci, urb_priv);
+		return ret;
+	}
+	if ((xhci->xhc_state & XHCI_STATE_DYING) ||
+			(xhci->xhc_state & XHCI_STATE_HALTED)) {
+		xhci_dbg(xhci, "Ep 0x%x: URB %p to be canceled on "
+				"non-responsive xHCI host.\n",
+				urb->ep->desc.bEndpointAddress, urb);
+		/* Let the stop endpoint command watchdog timer (which set this
+		 * state) finish cleaning up the endpoint TD lists.  We must
+		 * have caught it in the middle of dropping a lock and giving
+		 * back an URB.
+		 */
+		goto done;
+	}
+
+	ep_index = xhci_get_endpoint_index(&urb->ep->desc);
+	ep = &xhci->devs[urb->dev->slot_id]->eps[ep_index];
+	ep_ring = xhci_urb_to_transfer_ring(xhci, urb);
+	if (!ep_ring) {
+		ret = -EINVAL;
+		goto done;
+	}
+
+	urb_priv = urb->hcpriv;
+	i = urb_priv->td_cnt;
+	if (i < urb_priv->length)
+		xhci_dbg(xhci, "Cancel URB %p, dev %s, ep 0x%x, "
+				"starting at offset 0x%llx\n",
+				urb, urb->dev->devpath,
+				urb->ep->desc.bEndpointAddress,
+				(unsigned long long) xhci_trb_virt_to_dma(
+					urb_priv->td[i]->start_seg,
+					urb_priv->td[i]->first_trb));
+
+	for (; i < urb_priv->length; i++) {
+		td = urb_priv->td[i];
+		list_add_tail(&td->cancelled_td_list, &ep->cancelled_td_list);
+	}
+
+	/* Queue a stop endpoint command, but only if this is
+	 * the first cancellation to be handled.
+	 */
+	if (!(ep->ep_state & EP_HALT_PENDING)) {
+		ep->ep_state |= EP_HALT_PENDING;
+		ep->stop_cmds_pending++;
+		ep->stop_cmd_timer.expires = jiffies +
+			XHCI_STOP_EP_CMD_TIMEOUT * HZ;
+		add_timer(&ep->stop_cmd_timer);
+		xhci_queue_stop_endpoint(xhci, urb->dev->slot_id, ep_index, 0);
+		xhci_ring_cmd_db(xhci);
+	}
+done:
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return ret;
+}
+
+/* Drop an endpoint from a new bandwidth configuration for this device.
+ * Only one call to this function is allowed per endpoint before
+ * check_bandwidth() or reset_bandwidth() must be called.
+ * A call to xhci_drop_endpoint() followed by a call to xhci_add_endpoint() will
+ * add the endpoint to the schedule with possibly new parameters denoted by a
+ * different endpoint descriptor in usb_host_endpoint.
+ * A call to xhci_add_endpoint() followed by a call to xhci_drop_endpoint() is
+ * not allowed.
+ *
+ * The USB core will not allow URBs to be queued to an endpoint that is being
+ * disabled, so there's no need for mutual exclusion to protect
+ * the xhci->devs[slot_id] structure.
+ */
+int xhci_drop_endpoint(struct usb_hcd *hcd, struct usb_device *udev,
+		struct usb_host_endpoint *ep)
+{
+	struct xhci_hcd *xhci;
+	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;
+	u32 drop_flag;
+	u32 new_add_flags, new_drop_flags, new_slot_info;
+	int ret;
+
+	ret = xhci_check_args(hcd, udev, ep, 1, true, __func__);
+	if (ret <= 0)
+		return ret;
+	xhci = hcd_to_xhci(hcd);
+	if (xhci->xhc_state & XHCI_STATE_DYING)
+		return -ENODEV;
+
+	xhci_dbg(xhci, "%s called for udev %p\n", __func__, udev);
+	drop_flag = xhci_get_endpoint_flag(&ep->desc);
+	if (drop_flag == SLOT_FLAG || drop_flag == EP0_FLAG) {
+		xhci_dbg(xhci, "xHCI %s - can't drop slot or ep 0 %#x\n",
+				__func__, drop_flag);
+		return 0;
+	}
+
+	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_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 & cpu_to_le32(EP_STATE_MASK)) ==
+	     cpu_to_le32(EP_STATE_DISABLED)) ||
+	    le32_to_cpu(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;
+	}
+
+	ctrl_ctx->drop_flags |= cpu_to_le32(drop_flag);
+	new_drop_flags = le32_to_cpu(ctrl_ctx->drop_flags);
+
+	ctrl_ctx->add_flags &= cpu_to_le32(~drop_flag);
+	new_add_flags = le32_to_cpu(ctrl_ctx->add_flags);
+
+	last_ctx = xhci_last_valid_endpoint(le32_to_cpu(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 ((le32_to_cpu(slot_ctx->dev_info) & LAST_CTX_MASK) >
+	    LAST_CTX(last_ctx)) {
+		slot_ctx->dev_info &= cpu_to_le32(~LAST_CTX_MASK);
+		slot_ctx->dev_info |= cpu_to_le32(LAST_CTX(last_ctx));
+	}
+	new_slot_info = le32_to_cpu(slot_ctx->dev_info);
+
+	xhci_endpoint_zero(xhci, xhci->devs[udev->slot_id], ep);
+
+	xhci_dbg(xhci, "drop ep 0x%x, slot id %d, new drop flags = %#x, new add flags = %#x, new slot info = %#x\n",
+			(unsigned int) ep->desc.bEndpointAddress,
+			udev->slot_id,
+			(unsigned int) new_drop_flags,
+			(unsigned int) new_add_flags,
+			(unsigned int) new_slot_info);
+	return 0;
+}
+
+/* Add an endpoint to a new possible bandwidth configuration for this device.
+ * Only one call to this function is allowed per endpoint before
+ * check_bandwidth() or reset_bandwidth() must be called.
+ * A call to xhci_drop_endpoint() followed by a call to xhci_add_endpoint() will
+ * add the endpoint to the schedule with possibly new parameters denoted by a
+ * different endpoint descriptor in usb_host_endpoint.
+ * A call to xhci_add_endpoint() followed by a call to xhci_drop_endpoint() is
+ * not allowed.
+ *
+ * The USB core will not allow URBs to be queued to an endpoint until the
+ * configuration or alt setting is installed in the device, so there's no need
+ * for mutual exclusion to protect the xhci->devs[slot_id] structure.
+ */
+int xhci_add_endpoint(struct usb_hcd *hcd, struct usb_device *udev,
+		struct usb_host_endpoint *ep)
+{
+	struct xhci_hcd *xhci;
+	struct xhci_container_ctx *in_ctx, *out_ctx;
+	unsigned int ep_index;
+	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;
+	struct xhci_virt_device *virt_dev;
+	int ret = 0;
+
+	ret = xhci_check_args(hcd, udev, ep, 1, true, __func__);
+	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);
+	if (xhci->xhc_state & XHCI_STATE_DYING)
+		return -ENODEV;
+
+	added_ctxs = xhci_get_endpoint_flag(&ep->desc);
+	last_ctx = xhci_last_valid_endpoint(added_ctxs);
+	if (added_ctxs == SLOT_FLAG || added_ctxs == EP0_FLAG) {
+		/* FIXME when we have to issue an evaluate endpoint command to
+		 * deal with ep0 max packet size changing once we get the
+		 * descriptors
+		 */
+		xhci_dbg(xhci, "xHCI %s - can't add slot or ep 0 %#x\n",
+				__func__, added_ctxs);
+		return 0;
+	}
+
+	virt_dev = xhci->devs[udev->slot_id];
+	in_ctx = virt_dev->in_ctx;
+	out_ctx = virt_dev->out_ctx;
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, in_ctx);
+	ep_index = xhci_get_endpoint_index(&ep->desc);
+
+	/* If this endpoint is already in use, and the upper layers are trying
+	 * to add it again without dropping it, reject the addition.
+	 */
+	if (virt_dev->eps[ep_index].ring &&
+			!(le32_to_cpu(ctrl_ctx->drop_flags) &
+				xhci_get_endpoint_flag(&ep->desc))) {
+		xhci_warn(xhci, "Trying to add endpoint 0x%x "
+				"without dropping it.\n",
+				(unsigned int) ep->desc.bEndpointAddress);
+		return -EINVAL;
+	}
+
+	/* If the HCD has already noted the endpoint is enabled,
+	 * ignore this request.
+	 */
+	if (le32_to_cpu(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;
+	}
+
+	/*
+	 * Configuration and alternate setting changes must be done in
+	 * process context, not interrupt context (or so documenation
+	 * for usb_set_interface() and usb_set_configuration() claim).
+	 */
+	if (xhci_endpoint_init(xhci, virt_dev, udev, ep, GFP_NOIO) < 0) {
+		dev_dbg(&udev->dev, "%s - could not initialize ep %#x\n",
+				__func__, ep->desc.bEndpointAddress);
+		return -ENOMEM;
+	}
+
+	ctrl_ctx->add_flags |= cpu_to_le32(added_ctxs);
+	new_add_flags = le32_to_cpu(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,
+	 * this re-adds a new state for the endpoint from the new endpoint
+	 * descriptors.  We must drop and re-add this endpoint, so we leave the
+	 * drop flags alone.
+	 */
+	new_drop_flags = le32_to_cpu(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 ((le32_to_cpu(slot_ctx->dev_info) & LAST_CTX_MASK) <
+	    LAST_CTX(last_ctx)) {
+		slot_ctx->dev_info &= cpu_to_le32(~LAST_CTX_MASK);
+		slot_ctx->dev_info |= cpu_to_le32(LAST_CTX(last_ctx));
+	}
+	new_slot_info = le32_to_cpu(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,
+			udev->slot_id,
+			(unsigned int) new_drop_flags,
+			(unsigned int) new_add_flags,
+			(unsigned int) new_slot_info);
+	return 0;
+}
+
+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
+	 * configure endpoint command will leave that endpoint's state
+	 * untouched.  Make sure we don't leave any old state in the input
+	 * endpoint contexts.
+	 */
+	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 &= cpu_to_le32(~LAST_CTX_MASK);
+	/* Endpoint 0 is always valid */
+	slot_ctx->dev_info |= cpu_to_le32(LAST_CTX(1));
+	for (i = 1; i < 31; ++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;
+		ep_ctx->tx_info = 0;
+	}
+}
+
+static int xhci_configure_endpoint_result(struct xhci_hcd *xhci,
+		struct usb_device *udev, u32 *cmd_status)
+{
+	int ret;
+
+	switch (*cmd_status) {
+	case COMP_ENOMEM:
+		dev_warn(&udev->dev, "Not enough host controller resources "
+				"for new device state.\n");
+		ret = -ENOMEM;
+		/* FIXME: can we allocate more resources for the HC? */
+		break;
+	case COMP_BW_ERR:
+	case COMP_2ND_BW_ERR:
+		dev_warn(&udev->dev, "Not enough bandwidth "
+				"for new device state.\n");
+		ret = -ENOSPC;
+		/* FIXME: can we go back to the old state? */
+		break;
+	case COMP_TRB_ERR:
+		/* the HCD set up something wrong */
+		dev_warn(&udev->dev, "ERROR: Endpoint drop flag = 0, "
+				"add flag = 1, "
+				"and endpoint is not disabled.\n");
+		ret = -EINVAL;
+		break;
+	case COMP_DEV_ERR:
+		dev_warn(&udev->dev, "ERROR: Incompatible device for endpoint "
+				"configure command.\n");
+		ret = -ENODEV;
+		break;
+	case COMP_SUCCESS:
+		dev_dbg(&udev->dev, "Successful Endpoint Configure command\n");
+		ret = 0;
+		break;
+	default:
+		xhci_err(xhci, "ERROR: unexpected command completion "
+				"code 0x%x.\n", *cmd_status);
+		ret = -EINVAL;
+		break;
+	}
+	return ret;
+}
+
+static int xhci_evaluate_context_result(struct xhci_hcd *xhci,
+		struct usb_device *udev, u32 *cmd_status)
+{
+	int ret;
+	struct xhci_virt_device *virt_dev = xhci->devs[udev->slot_id];
+
+	switch (*cmd_status) {
+	case COMP_EINVAL:
+		dev_warn(&udev->dev, "WARN: xHCI driver setup invalid evaluate "
+				"context command.\n");
+		ret = -EINVAL;
+		break;
+	case COMP_EBADSLT:
+		dev_warn(&udev->dev, "WARN: slot not enabled for"
+				"evaluate context command.\n");
+		ret = -EINVAL;
+		break;
+	case COMP_CTX_STATE:
+		dev_warn(&udev->dev, "WARN: invalid context state for "
+				"evaluate context command.\n");
+		xhci_dbg_ctx(xhci, virt_dev->out_ctx, 1);
+		ret = -EINVAL;
+		break;
+	case COMP_DEV_ERR:
+		dev_warn(&udev->dev, "ERROR: Incompatible device for evaluate "
+				"context command.\n");
+		ret = -ENODEV;
+		break;
+	case COMP_MEL_ERR:
+		/* Max Exit Latency too large error */
+		dev_warn(&udev->dev, "WARN: Max Exit Latency too large\n");
+		ret = -EINVAL;
+		break;
+	case COMP_SUCCESS:
+		dev_dbg(&udev->dev, "Successful evaluate context command\n");
+		ret = 0;
+		break;
+	default:
+		xhci_err(xhci, "ERROR: unexpected command completion "
+				"code 0x%x.\n", *cmd_status);
+		ret = -EINVAL;
+		break;
+	}
+	return ret;
+}
+
+static u32 xhci_count_num_new_endpoints(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx)
+{
+	struct xhci_input_control_ctx *ctrl_ctx;
+	u32 valid_add_flags;
+	u32 valid_drop_flags;
+
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, in_ctx);
+	/* Ignore the slot flag (bit 0), and the default control endpoint flag
+	 * (bit 1).  The default control endpoint is added during the Address
+	 * Device command and is never removed until the slot is disabled.
+	 */
+	valid_add_flags = ctrl_ctx->add_flags >> 2;
+	valid_drop_flags = ctrl_ctx->drop_flags >> 2;
+
+	/* Use hweight32 to count the number of ones in the add flags, or
+	 * number of endpoints added.  Don't count endpoints that are changed
+	 * (both added and dropped).
+	 */
+	return hweight32(valid_add_flags) -
+		hweight32(valid_add_flags & valid_drop_flags);
+}
+
+static unsigned int xhci_count_num_dropped_endpoints(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx)
+{
+	struct xhci_input_control_ctx *ctrl_ctx;
+	u32 valid_add_flags;
+	u32 valid_drop_flags;
+
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, in_ctx);
+	valid_add_flags = ctrl_ctx->add_flags >> 2;
+	valid_drop_flags = ctrl_ctx->drop_flags >> 2;
+
+	return hweight32(valid_drop_flags) -
+		hweight32(valid_add_flags & valid_drop_flags);
+}
+
+/*
+ * We need to reserve the new number of endpoints before the configure endpoint
+ * command completes.  We can't subtract the dropped endpoints from the number
+ * of active endpoints until the command completes because we can oversubscribe
+ * the host in this case:
+ *
+ *  - the first configure endpoint command drops more endpoints than it adds
+ *  - a second configure endpoint command that adds more endpoints is queued
+ *  - the first configure endpoint command fails, so the config is unchanged
+ *  - the second command may succeed, even though there isn't enough resources
+ *
+ * Must be called with xhci->lock held.
+ */
+static int xhci_reserve_host_resources(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx)
+{
+	u32 added_eps;
+
+	added_eps = xhci_count_num_new_endpoints(xhci, in_ctx);
+	if (xhci->num_active_eps + added_eps > xhci->limit_active_eps) {
+		xhci_dbg(xhci, "Not enough ep ctxs: "
+				"%u active, need to add %u, limit is %u.\n",
+				xhci->num_active_eps, added_eps,
+				xhci->limit_active_eps);
+		return -ENOMEM;
+	}
+	xhci->num_active_eps += added_eps;
+	xhci_dbg(xhci, "Adding %u ep ctxs, %u now active.\n", added_eps,
+			xhci->num_active_eps);
+	return 0;
+}
+
+/*
+ * The configure endpoint was failed by the xHC for some other reason, so we
+ * need to revert the resources that failed configuration would have used.
+ *
+ * Must be called with xhci->lock held.
+ */
+static void xhci_free_host_resources(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx)
+{
+	u32 num_failed_eps;
+
+	num_failed_eps = xhci_count_num_new_endpoints(xhci, in_ctx);
+	xhci->num_active_eps -= num_failed_eps;
+	xhci_dbg(xhci, "Removing %u failed ep ctxs, %u now active.\n",
+			num_failed_eps,
+			xhci->num_active_eps);
+}
+
+/*
+ * Now that the command has completed, clean up the active endpoint count by
+ * subtracting out the endpoints that were dropped (but not changed).
+ *
+ * Must be called with xhci->lock held.
+ */
+static void xhci_finish_resource_reservation(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx)
+{
+	u32 num_dropped_eps;
+
+	num_dropped_eps = xhci_count_num_dropped_endpoints(xhci, in_ctx);
+	xhci->num_active_eps -= num_dropped_eps;
+	if (num_dropped_eps)
+		xhci_dbg(xhci, "Removing %u dropped ep ctxs, %u now active.\n",
+				num_dropped_eps,
+				xhci->num_active_eps);
+}
+
+static unsigned int xhci_get_block_size(struct usb_device *udev)
+{
+	switch (udev->speed) {
+	case USB_SPEED_LOW:
+	case USB_SPEED_FULL:
+		return FS_BLOCK;
+	case USB_SPEED_HIGH:
+		return HS_BLOCK;
+	case USB_SPEED_SUPER:
+		return SS_BLOCK;
+	case USB_SPEED_UNKNOWN:
+	case USB_SPEED_WIRELESS:
+	default:
+		/* Should never happen */
+		return 1;
+	}
+}
+
+static unsigned int
+xhci_get_largest_overhead(struct xhci_interval_bw *interval_bw)
+{
+	if (interval_bw->overhead[LS_OVERHEAD_TYPE])
+		return LS_OVERHEAD;
+	if (interval_bw->overhead[FS_OVERHEAD_TYPE])
+		return FS_OVERHEAD;
+	return HS_OVERHEAD;
+}
+
+/* If we are changing a LS/FS device under a HS hub,
+ * make sure (if we are activating a new TT) that the HS bus has enough
+ * bandwidth for this new TT.
+ */
+static int xhci_check_tt_bw_table(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		int old_active_eps)
+{
+	struct xhci_interval_bw_table *bw_table;
+	struct xhci_tt_bw_info *tt_info;
+
+	/* Find the bandwidth table for the root port this TT is attached to. */
+	bw_table = &xhci->rh_bw[virt_dev->real_port - 1].bw_table;
+	tt_info = virt_dev->tt_info;
+	/* If this TT already had active endpoints, the bandwidth for this TT
+	 * has already been added.  Removing all periodic endpoints (and thus
+	 * making the TT enactive) will only decrease the bandwidth used.
+	 */
+	if (old_active_eps)
+		return 0;
+	if (old_active_eps == 0 && tt_info->active_eps != 0) {
+		if (bw_table->bw_used + TT_HS_OVERHEAD > HS_BW_LIMIT)
+			return -ENOMEM;
+		return 0;
+	}
+	/* Not sure why we would have no new active endpoints...
+	 *
+	 * Maybe because of an Evaluate Context change for a hub update or a
+	 * control endpoint 0 max packet size change?
+	 * FIXME: skip the bandwidth calculation in that case.
+	 */
+	return 0;
+}
+
+static int xhci_check_ss_bw(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev)
+{
+	unsigned int bw_reserved;
+
+	bw_reserved = DIV_ROUND_UP(SS_BW_RESERVED*SS_BW_LIMIT_IN, 100);
+	if (virt_dev->bw_table->ss_bw_in > (SS_BW_LIMIT_IN - bw_reserved))
+		return -ENOMEM;
+
+	bw_reserved = DIV_ROUND_UP(SS_BW_RESERVED*SS_BW_LIMIT_OUT, 100);
+	if (virt_dev->bw_table->ss_bw_out > (SS_BW_LIMIT_OUT - bw_reserved))
+		return -ENOMEM;
+
+	return 0;
+}
+
+/*
+ * This algorithm is a very conservative estimate of the worst-case scheduling
+ * scenario for any one interval.  The hardware dynamically schedules the
+ * packets, so we can't tell which microframe could be the limiting factor in
+ * the bandwidth scheduling.  This only takes into account periodic endpoints.
+ *
+ * Obviously, we can't solve an NP complete problem to find the minimum worst
+ * case scenario.  Instead, we come up with an estimate that is no less than
+ * the worst case bandwidth used for any one microframe, but may be an
+ * over-estimate.
+ *
+ * We walk the requirements for each endpoint by interval, starting with the
+ * smallest interval, and place packets in the schedule where there is only one
+ * possible way to schedule packets for that interval.  In order to simplify
+ * this algorithm, we record the largest max packet size for each interval, and
+ * assume all packets will be that size.
+ *
+ * For interval 0, we obviously must schedule all packets for each interval.
+ * The bandwidth for interval 0 is just the amount of data to be transmitted
+ * (the sum of all max ESIT payload sizes, plus any overhead per packet times
+ * the number of packets).
+ *
+ * For interval 1, we have two possible microframes to schedule those packets
+ * in.  For this algorithm, if we can schedule the same number of packets for
+ * each possible scheduling opportunity (each microframe), we will do so.  The
+ * remaining number of packets will be saved to be transmitted in the gaps in
+ * the next interval's scheduling sequence.
+ *
+ * As we move those remaining packets to be scheduled with interval 2 packets,
+ * we have to double the number of remaining packets to transmit.  This is
+ * because the intervals are actually powers of 2, and we would be transmitting
+ * the previous interval's packets twice in this interval.  We also have to be
+ * sure that when we look at the largest max packet size for this interval, we
+ * also look at the largest max packet size for the remaining packets and take
+ * the greater of the two.
+ *
+ * The algorithm continues to evenly distribute packets in each scheduling
+ * opportunity, and push the remaining packets out, until we get to the last
+ * interval.  Then those packets and their associated overhead are just added
+ * to the bandwidth used.
+ */
+static int xhci_check_bw_table(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		int old_active_eps)
+{
+	unsigned int bw_reserved;
+	unsigned int max_bandwidth;
+	unsigned int bw_used;
+	unsigned int block_size;
+	struct xhci_interval_bw_table *bw_table;
+	unsigned int packet_size = 0;
+	unsigned int overhead = 0;
+	unsigned int packets_transmitted = 0;
+	unsigned int packets_remaining = 0;
+	unsigned int i;
+
+	if (virt_dev->udev->speed == USB_SPEED_SUPER)
+		return xhci_check_ss_bw(xhci, virt_dev);
+
+	if (virt_dev->udev->speed == USB_SPEED_HIGH) {
+		max_bandwidth = HS_BW_LIMIT;
+		/* Convert percent of bus BW reserved to blocks reserved */
+		bw_reserved = DIV_ROUND_UP(HS_BW_RESERVED * max_bandwidth, 100);
+	} else {
+		max_bandwidth = FS_BW_LIMIT;
+		bw_reserved = DIV_ROUND_UP(FS_BW_RESERVED * max_bandwidth, 100);
+	}
+
+	bw_table = virt_dev->bw_table;
+	/* We need to translate the max packet size and max ESIT payloads into
+	 * the units the hardware uses.
+	 */
+	block_size = xhci_get_block_size(virt_dev->udev);
+
+	/* If we are manipulating a LS/FS device under a HS hub, double check
+	 * that the HS bus has enough bandwidth if we are activing a new TT.
+	 */
+	if (virt_dev->tt_info) {
+		xhci_dbg(xhci, "Recalculating BW for rootport %u\n",
+				virt_dev->real_port);
+		if (xhci_check_tt_bw_table(xhci, virt_dev, old_active_eps)) {
+			xhci_warn(xhci, "Not enough bandwidth on HS bus for "
+					"newly activated TT.\n");
+			return -ENOMEM;
+		}
+		xhci_dbg(xhci, "Recalculating BW for TT slot %u port %u\n",
+				virt_dev->tt_info->slot_id,
+				virt_dev->tt_info->ttport);
+	} else {
+		xhci_dbg(xhci, "Recalculating BW for rootport %u\n",
+				virt_dev->real_port);
+	}
+
+	/* Add in how much bandwidth will be used for interval zero, or the
+	 * rounded max ESIT payload + number of packets * largest overhead.
+	 */
+	bw_used = DIV_ROUND_UP(bw_table->interval0_esit_payload, block_size) +
+		bw_table->interval_bw[0].num_packets *
+		xhci_get_largest_overhead(&bw_table->interval_bw[0]);
+
+	for (i = 1; i < XHCI_MAX_INTERVAL; i++) {
+		unsigned int bw_added;
+		unsigned int largest_mps;
+		unsigned int interval_overhead;
+
+		/*
+		 * How many packets could we transmit in this interval?
+		 * If packets didn't fit in the previous interval, we will need
+		 * to transmit that many packets twice within this interval.
+		 */
+		packets_remaining = 2 * packets_remaining +
+			bw_table->interval_bw[i].num_packets;
+
+		/* Find the largest max packet size of this or the previous
+		 * interval.
+		 */
+		if (list_empty(&bw_table->interval_bw[i].endpoints))
+			largest_mps = 0;
+		else {
+			struct xhci_virt_ep *virt_ep;
+			struct list_head *ep_entry;
+
+			ep_entry = bw_table->interval_bw[i].endpoints.next;
+			virt_ep = list_entry(ep_entry,
+					struct xhci_virt_ep, bw_endpoint_list);
+			/* Convert to blocks, rounding up */
+			largest_mps = DIV_ROUND_UP(
+					virt_ep->bw_info.max_packet_size,
+					block_size);
+		}
+		if (largest_mps > packet_size)
+			packet_size = largest_mps;
+
+		/* Use the larger overhead of this or the previous interval. */
+		interval_overhead = xhci_get_largest_overhead(
+				&bw_table->interval_bw[i]);
+		if (interval_overhead > overhead)
+			overhead = interval_overhead;
+
+		/* How many packets can we evenly distribute across
+		 * (1 << (i + 1)) possible scheduling opportunities?
+		 */
+		packets_transmitted = packets_remaining >> (i + 1);
+
+		/* Add in the bandwidth used for those scheduled packets */
+		bw_added = packets_transmitted * (overhead + packet_size);
+
+		/* How many packets do we have remaining to transmit? */
+		packets_remaining = packets_remaining % (1 << (i + 1));
+
+		/* What largest max packet size should those packets have? */
+		/* If we've transmitted all packets, don't carry over the
+		 * largest packet size.
+		 */
+		if (packets_remaining == 0) {
+			packet_size = 0;
+			overhead = 0;
+		} else if (packets_transmitted > 0) {
+			/* Otherwise if we do have remaining packets, and we've
+			 * scheduled some packets in this interval, take the
+			 * largest max packet size from endpoints with this
+			 * interval.
+			 */
+			packet_size = largest_mps;
+			overhead = interval_overhead;
+		}
+		/* Otherwise carry over packet_size and overhead from the last
+		 * time we had a remainder.
+		 */
+		bw_used += bw_added;
+		if (bw_used > max_bandwidth) {
+			xhci_warn(xhci, "Not enough bandwidth. "
+					"Proposed: %u, Max: %u\n",
+				bw_used, max_bandwidth);
+			return -ENOMEM;
+		}
+	}
+	/*
+	 * Ok, we know we have some packets left over after even-handedly
+	 * scheduling interval 15.  We don't know which microframes they will
+	 * fit into, so we over-schedule and say they will be scheduled every
+	 * microframe.
+	 */
+	if (packets_remaining > 0)
+		bw_used += overhead + packet_size;
+
+	if (!virt_dev->tt_info && virt_dev->udev->speed == USB_SPEED_HIGH) {
+		unsigned int port_index = virt_dev->real_port - 1;
+
+		/* OK, we're manipulating a HS device attached to a
+		 * root port bandwidth domain.  Include the number of active TTs
+		 * in the bandwidth used.
+		 */
+		bw_used += TT_HS_OVERHEAD *
+			xhci->rh_bw[port_index].num_active_tts;
+	}
+
+	xhci_dbg(xhci, "Final bandwidth: %u, Limit: %u, Reserved: %u, "
+		"Available: %u " "percent\n",
+		bw_used, max_bandwidth, bw_reserved,
+		(max_bandwidth - bw_used - bw_reserved) * 100 /
+		max_bandwidth);
+
+	bw_used += bw_reserved;
+	if (bw_used > max_bandwidth) {
+		xhci_warn(xhci, "Not enough bandwidth. Proposed: %u, Max: %u\n",
+				bw_used, max_bandwidth);
+		return -ENOMEM;
+	}
+
+	bw_table->bw_used = bw_used;
+	return 0;
+}
+
+static bool xhci_is_async_ep(unsigned int ep_type)
+{
+	return (ep_type != ISOC_OUT_EP && ep_type != INT_OUT_EP &&
+					ep_type != ISOC_IN_EP &&
+					ep_type != INT_IN_EP);
+}
+
+static bool xhci_is_sync_in_ep(unsigned int ep_type)
+{
+	return (ep_type == ISOC_IN_EP || ep_type == INT_IN_EP);
+}
+
+static unsigned int xhci_get_ss_bw_consumed(struct xhci_bw_info *ep_bw)
+{
+	unsigned int mps = DIV_ROUND_UP(ep_bw->max_packet_size, SS_BLOCK);
+
+	if (ep_bw->ep_interval == 0)
+		return SS_OVERHEAD_BURST +
+			(ep_bw->mult * ep_bw->num_packets *
+					(SS_OVERHEAD + mps));
+	return DIV_ROUND_UP(ep_bw->mult * ep_bw->num_packets *
+				(SS_OVERHEAD + mps + SS_OVERHEAD_BURST),
+				1 << ep_bw->ep_interval);
+
+}
+
+void xhci_drop_ep_from_interval_table(struct xhci_hcd *xhci,
+		struct xhci_bw_info *ep_bw,
+		struct xhci_interval_bw_table *bw_table,
+		struct usb_device *udev,
+		struct xhci_virt_ep *virt_ep,
+		struct xhci_tt_bw_info *tt_info)
+{
+	struct xhci_interval_bw	*interval_bw;
+	int normalized_interval;
+
+	if (xhci_is_async_ep(ep_bw->type))
+		return;
+
+	if (udev->speed == USB_SPEED_SUPER) {
+		if (xhci_is_sync_in_ep(ep_bw->type))
+			xhci->devs[udev->slot_id]->bw_table->ss_bw_in -=
+				xhci_get_ss_bw_consumed(ep_bw);
+		else
+			xhci->devs[udev->slot_id]->bw_table->ss_bw_out -=
+				xhci_get_ss_bw_consumed(ep_bw);
+		return;
+	}
+
+	/* SuperSpeed endpoints never get added to intervals in the table, so
+	 * this check is only valid for HS/FS/LS devices.
+	 */
+	if (list_empty(&virt_ep->bw_endpoint_list))
+		return;
+	/* For LS/FS devices, we need to translate the interval expressed in
+	 * microframes to frames.
+	 */
+	if (udev->speed == USB_SPEED_HIGH)
+		normalized_interval = ep_bw->ep_interval;
+	else
+		normalized_interval = ep_bw->ep_interval - 3;
+
+	if (normalized_interval == 0)
+		bw_table->interval0_esit_payload -= ep_bw->max_esit_payload;
+	interval_bw = &bw_table->interval_bw[normalized_interval];
+	interval_bw->num_packets -= ep_bw->num_packets;
+	switch (udev->speed) {
+	case USB_SPEED_LOW:
+		interval_bw->overhead[LS_OVERHEAD_TYPE] -= 1;
+		break;
+	case USB_SPEED_FULL:
+		interval_bw->overhead[FS_OVERHEAD_TYPE] -= 1;
+		break;
+	case USB_SPEED_HIGH:
+		interval_bw->overhead[HS_OVERHEAD_TYPE] -= 1;
+		break;
+	case USB_SPEED_SUPER:
+	case USB_SPEED_UNKNOWN:
+	case USB_SPEED_WIRELESS:
+		/* Should never happen because only LS/FS/HS endpoints will get
+		 * added to the endpoint list.
+		 */
+		return;
+	}
+	if (tt_info)
+		tt_info->active_eps -= 1;
+	list_del_init(&virt_ep->bw_endpoint_list);
+}
+
+static void xhci_add_ep_to_interval_table(struct xhci_hcd *xhci,
+		struct xhci_bw_info *ep_bw,
+		struct xhci_interval_bw_table *bw_table,
+		struct usb_device *udev,
+		struct xhci_virt_ep *virt_ep,
+		struct xhci_tt_bw_info *tt_info)
+{
+	struct xhci_interval_bw	*interval_bw;
+	struct xhci_virt_ep *smaller_ep;
+	int normalized_interval;
+
+	if (xhci_is_async_ep(ep_bw->type))
+		return;
+
+	if (udev->speed == USB_SPEED_SUPER) {
+		if (xhci_is_sync_in_ep(ep_bw->type))
+			xhci->devs[udev->slot_id]->bw_table->ss_bw_in +=
+				xhci_get_ss_bw_consumed(ep_bw);
+		else
+			xhci->devs[udev->slot_id]->bw_table->ss_bw_out +=
+				xhci_get_ss_bw_consumed(ep_bw);
+		return;
+	}
+
+	/* For LS/FS devices, we need to translate the interval expressed in
+	 * microframes to frames.
+	 */
+	if (udev->speed == USB_SPEED_HIGH)
+		normalized_interval = ep_bw->ep_interval;
+	else
+		normalized_interval = ep_bw->ep_interval - 3;
+
+	if (normalized_interval == 0)
+		bw_table->interval0_esit_payload += ep_bw->max_esit_payload;
+	interval_bw = &bw_table->interval_bw[normalized_interval];
+	interval_bw->num_packets += ep_bw->num_packets;
+	switch (udev->speed) {
+	case USB_SPEED_LOW:
+		interval_bw->overhead[LS_OVERHEAD_TYPE] += 1;
+		break;
+	case USB_SPEED_FULL:
+		interval_bw->overhead[FS_OVERHEAD_TYPE] += 1;
+		break;
+	case USB_SPEED_HIGH:
+		interval_bw->overhead[HS_OVERHEAD_TYPE] += 1;
+		break;
+	case USB_SPEED_SUPER:
+	case USB_SPEED_UNKNOWN:
+	case USB_SPEED_WIRELESS:
+		/* Should never happen because only LS/FS/HS endpoints will get
+		 * added to the endpoint list.
+		 */
+		return;
+	}
+
+	if (tt_info)
+		tt_info->active_eps += 1;
+	/* Insert the endpoint into the list, largest max packet size first. */
+	list_for_each_entry(smaller_ep, &interval_bw->endpoints,
+			bw_endpoint_list) {
+		if (ep_bw->max_packet_size >=
+				smaller_ep->bw_info.max_packet_size) {
+			/* Add the new ep before the smaller endpoint */
+			list_add_tail(&virt_ep->bw_endpoint_list,
+					&smaller_ep->bw_endpoint_list);
+			return;
+		}
+	}
+	/* Add the new endpoint at the end of the list. */
+	list_add_tail(&virt_ep->bw_endpoint_list,
+			&interval_bw->endpoints);
+}
+
+void xhci_update_tt_active_eps(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		int old_active_eps)
+{
+	struct xhci_root_port_bw_info *rh_bw_info;
+	if (!virt_dev->tt_info)
+		return;
+
+	rh_bw_info = &xhci->rh_bw[virt_dev->real_port - 1];
+	if (old_active_eps == 0 &&
+				virt_dev->tt_info->active_eps != 0) {
+		rh_bw_info->num_active_tts += 1;
+		rh_bw_info->bw_table.bw_used += TT_HS_OVERHEAD;
+	} else if (old_active_eps != 0 &&
+				virt_dev->tt_info->active_eps == 0) {
+		rh_bw_info->num_active_tts -= 1;
+		rh_bw_info->bw_table.bw_used -= TT_HS_OVERHEAD;
+	}
+}
+
+static int xhci_reserve_bandwidth(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		struct xhci_container_ctx *in_ctx)
+{
+	struct xhci_bw_info ep_bw_info[31];
+	int i;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	int old_active_eps = 0;
+
+	if (virt_dev->tt_info)
+		old_active_eps = virt_dev->tt_info->active_eps;
+
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, in_ctx);
+
+	for (i = 0; i < 31; i++) {
+		if (!EP_IS_ADDED(ctrl_ctx, i) && !EP_IS_DROPPED(ctrl_ctx, i))
+			continue;
+
+		/* Make a copy of the BW info in case we need to revert this */
+		memcpy(&ep_bw_info[i], &virt_dev->eps[i].bw_info,
+				sizeof(ep_bw_info[i]));
+		/* Drop the endpoint from the interval table if the endpoint is
+		 * being dropped or changed.
+		 */
+		if (EP_IS_DROPPED(ctrl_ctx, i))
+			xhci_drop_ep_from_interval_table(xhci,
+					&virt_dev->eps[i].bw_info,
+					virt_dev->bw_table,
+					virt_dev->udev,
+					&virt_dev->eps[i],
+					virt_dev->tt_info);
+	}
+	/* Overwrite the information stored in the endpoints' bw_info */
+	xhci_update_bw_info(xhci, virt_dev->in_ctx, ctrl_ctx, virt_dev);
+	for (i = 0; i < 31; i++) {
+		/* Add any changed or added endpoints to the interval table */
+		if (EP_IS_ADDED(ctrl_ctx, i))
+			xhci_add_ep_to_interval_table(xhci,
+					&virt_dev->eps[i].bw_info,
+					virt_dev->bw_table,
+					virt_dev->udev,
+					&virt_dev->eps[i],
+					virt_dev->tt_info);
+	}
+
+	if (!xhci_check_bw_table(xhci, virt_dev, old_active_eps)) {
+		/* Ok, this fits in the bandwidth we have.
+		 * Update the number of active TTs.
+		 */
+		xhci_update_tt_active_eps(xhci, virt_dev, old_active_eps);
+		return 0;
+	}
+
+	/* We don't have enough bandwidth for this, revert the stored info. */
+	for (i = 0; i < 31; i++) {
+		if (!EP_IS_ADDED(ctrl_ctx, i) && !EP_IS_DROPPED(ctrl_ctx, i))
+			continue;
+
+		/* Drop the new copies of any added or changed endpoints from
+		 * the interval table.
+		 */
+		if (EP_IS_ADDED(ctrl_ctx, i)) {
+			xhci_drop_ep_from_interval_table(xhci,
+					&virt_dev->eps[i].bw_info,
+					virt_dev->bw_table,
+					virt_dev->udev,
+					&virt_dev->eps[i],
+					virt_dev->tt_info);
+		}
+		/* Revert the endpoint back to its old information */
+		memcpy(&virt_dev->eps[i].bw_info, &ep_bw_info[i],
+				sizeof(ep_bw_info[i]));
+		/* Add any changed or dropped endpoints back into the table */
+		if (EP_IS_DROPPED(ctrl_ctx, i))
+			xhci_add_ep_to_interval_table(xhci,
+					&virt_dev->eps[i].bw_info,
+					virt_dev->bw_table,
+					virt_dev->udev,
+					&virt_dev->eps[i],
+					virt_dev->tt_info);
+	}
+	return -ENOMEM;
+}
+
+
+/* Issue a configure endpoint command or evaluate context command
+ * and wait for it to finish.
+ */
+static int xhci_configure_endpoint(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		struct xhci_command *command,
+		bool ctx_change, bool must_succeed)
+{
+	int ret;
+	int timeleft;
+	unsigned long flags;
+	struct xhci_container_ctx *in_ctx;
+	struct completion *cmd_completion;
+	u32 *cmd_status;
+	struct xhci_virt_device *virt_dev;
+	union xhci_trb *cmd_trb;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	virt_dev = xhci->devs[udev->slot_id];
+
+	if (command)
+		in_ctx = command->in_ctx;
+	else
+		in_ctx = virt_dev->in_ctx;
+
+	if ((xhci->quirks & XHCI_EP_LIMIT_QUIRK) &&
+			xhci_reserve_host_resources(xhci, in_ctx)) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		xhci_warn(xhci, "Not enough host resources, "
+				"active endpoint contexts = %u\n",
+				xhci->num_active_eps);
+		return -ENOMEM;
+	}
+	if ((xhci->quirks & XHCI_SW_BW_CHECKING) &&
+			xhci_reserve_bandwidth(xhci, virt_dev, in_ctx)) {
+		if ((xhci->quirks & XHCI_EP_LIMIT_QUIRK))
+			xhci_free_host_resources(xhci, in_ctx);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		xhci_warn(xhci, "Not enough bandwidth\n");
+		return -ENOMEM;
+	}
+
+	if (command) {
+		cmd_completion = command->completion;
+		cmd_status = &command->status;
+		command->command_trb = xhci->cmd_ring->enqueue;
+
+		/* Enqueue pointer can be left pointing to the link TRB,
+		 * we must handle that
+		 */
+		if (TRB_TYPE_LINK_LE32(command->command_trb->link.control))
+			command->command_trb =
+				xhci->cmd_ring->enq_seg->next->trbs;
+
+		list_add_tail(&command->cmd_list, &virt_dev->cmd_list);
+	} else {
+		cmd_completion = &virt_dev->cmd_completion;
+		cmd_status = &virt_dev->cmd_status;
+	}
+	init_completion(cmd_completion);
+
+	cmd_trb = xhci->cmd_ring->dequeue;
+	if (!ctx_change)
+		ret = xhci_queue_configure_endpoint(xhci, in_ctx->dma,
+				udev->slot_id, must_succeed);
+	else
+		ret = xhci_queue_evaluate_context(xhci, in_ctx->dma,
+				udev->slot_id, must_succeed);
+	if (ret < 0) {
+		if (command)
+			list_del(&command->cmd_list);
+		if ((xhci->quirks & XHCI_EP_LIMIT_QUIRK))
+			xhci_free_host_resources(xhci, in_ctx);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		xhci_dbg(xhci, "FIXME allocate a new ring segment\n");
+		return -ENOMEM;
+	}
+	xhci_ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* Wait for the configure endpoint command to complete */
+	timeleft = wait_for_completion_interruptible_timeout(
+			cmd_completion,
+			XHCI_CMD_DEFAULT_TIMEOUT);
+	if (timeleft <= 0) {
+		xhci_warn(xhci, "%s while waiting for %s command\n",
+				timeleft == 0 ? "Timeout" : "Signal",
+				ctx_change == 0 ?
+					"configure endpoint" :
+					"evaluate context");
+		/* cancel the configure endpoint command */
+		ret = xhci_cancel_cmd(xhci, command, cmd_trb);
+		if (ret < 0)
+			return ret;
+		return -ETIME;
+	}
+
+	if (!ctx_change)
+		ret = xhci_configure_endpoint_result(xhci, udev, cmd_status);
+	else
+		ret = xhci_evaluate_context_result(xhci, udev, cmd_status);
+
+	if ((xhci->quirks & XHCI_EP_LIMIT_QUIRK)) {
+		spin_lock_irqsave(&xhci->lock, flags);
+		/* If the command failed, remove the reserved resources.
+		 * Otherwise, clean up the estimate to include dropped eps.
+		 */
+		if (ret)
+			xhci_free_host_resources(xhci, in_ctx);
+		else
+			xhci_finish_resource_reservation(xhci, in_ctx);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+	}
+	return ret;
+}
+
+/* Called after one or more calls to xhci_add_endpoint() or
+ * xhci_drop_endpoint().  If this call fails, the USB core is expected
+ * to call xhci_reset_bandwidth().
+ *
+ * Since we are in the middle of changing either configuration or
+ * installing a new alt setting, the USB core won't allow URBs to be
+ * enqueued for any endpoint on the old config or interface.  Nothing
+ * else should be touching the xhci->devs[slot_id] structure, so we
+ * don't need to take the xhci->lock for manipulating that.
+ */
+int xhci_check_bandwidth(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	int i;
+	int ret = 0;
+	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, true, __func__);
+	if (ret <= 0)
+		return ret;
+	xhci = hcd_to_xhci(hcd);
+	if (xhci->xhc_state & XHCI_STATE_DYING)
+		return -ENODEV;
+
+	xhci_dbg(xhci, "%s called for udev %p\n", __func__, udev);
+	virt_dev = xhci->devs[udev->slot_id];
+
+	/* See section 4.6.6 - A0 = 1; A1 = D0 = D1 = 0 */
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, virt_dev->in_ctx);
+	ctrl_ctx->add_flags |= cpu_to_le32(SLOT_FLAG);
+	ctrl_ctx->add_flags &= cpu_to_le32(~EP0_FLAG);
+	ctrl_ctx->drop_flags &= cpu_to_le32(~(SLOT_FLAG | EP0_FLAG));
+
+	/* Don't issue the command if there's no endpoints to update. */
+	if (ctrl_ctx->add_flags == cpu_to_le32(SLOT_FLAG) &&
+			ctrl_ctx->drop_flags == 0)
+		return 0;
+
+	xhci_dbg(xhci, "New Input Control Context:\n");
+	slot_ctx = xhci_get_slot_ctx(xhci, virt_dev->in_ctx);
+	xhci_dbg_ctx(xhci, virt_dev->in_ctx,
+		     LAST_CTX_TO_EP_NUM(le32_to_cpu(slot_ctx->dev_info)));
+
+	ret = xhci_configure_endpoint(xhci, udev, NULL,
+			false, false);
+	if (ret) {
+		/* Callee should call reset_bandwidth() */
+		return ret;
+	}
+
+	xhci_dbg(xhci, "Output context after successful config ep cmd:\n");
+	xhci_dbg_ctx(xhci, virt_dev->out_ctx,
+		     LAST_CTX_TO_EP_NUM(le32_to_cpu(slot_ctx->dev_info)));
+
+	/* Free any rings that were dropped, but not changed. */
+	for (i = 1; i < 31; ++i) {
+		if ((le32_to_cpu(ctrl_ctx->drop_flags) & (1 << (i + 1))) &&
+		    !(le32_to_cpu(ctrl_ctx->add_flags) & (1 << (i + 1))))
+			xhci_free_or_cache_endpoint_ring(xhci, virt_dev, i);
+	}
+	xhci_zero_in_ctx(xhci, virt_dev);
+	/*
+	 * Install any rings for completely new endpoints or changed endpoints,
+	 * and free or cache any old rings from changed endpoints.
+	 */
+	for (i = 1; i < 31; ++i) {
+		if (!virt_dev->eps[i].new_ring)
+			continue;
+		/* Only cache or free the old ring if it exists.
+		 * It may not if this is the first add of an endpoint.
+		 */
+		if (virt_dev->eps[i].ring) {
+			xhci_free_or_cache_endpoint_ring(xhci, virt_dev, i);
+		}
+		virt_dev->eps[i].ring = virt_dev->eps[i].new_ring;
+		virt_dev->eps[i].new_ring = NULL;
+	}
+
+	return ret;
+}
+
+void xhci_reset_bandwidth(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	struct xhci_hcd *xhci;
+	struct xhci_virt_device	*virt_dev;
+	int i, ret;
+
+	ret = xhci_check_args(hcd, udev, NULL, 0, true, __func__);
+	if (ret <= 0)
+		return;
+	xhci = hcd_to_xhci(hcd);
+
+	xhci_dbg(xhci, "%s called for udev %p\n", __func__, udev);
+	virt_dev = xhci->devs[udev->slot_id];
+	/* Free any rings allocated for added endpoints */
+	for (i = 0; i < 31; ++i) {
+		if (virt_dev->eps[i].new_ring) {
+			xhci_ring_free(xhci, virt_dev->eps[i].new_ring);
+			virt_dev->eps[i].new_ring = NULL;
+		}
+	}
+	xhci_zero_in_ctx(xhci, virt_dev);
+}
+
+static void xhci_setup_input_ctx_for_config_ep(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx,
+		struct xhci_container_ctx *out_ctx,
+		u32 add_flags, u32 drop_flags)
+{
+	struct xhci_input_control_ctx *ctrl_ctx;
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, in_ctx);
+	ctrl_ctx->add_flags = cpu_to_le32(add_flags);
+	ctrl_ctx->drop_flags = cpu_to_le32(drop_flags);
+	xhci_slot_copy(xhci, in_ctx, out_ctx);
+	ctrl_ctx->add_flags |= cpu_to_le32(SLOT_FLAG);
+
+	xhci_dbg(xhci, "Input Context:\n");
+	xhci_dbg_ctx(xhci, in_ctx, xhci_last_valid_endpoint(add_flags));
+}
+
+static void xhci_setup_input_ctx_for_quirk(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		struct xhci_dequeue_state *deq_state)
+{
+	struct xhci_container_ctx *in_ctx;
+	struct xhci_ep_ctx *ep_ctx;
+	u32 added_ctxs;
+	dma_addr_t addr;
+
+	xhci_endpoint_copy(xhci, xhci->devs[slot_id]->in_ctx,
+			xhci->devs[slot_id]->out_ctx, ep_index);
+	in_ctx = xhci->devs[slot_id]->in_ctx;
+	ep_ctx = xhci_get_ep_ctx(xhci, in_ctx, ep_index);
+	addr = xhci_trb_virt_to_dma(deq_state->new_deq_seg,
+			deq_state->new_deq_ptr);
+	if (addr == 0) {
+		xhci_warn(xhci, "WARN Cannot submit config ep after "
+				"reset ep command\n");
+		xhci_warn(xhci, "WARN deq seg = %p, deq ptr = %p\n",
+				deq_state->new_deq_seg,
+				deq_state->new_deq_ptr);
+		return;
+	}
+	ep_ctx->deq = cpu_to_le64(addr | deq_state->new_cycle_state);
+
+	added_ctxs = xhci_get_endpoint_flag_from_index(ep_index);
+	xhci_setup_input_ctx_for_config_ep(xhci, xhci->devs[slot_id]->in_ctx,
+			xhci->devs[slot_id]->out_ctx, added_ctxs, added_ctxs);
+}
+
+void xhci_cleanup_stalled_ring(struct xhci_hcd *xhci,
+		struct usb_device *udev, unsigned int ep_index)
+{
+	struct xhci_dequeue_state deq_state;
+	struct xhci_virt_ep *ep;
+
+	xhci_dbg(xhci, "Cleaning up stalled endpoint ring\n");
+	ep = &xhci->devs[udev->slot_id]->eps[ep_index];
+	/* 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->stopped_stream, ep->stopped_td,
+			&deq_state);
+
+	/* HW with the reset endpoint quirk will use the saved dequeue state to
+	 * issue a configure endpoint command later.
+	 */
+	if (!(xhci->quirks & XHCI_RESET_EP_QUIRK)) {
+		xhci_dbg(xhci, "Queueing new dequeue state\n");
+		xhci_queue_new_dequeue_state(xhci, udev->slot_id,
+				ep_index, ep->stopped_stream, &deq_state);
+	} else {
+		/* Better hope no one uses the input context between now and the
+		 * reset endpoint completion!
+		 * XXX: No idea how this hardware will react when stream rings
+		 * are enabled.
+		 */
+		xhci_dbg(xhci, "Setting up input context for "
+				"configure endpoint command\n");
+		xhci_setup_input_ctx_for_quirk(xhci, udev->slot_id,
+				ep_index, &deq_state);
+	}
+}
+
+/* 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_virt_ep *virt_ep;
+
+	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);
+	virt_ep = &xhci->devs[udev->slot_id]->eps[ep_index];
+	if (!virt_ep->stopped_td) {
+		xhci_dbg(xhci, "Endpoint 0x%x not halted, refusing to reset.\n",
+				ep->desc.bEndpointAddress);
+		return;
+	}
+	if (usb_endpoint_xfer_control(&ep->desc)) {
+		xhci_dbg(xhci, "Control endpoint stall already handled.\n");
+		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_cleanup_stalled_ring(xhci, udev, ep_index);
+		kfree(virt_ep->stopped_td);
+		xhci_ring_cmd_db(xhci);
+	}
+	virt_ep->stopped_td = NULL;
+	virt_ep->stopped_trb = NULL;
+	virt_ep->stopped_stream = 0;
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	if (ret)
+		xhci_warn(xhci, "FIXME allocate a new ring segment\n");
+}
+
+static int xhci_check_streams_endpoint(struct xhci_hcd *xhci,
+		struct usb_device *udev, struct usb_host_endpoint *ep,
+		unsigned int slot_id)
+{
+	int ret;
+	unsigned int ep_index;
+	unsigned int ep_state;
+
+	if (!ep)
+		return -EINVAL;
+	ret = xhci_check_args(xhci_to_hcd(xhci), udev, ep, 1, true, __func__);
+	if (ret <= 0)
+		return -EINVAL;
+	if (ep->ss_ep_comp.bmAttributes == 0) {
+		xhci_warn(xhci, "WARN: SuperSpeed Endpoint Companion"
+				" descriptor for ep 0x%x does not support streams\n",
+				ep->desc.bEndpointAddress);
+		return -EINVAL;
+	}
+
+	ep_index = xhci_get_endpoint_index(&ep->desc);
+	ep_state = xhci->devs[slot_id]->eps[ep_index].ep_state;
+	if (ep_state & EP_HAS_STREAMS ||
+			ep_state & EP_GETTING_STREAMS) {
+		xhci_warn(xhci, "WARN: SuperSpeed bulk endpoint 0x%x "
+				"already has streams set up.\n",
+				ep->desc.bEndpointAddress);
+		xhci_warn(xhci, "Send email to xHCI maintainer and ask for "
+				"dynamic stream context array reallocation.\n");
+		return -EINVAL;
+	}
+	if (!list_empty(&xhci->devs[slot_id]->eps[ep_index].ring->td_list)) {
+		xhci_warn(xhci, "Cannot setup streams for SuperSpeed bulk "
+				"endpoint 0x%x; URBs are pending.\n",
+				ep->desc.bEndpointAddress);
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static void xhci_calculate_streams_entries(struct xhci_hcd *xhci,
+		unsigned int *num_streams, unsigned int *num_stream_ctxs)
+{
+	unsigned int max_streams;
+
+	/* The stream context array size must be a power of two */
+	*num_stream_ctxs = roundup_pow_of_two(*num_streams);
+	/*
+	 * Find out how many primary stream array entries the host controller
+	 * supports.  Later we may use secondary stream arrays (similar to 2nd
+	 * level page entries), but that's an optional feature for xHCI host
+	 * controllers. xHCs must support at least 4 stream IDs.
+	 */
+	max_streams = HCC_MAX_PSA(xhci->hcc_params);
+	if (*num_stream_ctxs > max_streams) {
+		xhci_dbg(xhci, "xHCI HW only supports %u stream ctx entries.\n",
+				max_streams);
+		*num_stream_ctxs = max_streams;
+		*num_streams = max_streams;
+	}
+}
+
+/* Returns an error code if one of the endpoint already has streams.
+ * This does not change any data structures, it only checks and gathers
+ * information.
+ */
+static int xhci_calculate_streams_and_bitmask(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		struct usb_host_endpoint **eps, unsigned int num_eps,
+		unsigned int *num_streams, u32 *changed_ep_bitmask)
+{
+	unsigned int max_streams;
+	unsigned int endpoint_flag;
+	int i;
+	int ret;
+
+	for (i = 0; i < num_eps; i++) {
+		ret = xhci_check_streams_endpoint(xhci, udev,
+				eps[i], udev->slot_id);
+		if (ret < 0)
+			return ret;
+
+		max_streams = usb_ss_max_streams(&eps[i]->ss_ep_comp);
+		if (max_streams < (*num_streams - 1)) {
+			xhci_dbg(xhci, "Ep 0x%x only supports %u stream IDs.\n",
+					eps[i]->desc.bEndpointAddress,
+					max_streams);
+			*num_streams = max_streams+1;
+		}
+
+		endpoint_flag = xhci_get_endpoint_flag(&eps[i]->desc);
+		if (*changed_ep_bitmask & endpoint_flag)
+			return -EINVAL;
+		*changed_ep_bitmask |= endpoint_flag;
+	}
+	return 0;
+}
+
+static u32 xhci_calculate_no_streams_bitmask(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		struct usb_host_endpoint **eps, unsigned int num_eps)
+{
+	u32 changed_ep_bitmask = 0;
+	unsigned int slot_id;
+	unsigned int ep_index;
+	unsigned int ep_state;
+	int i;
+
+	slot_id = udev->slot_id;
+	if (!xhci->devs[slot_id])
+		return 0;
+
+	for (i = 0; i < num_eps; i++) {
+		ep_index = xhci_get_endpoint_index(&eps[i]->desc);
+		ep_state = xhci->devs[slot_id]->eps[ep_index].ep_state;
+		/* Are streams already being freed for the endpoint? */
+		if (ep_state & EP_GETTING_NO_STREAMS) {
+			xhci_warn(xhci, "WARN Can't disable streams for "
+					"endpoint 0x%x\n, "
+					"streams are being disabled already.",
+					eps[i]->desc.bEndpointAddress);
+			return 0;
+		}
+		/* Are there actually any streams to free? */
+		if (!(ep_state & EP_HAS_STREAMS) &&
+				!(ep_state & EP_GETTING_STREAMS)) {
+			xhci_warn(xhci, "WARN Can't disable streams for "
+					"endpoint 0x%x\n, "
+					"streams are already disabled!",
+					eps[i]->desc.bEndpointAddress);
+			xhci_warn(xhci, "WARN xhci_free_streams() called "
+					"with non-streams endpoint\n");
+			return 0;
+		}
+		changed_ep_bitmask |= xhci_get_endpoint_flag(&eps[i]->desc);
+	}
+	return changed_ep_bitmask;
+}
+
+/*
+ * The USB device drivers use this function (though the HCD interface in USB
+ * core) to prepare a set of bulk endpoints to use streams.  Streams are used to
+ * coordinate mass storage command queueing across multiple endpoints (basically
+ * a stream ID == a task ID).
+ *
+ * Setting up streams involves allocating the same size stream context array
+ * for each endpoint and issuing a configure endpoint command for all endpoints.
+ *
+ * Don't allow the call to succeed if one endpoint only supports one stream
+ * (which means it doesn't support streams at all).
+ *
+ * Drivers may get less stream IDs than they asked for, if the host controller
+ * hardware or endpoints claim they can't support the number of requested
+ * stream IDs.
+ */
+int xhci_alloc_streams(struct usb_hcd *hcd, struct usb_device *udev,
+		struct usb_host_endpoint **eps, unsigned int num_eps,
+		unsigned int num_streams, gfp_t mem_flags)
+{
+	int i, ret;
+	struct xhci_hcd *xhci;
+	struct xhci_virt_device *vdev;
+	struct xhci_command *config_cmd;
+	unsigned int ep_index;
+	unsigned int num_stream_ctxs;
+	unsigned long flags;
+	u32 changed_ep_bitmask = 0;
+
+	if (!eps)
+		return -EINVAL;
+
+	/* Add one to the number of streams requested to account for
+	 * stream 0 that is reserved for xHCI usage.
+	 */
+	num_streams += 1;
+	xhci = hcd_to_xhci(hcd);
+	xhci_dbg(xhci, "Driver wants %u stream IDs (including stream 0).\n",
+			num_streams);
+
+	config_cmd = xhci_alloc_command(xhci, true, true, mem_flags);
+	if (!config_cmd) {
+		xhci_dbg(xhci, "Could not allocate xHCI command structure.\n");
+		return -ENOMEM;
+	}
+
+	/* Check to make sure all endpoints are not already configured for
+	 * streams.  While we're at it, find the maximum number of streams that
+	 * all the endpoints will support and check for duplicate endpoints.
+	 */
+	spin_lock_irqsave(&xhci->lock, flags);
+	ret = xhci_calculate_streams_and_bitmask(xhci, udev, eps,
+			num_eps, &num_streams, &changed_ep_bitmask);
+	if (ret < 0) {
+		xhci_free_command(xhci, config_cmd);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return ret;
+	}
+	if (num_streams <= 1) {
+		xhci_warn(xhci, "WARN: endpoints can't handle "
+				"more than one stream.\n");
+		xhci_free_command(xhci, config_cmd);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return -EINVAL;
+	}
+	vdev = xhci->devs[udev->slot_id];
+	/* Mark each endpoint as being in transition, so
+	 * xhci_urb_enqueue() will reject all URBs.
+	 */
+	for (i = 0; i < num_eps; i++) {
+		ep_index = xhci_get_endpoint_index(&eps[i]->desc);
+		vdev->eps[ep_index].ep_state |= EP_GETTING_STREAMS;
+	}
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* Setup internal data structures and allocate HW data structures for
+	 * streams (but don't install the HW structures in the input context
+	 * until we're sure all memory allocation succeeded).
+	 */
+	xhci_calculate_streams_entries(xhci, &num_streams, &num_stream_ctxs);
+	xhci_dbg(xhci, "Need %u stream ctx entries for %u stream IDs.\n",
+			num_stream_ctxs, num_streams);
+
+	for (i = 0; i < num_eps; i++) {
+		ep_index = xhci_get_endpoint_index(&eps[i]->desc);
+		vdev->eps[ep_index].stream_info = xhci_alloc_stream_info(xhci,
+				num_stream_ctxs,
+				num_streams, mem_flags);
+		if (!vdev->eps[ep_index].stream_info)
+			goto cleanup;
+		/* Set maxPstreams in endpoint context and update deq ptr to
+		 * point to stream context array. FIXME
+		 */
+	}
+
+	/* Set up the input context for a configure endpoint command. */
+	for (i = 0; i < num_eps; i++) {
+		struct xhci_ep_ctx *ep_ctx;
+
+		ep_index = xhci_get_endpoint_index(&eps[i]->desc);
+		ep_ctx = xhci_get_ep_ctx(xhci, config_cmd->in_ctx, ep_index);
+
+		xhci_endpoint_copy(xhci, config_cmd->in_ctx,
+				vdev->out_ctx, ep_index);
+		xhci_setup_streams_ep_input_ctx(xhci, ep_ctx,
+				vdev->eps[ep_index].stream_info);
+	}
+	/* Tell the HW to drop its old copy of the endpoint context info
+	 * and add the updated copy from the input context.
+	 */
+	xhci_setup_input_ctx_for_config_ep(xhci, config_cmd->in_ctx,
+			vdev->out_ctx, changed_ep_bitmask, changed_ep_bitmask);
+
+	/* Issue and wait for the configure endpoint command */
+	ret = xhci_configure_endpoint(xhci, udev, config_cmd,
+			false, false);
+
+	/* xHC rejected the configure endpoint command for some reason, so we
+	 * leave the old ring intact and free our internal streams data
+	 * structure.
+	 */
+	if (ret < 0)
+		goto cleanup;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	for (i = 0; i < num_eps; i++) {
+		ep_index = xhci_get_endpoint_index(&eps[i]->desc);
+		vdev->eps[ep_index].ep_state &= ~EP_GETTING_STREAMS;
+		xhci_dbg(xhci, "Slot %u ep ctx %u now has streams.\n",
+			 udev->slot_id, ep_index);
+		vdev->eps[ep_index].ep_state |= EP_HAS_STREAMS;
+	}
+	xhci_free_command(xhci, config_cmd);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* Subtract 1 for stream 0, which drivers can't use */
+	return num_streams - 1;
+
+cleanup:
+	/* If it didn't work, free the streams! */
+	for (i = 0; i < num_eps; i++) {
+		ep_index = xhci_get_endpoint_index(&eps[i]->desc);
+		xhci_free_stream_info(xhci, vdev->eps[ep_index].stream_info);
+		vdev->eps[ep_index].stream_info = NULL;
+		/* FIXME Unset maxPstreams in endpoint context and
+		 * update deq ptr to point to normal string ring.
+		 */
+		vdev->eps[ep_index].ep_state &= ~EP_GETTING_STREAMS;
+		vdev->eps[ep_index].ep_state &= ~EP_HAS_STREAMS;
+		xhci_endpoint_zero(xhci, vdev, eps[i]);
+	}
+	xhci_free_command(xhci, config_cmd);
+	return -ENOMEM;
+}
+
+/* Transition the endpoint from using streams to being a "normal" endpoint
+ * without streams.
+ *
+ * Modify the endpoint context state, submit a configure endpoint command,
+ * and free all endpoint rings for streams if that completes successfully.
+ */
+int xhci_free_streams(struct usb_hcd *hcd, struct usb_device *udev,
+		struct usb_host_endpoint **eps, unsigned int num_eps,
+		gfp_t mem_flags)
+{
+	int i, ret;
+	struct xhci_hcd *xhci;
+	struct xhci_virt_device *vdev;
+	struct xhci_command *command;
+	unsigned int ep_index;
+	unsigned long flags;
+	u32 changed_ep_bitmask;
+
+	xhci = hcd_to_xhci(hcd);
+	vdev = xhci->devs[udev->slot_id];
+
+	/* Set up a configure endpoint command to remove the streams rings */
+	spin_lock_irqsave(&xhci->lock, flags);
+	changed_ep_bitmask = xhci_calculate_no_streams_bitmask(xhci,
+			udev, eps, num_eps);
+	if (changed_ep_bitmask == 0) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return -EINVAL;
+	}
+
+	/* Use the xhci_command structure from the first endpoint.  We may have
+	 * allocated too many, but the driver may call xhci_free_streams() for
+	 * each endpoint it grouped into one call to xhci_alloc_streams().
+	 */
+	ep_index = xhci_get_endpoint_index(&eps[0]->desc);
+	command = vdev->eps[ep_index].stream_info->free_streams_command;
+	for (i = 0; i < num_eps; i++) {
+		struct xhci_ep_ctx *ep_ctx;
+
+		ep_index = xhci_get_endpoint_index(&eps[i]->desc);
+		ep_ctx = xhci_get_ep_ctx(xhci, command->in_ctx, ep_index);
+		xhci->devs[udev->slot_id]->eps[ep_index].ep_state |=
+			EP_GETTING_NO_STREAMS;
+
+		xhci_endpoint_copy(xhci, command->in_ctx,
+				vdev->out_ctx, ep_index);
+		xhci_setup_no_streams_ep_input_ctx(xhci, ep_ctx,
+				&vdev->eps[ep_index]);
+	}
+	xhci_setup_input_ctx_for_config_ep(xhci, command->in_ctx,
+			vdev->out_ctx, changed_ep_bitmask, changed_ep_bitmask);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* Issue and wait for the configure endpoint command,
+	 * which must succeed.
+	 */
+	ret = xhci_configure_endpoint(xhci, udev, command,
+			false, true);
+
+	/* xHC rejected the configure endpoint command for some reason, so we
+	 * leave the streams rings intact.
+	 */
+	if (ret < 0)
+		return ret;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	for (i = 0; i < num_eps; i++) {
+		ep_index = xhci_get_endpoint_index(&eps[i]->desc);
+		xhci_free_stream_info(xhci, vdev->eps[ep_index].stream_info);
+		vdev->eps[ep_index].stream_info = NULL;
+		/* FIXME Unset maxPstreams in endpoint context and
+		 * update deq ptr to point to normal string ring.
+		 */
+		vdev->eps[ep_index].ep_state &= ~EP_GETTING_NO_STREAMS;
+		vdev->eps[ep_index].ep_state &= ~EP_HAS_STREAMS;
+	}
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	return 0;
+}
+
+/*
+ * Deletes endpoint resources for endpoints that were active before a Reset
+ * Device command, or a Disable Slot command.  The Reset Device command leaves
+ * the control endpoint intact, whereas the Disable Slot command deletes it.
+ *
+ * Must be called with xhci->lock held.
+ */
+void xhci_free_device_endpoint_resources(struct xhci_hcd *xhci,
+	struct xhci_virt_device *virt_dev, bool drop_control_ep)
+{
+	int i;
+	unsigned int num_dropped_eps = 0;
+	unsigned int drop_flags = 0;
+
+	for (i = (drop_control_ep ? 0 : 1); i < 31; i++) {
+		if (virt_dev->eps[i].ring) {
+			drop_flags |= 1 << i;
+			num_dropped_eps++;
+		}
+	}
+	xhci->num_active_eps -= num_dropped_eps;
+	if (num_dropped_eps)
+		xhci_dbg(xhci, "Dropped %u ep ctxs, flags = 0x%x, "
+				"%u now active.\n",
+				num_dropped_eps, drop_flags,
+				xhci->num_active_eps);
+}
+
+/*
+ * This submits a Reset Device Command, which will set the device state to 0,
+ * set the device address to 0, and disable all the endpoints except the default
+ * control endpoint.  The USB core should come back and call
+ * xhci_address_device(), and then re-set up the configuration.  If this is
+ * called because of a usb_reset_and_verify_device(), then the old alternate
+ * settings will be re-installed through the normal bandwidth allocation
+ * functions.
+ *
+ * Wait for the Reset Device command to finish.  Remove all structures
+ * associated with the endpoints that were disabled.  Clear the input device
+ * structure?  Cache the rings?  Reset the control endpoint 0 max packet size?
+ *
+ * If the virt_dev to be reset does not exist or does not match the udev,
+ * it means the device is lost, possibly due to the xHC restore error and
+ * re-initialization during S3/S4. In this case, call xhci_alloc_dev() to
+ * re-allocate the device.
+ */
+int xhci_discover_or_reset_device(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	int ret, i;
+	unsigned long flags;
+	struct xhci_hcd *xhci;
+	unsigned int slot_id;
+	struct xhci_virt_device *virt_dev;
+	struct xhci_command *reset_device_cmd;
+	int timeleft;
+	int last_freed_endpoint;
+	struct xhci_slot_ctx *slot_ctx;
+	int old_active_eps = 0;
+
+	ret = xhci_check_args(hcd, udev, NULL, 0, false, __func__);
+	if (ret <= 0)
+		return ret;
+	xhci = hcd_to_xhci(hcd);
+	slot_id = udev->slot_id;
+	virt_dev = xhci->devs[slot_id];
+	if (!virt_dev) {
+		xhci_dbg(xhci, "The device to be reset with slot ID %u does "
+				"not exist. Re-allocate the device\n", slot_id);
+		ret = xhci_alloc_dev(hcd, udev);
+		if (ret == 1)
+			return 0;
+		else
+			return -EINVAL;
+	}
+
+	if (virt_dev->udev != udev) {
+		/* If the virt_dev and the udev does not match, this virt_dev
+		 * may belong to another udev.
+		 * Re-allocate the device.
+		 */
+		xhci_dbg(xhci, "The device to be reset with slot ID %u does "
+				"not match the udev. Re-allocate the device\n",
+				slot_id);
+		ret = xhci_alloc_dev(hcd, udev);
+		if (ret == 1)
+			return 0;
+		else
+			return -EINVAL;
+	}
+
+	/* If device is not setup, there is no point in resetting it */
+	slot_ctx = xhci_get_slot_ctx(xhci, virt_dev->out_ctx);
+	if (GET_SLOT_STATE(le32_to_cpu(slot_ctx->dev_state)) ==
+						SLOT_STATE_DISABLED)
+		return 0;
+
+	xhci_dbg(xhci, "Resetting device with slot ID %u\n", slot_id);
+	/* Allocate the command structure that holds the struct completion.
+	 * Assume we're in process context, since the normal device reset
+	 * process has to wait for the device anyway.  Storage devices are
+	 * reset as part of error handling, so use GFP_NOIO instead of
+	 * GFP_KERNEL.
+	 */
+	reset_device_cmd = xhci_alloc_command(xhci, false, true, GFP_NOIO);
+	if (!reset_device_cmd) {
+		xhci_dbg(xhci, "Couldn't allocate command structure.\n");
+		return -ENOMEM;
+	}
+
+	/* Attempt to submit the Reset Device command to the command ring */
+	spin_lock_irqsave(&xhci->lock, flags);
+	reset_device_cmd->command_trb = xhci->cmd_ring->enqueue;
+
+	/* Enqueue pointer can be left pointing to the link TRB,
+	 * we must handle that
+	 */
+	if (TRB_TYPE_LINK_LE32(reset_device_cmd->command_trb->link.control))
+		reset_device_cmd->command_trb =
+			xhci->cmd_ring->enq_seg->next->trbs;
+
+	list_add_tail(&reset_device_cmd->cmd_list, &virt_dev->cmd_list);
+	ret = xhci_queue_reset_device(xhci, slot_id);
+	if (ret) {
+		xhci_dbg(xhci, "FIXME: allocate a command ring segment\n");
+		list_del(&reset_device_cmd->cmd_list);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		goto command_cleanup;
+	}
+	xhci_ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* Wait for the Reset Device command to finish */
+	timeleft = wait_for_completion_interruptible_timeout(
+			reset_device_cmd->completion,
+			USB_CTRL_SET_TIMEOUT);
+	if (timeleft <= 0) {
+		xhci_warn(xhci, "%s while waiting for reset device command\n",
+				timeleft == 0 ? "Timeout" : "Signal");
+		spin_lock_irqsave(&xhci->lock, flags);
+		/* The timeout might have raced with the event ring handler, so
+		 * only delete from the list if the item isn't poisoned.
+		 */
+		if (reset_device_cmd->cmd_list.next != LIST_POISON1)
+			list_del(&reset_device_cmd->cmd_list);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		ret = -ETIME;
+		goto command_cleanup;
+	}
+
+	/* The Reset Device command can't fail, according to the 0.95/0.96 spec,
+	 * unless we tried to reset a slot ID that wasn't enabled,
+	 * or the device wasn't in the addressed or configured state.
+	 */
+	ret = reset_device_cmd->status;
+	switch (ret) {
+	case COMP_EBADSLT: /* 0.95 completion code for bad slot ID */
+	case COMP_CTX_STATE: /* 0.96 completion code for same thing */
+		xhci_info(xhci, "Can't reset device (slot ID %u) in %s state\n",
+				slot_id,
+				xhci_get_slot_state(xhci, virt_dev->out_ctx));
+		xhci_info(xhci, "Not freeing device rings.\n");
+		/* Don't treat this as an error.  May change my mind later. */
+		ret = 0;
+		goto command_cleanup;
+	case COMP_SUCCESS:
+		xhci_dbg(xhci, "Successful reset device command.\n");
+		break;
+	default:
+		if (xhci_is_vendor_info_code(xhci, ret))
+			break;
+		xhci_warn(xhci, "Unknown completion code %u for "
+				"reset device command.\n", ret);
+		ret = -EINVAL;
+		goto command_cleanup;
+	}
+
+	/* Free up host controller endpoint resources */
+	if ((xhci->quirks & XHCI_EP_LIMIT_QUIRK)) {
+		spin_lock_irqsave(&xhci->lock, flags);
+		/* Don't delete the default control endpoint resources */
+		xhci_free_device_endpoint_resources(xhci, virt_dev, false);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+	}
+
+	/* Everything but endpoint 0 is disabled, so free or cache the rings. */
+	last_freed_endpoint = 1;
+	for (i = 1; i < 31; ++i) {
+		struct xhci_virt_ep *ep = &virt_dev->eps[i];
+
+		if (ep->ep_state & EP_HAS_STREAMS) {
+			xhci_free_stream_info(xhci, ep->stream_info);
+			ep->stream_info = NULL;
+			ep->ep_state &= ~EP_HAS_STREAMS;
+		}
+
+		if (ep->ring) {
+			xhci_free_or_cache_endpoint_ring(xhci, virt_dev, i);
+			last_freed_endpoint = i;
+		}
+		if (!list_empty(&virt_dev->eps[i].bw_endpoint_list))
+			xhci_drop_ep_from_interval_table(xhci,
+					&virt_dev->eps[i].bw_info,
+					virt_dev->bw_table,
+					udev,
+					&virt_dev->eps[i],
+					virt_dev->tt_info);
+		xhci_clear_endpoint_bw_info(&virt_dev->eps[i].bw_info);
+	}
+	/* If necessary, update the number of active TTs on this root port */
+	xhci_update_tt_active_eps(xhci, virt_dev, old_active_eps);
+
+	xhci_dbg(xhci, "Output context after successful reset device cmd:\n");
+	xhci_dbg_ctx(xhci, virt_dev->out_ctx, last_freed_endpoint);
+	ret = 0;
+
+command_cleanup:
+	xhci_free_command(xhci, reset_device_cmd);
+	return ret;
+}
+
+/*
+ * At this point, the struct usb_device is about to go away, the device has
+ * disconnected, and all traffic has been stopped and the endpoints have been
+ * disabled.  Free any HC data structures associated with that device.
+ */
+void xhci_free_dev(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	struct xhci_virt_device *virt_dev;
+	unsigned long flags;
+	u32 state;
+	int i, ret;
+
+	ret = xhci_check_args(hcd, udev, NULL, 0, true, __func__);
+	/* If the host is halted due to driver unload, we still need to free the
+	 * device.
+	 */
+	if (ret <= 0 && ret != -ENODEV)
+		return;
+
+	virt_dev = xhci->devs[udev->slot_id];
+
+	/* Stop any wayward timer functions (which may grab the lock) */
+	for (i = 0; i < 31; ++i) {
+		virt_dev->eps[i].ep_state &= ~EP_HALT_PENDING;
+		del_timer_sync(&virt_dev->eps[i].stop_cmd_timer);
+	}
+
+	if (udev->usb2_hw_lpm_enabled) {
+		xhci_set_usb2_hardware_lpm(hcd, udev, 0);
+		udev->usb2_hw_lpm_enabled = 0;
+	}
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	/* Don't disable the slot if the host controller is dead. */
+	state = xhci_readl(xhci, &xhci->op_regs->status);
+	if (state == 0xffffffff || (xhci->xhc_state & XHCI_STATE_DYING) ||
+			(xhci->xhc_state & XHCI_STATE_HALTED)) {
+		xhci_free_virt_device(xhci, udev->slot_id);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return;
+	}
+
+	if (xhci_queue_slot_control(xhci, TRB_DISABLE_SLOT, udev->slot_id)) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		xhci_dbg(xhci, "FIXME: allocate a command ring segment\n");
+		return;
+	}
+	xhci_ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	/*
+	 * Event command completion handler will free any data structures
+	 * associated with the slot.  XXX Can free sleep?
+	 */
+}
+
+/*
+ * Checks if we have enough host controller resources for the default control
+ * endpoint.
+ *
+ * Must be called with xhci->lock held.
+ */
+static int xhci_reserve_host_control_ep_resources(struct xhci_hcd *xhci)
+{
+	if (xhci->num_active_eps + 1 > xhci->limit_active_eps) {
+		xhci_dbg(xhci, "Not enough ep ctxs: "
+				"%u active, need to add 1, limit is %u.\n",
+				xhci->num_active_eps, xhci->limit_active_eps);
+		return -ENOMEM;
+	}
+	xhci->num_active_eps += 1;
+	xhci_dbg(xhci, "Adding 1 ep ctx, %u now active.\n",
+			xhci->num_active_eps);
+	return 0;
+}
+
+
+/*
+ * Returns 0 if the xHC ran out of device slots, the Enable Slot command
+ * timed out, or allocating memory failed.  Returns 1 on success.
+ */
+int xhci_alloc_dev(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	unsigned long flags;
+	int timeleft;
+	int ret;
+	union xhci_trb *cmd_trb;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	cmd_trb = xhci->cmd_ring->dequeue;
+	ret = xhci_queue_slot_control(xhci, TRB_ENABLE_SLOT, 0);
+	if (ret) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		xhci_dbg(xhci, "FIXME: allocate a command ring segment\n");
+		return 0;
+	}
+	xhci_ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* XXX: how much time for xHC slot assignment? */
+	timeleft = wait_for_completion_interruptible_timeout(&xhci->addr_dev,
+			XHCI_CMD_DEFAULT_TIMEOUT);
+	if (timeleft <= 0) {
+		xhci_warn(xhci, "%s while waiting for a slot\n",
+				timeleft == 0 ? "Timeout" : "Signal");
+		/* cancel the enable slot request */
+		return xhci_cancel_cmd(xhci, NULL, cmd_trb);
+	}
+
+	if (!xhci->slot_id) {
+		xhci_err(xhci, "Error while assigning device slot ID\n");
+		return 0;
+	}
+
+	if ((xhci->quirks & XHCI_EP_LIMIT_QUIRK)) {
+		spin_lock_irqsave(&xhci->lock, flags);
+		ret = xhci_reserve_host_control_ep_resources(xhci);
+		if (ret) {
+			spin_unlock_irqrestore(&xhci->lock, flags);
+			xhci_warn(xhci, "Not enough host resources, "
+					"active endpoint contexts = %u\n",
+					xhci->num_active_eps);
+			goto disable_slot;
+		}
+		spin_unlock_irqrestore(&xhci->lock, flags);
+	}
+	/* Use GFP_NOIO, since this function can be called from
+	 * xhci_discover_or_reset_device(), which may be called as part of
+	 * mass storage driver error handling.
+	 */
+	if (!xhci_alloc_virt_device(xhci, xhci->slot_id, udev, GFP_NOIO)) {
+		xhci_warn(xhci, "Could not allocate xHCI USB device data structures\n");
+		goto disable_slot;
+	}
+	udev->slot_id = xhci->slot_id;
+	/* Is this a LS or FS device under a HS hub? */
+	/* Hub or peripherial? */
+	return 1;
+
+disable_slot:
+	/* Disable slot, if we can do it without mem alloc */
+	spin_lock_irqsave(&xhci->lock, flags);
+	if (!xhci_queue_slot_control(xhci, TRB_DISABLE_SLOT, udev->slot_id))
+		xhci_ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return 0;
+}
+
+/*
+ * Issue an Address Device command (which will issue a SetAddress request to
+ * the device).
+ * We should be protected by the usb_address0_mutex in khubd's hub_port_init, so
+ * we should only issue and wait on one address command at the same time.
+ *
+ * We add one to the device address issued by the hardware because the USB core
+ * uses address 1 for the root hubs (even though they're not really devices).
+ */
+int xhci_address_device(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	unsigned long flags;
+	int timeleft;
+	struct xhci_virt_device *virt_dev;
+	int ret = 0;
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	struct xhci_slot_ctx *slot_ctx;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	u64 temp_64;
+	union xhci_trb *cmd_trb;
+
+	if (!udev->slot_id) {
+		xhci_dbg(xhci, "Bad Slot ID %d\n", udev->slot_id);
+		return -EINVAL;
+	}
+
+	virt_dev = xhci->devs[udev->slot_id];
+
+	if (WARN_ON(!virt_dev)) {
+		/*
+		 * In plug/unplug torture test with an NEC controller,
+		 * a zero-dereference was observed once due to virt_dev = 0.
+		 * Print useful debug rather than crash if it is observed again!
+		 */
+		xhci_warn(xhci, "Virt dev invalid for slot_id 0x%x!\n",
+			udev->slot_id);
+		return -EINVAL;
+	}
+
+	slot_ctx = xhci_get_slot_ctx(xhci, virt_dev->in_ctx);
+	/*
+	 * If this is the first Set Address since device plug-in or
+	 * virt_device realloaction after a resume with an xHCI power loss,
+	 * then set up the slot context.
+	 */
+	if (!slot_ctx->dev_info)
+		xhci_setup_addressable_virt_dev(xhci, udev);
+	/* Otherwise, update the control endpoint ring enqueue pointer. */
+	else
+		xhci_copy_ep0_dequeue_into_input_ctx(xhci, udev);
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, virt_dev->in_ctx);
+	ctrl_ctx->add_flags = cpu_to_le32(SLOT_FLAG | EP0_FLAG);
+	ctrl_ctx->drop_flags = 0;
+
+	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);
+	cmd_trb = xhci->cmd_ring->dequeue;
+	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");
+		return ret;
+	}
+	xhci_ring_cmd_db(xhci);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	/* ctrl tx can take up to 5 sec; XXX: need more time for xHC? */
+	timeleft = wait_for_completion_interruptible_timeout(&xhci->addr_dev,
+			XHCI_CMD_DEFAULT_TIMEOUT);
+	/* FIXME: From section 4.3.4: "Software shall be responsible for timing
+	 * the SetAddress() "recovery interval" required by USB and aborting the
+	 * command on a timeout.
+	 */
+	if (timeleft <= 0) {
+		xhci_warn(xhci, "%s while waiting for address device command\n",
+				timeleft == 0 ? "Timeout" : "Signal");
+		/* cancel the address device command */
+		ret = xhci_cancel_cmd(xhci, NULL, cmd_trb);
+		if (ret < 0)
+			return ret;
+		return -ETIME;
+	}
+
+	switch (virt_dev->cmd_status) {
+	case COMP_CTX_STATE:
+	case COMP_EBADSLT:
+		xhci_err(xhci, "Setup ERROR: address device command for slot %d.\n",
+				udev->slot_id);
+		ret = -EINVAL;
+		break;
+	case COMP_TX_ERR:
+		dev_warn(&udev->dev, "Device not responding to set address.\n");
+		ret = -EPROTO;
+		break;
+	case COMP_DEV_ERR:
+		dev_warn(&udev->dev, "ERROR: Incompatible device for address "
+				"device command.\n");
+		ret = -ENODEV;
+		break;
+	case COMP_SUCCESS:
+		xhci_dbg(xhci, "Successful Address Device command\n");
+		break;
+	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_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[udev->slot_id],
+		 (unsigned long long)
+		 le64_to_cpu(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);
+	xhci_dbg(xhci, "Slot ID %d Input Context:\n", udev->slot_id);
+	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, 2);
+	/*
+	 * USB core uses address 1 for the roothubs, so we add one to the
+	 * address given back to us by the HC.
+	 */
+	slot_ctx = xhci_get_slot_ctx(xhci, virt_dev->out_ctx);
+	/* Use kernel assigned address for devices; store xHC assigned
+	 * address locally. */
+	virt_dev->address = (le32_to_cpu(slot_ctx->dev_state) & DEV_ADDR_MASK)
+		+ 1;
+	/* Zero the input context control for later use */
+	ctrl_ctx->add_flags = 0;
+	ctrl_ctx->drop_flags = 0;
+
+	xhci_dbg(xhci, "Internal device address = %d\n", virt_dev->address);
+
+	return 0;
+}
+
+/*
+ * Transfer the port index into real index in the HW port status
+ * registers. Caculate offset between the port's PORTSC register
+ * and port status base. Divide the number of per port register
+ * to get the real index. The raw port number bases 1.
+ */
+int xhci_find_raw_port_number(struct usb_hcd *hcd, int port1)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	__le32 __iomem *base_addr = &xhci->op_regs->port_status_base;
+	__le32 __iomem *addr;
+	int raw_port;
+
+	if (hcd->speed != HCD_USB3)
+		addr = xhci->usb2_ports[port1 - 1];
+	else
+		addr = xhci->usb3_ports[port1 - 1];
+
+	raw_port = (addr - base_addr)/NUM_PORT_REGS + 1;
+	return raw_port;
+}
+
+#ifdef CONFIG_PM_RUNTIME
+
+/* BESL to HIRD Encoding array for USB2 LPM */
+static int xhci_besl_encoding[16] = {125, 150, 200, 300, 400, 500, 1000, 2000,
+	3000, 4000, 5000, 6000, 7000, 8000, 9000, 10000};
+
+/* Calculate HIRD/BESL for USB2 PORTPMSC*/
+static int xhci_calculate_hird_besl(struct xhci_hcd *xhci,
+					struct usb_device *udev)
+{
+	int u2del, besl, besl_host;
+	int besl_device = 0;
+	u32 field;
+
+	u2del = HCS_U2_LATENCY(xhci->hcs_params3);
+	field = le32_to_cpu(udev->bos->ext_cap->bmAttributes);
+
+	if (field & USB_BESL_SUPPORT) {
+		for (besl_host = 0; besl_host < 16; besl_host++) {
+			if (xhci_besl_encoding[besl_host] >= u2del)
+				break;
+		}
+		/* Use baseline BESL value as default */
+		if (field & USB_BESL_BASELINE_VALID)
+			besl_device = USB_GET_BESL_BASELINE(field);
+		else if (field & USB_BESL_DEEP_VALID)
+			besl_device = USB_GET_BESL_DEEP(field);
+	} else {
+		if (u2del <= 50)
+			besl_host = 0;
+		else
+			besl_host = (u2del - 51) / 75 + 1;
+	}
+
+	besl = besl_host + besl_device;
+	if (besl > 15)
+		besl = 15;
+
+	return besl;
+}
+
+static int xhci_usb2_software_lpm_test(struct usb_hcd *hcd,
+					struct usb_device *udev)
+{
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+	struct dev_info	*dev_info;
+	__le32 __iomem	**port_array;
+	__le32 __iomem	*addr, *pm_addr;
+	u32		temp, dev_id;
+	unsigned int	port_num;
+	unsigned long	flags;
+	int		hird;
+	int		ret;
+
+	if (hcd->speed == HCD_USB3 || !xhci->sw_lpm_support ||
+			!udev->lpm_capable)
+		return -EINVAL;
+
+	/* we only support lpm for non-hub device connected to root hub yet */
+	if (!udev->parent || udev->parent->parent ||
+			udev->descriptor.bDeviceClass == USB_CLASS_HUB)
+		return -EINVAL;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+
+	/* Look for devices in lpm_failed_devs list */
+	dev_id = le16_to_cpu(udev->descriptor.idVendor) << 16 |
+			le16_to_cpu(udev->descriptor.idProduct);
+	list_for_each_entry(dev_info, &xhci->lpm_failed_devs, list) {
+		if (dev_info->dev_id == dev_id) {
+			ret = -EINVAL;
+			goto finish;
+		}
+	}
+
+	port_array = xhci->usb2_ports;
+	port_num = udev->portnum - 1;
+
+	if (port_num > HCS_MAX_PORTS(xhci->hcs_params1)) {
+		xhci_dbg(xhci, "invalid port number %d\n", udev->portnum);
+		ret = -EINVAL;
+		goto finish;
+	}
+
+	/*
+	 * Test USB 2.0 software LPM.
+	 * FIXME: some xHCI 1.0 hosts may implement a new register to set up
+	 * hardware-controlled USB 2.0 LPM. See section 5.4.11 and 4.23.5.1.1.1
+	 * in the June 2011 errata release.
+	 */
+	xhci_dbg(xhci, "test port %d software LPM\n", port_num);
+	/*
+	 * Set L1 Device Slot and HIRD/BESL.
+	 * Check device's USB 2.0 extension descriptor to determine whether
+	 * HIRD or BESL shoule be used. See USB2.0 LPM errata.
+	 */
+	pm_addr = port_array[port_num] + 1;
+	hird = xhci_calculate_hird_besl(xhci, udev);
+	temp = PORT_L1DS(udev->slot_id) | PORT_HIRD(hird);
+	xhci_writel(xhci, temp, pm_addr);
+
+	/* Set port link state to U2(L1) */
+	addr = port_array[port_num];
+	xhci_set_link_state(xhci, port_array, port_num, XDEV_U2);
+
+	/* wait for ACK */
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	msleep(10);
+	spin_lock_irqsave(&xhci->lock, flags);
+
+	/* Check L1 Status */
+	ret = xhci_handshake(xhci, pm_addr,
+			PORT_L1S_MASK, PORT_L1S_SUCCESS, 125);
+	if (ret != -ETIMEDOUT) {
+		/* enter L1 successfully */
+		temp = xhci_readl(xhci, addr);
+		xhci_dbg(xhci, "port %d entered L1 state, port status 0x%x\n",
+				port_num, temp);
+		ret = 0;
+	} else {
+		temp = xhci_readl(xhci, pm_addr);
+		xhci_dbg(xhci, "port %d software lpm failed, L1 status %d\n",
+				port_num, temp & PORT_L1S_MASK);
+		ret = -EINVAL;
+	}
+
+	/* Resume the port */
+	xhci_set_link_state(xhci, port_array, port_num, XDEV_U0);
+
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	msleep(10);
+	spin_lock_irqsave(&xhci->lock, flags);
+
+	/* Clear PLC */
+	xhci_test_and_clear_bit(xhci, port_array, port_num, PORT_PLC);
+
+	/* Check PORTSC to make sure the device is in the right state */
+	if (!ret) {
+		temp = xhci_readl(xhci, addr);
+		xhci_dbg(xhci, "resumed port %d status 0x%x\n",	port_num, temp);
+		if (!(temp & PORT_CONNECT) || !(temp & PORT_PE) ||
+				(temp & PORT_PLS_MASK) != XDEV_U0) {
+			xhci_dbg(xhci, "port L1 resume fail\n");
+			ret = -EINVAL;
+		}
+	}
+
+	if (ret) {
+		/* Insert dev to lpm_failed_devs list */
+		xhci_warn(xhci, "device LPM test failed, may disconnect and "
+				"re-enumerate\n");
+		dev_info = kzalloc(sizeof(struct dev_info), GFP_ATOMIC);
+		if (!dev_info) {
+			ret = -ENOMEM;
+			goto finish;
+		}
+		dev_info->dev_id = dev_id;
+		INIT_LIST_HEAD(&dev_info->list);
+		list_add(&dev_info->list, &xhci->lpm_failed_devs);
+	} else {
+		xhci_ring_device(xhci, udev->slot_id);
+	}
+
+finish:
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return ret;
+}
+
+int xhci_set_usb2_hardware_lpm(struct usb_hcd *hcd,
+			struct usb_device *udev, int enable)
+{
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+	__le32 __iomem	**port_array;
+	__le32 __iomem	*pm_addr;
+	u32		temp;
+	unsigned int	port_num;
+	unsigned long	flags;
+	int		hird;
+
+	if (hcd->speed == HCD_USB3 || !xhci->hw_lpm_support ||
+			!udev->lpm_capable)
+		return -EPERM;
+
+	if (!udev->parent || udev->parent->parent ||
+			udev->descriptor.bDeviceClass == USB_CLASS_HUB)
+		return -EPERM;
+
+	if (udev->usb2_hw_lpm_capable != 1)
+		return -EPERM;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+
+	port_array = xhci->usb2_ports;
+	port_num = udev->portnum - 1;
+	pm_addr = port_array[port_num] + 1;
+	temp = xhci_readl(xhci, pm_addr);
+
+	xhci_dbg(xhci, "%s port %d USB2 hardware LPM\n",
+			enable ? "enable" : "disable", port_num);
+
+	hird = xhci_calculate_hird_besl(xhci, udev);
+
+	if (enable) {
+		temp &= ~PORT_HIRD_MASK;
+		temp |= PORT_HIRD(hird) | PORT_RWE;
+		xhci_writel(xhci, temp, pm_addr);
+		temp = xhci_readl(xhci, pm_addr);
+		temp |= PORT_HLE;
+		xhci_writel(xhci, temp, pm_addr);
+	} else {
+		temp &= ~(PORT_HLE | PORT_RWE | PORT_HIRD_MASK);
+		xhci_writel(xhci, temp, pm_addr);
+	}
+
+	spin_unlock_irqrestore(&xhci->lock, flags);
+	return 0;
+}
+
+int xhci_update_device(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
+	int		ret;
+
+	ret = xhci_usb2_software_lpm_test(hcd, udev);
+	if (!ret) {
+		xhci_dbg(xhci, "software LPM test succeed\n");
+		if (xhci->hw_lpm_support == 1) {
+			udev->usb2_hw_lpm_capable = 1;
+			ret = xhci_set_usb2_hardware_lpm(hcd, udev, 1);
+			if (!ret)
+				udev->usb2_hw_lpm_enabled = 1;
+		}
+	}
+
+	return 0;
+}
+
+#else
+
+int xhci_set_usb2_hardware_lpm(struct usb_hcd *hcd,
+				struct usb_device *udev, int enable)
+{
+	return 0;
+}
+
+int xhci_update_device(struct usb_hcd *hcd, struct usb_device *udev)
+{
+	return 0;
+}
+
+#endif /* CONFIG_PM_RUNTIME */
+
+/*---------------------- USB 3.0 Link PM functions ------------------------*/
+
+#ifdef CONFIG_PM
+/* Service interval in nanoseconds = 2^(bInterval - 1) * 125us * 1000ns / 1us */
+static unsigned long long xhci_service_interval_to_ns(
+		struct usb_endpoint_descriptor *desc)
+{
+	return (1ULL << (desc->bInterval - 1)) * 125 * 1000;
+}
+
+static u16 xhci_get_timeout_no_hub_lpm(struct usb_device *udev,
+		enum usb3_link_state state)
+{
+	unsigned long long sel;
+	unsigned long long pel;
+	unsigned int max_sel_pel;
+	char *state_name;
+
+	switch (state) {
+	case USB3_LPM_U1:
+		/* Convert SEL and PEL stored in nanoseconds to microseconds */
+		sel = DIV_ROUND_UP(udev->u1_params.sel, 1000);
+		pel = DIV_ROUND_UP(udev->u1_params.pel, 1000);
+		max_sel_pel = USB3_LPM_MAX_U1_SEL_PEL;
+		state_name = "U1";
+		break;
+	case USB3_LPM_U2:
+		sel = DIV_ROUND_UP(udev->u2_params.sel, 1000);
+		pel = DIV_ROUND_UP(udev->u2_params.pel, 1000);
+		max_sel_pel = USB3_LPM_MAX_U2_SEL_PEL;
+		state_name = "U2";
+		break;
+	default:
+		dev_warn(&udev->dev, "%s: Can't get timeout for non-U1 or U2 state.\n",
+				__func__);
+		return USB3_LPM_DISABLED;
+	}
+
+	if (sel <= max_sel_pel && pel <= max_sel_pel)
+		return USB3_LPM_DEVICE_INITIATED;
+
+	if (sel > max_sel_pel)
+		dev_dbg(&udev->dev, "Device-initiated %s disabled "
+				"due to long SEL %llu ms\n",
+				state_name, sel);
+	else
+		dev_dbg(&udev->dev, "Device-initiated %s disabled "
+				"due to long PEL %llu\n ms",
+				state_name, pel);
+	return USB3_LPM_DISABLED;
+}
+
+/* Returns the hub-encoded U1 timeout value.
+ * The U1 timeout should be the maximum of the following values:
+ *  - For control endpoints, U1 system exit latency (SEL) * 3
+ *  - For bulk endpoints, U1 SEL * 5
+ *  - For interrupt endpoints:
+ *    - Notification EPs, U1 SEL * 3
+ *    - Periodic EPs, max(105% of bInterval, U1 SEL * 2)
+ *  - For isochronous endpoints, max(105% of bInterval, U1 SEL * 2)
+ */
+static u16 xhci_calculate_intel_u1_timeout(struct usb_device *udev,
+		struct usb_endpoint_descriptor *desc)
+{
+	unsigned long long timeout_ns;
+	int ep_type;
+	int intr_type;
+
+	ep_type = usb_endpoint_type(desc);
+	switch (ep_type) {
+	case USB_ENDPOINT_XFER_CONTROL:
+		timeout_ns = udev->u1_params.sel * 3;
+		break;
+	case USB_ENDPOINT_XFER_BULK:
+		timeout_ns = udev->u1_params.sel * 5;
+		break;
+	case USB_ENDPOINT_XFER_INT:
+		intr_type = usb_endpoint_interrupt_type(desc);
+		if (intr_type == USB_ENDPOINT_INTR_NOTIFICATION) {
+			timeout_ns = udev->u1_params.sel * 3;
+			break;
+		}
+		/* Otherwise the calculation is the same as isoc eps */
+	case USB_ENDPOINT_XFER_ISOC:
+		timeout_ns = xhci_service_interval_to_ns(desc);
+		timeout_ns = DIV_ROUND_UP_ULL(timeout_ns * 105, 100);
+		if (timeout_ns < udev->u1_params.sel * 2)
+			timeout_ns = udev->u1_params.sel * 2;
+		break;
+	default:
+		return 0;
+	}
+
+	/* The U1 timeout is encoded in 1us intervals. */
+	timeout_ns = DIV_ROUND_UP_ULL(timeout_ns, 1000);
+	/* Don't return a timeout of zero, because that's USB3_LPM_DISABLED. */
+	if (timeout_ns == USB3_LPM_DISABLED)
+		timeout_ns++;
+
+	/* If the necessary timeout value is bigger than what we can set in the
+	 * USB 3.0 hub, we have to disable hub-initiated U1.
+	 */
+	if (timeout_ns <= USB3_LPM_U1_MAX_TIMEOUT)
+		return timeout_ns;
+	dev_dbg(&udev->dev, "Hub-initiated U1 disabled "
+			"due to long timeout %llu ms\n", timeout_ns);
+	return xhci_get_timeout_no_hub_lpm(udev, USB3_LPM_U1);
+}
+
+/* Returns the hub-encoded U2 timeout value.
+ * The U2 timeout should be the maximum of:
+ *  - 10 ms (to avoid the bandwidth impact on the scheduler)
+ *  - largest bInterval of any active periodic endpoint (to avoid going
+ *    into lower power link states between intervals).
+ *  - the U2 Exit Latency of the device
+ */
+static u16 xhci_calculate_intel_u2_timeout(struct usb_device *udev,
+		struct usb_endpoint_descriptor *desc)
+{
+	unsigned long long timeout_ns;
+	unsigned long long u2_del_ns;
+
+	timeout_ns = 10 * 1000 * 1000;
+
+	if ((usb_endpoint_xfer_int(desc) || usb_endpoint_xfer_isoc(desc)) &&
+			(xhci_service_interval_to_ns(desc) > timeout_ns))
+		timeout_ns = xhci_service_interval_to_ns(desc);
+
+	u2_del_ns = le16_to_cpu(udev->bos->ss_cap->bU2DevExitLat) * 1000ULL;
+	if (u2_del_ns > timeout_ns)
+		timeout_ns = u2_del_ns;
+
+	/* The U2 timeout is encoded in 256us intervals */
+	timeout_ns = DIV_ROUND_UP_ULL(timeout_ns, 256 * 1000);
+	/* If the necessary timeout value is bigger than what we can set in the
+	 * USB 3.0 hub, we have to disable hub-initiated U2.
+	 */
+	if (timeout_ns <= USB3_LPM_U2_MAX_TIMEOUT)
+		return timeout_ns;
+	dev_dbg(&udev->dev, "Hub-initiated U2 disabled "
+			"due to long timeout %llu ms\n", timeout_ns);
+	return xhci_get_timeout_no_hub_lpm(udev, USB3_LPM_U2);
+}
+
+static u16 xhci_call_host_update_timeout_for_endpoint(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		struct usb_endpoint_descriptor *desc,
+		enum usb3_link_state state,
+		u16 *timeout)
+{
+	if (state == USB3_LPM_U1) {
+		if (xhci->quirks & XHCI_INTEL_HOST)
+			return xhci_calculate_intel_u1_timeout(udev, desc);
+	} else {
+		if (xhci->quirks & XHCI_INTEL_HOST)
+			return xhci_calculate_intel_u2_timeout(udev, desc);
+	}
+
+	return USB3_LPM_DISABLED;
+}
+
+static int xhci_update_timeout_for_endpoint(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		struct usb_endpoint_descriptor *desc,
+		enum usb3_link_state state,
+		u16 *timeout)
+{
+	u16 alt_timeout;
+
+	alt_timeout = xhci_call_host_update_timeout_for_endpoint(xhci, udev,
+		desc, state, timeout);
+
+	/* If we found we can't enable hub-initiated LPM, or
+	 * the U1 or U2 exit latency was too high to allow
+	 * device-initiated LPM as well, just stop searching.
+	 */
+	if (alt_timeout == USB3_LPM_DISABLED ||
+			alt_timeout == USB3_LPM_DEVICE_INITIATED) {
+		*timeout = alt_timeout;
+		return -E2BIG;
+	}
+	if (alt_timeout > *timeout)
+		*timeout = alt_timeout;
+	return 0;
+}
+
+static int xhci_update_timeout_for_interface(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		struct usb_host_interface *alt,
+		enum usb3_link_state state,
+		u16 *timeout)
+{
+	int j;
+
+	for (j = 0; j < alt->desc.bNumEndpoints; j++) {
+		if (xhci_update_timeout_for_endpoint(xhci, udev,
+					&alt->endpoint[j].desc, state, timeout))
+			return -E2BIG;
+		continue;
+	}
+	return 0;
+}
+
+static int xhci_check_intel_tier_policy(struct usb_device *udev,
+		enum usb3_link_state state)
+{
+	struct usb_device *parent;
+	unsigned int num_hubs;
+
+	if (state == USB3_LPM_U2)
+		return 0;
+
+	/* Don't enable U1 if the device is on a 2nd tier hub or lower. */
+	for (parent = udev->parent, num_hubs = 0; parent->parent;
+			parent = parent->parent)
+		num_hubs++;
+
+	if (num_hubs < 2)
+		return 0;
+
+	dev_dbg(&udev->dev, "Disabling U1 link state for device"
+			" below second-tier hub.\n");
+	dev_dbg(&udev->dev, "Plug device into first-tier hub "
+			"to decrease power consumption.\n");
+	return -E2BIG;
+}
+
+static int xhci_check_tier_policy(struct xhci_hcd *xhci,
+		struct usb_device *udev,
+		enum usb3_link_state state)
+{
+	if (xhci->quirks & XHCI_INTEL_HOST)
+		return xhci_check_intel_tier_policy(udev, state);
+	return -EINVAL;
+}
+
+/* Returns the U1 or U2 timeout that should be enabled.
+ * If the tier check or timeout setting functions return with a non-zero exit
+ * code, that means the timeout value has been finalized and we shouldn't look
+ * at any more endpoints.
+ */
+static u16 xhci_calculate_lpm_timeout(struct usb_hcd *hcd,
+			struct usb_device *udev, enum usb3_link_state state)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	struct usb_host_config *config;
+	char *state_name;
+	int i;
+	u16 timeout = USB3_LPM_DISABLED;
+
+	if (state == USB3_LPM_U1)
+		state_name = "U1";
+	else if (state == USB3_LPM_U2)
+		state_name = "U2";
+	else {
+		dev_warn(&udev->dev, "Can't enable unknown link state %i\n",
+				state);
+		return timeout;
+	}
+
+	if (xhci_check_tier_policy(xhci, udev, state) < 0)
+		return timeout;
+
+	/* Gather some information about the currently installed configuration
+	 * and alternate interface settings.
+	 */
+	if (xhci_update_timeout_for_endpoint(xhci, udev, &udev->ep0.desc,
+			state, &timeout))
+		return timeout;
+
+	config = udev->actconfig;
+	if (!config)
+		return timeout;
+
+	for (i = 0; i < USB_MAXINTERFACES; i++) {
+		struct usb_driver *driver;
+		struct usb_interface *intf = config->interface[i];
+
+		if (!intf)
+			continue;
+
+		/* Check if any currently bound drivers want hub-initiated LPM
+		 * disabled.
+		 */
+		if (intf->dev.driver) {
+			driver = to_usb_driver(intf->dev.driver);
+			if (driver && driver->disable_hub_initiated_lpm) {
+				dev_dbg(&udev->dev, "Hub-initiated %s disabled "
+						"at request of driver %s\n",
+						state_name, driver->name);
+				return xhci_get_timeout_no_hub_lpm(udev, state);
+			}
+		}
+
+		/* Not sure how this could happen... */
+		if (!intf->cur_altsetting)
+			continue;
+
+		if (xhci_update_timeout_for_interface(xhci, udev,
+					intf->cur_altsetting,
+					state, &timeout))
+			return timeout;
+	}
+	return timeout;
+}
+
+/*
+ * Issue an Evaluate Context command to change the Maximum Exit Latency in the
+ * slot context.  If that succeeds, store the new MEL in the xhci_virt_device.
+ */
+static int xhci_change_max_exit_latency(struct xhci_hcd *xhci,
+			struct usb_device *udev, u16 max_exit_latency)
+{
+	struct xhci_virt_device *virt_dev;
+	struct xhci_command *command;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	struct xhci_slot_ctx *slot_ctx;
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	if (max_exit_latency == xhci->devs[udev->slot_id]->current_mel) {
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return 0;
+	}
+
+	/* Attempt to issue an Evaluate Context command to change the MEL. */
+	virt_dev = xhci->devs[udev->slot_id];
+	command = xhci->lpm_command;
+	xhci_slot_copy(xhci, command->in_ctx, virt_dev->out_ctx);
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, command->in_ctx);
+	ctrl_ctx->add_flags |= cpu_to_le32(SLOT_FLAG);
+	slot_ctx = xhci_get_slot_ctx(xhci, command->in_ctx);
+	slot_ctx->dev_info2 &= cpu_to_le32(~((u32) MAX_EXIT));
+	slot_ctx->dev_info2 |= cpu_to_le32(max_exit_latency);
+
+	xhci_dbg(xhci, "Set up evaluate context for LPM MEL change.\n");
+	xhci_dbg(xhci, "Slot %u Input Context:\n", udev->slot_id);
+	xhci_dbg_ctx(xhci, command->in_ctx, 0);
+
+	/* Issue and wait for the evaluate context command. */
+	ret = xhci_configure_endpoint(xhci, udev, command,
+			true, true);
+	xhci_dbg(xhci, "Slot %u Output Context:\n", udev->slot_id);
+	xhci_dbg_ctx(xhci, virt_dev->out_ctx, 0);
+
+	if (!ret) {
+		spin_lock_irqsave(&xhci->lock, flags);
+		virt_dev->current_mel = max_exit_latency;
+		spin_unlock_irqrestore(&xhci->lock, flags);
+	}
+	return ret;
+}
+
+static int calculate_max_exit_latency(struct usb_device *udev,
+		enum usb3_link_state state_changed,
+		u16 hub_encoded_timeout)
+{
+	unsigned long long u1_mel_us = 0;
+	unsigned long long u2_mel_us = 0;
+	unsigned long long mel_us = 0;
+	bool disabling_u1;
+	bool disabling_u2;
+	bool enabling_u1;
+	bool enabling_u2;
+
+	disabling_u1 = (state_changed == USB3_LPM_U1 &&
+			hub_encoded_timeout == USB3_LPM_DISABLED);
+	disabling_u2 = (state_changed == USB3_LPM_U2 &&
+			hub_encoded_timeout == USB3_LPM_DISABLED);
+
+	enabling_u1 = (state_changed == USB3_LPM_U1 &&
+			hub_encoded_timeout != USB3_LPM_DISABLED);
+	enabling_u2 = (state_changed == USB3_LPM_U2 &&
+			hub_encoded_timeout != USB3_LPM_DISABLED);
+
+	/* If U1 was already enabled and we're not disabling it,
+	 * or we're going to enable U1, account for the U1 max exit latency.
+	 */
+	if ((udev->u1_params.timeout != USB3_LPM_DISABLED && !disabling_u1) ||
+			enabling_u1)
+		u1_mel_us = DIV_ROUND_UP(udev->u1_params.mel, 1000);
+	if ((udev->u2_params.timeout != USB3_LPM_DISABLED && !disabling_u2) ||
+			enabling_u2)
+		u2_mel_us = DIV_ROUND_UP(udev->u2_params.mel, 1000);
+
+	if (u1_mel_us > u2_mel_us)
+		mel_us = u1_mel_us;
+	else
+		mel_us = u2_mel_us;
+	/* xHCI host controller max exit latency field is only 16 bits wide. */
+	if (mel_us > MAX_EXIT) {
+		dev_warn(&udev->dev, "Link PM max exit latency of %lluus "
+				"is too big.\n", mel_us);
+		return -E2BIG;
+	}
+	return mel_us;
+}
+
+/* Returns the USB3 hub-encoded value for the U1/U2 timeout. */
+int xhci_enable_usb3_lpm_timeout(struct usb_hcd *hcd,
+			struct usb_device *udev, enum usb3_link_state state)
+{
+	struct xhci_hcd	*xhci;
+	u16 hub_encoded_timeout;
+	int mel;
+	int ret;
+
+	xhci = hcd_to_xhci(hcd);
+	/* The LPM timeout values are pretty host-controller specific, so don't
+	 * enable hub-initiated timeouts unless the vendor has provided
+	 * information about their timeout algorithm.
+	 */
+	if (!xhci || !(xhci->quirks & XHCI_LPM_SUPPORT) ||
+			!xhci->devs[udev->slot_id])
+		return USB3_LPM_DISABLED;
+
+	hub_encoded_timeout = xhci_calculate_lpm_timeout(hcd, udev, state);
+	mel = calculate_max_exit_latency(udev, state, hub_encoded_timeout);
+	if (mel < 0) {
+		/* Max Exit Latency is too big, disable LPM. */
+		hub_encoded_timeout = USB3_LPM_DISABLED;
+		mel = 0;
+	}
+
+	ret = xhci_change_max_exit_latency(xhci, udev, mel);
+	if (ret)
+		return ret;
+	return hub_encoded_timeout;
+}
+
+int xhci_disable_usb3_lpm_timeout(struct usb_hcd *hcd,
+			struct usb_device *udev, enum usb3_link_state state)
+{
+	struct xhci_hcd	*xhci;
+	u16 mel;
+	int ret;
+
+	xhci = hcd_to_xhci(hcd);
+	if (!xhci || !(xhci->quirks & XHCI_LPM_SUPPORT) ||
+			!xhci->devs[udev->slot_id])
+		return 0;
+
+	mel = calculate_max_exit_latency(udev, state, USB3_LPM_DISABLED);
+	ret = xhci_change_max_exit_latency(xhci, udev, mel);
+	if (ret)
+		return ret;
+	return 0;
+}
+#else /* CONFIG_PM */
+
+int xhci_enable_usb3_lpm_timeout(struct usb_hcd *hcd,
+			struct usb_device *udev, enum usb3_link_state state)
+{
+	return USB3_LPM_DISABLED;
+}
+
+int xhci_disable_usb3_lpm_timeout(struct usb_hcd *hcd,
+			struct usb_device *udev, enum usb3_link_state state)
+{
+	return 0;
+}
+#endif	/* CONFIG_PM */
+
+/*-------------------------------------------------------------------------*/
+
+/* Once a hub descriptor is fetched for a device, we need to update the xHC's
+ * internal data structures for the device.
+ */
+int xhci_update_hub_device(struct usb_hcd *hcd, struct usb_device *hdev,
+			struct usb_tt *tt, gfp_t mem_flags)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	struct xhci_virt_device *vdev;
+	struct xhci_command *config_cmd;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	struct xhci_slot_ctx *slot_ctx;
+	unsigned long flags;
+	unsigned think_time;
+	int ret;
+
+	/* Ignore root hubs */
+	if (!hdev->parent)
+		return 0;
+
+	vdev = xhci->devs[hdev->slot_id];
+	if (!vdev) {
+		xhci_warn(xhci, "Cannot update hub desc for unknown device.\n");
+		return -EINVAL;
+	}
+	config_cmd = xhci_alloc_command(xhci, true, true, mem_flags);
+	if (!config_cmd) {
+		xhci_dbg(xhci, "Could not allocate xHCI command structure.\n");
+		return -ENOMEM;
+	}
+
+	spin_lock_irqsave(&xhci->lock, flags);
+	if (hdev->speed == USB_SPEED_HIGH &&
+			xhci_alloc_tt_info(xhci, vdev, hdev, tt, GFP_ATOMIC)) {
+		xhci_dbg(xhci, "Could not allocate xHCI TT structure.\n");
+		xhci_free_command(xhci, config_cmd);
+		spin_unlock_irqrestore(&xhci->lock, flags);
+		return -ENOMEM;
+	}
+
+	xhci_slot_copy(xhci, config_cmd->in_ctx, vdev->out_ctx);
+	ctrl_ctx = xhci_get_input_control_ctx(xhci, config_cmd->in_ctx);
+	ctrl_ctx->add_flags |= cpu_to_le32(SLOT_FLAG);
+	slot_ctx = xhci_get_slot_ctx(xhci, config_cmd->in_ctx);
+	slot_ctx->dev_info |= cpu_to_le32(DEV_HUB);
+	if (tt->multi)
+		slot_ctx->dev_info |= cpu_to_le32(DEV_MTT);
+	if (xhci->hci_version > 0x95) {
+		xhci_dbg(xhci, "xHCI version %x needs hub "
+				"TT think time and number of ports\n",
+				(unsigned int) xhci->hci_version);
+		slot_ctx->dev_info2 |= cpu_to_le32(XHCI_MAX_PORTS(hdev->maxchild));
+		/* Set TT think time - convert from ns to FS bit times.
+		 * 0 = 8 FS bit times, 1 = 16 FS bit times,
+		 * 2 = 24 FS bit times, 3 = 32 FS bit times.
+		 *
+		 * xHCI 1.0: this field shall be 0 if the device is not a
+		 * High-spped hub.
+		 */
+		think_time = tt->think_time;
+		if (think_time != 0)
+			think_time = (think_time / 666) - 1;
+		if (xhci->hci_version < 0x100 || hdev->speed == USB_SPEED_HIGH)
+			slot_ctx->tt_info |=
+				cpu_to_le32(TT_THINK_TIME(think_time));
+	} else {
+		xhci_dbg(xhci, "xHCI version %x doesn't need hub "
+				"TT think time or number of ports\n",
+				(unsigned int) xhci->hci_version);
+	}
+	slot_ctx->dev_state = 0;
+	spin_unlock_irqrestore(&xhci->lock, flags);
+
+	xhci_dbg(xhci, "Set up %s for hub device.\n",
+			(xhci->hci_version > 0x95) ?
+			"configure endpoint" : "evaluate context");
+	xhci_dbg(xhci, "Slot %u Input Context:\n", hdev->slot_id);
+	xhci_dbg_ctx(xhci, config_cmd->in_ctx, 0);
+
+	/* Issue and wait for the configure endpoint or
+	 * evaluate context command.
+	 */
+	if (xhci->hci_version > 0x95)
+		ret = xhci_configure_endpoint(xhci, hdev, config_cmd,
+				false, false);
+	else
+		ret = xhci_configure_endpoint(xhci, hdev, config_cmd,
+				true, false);
+
+	xhci_dbg(xhci, "Slot %u Output Context:\n", hdev->slot_id);
+	xhci_dbg_ctx(xhci, vdev->out_ctx, 0);
+
+	xhci_free_command(xhci, config_cmd);
+	return ret;
+}
+
+int xhci_get_frame(struct usb_hcd *hcd)
+{
+	struct xhci_hcd *xhci = hcd_to_xhci(hcd);
+	/* EHCI mods by the periodic size.  Why? */
+	return xhci_readl(xhci, &xhci->run_regs->microframe_index) >> 3;
+}
+
+int xhci_gen_setup(struct usb_hcd *hcd, xhci_get_quirks_t get_quirks)
+{
+	struct xhci_hcd		*xhci;
+	struct device		*dev = hcd->self.controller;
+	int			retval;
+	u32			temp;
+
+	/* Accept arbitrarily long scatter-gather lists */
+	hcd->self.sg_tablesize = ~0;
+	/* XHCI controllers don't stop the ep queue on short packets :| */
+	hcd->self.no_stop_on_short = 1;
+
+	if (usb_hcd_is_primary_hcd(hcd)) {
+		xhci = kzalloc(sizeof(struct xhci_hcd), GFP_KERNEL);
+		if (!xhci)
+			return -ENOMEM;
+		*((struct xhci_hcd **) hcd->hcd_priv) = xhci;
+		xhci->main_hcd = hcd;
+		/* Mark the first roothub as being USB 2.0.
+		 * The xHCI driver will register the USB 3.0 roothub.
+		 */
+		hcd->speed = HCD_USB2;
+		hcd->self.root_hub->speed = USB_SPEED_HIGH;
+		/*
+		 * USB 2.0 roothub under xHCI has an integrated TT,
+		 * (rate matching hub) as opposed to having an OHCI/UHCI
+		 * companion controller.
+		 */
+		hcd->has_tt = 1;
+	} else {
+		/* xHCI private pointer was set in xhci_pci_probe for the second
+		 * registered roothub.
+		 */
+		xhci = hcd_to_xhci(hcd);
+		temp = xhci_readl(xhci, &xhci->cap_regs->hcc_params);
+		if (HCC_64BIT_ADDR(temp)) {
+			xhci_dbg(xhci, "Enabling 64-bit DMA addresses.\n");
+			dma_set_mask(hcd->self.controller, DMA_BIT_MASK(64));
+		} else {
+			dma_set_mask(hcd->self.controller, DMA_BIT_MASK(32));
+		}
+		return 0;
+	}
+
+	xhci->cap_regs = hcd->regs;
+	xhci->op_regs = hcd->regs +
+		HC_LENGTH(xhci_readl(xhci, &xhci->cap_regs->hc_capbase));
+	xhci->run_regs = hcd->regs +
+		(xhci_readl(xhci, &xhci->cap_regs->run_regs_off) & RTSOFF_MASK);
+	/* Cache read-only capability registers */
+	xhci->hcs_params1 = xhci_readl(xhci, &xhci->cap_regs->hcs_params1);
+	xhci->hcs_params2 = xhci_readl(xhci, &xhci->cap_regs->hcs_params2);
+	xhci->hcs_params3 = xhci_readl(xhci, &xhci->cap_regs->hcs_params3);
+	xhci->hcc_params = xhci_readl(xhci, &xhci->cap_regs->hc_capbase);
+	xhci->hci_version = HC_VERSION(xhci->hcc_params);
+	xhci->hcc_params = xhci_readl(xhci, &xhci->cap_regs->hcc_params);
+	xhci_print_registers(xhci);
+
+	get_quirks(dev, xhci);
+
+	/* Make sure the HC is halted. */
+	retval = xhci_halt(xhci);
+	if (retval)
+		goto error;
+
+	xhci_dbg(xhci, "Resetting HCD\n");
+	/* Reset the internal HC memory state and registers. */
+	retval = xhci_reset(xhci);
+	if (retval)
+		goto error;
+	xhci_dbg(xhci, "Reset complete\n");
+
+	temp = xhci_readl(xhci, &xhci->cap_regs->hcc_params);
+	if (HCC_64BIT_ADDR(temp)) {
+		xhci_dbg(xhci, "Enabling 64-bit DMA addresses.\n");
+		dma_set_mask(hcd->self.controller, DMA_BIT_MASK(64));
+	} else {
+		dma_set_mask(hcd->self.controller, DMA_BIT_MASK(32));
+	}
+
+	xhci_dbg(xhci, "Calling HCD init\n");
+	/* Initialize HCD and host controller data structures. */
+	retval = xhci_init(hcd);
+	if (retval)
+		goto error;
+	xhci_dbg(xhci, "Called HCD init\n");
+	return 0;
+error:
+	kfree(xhci);
+	return retval;
+}
+
+MODULE_DESCRIPTION(DRIVER_DESC);
+MODULE_AUTHOR(DRIVER_AUTHOR);
+MODULE_LICENSE("GPL");
+
+static int __init xhci_hcd_init(void)
+{
+	int retval;
+
+	retval = xhci_register_pci();
+	if (retval < 0) {
+		printk(KERN_DEBUG "Problem registering PCI driver.");
+		return retval;
+	}
+	retval = xhci_register_plat();
+	if (retval < 0) {
+		printk(KERN_DEBUG "Problem registering platform driver.");
+		goto unreg_pci;
+	}
+	/*
+	 * Check the compiler generated sizes of structures that must be laid
+	 * out in specific ways for hardware access.
+	 */
+	BUILD_BUG_ON(sizeof(struct xhci_doorbell_array) != 256*32/8);
+	BUILD_BUG_ON(sizeof(struct xhci_slot_ctx) != 8*32/8);
+	BUILD_BUG_ON(sizeof(struct xhci_ep_ctx) != 8*32/8);
+	/* xhci_device_control has eight fields, and also
+	 * embeds one xhci_slot_ctx and 31 xhci_ep_ctx
+	 */
+	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);
+	BUILD_BUG_ON(sizeof(struct xhci_cap_regs) != 7*32/8);
+	BUILD_BUG_ON(sizeof(struct xhci_intr_reg) != 8*32/8);
+	/* xhci_run_regs has eight fields and embeds 128 xhci_intr_regs */
+	BUILD_BUG_ON(sizeof(struct xhci_run_regs) != (8+8*128)*32/8);
+	return 0;
+unreg_pci:
+	xhci_unregister_pci();
+	return retval;
+}
+module_init(xhci_hcd_init);
+
+static void __exit xhci_hcd_cleanup(void)
+{
+	xhci_unregister_pci();
+	xhci_unregister_plat();
+}
+module_exit(xhci_hcd_cleanup);
diff --git a/drivers/usb/host/xhci.h b/drivers/usb/host/xhci.h
new file mode 100644
index 0000000..29c978e
--- /dev/null
+++ b/drivers/usb/host/xhci.h
@@ -0,0 +1,1856 @@
+/*
+ * xHCI host controller driver
+ *
+ * Copyright (C) 2008 Intel Corp.
+ *
+ * Author: Sarah Sharp
+ * Some code borrowed from the Linux EHCI driver.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef __LINUX_XHCI_HCD_H
+#define __LINUX_XHCI_HCD_H
+
+#include <linux/usb.h>
+#include <linux/timer.h>
+#include <linux/kernel.h>
+#include <linux/usb/hcd.h>
+
+/* Code sharing between pci-quirks and xhci hcd */
+#include	"xhci-ext-caps.h"
+#include "pci-quirks.h"
+
+/* xHCI PCI Configuration Registers */
+#define XHCI_SBRN_OFFSET	(0x60)
+
+/* Max number of USB devices for any host controller - limit in section 6.1 */
+#define MAX_HC_SLOTS		256
+/* Section 5.3.3 - MaxPorts */
+#define MAX_HC_PORTS		127
+
+/*
+ * xHCI register interface.
+ * This corresponds to the eXtensible Host Controller Interface (xHCI)
+ * Revision 0.95 specification
+ */
+
+/**
+ * struct xhci_cap_regs - xHCI Host Controller Capability Registers.
+ * @hc_capbase:		length of the capabilities register and HC version number
+ * @hcs_params1:	HCSPARAMS1 - Structural Parameters 1
+ * @hcs_params2:	HCSPARAMS2 - Structural Parameters 2
+ * @hcs_params3:	HCSPARAMS3 - Structural Parameters 3
+ * @hcc_params:		HCCPARAMS - Capability Parameters
+ * @db_off:		DBOFF - Doorbell array offset
+ * @run_regs_off:	RTSOFF - Runtime register space offset
+ */
+struct xhci_cap_regs {
+	__le32	hc_capbase;
+	__le32	hcs_params1;
+	__le32	hcs_params2;
+	__le32	hcs_params3;
+	__le32	hcc_params;
+	__le32	db_off;
+	__le32	run_regs_off;
+	/* Reserved up to (CAPLENGTH - 0x1C) */
+};
+
+/* hc_capbase bitmasks */
+/* bits 7:0 - how long is the Capabilities register */
+#define HC_LENGTH(p)		XHCI_HC_LENGTH(p)
+/* bits 31:16	*/
+#define HC_VERSION(p)		(((p) >> 16) & 0xffff)
+
+/* HCSPARAMS1 - hcs_params1 - bitmasks */
+/* bits 0:7, Max Device Slots */
+#define HCS_MAX_SLOTS(p)	(((p) >> 0) & 0xff)
+#define HCS_SLOTS_MASK		0xff
+/* bits 8:18, Max Interrupters */
+#define HCS_MAX_INTRS(p)	(((p) >> 8) & 0x7ff)
+/* bits 24:31, Max Ports - max value is 0x7F = 127 ports */
+#define HCS_MAX_PORTS(p)	(((p) >> 24) & 0x7f)
+
+/* HCSPARAMS2 - hcs_params2 - bitmasks */
+/* bits 0:3, frames or uframes that SW needs to queue transactions
+ * ahead of the HW to meet periodic deadlines */
+#define HCS_IST(p)		(((p) >> 0) & 0xf)
+/* bits 4:7, max number of Event Ring segments */
+#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 */
+#define HCS_U1_LATENCY(p)	(((p) >> 0) & 0xff)
+/* bits 16:31, Max U2 to U0 latency for the roothub ports */
+#define HCS_U2_LATENCY(p)	(((p) >> 16) & 0xffff)
+
+/* HCCPARAMS - hcc_params - bitmasks */
+/* true: HC can use 64-bit address pointers */
+#define HCC_64BIT_ADDR(p)	((p) & (1 << 0))
+/* true: HC can do bandwidth negotiation */
+#define HCC_BANDWIDTH_NEG(p)	((p) & (1 << 1))
+/* true: HC uses 64-byte Device Context structures
+ * FIXME 64-byte context structures aren't supported yet.
+ */
+#define HCC_64BYTE_CONTEXT(p)	((p) & (1 << 2))
+/* true: HC has port power switches */
+#define HCC_PPC(p)		((p) & (1 << 3))
+/* true: HC has port indicators */
+#define HCS_INDICATOR(p)	((p) & (1 << 4))
+/* true: HC has Light HC Reset Capability */
+#define HCC_LIGHT_RESET(p)	((p) & (1 << 5))
+/* true: HC supports latency tolerance messaging */
+#define HCC_LTC(p)		((p) & (1 << 6))
+/* true: no secondary Stream ID Support */
+#define HCC_NSS(p)		((p) & (1 << 7))
+/* Max size for Primary Stream Arrays - 2^(n+1), where n is bits 12:15 */
+#define HCC_MAX_PSA(p)		(1 << ((((p) >> 12) & 0xf) + 1))
+/* Extended Capabilities pointer from PCI base - section 5.3.6 */
+#define HCC_EXT_CAPS(p)		XHCI_HCC_EXT_CAPS(p)
+
+/* db_off bitmask - bits 0:1 reserved */
+#define	DBOFF_MASK	(~0x3)
+
+/* run_regs_off bitmask - bits 0:4 reserved */
+#define	RTSOFF_MASK	(~0x1f)
+
+
+/* Number of registers per port */
+#define	NUM_PORT_REGS	4
+
+/**
+ * struct xhci_op_regs - xHCI Host Controller Operational Registers.
+ * @command:		USBCMD - xHC command register
+ * @status:		USBSTS - xHC status register
+ * @page_size:		This indicates the page size that the host controller
+ * 			supports.  If bit n is set, the HC supports a page size
+ * 			of 2^(n+12), up to a 128MB page size.
+ * 			4K is the minimum page size.
+ * @cmd_ring:		CRP - 64-bit Command Ring Pointer
+ * @dcbaa_ptr:		DCBAAP - 64-bit Device Context Base Address Array Pointer
+ * @config_reg:		CONFIG - Configure Register
+ * @port_status_base:	PORTSCn - base address for Port Status and Control
+ * 			Each port has a Port Status and Control register,
+ * 			followed by a Port Power Management Status and Control
+ * 			register, a Port Link Info register, and a reserved
+ * 			register.
+ * @port_power_base:	PORTPMSCn - base address for
+ * 			Port Power Management Status and Control
+ * @port_link_base:	PORTLIn - base address for Port Link Info (current
+ * 			Link PM state and control) for USB 2.1 and USB 3.0
+ * 			devices.
+ */
+struct xhci_op_regs {
+	__le32	command;
+	__le32	status;
+	__le32	page_size;
+	__le32	reserved1;
+	__le32	reserved2;
+	__le32	dev_notification;
+	__le64	cmd_ring;
+	/* rsvd: offset 0x20-2F */
+	__le32	reserved3[4];
+	__le64	dcbaa_ptr;
+	__le32	config_reg;
+	/* rsvd: offset 0x3C-3FF */
+	__le32	reserved4[241];
+	/* port 1 registers, which serve as a base address for other ports */
+	__le32	port_status_base;
+	__le32	port_power_base;
+	__le32	port_link_base;
+	__le32	reserved5;
+	/* registers for ports 2-255 */
+	__le32	reserved6[NUM_PORT_REGS*254];
+};
+
+/* USBCMD - USB command - command bitmasks */
+/* start/stop HC execution - do not write unless HC is halted*/
+#define CMD_RUN		XHCI_CMD_RUN
+/* Reset HC - resets internal HC state machine and all registers (except
+ * PCI config regs).  HC does NOT drive a USB reset on the downstream ports.
+ * The xHCI driver must reinitialize the xHC after setting this bit.
+ */
+#define CMD_RESET	(1 << 1)
+/* Event Interrupt Enable - a '1' allows interrupts from the host controller */
+#define CMD_EIE		XHCI_CMD_EIE
+/* Host System Error Interrupt Enable - get out-of-band signal for HC errors */
+#define CMD_HSEIE	XHCI_CMD_HSEIE
+/* bits 4:6 are reserved (and should be preserved on writes). */
+/* light reset (port status stays unchanged) - reset completed when this is 0 */
+#define CMD_LRESET	(1 << 7)
+/* host controller save/restore state. */
+#define CMD_CSS		(1 << 8)
+#define CMD_CRS		(1 << 9)
+/* Enable Wrap Event - '1' means xHC generates an event when MFINDEX wraps. */
+#define CMD_EWE		XHCI_CMD_EWE
+/* MFINDEX power management - '1' means xHC can stop MFINDEX counter if all root
+ * hubs are in U3 (selective suspend), disconnect, disabled, or powered-off.
+ * '0' means the xHC can power it off if all ports are in the disconnect,
+ * disabled, or powered-off state.
+ */
+#define CMD_PM_INDEX	(1 << 11)
+/* bits 12:31 are reserved (and should be preserved on writes). */
+
+/* IMAN - Interrupt Management Register */
+#define IMAN_IE		(1 << 1)
+#define IMAN_IP		(1 << 0)
+
+/* USBSTS - USB status - status bitmasks */
+/* HC not running - set to 1 when run/stop bit is cleared. */
+#define STS_HALT	XHCI_STS_HALT
+/* serious error, e.g. PCI parity error.  The HC will clear the run/stop bit. */
+#define STS_FATAL	(1 << 2)
+/* event interrupt - clear this prior to clearing any IP flags in IR set*/
+#define STS_EINT	(1 << 3)
+/* port change detect */
+#define STS_PORT	(1 << 4)
+/* bits 5:7 reserved and zeroed */
+/* save state status - '1' means xHC is saving state */
+#define STS_SAVE	(1 << 8)
+/* restore state status - '1' means xHC is restoring state */
+#define STS_RESTORE	(1 << 9)
+/* true: save or restore error */
+#define STS_SRE		(1 << 10)
+/* true: Controller Not Ready to accept doorbell or op reg writes after reset */
+#define STS_CNR		XHCI_STS_CNR
+/* true: internal Host Controller Error - SW needs to reset and reinitialize */
+#define STS_HCE		(1 << 12)
+/* bits 13:31 reserved and should be preserved */
+
+/*
+ * DNCTRL - Device Notification Control Register - dev_notification bitmasks
+ * Generate a device notification event when the HC sees a transaction with a
+ * notification type that matches a bit set in this bit field.
+ */
+#define	DEV_NOTE_MASK		(0xffff)
+#define ENABLE_DEV_NOTE(x)	(1 << (x))
+/* Most of the device notification types should only be used for debug.
+ * SW does need to pay attention to function wake notifications.
+ */
+#define	DEV_NOTE_FWAKE		ENABLE_DEV_NOTE(1)
+
+/* CRCR - Command Ring Control Register - cmd_ring bitmasks */
+/* bit 0 is the command ring cycle state */
+/* stop ring operation after completion of the currently executing command */
+#define CMD_RING_PAUSE		(1 << 1)
+/* stop ring immediately - abort the currently executing command */
+#define CMD_RING_ABORT		(1 << 2)
+/* true: command ring is running */
+#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_RSVD_BITS	(0x3f)
+
+/* CONFIG - Configure Register - config_reg bitmasks */
+/* bits 0:7 - maximum number of device slots enabled (NumSlotsEn) */
+#define MAX_DEVS(p)	((p) & 0xff)
+/* bits 8:31 - reserved and should be preserved */
+
+/* PORTSC - Port Status and Control Register - port_status_base bitmasks */
+/* true: device connected */
+#define PORT_CONNECT	(1 << 0)
+/* true: port enabled */
+#define PORT_PE		(1 << 1)
+/* bit 2 reserved and zeroed */
+/* true: port has an over-current condition */
+#define PORT_OC		(1 << 3)
+/* true: port reset signaling asserted */
+#define PORT_RESET	(1 << 4)
+/* Port Link State - bits 5:8
+ * A read gives the current link PM state of the port,
+ * a write with Link State Write Strobe set sets the link state.
+ */
+#define PORT_PLS_MASK	(0xf << 5)
+#define XDEV_U0		(0x0 << 5)
+#define XDEV_U2		(0x2 << 5)
+#define XDEV_U3		(0x3 << 5)
+#define XDEV_RESUME	(0xf << 5)
+/* true: port has power (see HCC_PPC) */
+#define PORT_POWER	(1 << 9)
+/* bits 10:13 indicate device speed:
+ * 0 - undefined speed - port hasn't be initialized by a reset yet
+ * 1 - full speed
+ * 2 - low speed
+ * 3 - high speed
+ * 4 - super speed
+ * 5-15 reserved
+ */
+#define DEV_SPEED_MASK		(0xf << 10)
+#define	XDEV_FS			(0x1 << 10)
+#define	XDEV_LS			(0x2 << 10)
+#define	XDEV_HS			(0x3 << 10)
+#define	XDEV_SS			(0x4 << 10)
+#define DEV_UNDEFSPEED(p)	(((p) & DEV_SPEED_MASK) == (0x0<<10))
+#define DEV_FULLSPEED(p)	(((p) & DEV_SPEED_MASK) == XDEV_FS)
+#define DEV_LOWSPEED(p)		(((p) & DEV_SPEED_MASK) == XDEV_LS)
+#define DEV_HIGHSPEED(p)	(((p) & DEV_SPEED_MASK) == XDEV_HS)
+#define DEV_SUPERSPEED(p)	(((p) & DEV_SPEED_MASK) == XDEV_SS)
+/* Bits 20:23 in the Slot Context are the speed for the device */
+#define	SLOT_SPEED_FS		(XDEV_FS << 10)
+#define	SLOT_SPEED_LS		(XDEV_LS << 10)
+#define	SLOT_SPEED_HS		(XDEV_HS << 10)
+#define	SLOT_SPEED_SS		(XDEV_SS << 10)
+/* Port Indicator Control */
+#define PORT_LED_OFF	(0 << 14)
+#define PORT_LED_AMBER	(1 << 14)
+#define PORT_LED_GREEN	(2 << 14)
+#define PORT_LED_MASK	(3 << 14)
+/* Port Link State Write Strobe - set this when changing link state */
+#define PORT_LINK_STROBE	(1 << 16)
+/* true: connect status change */
+#define PORT_CSC	(1 << 17)
+/* true: port enable change */
+#define PORT_PEC	(1 << 18)
+/* true: warm reset for a USB 3.0 device is done.  A "hot" reset puts the port
+ * into an enabled state, and the device into the default state.  A "warm" reset
+ * also resets the link, forcing the device through the link training sequence.
+ * SW can also look at the Port Reset register to see when warm reset is done.
+ */
+#define PORT_WRC	(1 << 19)
+/* true: over-current change */
+#define PORT_OCC	(1 << 20)
+/* true: reset change - 1 to 0 transition of PORT_RESET */
+#define PORT_RC		(1 << 21)
+/* port link status change - set on some port link state transitions:
+ *  Transition				Reason
+ *  ------------------------------------------------------------------------------
+ *  - U3 to Resume			Wakeup signaling from a device
+ *  - Resume to Recovery to U0		USB 3.0 device resume
+ *  - Resume to U0			USB 2.0 device resume
+ *  - U3 to Recovery to U0		Software resume of USB 3.0 device complete
+ *  - U3 to U0				Software resume of USB 2.0 device complete
+ *  - U2 to U0				L1 resume of USB 2.1 device complete
+ *  - U0 to U0 (???)			L1 entry rejection by USB 2.1 device
+ *  - U0 to disabled			L1 entry error with USB 2.1 device
+ *  - Any state to inactive		Error on USB 3.0 port
+ */
+#define PORT_PLC	(1 << 22)
+/* port configure error change - port failed to configure its link partner */
+#define PORT_CEC	(1 << 23)
+/* Cold Attach Status - xHC can set this bit to report device attached during
+ * Sx state. Warm port reset should be perfomed to clear this bit and move port
+ * to connected state.
+ */
+#define PORT_CAS	(1 << 24)
+/* wake on connect (enable) */
+#define PORT_WKCONN_E	(1 << 25)
+/* wake on disconnect (enable) */
+#define PORT_WKDISC_E	(1 << 26)
+/* wake on over-current (enable) */
+#define PORT_WKOC_E	(1 << 27)
+/* bits 28:29 reserved */
+/* true: device is removable - for USB 3.0 roothub emulation */
+#define PORT_DEV_REMOVE	(1 << 30)
+/* Initiate a warm port reset - complete when PORT_WRC is '1' */
+#define PORT_WR		(1 << 31)
+
+/* We mark duplicate entries with -1 */
+#define DUPLICATE_ENTRY ((u8)(-1))
+
+/* Port Power Management Status and Control - port_power_base bitmasks */
+/* Inactivity timer value for transitions into U1, in microseconds.
+ * Timeout can be up to 127us.  0xFF means an infinite timeout.
+ */
+#define PORT_U1_TIMEOUT(p)	((p) & 0xff)
+#define PORT_U1_TIMEOUT_MASK	0xff
+/* Inactivity timer value for transitions into U2 */
+#define PORT_U2_TIMEOUT(p)	(((p) & 0xff) << 8)
+#define PORT_U2_TIMEOUT_MASK	(0xff << 8)
+/* Bits 24:31 for port testing */
+
+/* USB2 Protocol PORTSPMSC */
+#define	PORT_L1S_MASK		7
+#define	PORT_L1S_SUCCESS	1
+#define	PORT_RWE		(1 << 3)
+#define	PORT_HIRD(p)		(((p) & 0xf) << 4)
+#define	PORT_HIRD_MASK		(0xf << 4)
+#define	PORT_L1DS(p)		(((p) & 0xff) << 8)
+#define	PORT_HLE		(1 << 16)
+
+/**
+ * struct xhci_intr_reg - Interrupt Register Set
+ * @irq_pending:	IMAN - Interrupt Management Register.  Used to enable
+ *			interrupts and check for pending interrupts.
+ * @irq_control:	IMOD - Interrupt Moderation Register.
+ * 			Used to throttle interrupts.
+ * @erst_size:		Number of segments in the Event Ring Segment Table (ERST).
+ * @erst_base:		ERST base address.
+ * @erst_dequeue:	Event ring dequeue pointer.
+ *
+ * Each interrupter (defined by a MSI-X vector) has an event ring and an Event
+ * Ring Segment Table (ERST) associated with it.  The event ring is comprised of
+ * multiple segments of the same size.  The HC places events on the ring and
+ * "updates the Cycle bit in the TRBs to indicate to software the current
+ * position of the Enqueue Pointer." The HCD (Linux) processes those events and
+ * updates the dequeue pointer.
+ */
+struct xhci_intr_reg {
+	__le32	irq_pending;
+	__le32	irq_control;
+	__le32	erst_size;
+	__le32	rsvd;
+	__le64	erst_base;
+	__le64	erst_dequeue;
+};
+
+/* irq_pending bitmasks */
+#define	ER_IRQ_PENDING(p)	((p) & 0x1)
+/* bits 2:31 need to be preserved */
+/* THIS IS BUGGY - FIXME - IP IS WRITE 1 TO CLEAR */
+#define	ER_IRQ_CLEAR(p)		((p) & 0xfffffffe)
+#define	ER_IRQ_ENABLE(p)	((ER_IRQ_CLEAR(p)) | 0x2)
+#define	ER_IRQ_DISABLE(p)	((ER_IRQ_CLEAR(p)) & ~(0x2))
+
+/* irq_control bitmasks */
+/* Minimum interval between interrupts (in 250ns intervals).  The interval
+ * between interrupts will be longer if there are no events on the event ring.
+ * Default is 4000 (1 ms).
+ */
+#define ER_IRQ_INTERVAL_MASK	(0xffff)
+/* Counter used to count down the time to the next interrupt - HW use only */
+#define ER_IRQ_COUNTER_MASK	(0xffff << 16)
+
+/* erst_size bitmasks */
+/* Preserve bits 16:31 of erst_size */
+#define	ERST_SIZE_MASK		(0xffff << 16)
+
+/* erst_dequeue bitmasks */
+/* Dequeue ERST Segment Index (DESI) - Segment number (or alias)
+ * where the current dequeue pointer lies.  This is an optional HW hint.
+ */
+#define ERST_DESI_MASK		(0x7)
+/* Event Handler Busy (EHB) - is the event ring scheduled to be serviced by
+ * a work queue (or delayed service routine)?
+ */
+#define ERST_EHB		(1 << 3)
+#define ERST_PTR_MASK		(0xf)
+
+/**
+ * struct xhci_run_regs
+ * @microframe_index:
+ * 		MFINDEX - current microframe number
+ *
+ * Section 5.5 Host Controller Runtime Registers:
+ * "Software should read and write these registers using only Dword (32 bit)
+ * or larger accesses"
+ */
+struct xhci_run_regs {
+	__le32			microframe_index;
+	__le32			rsvd[7];
+	struct xhci_intr_reg	ir_set[128];
+};
+
+/**
+ * struct doorbell_array
+ *
+ * Bits  0 -  7: Endpoint target
+ * Bits  8 - 15: RsvdZ
+ * Bits 16 - 31: Stream ID
+ *
+ * Section 5.6
+ */
+struct xhci_doorbell_array {
+	__le32	doorbell[256];
+};
+
+#define DB_VALUE(ep, stream)	((((ep) + 1) & 0xff) | ((stream) << 16))
+#define DB_VALUE_HOST		0x00000000
+
+/**
+ * struct xhci_protocol_caps
+ * @revision:		major revision, minor revision, capability ID,
+ *			and next capability pointer.
+ * @name_string:	Four ASCII characters to say which spec this xHC
+ *			follows, typically "USB ".
+ * @port_info:		Port offset, count, and protocol-defined information.
+ */
+struct xhci_protocol_caps {
+	u32	revision;
+	u32	name_string;
+	u32	port_info;
+};
+
+#define	XHCI_EXT_PORT_MAJOR(x)	(((x) >> 24) & 0xff)
+#define	XHCI_EXT_PORT_OFF(x)	((x) & 0xff)
+#define	XHCI_EXT_PORT_COUNT(x)	(((x) >> 8) & 0xff)
+
+/**
+ * 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
+ * @tt_info:	tt_info is used to construct split transaction tokens
+ * @dev_state:	slot state and device address
+ *
+ * Slot Context - section 6.2.1.1.  This assumes the HC uses 32-byte context
+ * structures.  If the HC uses 64-byte contexts, there is an additional 32 bytes
+ * reserved at the end of the slot context for HC internal use.
+ */
+struct xhci_slot_ctx {
+	__le32	dev_info;
+	__le32	dev_info2;
+	__le32	tt_info;
+	__le32	dev_state;
+	/* offset 0x10 to 0x1f reserved for HC internal use */
+	__le32	reserved[4];
+};
+
+/* dev_info bitmasks */
+/* Route String - 0:19 */
+#define ROUTE_STRING_MASK	(0xfffff)
+/* Device speed - values defined by PORTSC Device Speed field - 20:23 */
+#define DEV_SPEED	(0xf << 20)
+/* bit 24 reserved */
+/* Is this LS/FS device connected through a HS hub? - bit 25 */
+#define DEV_MTT		(0x1 << 25)
+/* Set if the device is a hub - bit 26 */
+#define DEV_HUB		(0x1 << 26)
+/* Index of the last valid endpoint context in this device context - 27:31 */
+#define LAST_CTX_MASK	(0x1f << 27)
+#define LAST_CTX(p)	((p) << 27)
+#define LAST_CTX_TO_EP_NUM(p)	(((p) >> 27) - 1)
+#define SLOT_FLAG	(1 << 0)
+#define EP0_FLAG	(1 << 1)
+
+/* dev_info2 bitmasks */
+/* Max Exit Latency (ms) - worst case time to wake up all links in dev path */
+#define MAX_EXIT	(0xffff)
+/* Root hub port number that is needed to access the USB device */
+#define ROOT_HUB_PORT(p)	(((p) & 0xff) << 16)
+#define DEVINFO_TO_ROOT_HUB_PORT(p)	(((p) >> 16) & 0xff)
+/* Maximum number of ports under a hub device */
+#define XHCI_MAX_PORTS(p)	(((p) & 0xff) << 24)
+
+/* tt_info bitmasks */
+/*
+ * TT Hub Slot ID - for low or full speed devices attached to a high-speed hub
+ * The Slot ID of the hub that isolates the high speed signaling from
+ * this low or full-speed device.  '0' if attached to root hub port.
+ */
+#define TT_SLOT		(0xff)
+/*
+ * The number of the downstream facing port of the high-speed hub
+ * '0' if the device is not low or full speed.
+ */
+#define TT_PORT		(0xff << 8)
+#define TT_THINK_TIME(p)	(((p) & 0x3) << 16)
+
+/* dev_state bitmasks */
+/* USB device address - assigned by the HC */
+#define DEV_ADDR_MASK	(0xff)
+/* bits 8:26 reserved */
+/* Slot state */
+#define SLOT_STATE	(0x1f << 27)
+#define GET_SLOT_STATE(p)	(((p) & (0x1f << 27)) >> 27)
+
+#define SLOT_STATE_DISABLED	0
+#define SLOT_STATE_ENABLED	SLOT_STATE_DISABLED
+#define SLOT_STATE_DEFAULT	1
+#define SLOT_STATE_ADDRESSED	2
+#define SLOT_STATE_CONFIGURED	3
+
+/**
+ * struct xhci_ep_ctx
+ * @ep_info:	endpoint state, streams, mult, and interval information.
+ * @ep_info2:	information on endpoint type, max packet size, max burst size,
+ * 		error count, and whether the HC will force an event for all
+ * 		transactions.
+ * @deq:	64-bit ring dequeue pointer address.  If the endpoint only
+ * 		defines one stream, this points to the endpoint transfer ring.
+ * 		Otherwise, it points to a stream context array, which has a
+ * 		ring pointer for each flow.
+ * @tx_info:
+ * 		Average TRB lengths for the endpoint ring and
+ * 		max payload within an Endpoint Service Interval Time (ESIT).
+ *
+ * Endpoint Context - section 6.2.1.2.  This assumes the HC uses 32-byte context
+ * structures.  If the HC uses 64-byte contexts, there is an additional 32 bytes
+ * reserved at the end of the endpoint context for HC internal use.
+ */
+struct xhci_ep_ctx {
+	__le32	ep_info;
+	__le32	ep_info2;
+	__le64	deq;
+	__le32	tx_info;
+	/* offset 0x14 - 0x1f reserved for HC internal use */
+	__le32	reserved[3];
+};
+
+/* ep_info bitmasks */
+/*
+ * Endpoint State - bits 0:2
+ * 0 - disabled
+ * 1 - running
+ * 2 - halted due to halt condition - ok to manipulate endpoint ring
+ * 3 - stopped
+ * 4 - TRB error
+ * 5-7 - reserved
+ */
+#define EP_STATE_MASK		(0xf)
+#define EP_STATE_DISABLED	0
+#define EP_STATE_RUNNING	1
+#define EP_STATE_HALTED		2
+#define EP_STATE_STOPPED	3
+#define EP_STATE_ERROR		4
+/* Mult - Max number of burtst within an interval, in EP companion desc. */
+#define EP_MULT(p)		(((p) & 0x3) << 8)
+#define CTX_TO_EP_MULT(p)	(((p) >> 8) & 0x3)
+/* bits 10:14 are Max Primary Streams */
+/* bit 15 is Linear Stream Array */
+/* Interval - period between requests to an endpoint - 125u increments. */
+#define EP_INTERVAL(p)		(((p) & 0xff) << 16)
+#define EP_INTERVAL_TO_UFRAMES(p)		(1 << (((p) >> 16) & 0xff))
+#define CTX_TO_EP_INTERVAL(p)	(((p) >> 16) & 0xff)
+#define EP_MAXPSTREAMS_MASK	(0x1f << 10)
+#define EP_MAXPSTREAMS(p)	(((p) << 10) & EP_MAXPSTREAMS_MASK)
+/* Endpoint is set up with a Linear Stream Array (vs. Secondary Stream Array) */
+#define	EP_HAS_LSA		(1 << 15)
+
+/* ep_info2 bitmasks */
+/*
+ * Force Event - generate transfer events for all TRBs for this endpoint
+ * This will tell the HC to ignore the IOC and ISP flags (for debugging only).
+ */
+#define	FORCE_EVENT	(0x1)
+#define ERROR_COUNT(p)	(((p) & 0x3) << 1)
+#define CTX_TO_EP_TYPE(p)	(((p) >> 3) & 0x7)
+#define EP_TYPE(p)	((p) << 3)
+#define ISOC_OUT_EP	1
+#define BULK_OUT_EP	2
+#define INT_OUT_EP	3
+#define CTRL_EP		4
+#define ISOC_IN_EP	5
+#define BULK_IN_EP	6
+#define INT_IN_EP	7
+/* bit 6 reserved */
+/* bit 7 is Host Initiate Disable - for disabling stream selection */
+#define MAX_BURST(p)	(((p)&0xff) << 8)
+#define CTX_TO_MAX_BURST(p)	(((p) >> 8) & 0xff)
+#define MAX_PACKET(p)	(((p)&0xffff) << 16)
+#define MAX_PACKET_MASK		(0xffff << 16)
+#define MAX_PACKET_DECODED(p)	(((p) >> 16) & 0xffff)
+
+/* Get max packet size from ep desc. Bit 10..0 specify the max packet size.
+ * USB2.0 spec 9.6.6.
+ */
+#define GET_MAX_PACKET(p)	((p) & 0x7ff)
+
+/* tx_info bitmasks */
+#define AVG_TRB_LENGTH_FOR_EP(p)	((p) & 0xffff)
+#define MAX_ESIT_PAYLOAD_FOR_EP(p)	(((p) & 0xffff) << 16)
+#define CTX_TO_MAX_ESIT_PAYLOAD(p)	(((p) >> 16) & 0xffff)
+
+/* deq bitmasks */
+#define EP_CTX_CYCLE_MASK		(1 << 0)
+
+
+/**
+ * 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_input_control_ctx {
+	__le32	drop_flags;
+	__le32	add_flags;
+	__le32	rsvd2[6];
+};
+
+#define	EP_IS_ADDED(ctrl_ctx, i) \
+	(le32_to_cpu(ctrl_ctx->add_flags) & (1 << (i + 1)))
+#define	EP_IS_DROPPED(ctrl_ctx, i)       \
+	(le32_to_cpu(ctrl_ctx->drop_flags) & (1 << (i + 1)))
+
+/* Represents everything that is needed to issue a command on the command ring.
+ * It's useful to pre-allocate these for commands that cannot fail due to
+ * out-of-memory errors, like freeing streams.
+ */
+struct xhci_command {
+	/* Input context for changing device state */
+	struct xhci_container_ctx	*in_ctx;
+	u32				status;
+	/* If completion is null, no one is waiting on this command
+	 * and the structure can be freed after the command completes.
+	 */
+	struct completion		*completion;
+	union xhci_trb			*command_trb;
+	struct list_head		cmd_list;
+};
+
+/* drop context bitmasks */
+#define	DROP_EP(x)	(0x1 << x)
+/* add context bitmasks */
+#define	ADD_EP(x)	(0x1 << x)
+
+struct xhci_stream_ctx {
+	/* 64-bit stream ring address, cycle state, and stream type */
+	__le64	stream_ring;
+	/* offset 0x14 - 0x1f reserved for HC internal use */
+	__le32	reserved[2];
+};
+
+/* Stream Context Types (section 6.4.1) - bits 3:1 of stream ctx deq ptr */
+#define	SCT_FOR_CTX(p)		(((p) << 1) & 0x7)
+/* Secondary stream array type, dequeue pointer is to a transfer ring */
+#define	SCT_SEC_TR		0
+/* Primary stream array type, dequeue pointer is to a transfer ring */
+#define	SCT_PRI_TR		1
+/* Dequeue pointer is for a secondary stream array (SSA) with 8 entries */
+#define SCT_SSA_8		2
+#define SCT_SSA_16		3
+#define SCT_SSA_32		4
+#define SCT_SSA_64		5
+#define SCT_SSA_128		6
+#define SCT_SSA_256		7
+
+/* Assume no secondary streams for now */
+struct xhci_stream_info {
+	struct xhci_ring		**stream_rings;
+	/* Number of streams, including stream 0 (which drivers can't use) */
+	unsigned int			num_streams;
+	/* The stream context array may be bigger than
+	 * the number of streams the driver asked for
+	 */
+	struct xhci_stream_ctx		*stream_ctx_array;
+	unsigned int			num_stream_ctxs;
+	dma_addr_t			ctx_array_dma;
+	/* For mapping physical TRB addresses to segments in stream rings */
+	struct radix_tree_root		trb_address_map;
+	struct xhci_command		*free_streams_command;
+};
+
+#define	SMALL_STREAM_ARRAY_SIZE		256
+#define	MEDIUM_STREAM_ARRAY_SIZE	1024
+
+/* Some Intel xHCI host controllers need software to keep track of the bus
+ * bandwidth.  Keep track of endpoint info here.  Each root port is allocated
+ * the full bus bandwidth.  We must also treat TTs (including each port under a
+ * multi-TT hub) as a separate bandwidth domain.  The direct memory interface
+ * (DMI) also limits the total bandwidth (across all domains) that can be used.
+ */
+struct xhci_bw_info {
+	/* ep_interval is zero-based */
+	unsigned int		ep_interval;
+	/* mult and num_packets are one-based */
+	unsigned int		mult;
+	unsigned int		num_packets;
+	unsigned int		max_packet_size;
+	unsigned int		max_esit_payload;
+	unsigned int		type;
+};
+
+/* "Block" sizes in bytes the hardware uses for different device speeds.
+ * The logic in this part of the hardware limits the number of bits the hardware
+ * can use, so must represent bandwidth in a less precise manner to mimic what
+ * the scheduler hardware computes.
+ */
+#define	FS_BLOCK	1
+#define	HS_BLOCK	4
+#define	SS_BLOCK	16
+#define	DMI_BLOCK	32
+
+/* Each device speed has a protocol overhead (CRC, bit stuffing, etc) associated
+ * with each byte transferred.  SuperSpeed devices have an initial overhead to
+ * set up bursts.  These are in blocks, see above.  LS overhead has already been
+ * translated into FS blocks.
+ */
+#define DMI_OVERHEAD 8
+#define DMI_OVERHEAD_BURST 4
+#define SS_OVERHEAD 8
+#define SS_OVERHEAD_BURST 32
+#define HS_OVERHEAD 26
+#define FS_OVERHEAD 20
+#define LS_OVERHEAD 128
+/* The TTs need to claim roughly twice as much bandwidth (94 bytes per
+ * microframe ~= 24Mbps) of the HS bus as the devices can actually use because
+ * of overhead associated with split transfers crossing microframe boundaries.
+ * 31 blocks is pure protocol overhead.
+ */
+#define TT_HS_OVERHEAD (31 + 94)
+#define TT_DMI_OVERHEAD (25 + 12)
+
+/* Bandwidth limits in blocks */
+#define FS_BW_LIMIT		1285
+#define TT_BW_LIMIT		1320
+#define HS_BW_LIMIT		1607
+#define SS_BW_LIMIT_IN		3906
+#define DMI_BW_LIMIT_IN		3906
+#define SS_BW_LIMIT_OUT		3906
+#define DMI_BW_LIMIT_OUT	3906
+
+/* Percentage of bus bandwidth reserved for non-periodic transfers */
+#define FS_BW_RESERVED		10
+#define HS_BW_RESERVED		20
+#define SS_BW_RESERVED		10
+
+struct xhci_virt_ep {
+	struct xhci_ring		*ring;
+	/* Related to endpoints that are configured to use stream IDs only */
+	struct xhci_stream_info		*stream_info;
+	/* Temporary storage in case the configure endpoint command fails and we
+	 * have to restore the device state to the previous state
+	 */
+	struct xhci_ring		*new_ring;
+	unsigned int			ep_state;
+#define SET_DEQ_PENDING		(1 << 0)
+#define EP_HALTED		(1 << 1)	/* For stall handling */
+#define EP_HALT_PENDING		(1 << 2)	/* For URB cancellation */
+/* Transitioning the endpoint to using streams, don't enqueue URBs */
+#define EP_GETTING_STREAMS	(1 << 3)
+#define EP_HAS_STREAMS		(1 << 4)
+/* Transitioning the endpoint to not using streams, don't enqueue URBs */
+#define EP_GETTING_NO_STREAMS	(1 << 5)
+	/* ----  Related to URB cancellation ---- */
+	struct list_head	cancelled_td_list;
+	/* The TRB that was last reported in a stopped endpoint ring */
+	union xhci_trb		*stopped_trb;
+	struct xhci_td		*stopped_td;
+	unsigned int		stopped_stream;
+	/* Watchdog timer for stop endpoint command to cancel URBs */
+	struct timer_list	stop_cmd_timer;
+	int			stop_cmds_pending;
+	struct xhci_hcd		*xhci;
+	/* Dequeue pointer and dequeue segment for a submitted Set TR Dequeue
+	 * command.  We'll need to update the ring's dequeue segment and dequeue
+	 * pointer after the command completes.
+	 */
+	struct xhci_segment	*queued_deq_seg;
+	union xhci_trb		*queued_deq_ptr;
+	/*
+	 * Sometimes the xHC can not process isochronous endpoint ring quickly
+	 * enough, and it will miss some isoc tds on the ring and generate
+	 * a Missed Service Error Event.
+	 * Set skip flag when receive a Missed Service Error Event and
+	 * process the missed tds on the endpoint ring.
+	 */
+	bool			skip;
+	/* Bandwidth checking storage */
+	struct xhci_bw_info	bw_info;
+	struct list_head	bw_endpoint_list;
+};
+
+enum xhci_overhead_type {
+	LS_OVERHEAD_TYPE = 0,
+	FS_OVERHEAD_TYPE,
+	HS_OVERHEAD_TYPE,
+};
+
+struct xhci_interval_bw {
+	unsigned int		num_packets;
+	/* Sorted by max packet size.
+	 * Head of the list is the greatest max packet size.
+	 */
+	struct list_head	endpoints;
+	/* How many endpoints of each speed are present. */
+	unsigned int		overhead[3];
+};
+
+#define	XHCI_MAX_INTERVAL	16
+
+struct xhci_interval_bw_table {
+	unsigned int		interval0_esit_payload;
+	struct xhci_interval_bw	interval_bw[XHCI_MAX_INTERVAL];
+	/* Includes reserved bandwidth for async endpoints */
+	unsigned int		bw_used;
+	unsigned int		ss_bw_in;
+	unsigned int		ss_bw_out;
+};
+
+
+struct xhci_virt_device {
+	struct usb_device		*udev;
+	/*
+	 * Commands to the hardware are passed an "input context" that
+	 * tells the hardware what to change in its data structures.
+	 * The hardware will return changes in an "output context" that
+	 * software must allocate for the hardware.  We need to keep
+	 * track of input and output contexts separately because
+	 * these commands might fail and we don't trust the hardware.
+	 */
+	struct xhci_container_ctx       *out_ctx;
+	/* Used for addressing devices and configuration changes */
+	struct xhci_container_ctx       *in_ctx;
+	/* Rings saved to ensure old alt settings can be re-instated */
+	struct xhci_ring		**ring_cache;
+	int				num_rings_cached;
+	/* Store xHC assigned device address */
+	int				address;
+#define	XHCI_MAX_RINGS_CACHED	31
+	struct xhci_virt_ep		eps[31];
+	struct completion		cmd_completion;
+	/* Status of the last command issued for this device */
+	u32				cmd_status;
+	struct list_head		cmd_list;
+	u8				fake_port;
+	u8				real_port;
+	struct xhci_interval_bw_table	*bw_table;
+	struct xhci_tt_bw_info		*tt_info;
+	/* The current max exit latency for the enabled USB3 link states. */
+	u16				current_mel;
+};
+
+/*
+ * For each roothub, keep track of the bandwidth information for each periodic
+ * interval.
+ *
+ * If a high speed hub is attached to the roothub, each TT associated with that
+ * hub is a separate bandwidth domain.  The interval information for the
+ * endpoints on the devices under that TT will appear in the TT structure.
+ */
+struct xhci_root_port_bw_info {
+	struct list_head		tts;
+	unsigned int			num_active_tts;
+	struct xhci_interval_bw_table	bw_table;
+};
+
+struct xhci_tt_bw_info {
+	struct list_head		tt_list;
+	int				slot_id;
+	int				ttport;
+	struct xhci_interval_bw_table	bw_table;
+	int				active_eps;
+};
+
+
+/**
+ * struct xhci_device_context_array
+ * @dev_context_ptr	array of 64-bit DMA addresses for device contexts
+ */
+struct xhci_device_context_array {
+	/* 64-bit device addresses; we only write 32-bit addresses */
+	__le64			dev_context_ptrs[MAX_HC_SLOTS];
+	/* private xHCD pointers */
+	dma_addr_t	dma;
+};
+/* TODO: write function to set the 64-bit device DMA address */
+/*
+ * TODO: change this to be dynamically sized at HC mem init time since the HC
+ * might not be able to handle the maximum number of devices possible.
+ */
+
+
+struct xhci_transfer_event {
+	/* 64-bit buffer address, or immediate data */
+	__le64	buffer;
+	__le32	transfer_len;
+	/* This field is interpreted differently based on the type of TRB */
+	__le32	flags;
+};
+
+/* Transfer event TRB length bit mask */
+/* bits 0:23 */
+#define	EVENT_TRB_LEN(p)		((p) & 0xffffff)
+
+/** Transfer Event bit fields **/
+#define	TRB_TO_EP_ID(p)	(((p) >> 16) & 0x1f)
+
+/* Completion Code - only applicable for some types of TRBs */
+#define	COMP_CODE_MASK		(0xff << 24)
+#define GET_COMP_CODE(p)	(((p) & COMP_CODE_MASK) >> 24)
+#define COMP_SUCCESS	1
+/* Data Buffer Error */
+#define COMP_DB_ERR	2
+/* Babble Detected Error */
+#define COMP_BABBLE	3
+/* USB Transaction Error */
+#define COMP_TX_ERR	4
+/* TRB Error - some TRB field is invalid */
+#define COMP_TRB_ERR	5
+/* Stall Error - USB device is stalled */
+#define COMP_STALL	6
+/* Resource Error - HC doesn't have memory for that device configuration */
+#define COMP_ENOMEM	7
+/* Bandwidth Error - not enough room in schedule for this dev config */
+#define COMP_BW_ERR	8
+/* No Slots Available Error - HC ran out of device slots */
+#define COMP_ENOSLOTS	9
+/* Invalid Stream Type Error */
+#define COMP_STREAM_ERR	10
+/* Slot Not Enabled Error - doorbell rung for disabled device slot */
+#define COMP_EBADSLT	11
+/* Endpoint Not Enabled Error */
+#define COMP_EBADEP	12
+/* Short Packet */
+#define COMP_SHORT_TX	13
+/* Ring Underrun - doorbell rung for an empty isoc OUT ep ring */
+#define COMP_UNDERRUN	14
+/* Ring Overrun - isoc IN ep ring is empty when ep is scheduled to RX */
+#define COMP_OVERRUN	15
+/* Virtual Function Event Ring Full Error */
+#define COMP_VF_FULL	16
+/* Parameter Error - Context parameter is invalid */
+#define COMP_EINVAL	17
+/* Bandwidth Overrun Error - isoc ep exceeded its allocated bandwidth */
+#define COMP_BW_OVER	18
+/* Context State Error - illegal context state transition requested */
+#define COMP_CTX_STATE	19
+/* No Ping Response Error - HC didn't get PING_RESPONSE in time to TX */
+#define COMP_PING_ERR	20
+/* Event Ring is full */
+#define COMP_ER_FULL	21
+/* Incompatible Device Error */
+#define COMP_DEV_ERR	22
+/* Missed Service Error - HC couldn't service an isoc ep within interval */
+#define COMP_MISSED_INT	23
+/* Successfully stopped command ring */
+#define COMP_CMD_STOP	24
+/* Successfully aborted current command and stopped command ring */
+#define COMP_CMD_ABORT	25
+/* Stopped - transfer was terminated by a stop endpoint command */
+#define COMP_STOP	26
+/* Same as COMP_EP_STOPPED, but the transferred length in the event is invalid */
+#define COMP_STOP_INVAL	27
+/* Control Abort Error - Debug Capability - control pipe aborted */
+#define COMP_DBG_ABORT	28
+/* Max Exit Latency Too Large Error */
+#define COMP_MEL_ERR	29
+/* TRB type 30 reserved */
+/* Isoc Buffer Overrun - an isoc IN ep sent more data than could fit in TD */
+#define COMP_BUFF_OVER	31
+/* Event Lost Error - xHC has an "internal event overrun condition" */
+#define COMP_ISSUES	32
+/* Undefined Error - reported when other error codes don't apply */
+#define COMP_UNKNOWN	33
+/* Invalid Stream ID Error */
+#define COMP_STRID_ERR	34
+/* Secondary Bandwidth Error - may be returned by a Configure Endpoint cmd */
+#define COMP_2ND_BW_ERR	35
+/* Split Transaction Error */
+#define	COMP_SPLIT_ERR	36
+
+struct xhci_link_trb {
+	/* 64-bit segment pointer*/
+	__le64 segment_ptr;
+	__le32 intr_target;
+	__le32 control;
+};
+
+/* control bitfields */
+#define LINK_TOGGLE	(0x1<<1)
+
+/* Command completion event TRB */
+struct xhci_event_cmd {
+	/* Pointer to command TRB, or the value passed by the event data trb */
+	__le64 cmd_trb;
+	__le32 status;
+	__le32 flags;
+};
+
+/* flags bitmasks */
+/* bits 16:23 are the virtual function ID */
+/* bits 24:31 are the slot ID */
+#define TRB_TO_SLOT_ID(p)	(((p) & (0xff<<24)) >> 24)
+#define SLOT_ID_FOR_TRB(p)	(((p) & 0xff) << 24)
+
+/* Stop Endpoint TRB - ep_index to endpoint ID for this TRB */
+#define TRB_TO_EP_INDEX(p)		((((p) & (0x1f << 16)) >> 16) - 1)
+#define	EP_ID_FOR_TRB(p)		((((p) + 1) & 0x1f) << 16)
+
+#define SUSPEND_PORT_FOR_TRB(p)		(((p) & 1) << 23)
+#define TRB_TO_SUSPEND_PORT(p)		(((p) & (1 << 23)) >> 23)
+#define LAST_EP_INDEX			30
+
+/* Set TR Dequeue Pointer command TRB fields */
+#define TRB_TO_STREAM_ID(p)		((((p) & (0xffff << 16)) >> 16))
+#define STREAM_ID_FOR_TRB(p)		((((p)) & 0xffff) << 16)
+
+
+/* Port Status Change Event TRB fields */
+/* Port ID - bits 31:24 */
+#define GET_PORT_ID(p)		(((p) & (0xff << 24)) >> 24)
+
+/* Normal TRB fields */
+/* transfer_len bitmasks - bits 0:16 */
+#define	TRB_LEN(p)		((p) & 0x1ffff)
+/* Interrupter Target - which MSI-X vector to target the completion event at */
+#define TRB_INTR_TARGET(p)	(((p) & 0x3ff) << 22)
+#define GET_INTR_TARGET(p)	(((p) >> 22) & 0x3ff)
+#define TRB_TBC(p)		(((p) & 0x3) << 7)
+#define TRB_TLBPC(p)		(((p) & 0xf) << 16)
+
+/* Cycle bit - indicates TRB ownership by HC or HCD */
+#define TRB_CYCLE		(1<<0)
+/*
+ * Force next event data TRB to be evaluated before task switch.
+ * Used to pass OS data back after a TD completes.
+ */
+#define TRB_ENT			(1<<1)
+/* Interrupt on short packet */
+#define TRB_ISP			(1<<2)
+/* Set PCIe no snoop attribute */
+#define TRB_NO_SNOOP		(1<<3)
+/* Chain multiple TRBs into a TD */
+#define TRB_CHAIN		(1<<4)
+/* Interrupt on completion */
+#define TRB_IOC			(1<<5)
+/* The buffer pointer contains immediate data */
+#define TRB_IDT			(1<<6)
+
+/* Block Event Interrupt */
+#define	TRB_BEI			(1<<9)
+
+/* Control transfer TRB specific fields */
+#define TRB_DIR_IN		(1<<16)
+#define	TRB_TX_TYPE(p)		((p) << 16)
+#define	TRB_DATA_OUT		2
+#define	TRB_DATA_IN		3
+
+/* Isochronous TRB specific fields */
+#define TRB_SIA			(1<<31)
+
+struct xhci_generic_trb {
+	__le32 field[4];
+};
+
+union xhci_trb {
+	struct xhci_link_trb		link;
+	struct xhci_transfer_event	trans_event;
+	struct xhci_event_cmd		event_cmd;
+	struct xhci_generic_trb		generic;
+};
+
+/* TRB bit mask */
+#define	TRB_TYPE_BITMASK	(0xfc00)
+#define TRB_TYPE(p)		((p) << 10)
+#define TRB_FIELD_TO_TYPE(p)	(((p) & TRB_TYPE_BITMASK) >> 10)
+/* TRB type IDs */
+/* bulk, interrupt, isoc scatter/gather, and control data stage */
+#define TRB_NORMAL		1
+/* setup stage for control transfers */
+#define TRB_SETUP		2
+/* data stage for control transfers */
+#define TRB_DATA		3
+/* status stage for control transfers */
+#define TRB_STATUS		4
+/* isoc transfers */
+#define TRB_ISOC		5
+/* TRB for linking ring segments */
+#define TRB_LINK		6
+#define TRB_EVENT_DATA		7
+/* Transfer Ring No-op (not for the command ring) */
+#define TRB_TR_NOOP		8
+/* Command TRBs */
+/* Enable Slot Command */
+#define TRB_ENABLE_SLOT		9
+/* Disable Slot Command */
+#define TRB_DISABLE_SLOT	10
+/* Address Device Command */
+#define TRB_ADDR_DEV		11
+/* Configure Endpoint Command */
+#define TRB_CONFIG_EP		12
+/* Evaluate Context Command */
+#define TRB_EVAL_CONTEXT	13
+/* Reset Endpoint Command */
+#define TRB_RESET_EP		14
+/* Stop Transfer Ring Command */
+#define TRB_STOP_RING		15
+/* Set Transfer Ring Dequeue Pointer Command */
+#define TRB_SET_DEQ		16
+/* Reset Device Command */
+#define TRB_RESET_DEV		17
+/* Force Event Command (opt) */
+#define TRB_FORCE_EVENT		18
+/* Negotiate Bandwidth Command (opt) */
+#define TRB_NEG_BANDWIDTH	19
+/* Set Latency Tolerance Value Command (opt) */
+#define TRB_SET_LT		20
+/* Get port bandwidth Command */
+#define TRB_GET_BW		21
+/* Force Header Command - generate a transaction or link management packet */
+#define TRB_FORCE_HEADER	22
+/* No-op Command - not for transfer rings */
+#define TRB_CMD_NOOP		23
+/* TRB IDs 24-31 reserved */
+/* Event TRBS */
+/* Transfer Event */
+#define TRB_TRANSFER		32
+/* Command Completion Event */
+#define TRB_COMPLETION		33
+/* Port Status Change Event */
+#define TRB_PORT_STATUS		34
+/* Bandwidth Request Event (opt) */
+#define TRB_BANDWIDTH_EVENT	35
+/* Doorbell Event (opt) */
+#define TRB_DOORBELL		36
+/* Host Controller Event */
+#define TRB_HC_EVENT		37
+/* Device Notification Event - device sent function wake notification */
+#define TRB_DEV_NOTE		38
+/* MFINDEX Wrap Event - microframe counter wrapped */
+#define TRB_MFINDEX_WRAP	39
+/* TRB IDs 40-47 reserved, 48-63 is vendor-defined */
+
+/* Nec vendor-specific command completion event. */
+#define	TRB_NEC_CMD_COMP	48
+/* Get NEC firmware revision. */
+#define	TRB_NEC_GET_FW		49
+
+#define TRB_TYPE_LINK(x)	(((x) & TRB_TYPE_BITMASK) == TRB_TYPE(TRB_LINK))
+/* Above, but for __le32 types -- can avoid work by swapping constants: */
+#define TRB_TYPE_LINK_LE32(x)	(((x) & cpu_to_le32(TRB_TYPE_BITMASK)) == \
+				 cpu_to_le32(TRB_TYPE(TRB_LINK)))
+#define TRB_TYPE_NOOP_LE32(x)	(((x) & cpu_to_le32(TRB_TYPE_BITMASK)) == \
+				 cpu_to_le32(TRB_TYPE(TRB_TR_NOOP)))
+
+#define NEC_FW_MINOR(p)		(((p) >> 0) & 0xff)
+#define NEC_FW_MAJOR(p)		(((p) >> 8) & 0xff)
+
+/*
+ * TRBS_PER_SEGMENT must be a multiple of 4,
+ * since the command ring is 64-byte aligned.
+ * It must also be greater than 16.
+ */
+#define TRBS_PER_SEGMENT	64
+/* Allow two commands + a link TRB, along with any reserved command TRBs */
+#define MAX_RSVD_CMD_TRBS	(TRBS_PER_SEGMENT - 3)
+#define TRB_SEGMENT_SIZE	(TRBS_PER_SEGMENT*16)
+#define TRB_SEGMENT_SHIFT	(ilog2(TRB_SEGMENT_SIZE))
+/* TRB buffer pointers can't cross 64KB boundaries */
+#define TRB_MAX_BUFF_SHIFT		16
+#define TRB_MAX_BUFF_SIZE	(1 << TRB_MAX_BUFF_SHIFT)
+
+struct xhci_segment {
+	union xhci_trb		*trbs;
+	/* private to HCD */
+	struct xhci_segment	*next;
+	dma_addr_t		dma;
+};
+
+struct xhci_td {
+	struct list_head	td_list;
+	struct list_head	cancelled_td_list;
+	struct urb		*urb;
+	struct xhci_segment	*start_seg;
+	union xhci_trb		*first_trb;
+	union xhci_trb		*last_trb;
+};
+
+/* xHCI command default timeout value */
+#define XHCI_CMD_DEFAULT_TIMEOUT	(5 * HZ)
+
+/* command descriptor */
+struct xhci_cd {
+	struct list_head	cancel_cmd_list;
+	struct xhci_command	*command;
+	union xhci_trb		*cmd_trb;
+};
+
+struct xhci_dequeue_state {
+	struct xhci_segment *new_deq_seg;
+	union xhci_trb *new_deq_ptr;
+	int new_cycle_state;
+};
+
+enum xhci_ring_type {
+	TYPE_CTRL = 0,
+	TYPE_ISOC,
+	TYPE_BULK,
+	TYPE_INTR,
+	TYPE_STREAM,
+	TYPE_COMMAND,
+	TYPE_EVENT,
+};
+
+struct xhci_ring {
+	struct xhci_segment	*first_seg;
+	struct xhci_segment	*last_seg;
+	union  xhci_trb		*enqueue;
+	struct xhci_segment	*enq_seg;
+	unsigned int		enq_updates;
+	union  xhci_trb		*dequeue;
+	struct xhci_segment	*deq_seg;
+	unsigned int		deq_updates;
+	struct list_head	td_list;
+	/*
+	 * Write the cycle state into the TRB cycle field to give ownership of
+	 * the TRB to the host controller (if we are the producer), or to check
+	 * if we own the TRB (if we are the consumer).  See section 4.9.1.
+	 */
+	u32			cycle_state;
+	unsigned int		stream_id;
+	unsigned int		num_segs;
+	unsigned int		num_trbs_free;
+	unsigned int		num_trbs_free_temp;
+	enum xhci_ring_type	type;
+	bool			last_td_was_short;
+};
+
+struct xhci_erst_entry {
+	/* 64-bit event ring segment address */
+	__le64	seg_addr;
+	__le32	seg_size;
+	/* Set to zero */
+	__le32	rsvd;
+};
+
+struct xhci_erst {
+	struct xhci_erst_entry	*entries;
+	unsigned int		num_entries;
+	/* xhci->event_ring keeps track of segment dma addresses */
+	dma_addr_t		erst_dma_addr;
+	/* Num entries the ERST can contain */
+	unsigned int		erst_size;
+};
+
+struct xhci_scratchpad {
+	u64 *sp_array;
+	dma_addr_t sp_dma;
+	void **sp_buffers;
+	dma_addr_t *sp_dma_buffers;
+};
+
+struct urb_priv {
+	int	length;
+	int	td_cnt;
+	struct	xhci_td	*td[0];
+};
+
+/*
+ * 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,
+ * meaning 64 ring segments.
+ * Initial allocated size of the ERST, in number of entries */
+#define	ERST_NUM_SEGS	1
+/* Initial allocated size of the ERST, in number of entries */
+#define	ERST_SIZE	64
+/* Initial number of event segment rings allocated */
+#define	ERST_ENTRIES	1
+/* Poll every 60 seconds */
+#define	POLL_TIMEOUT	60
+/* Stop endpoint command timeout (secs) for URB cancellation watchdog timer */
+#define XHCI_STOP_EP_CMD_TIMEOUT	5
+/* XXX: Make these module parameters */
+
+struct s3_save {
+	u32	command;
+	u32	dev_nt;
+	u64	dcbaa_ptr;
+	u32	config_reg;
+	u32	irq_pending;
+	u32	irq_control;
+	u32	erst_size;
+	u64	erst_base;
+	u64	erst_dequeue;
+};
+
+/* Use for lpm */
+struct dev_info {
+	u32			dev_id;
+	struct	list_head	list;
+};
+
+struct xhci_bus_state {
+	unsigned long		bus_suspended;
+	unsigned long		next_statechange;
+
+	/* Port suspend arrays are indexed by the portnum of the fake roothub */
+	/* ports suspend status arrays - max 31 ports for USB2, 15 for USB3 */
+	u32			port_c_suspend;
+	u32			suspended_ports;
+	u32			port_remote_wakeup;
+	unsigned long		resume_done[USB_MAXCHILDREN];
+	/* which ports have started to resume */
+	unsigned long		resuming_ports;
+};
+
+static inline unsigned int hcd_index(struct usb_hcd *hcd)
+{
+	if (hcd->speed == HCD_USB3)
+		return 0;
+	else
+		return 1;
+}
+
+/* There is one xhci_hcd structure per controller */
+struct xhci_hcd {
+	struct usb_hcd *main_hcd;
+	struct usb_hcd *shared_hcd;
+	/* glue to PCI and HCD framework */
+	struct xhci_cap_regs __iomem *cap_regs;
+	struct xhci_op_regs __iomem *op_regs;
+	struct xhci_run_regs __iomem *run_regs;
+	struct xhci_doorbell_array __iomem *dba;
+	/* Our HCD's current interrupter register set */
+	struct	xhci_intr_reg __iomem *ir_set;
+
+	/* Cached register copies of read-only HC data */
+	__u32		hcs_params1;
+	__u32		hcs_params2;
+	__u32		hcs_params3;
+	__u32		hcc_params;
+
+	spinlock_t	lock;
+
+	/* packed release number */
+	u8		sbrn;
+	u16		hci_version;
+	u8		max_slots;
+	u8		max_interrupters;
+	u8		max_ports;
+	u8		isoc_threshold;
+	int		event_ring_max;
+	int		addr_64;
+	/* 4KB min, 128MB max */
+	int		page_size;
+	/* Valid values are 12 to 20, inclusive */
+	int		page_shift;
+	/* msi-x vectors */
+	int		msix_count;
+	struct msix_entry	*msix_entries;
+	/* data structures */
+	struct xhci_device_context_array *dcbaa;
+	struct xhci_ring	*cmd_ring;
+	unsigned int            cmd_ring_state;
+#define CMD_RING_STATE_RUNNING         (1 << 0)
+#define CMD_RING_STATE_ABORTED         (1 << 1)
+#define CMD_RING_STATE_STOPPED         (1 << 2)
+	struct list_head        cancel_cmd_list;
+	unsigned int		cmd_ring_reserved_trbs;
+	struct xhci_ring	*event_ring;
+	struct xhci_erst	erst;
+	/* Scratchpad */
+	struct xhci_scratchpad  *scratchpad;
+	/* Store LPM test failed devices' information */
+	struct list_head	lpm_failed_devs;
+
+	/* slot enabling and address device helpers */
+	struct completion	addr_dev;
+	int slot_id;
+	/* For USB 3.0 LPM enable/disable. */
+	struct xhci_command		*lpm_command;
+	/* Internal mirror of the HW's dcbaa */
+	struct xhci_virt_device	*devs[MAX_HC_SLOTS];
+	/* For keeping track of bandwidth domains per roothub. */
+	struct xhci_root_port_bw_info	*rh_bw;
+
+	/* DMA pools */
+	struct dma_pool	*device_pool;
+	struct dma_pool	*segment_pool;
+	struct dma_pool	*small_streams_pool;
+	struct dma_pool	*medium_streams_pool;
+
+#ifdef CONFIG_USB_XHCI_HCD_DEBUGGING
+	/* Poll the rings - for debugging */
+	struct timer_list	event_ring_timer;
+	int			zombie;
+#endif
+	/* Host controller watchdog timer structures */
+	unsigned int		xhc_state;
+
+	u32			command;
+	struct s3_save		s3;
+/* Host controller is dying - not responding to commands. "I'm not dead yet!"
+ *
+ * xHC interrupts have been disabled and a watchdog timer will (or has already)
+ * halt the xHCI host, and complete all URBs with an -ESHUTDOWN code.  Any code
+ * that sees this status (other than the timer that set it) should stop touching
+ * hardware immediately.  Interrupt handlers should return immediately when
+ * they see this status (any time they drop and re-acquire xhci->lock).
+ * xhci_urb_dequeue() should call usb_hcd_check_unlink_urb() and return without
+ * putting the TD on the canceled list, etc.
+ *
+ * There are no reports of xHCI host controllers that display this issue.
+ */
+#define XHCI_STATE_DYING	(1 << 0)
+#define XHCI_STATE_HALTED	(1 << 1)
+	/* Statistics */
+	int			error_bitmask;
+	unsigned int		quirks;
+#define	XHCI_LINK_TRB_QUIRK	(1 << 0)
+#define XHCI_RESET_EP_QUIRK	(1 << 1)
+#define XHCI_NEC_HOST		(1 << 2)
+#define XHCI_AMD_PLL_FIX	(1 << 3)
+#define XHCI_SPURIOUS_SUCCESS	(1 << 4)
+/*
+ * Certain Intel host controllers have a limit to the number of endpoint
+ * contexts they can handle.  Ideally, they would signal that they can't handle
+ * anymore endpoint contexts by returning a Resource Error for the Configure
+ * Endpoint command, but they don't.  Instead they expect software to keep track
+ * of the number of active endpoints for them, across configure endpoint
+ * commands, reset device commands, disable slot commands, and address device
+ * commands.
+ */
+#define XHCI_EP_LIMIT_QUIRK	(1 << 5)
+#define XHCI_BROKEN_MSI		(1 << 6)
+#define XHCI_RESET_ON_RESUME	(1 << 7)
+#define	XHCI_SW_BW_CHECKING	(1 << 8)
+#define XHCI_AMD_0x96_HOST	(1 << 9)
+#define XHCI_TRUST_TX_LENGTH	(1 << 10)
+#define XHCI_LPM_SUPPORT	(1 << 11)
+#define XHCI_INTEL_HOST		(1 << 12)
+#define XHCI_SPURIOUS_REBOOT	(1 << 13)
+#define XHCI_COMP_MODE_QUIRK	(1 << 14)
+#define XHCI_AVOID_BEI		(1 << 15)
+	unsigned int		num_active_eps;
+	unsigned int		limit_active_eps;
+	/* There are two roothubs to keep track of bus suspend info for */
+	struct xhci_bus_state   bus_state[2];
+	/* Is each xHCI roothub port a USB 3.0, USB 2.0, or USB 1.1 port? */
+	u8			*port_array;
+	/* Array of pointers to USB 3.0 PORTSC registers */
+	__le32 __iomem		**usb3_ports;
+	unsigned int		num_usb3_ports;
+	/* Array of pointers to USB 2.0 PORTSC registers */
+	__le32 __iomem		**usb2_ports;
+	unsigned int		num_usb2_ports;
+	/* support xHCI 0.96 spec USB2 software LPM */
+	unsigned		sw_lpm_support:1;
+	/* support xHCI 1.0 spec USB2 hardware LPM */
+	unsigned		hw_lpm_support:1;
+	/* Compliance Mode Recovery Data */
+	struct timer_list	comp_mode_recovery_timer;
+	u32			port_status_u0;
+/* Compliance Mode Timer Triggered every 2 seconds */
+#define COMP_MODE_RCVRY_MSECS 2000
+};
+
+/* convert between an HCD pointer and the corresponding EHCI_HCD */
+static inline struct xhci_hcd *hcd_to_xhci(struct usb_hcd *hcd)
+{
+	return *((struct xhci_hcd **) (hcd->hcd_priv));
+}
+
+static inline struct usb_hcd *xhci_to_hcd(struct xhci_hcd *xhci)
+{
+	return xhci->main_hcd;
+}
+
+#ifdef CONFIG_USB_XHCI_HCD_DEBUGGING
+#define XHCI_DEBUG	1
+#else
+#define XHCI_DEBUG	0
+#endif
+
+#define xhci_dbg(xhci, fmt, args...) \
+	do { if (XHCI_DEBUG) dev_dbg(xhci_to_hcd(xhci)->self.controller , fmt , ## args); } while (0)
+#define xhci_info(xhci, fmt, args...) \
+	do { if (XHCI_DEBUG) dev_info(xhci_to_hcd(xhci)->self.controller , fmt , ## args); } while (0)
+#define xhci_err(xhci, fmt, args...) \
+	dev_err(xhci_to_hcd(xhci)->self.controller , fmt , ## args)
+#define xhci_warn(xhci, fmt, args...) \
+	dev_warn(xhci_to_hcd(xhci)->self.controller , fmt , ## args)
+#define xhci_warn_ratelimited(xhci, fmt, args...) \
+	dev_warn_ratelimited(xhci_to_hcd(xhci)->self.controller , fmt , ## args)
+
+/* TODO: copied from ehci.h - can be refactored? */
+/* xHCI spec says all registers are little endian */
+static inline unsigned int xhci_readl(const struct xhci_hcd *xhci,
+		__le32 __iomem *regs)
+{
+	return readl(regs);
+}
+static inline void xhci_writel(struct xhci_hcd *xhci,
+		const unsigned int val, __le32 __iomem *regs)
+{
+	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,
+		__le64 __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, __le64 __iomem *regs)
+{
+	__u32 __iomem *ptr = (__u32 __iomem *) regs;
+	u32 val_lo = lower_32_bits(val);
+	u32 val_hi = upper_32_bits(val);
+
+	writel(val_lo, ptr);
+	writel(val_hi, ptr + 1);
+}
+
+static inline int xhci_link_trb_quirk(struct xhci_hcd *xhci)
+{
+	return xhci->quirks & XHCI_LINK_TRB_QUIRK;
+}
+
+/* xHCI debugging */
+void xhci_print_ir_set(struct xhci_hcd *xhci, int set_num);
+void xhci_print_registers(struct xhci_hcd *xhci);
+void xhci_dbg_regs(struct xhci_hcd *xhci);
+void xhci_print_run_regs(struct xhci_hcd *xhci);
+void xhci_print_trb_offsets(struct xhci_hcd *xhci, union xhci_trb *trb);
+void xhci_debug_trb(struct xhci_hcd *xhci, union xhci_trb *trb);
+void xhci_debug_segment(struct xhci_hcd *xhci, struct xhci_segment *seg);
+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_container_ctx *ctx, unsigned int last_ep);
+char *xhci_get_slot_state(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *ctx);
+void xhci_dbg_ep_rings(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		struct xhci_virt_ep *ep);
+
+/* xHCI memory management */
+void xhci_mem_cleanup(struct xhci_hcd *xhci);
+int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags);
+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);
+int xhci_setup_addressable_virt_dev(struct xhci_hcd *xhci, struct usb_device *udev);
+void xhci_copy_ep0_dequeue_into_input_ctx(struct xhci_hcd *xhci,
+		struct usb_device *udev);
+unsigned int xhci_get_endpoint_index(struct usb_endpoint_descriptor *desc);
+unsigned int xhci_get_endpoint_flag(struct usb_endpoint_descriptor *desc);
+unsigned int xhci_get_endpoint_flag_from_index(unsigned int ep_index);
+unsigned int xhci_last_valid_endpoint(u32 added_ctxs);
+void xhci_endpoint_zero(struct xhci_hcd *xhci, struct xhci_virt_device *virt_dev, struct usb_host_endpoint *ep);
+void xhci_drop_ep_from_interval_table(struct xhci_hcd *xhci,
+		struct xhci_bw_info *ep_bw,
+		struct xhci_interval_bw_table *bw_table,
+		struct usb_device *udev,
+		struct xhci_virt_ep *virt_ep,
+		struct xhci_tt_bw_info *tt_info);
+void xhci_update_tt_active_eps(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		int old_active_eps);
+void xhci_clear_endpoint_bw_info(struct xhci_bw_info *bw_info);
+void xhci_update_bw_info(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx,
+		struct xhci_input_control_ctx *ctrl_ctx,
+		struct xhci_virt_device *virt_dev);
+void xhci_endpoint_copy(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx,
+		struct xhci_container_ctx *out_ctx,
+		unsigned int ep_index);
+void xhci_slot_copy(struct xhci_hcd *xhci,
+		struct xhci_container_ctx *in_ctx,
+		struct xhci_container_ctx *out_ctx);
+int xhci_endpoint_init(struct xhci_hcd *xhci, struct xhci_virt_device *virt_dev,
+		struct usb_device *udev, struct usb_host_endpoint *ep,
+		gfp_t mem_flags);
+void xhci_ring_free(struct xhci_hcd *xhci, struct xhci_ring *ring);
+int xhci_ring_expansion(struct xhci_hcd *xhci, struct xhci_ring *ring,
+				unsigned int num_trbs, gfp_t flags);
+void xhci_free_or_cache_endpoint_ring(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		unsigned int ep_index);
+struct xhci_stream_info *xhci_alloc_stream_info(struct xhci_hcd *xhci,
+		unsigned int num_stream_ctxs,
+		unsigned int num_streams, gfp_t flags);
+void xhci_free_stream_info(struct xhci_hcd *xhci,
+		struct xhci_stream_info *stream_info);
+void xhci_setup_streams_ep_input_ctx(struct xhci_hcd *xhci,
+		struct xhci_ep_ctx *ep_ctx,
+		struct xhci_stream_info *stream_info);
+void xhci_setup_no_streams_ep_input_ctx(struct xhci_hcd *xhci,
+		struct xhci_ep_ctx *ep_ctx,
+		struct xhci_virt_ep *ep);
+void xhci_free_device_endpoint_resources(struct xhci_hcd *xhci,
+	struct xhci_virt_device *virt_dev, bool drop_control_ep);
+struct xhci_ring *xhci_dma_to_transfer_ring(
+		struct xhci_virt_ep *ep,
+		u64 address);
+struct xhci_ring *xhci_stream_id_to_ring(
+		struct xhci_virt_device *dev,
+		unsigned int ep_index,
+		unsigned int stream_id);
+struct xhci_command *xhci_alloc_command(struct xhci_hcd *xhci,
+		bool allocate_in_ctx, bool allocate_completion,
+		gfp_t mem_flags);
+void xhci_urb_free_priv(struct xhci_hcd *xhci, struct urb_priv *urb_priv);
+void xhci_free_command(struct xhci_hcd *xhci,
+		struct xhci_command *command);
+
+#ifdef CONFIG_PCI
+/* xHCI PCI glue */
+int xhci_register_pci(void);
+void xhci_unregister_pci(void);
+#else
+static inline int xhci_register_pci(void) { return 0; }
+static inline void xhci_unregister_pci(void) {}
+#endif
+
+#if defined(CONFIG_USB_XHCI_PLATFORM) \
+	|| defined(CONFIG_USB_XHCI_PLATFORM_MODULE)
+int xhci_register_plat(void);
+void xhci_unregister_plat(void);
+#else
+static inline int xhci_register_plat(void)
+{ return 0; }
+static inline void xhci_unregister_plat(void)
+{  }
+#endif
+
+/* xHCI host controller glue */
+typedef void (*xhci_get_quirks_t)(struct device *, struct xhci_hcd *);
+int xhci_handshake(struct xhci_hcd *xhci, void __iomem *ptr,
+		u32 mask, u32 done, int usec);
+void xhci_quiesce(struct xhci_hcd *xhci);
+int xhci_halt(struct xhci_hcd *xhci);
+int xhci_reset(struct xhci_hcd *xhci);
+int xhci_init(struct usb_hcd *hcd);
+int xhci_run(struct usb_hcd *hcd);
+void xhci_stop(struct usb_hcd *hcd);
+void xhci_shutdown(struct usb_hcd *hcd);
+int xhci_gen_setup(struct usb_hcd *hcd, xhci_get_quirks_t get_quirks);
+
+#ifdef	CONFIG_PM
+int xhci_suspend(struct xhci_hcd *xhci);
+int xhci_resume(struct xhci_hcd *xhci, bool hibernated);
+#else
+#define	xhci_suspend	NULL
+#define	xhci_resume	NULL
+#endif
+
+int xhci_get_frame(struct usb_hcd *hcd);
+irqreturn_t xhci_irq(struct usb_hcd *hcd);
+irqreturn_t xhci_msi_irq(int irq, struct usb_hcd *hcd);
+int xhci_alloc_dev(struct usb_hcd *hcd, struct usb_device *udev);
+void xhci_free_dev(struct usb_hcd *hcd, struct usb_device *udev);
+int xhci_alloc_tt_info(struct xhci_hcd *xhci,
+		struct xhci_virt_device *virt_dev,
+		struct usb_device *hdev,
+		struct usb_tt *tt, gfp_t mem_flags);
+int xhci_alloc_streams(struct usb_hcd *hcd, struct usb_device *udev,
+		struct usb_host_endpoint **eps, unsigned int num_eps,
+		unsigned int num_streams, gfp_t mem_flags);
+int xhci_free_streams(struct usb_hcd *hcd, struct usb_device *udev,
+		struct usb_host_endpoint **eps, unsigned int num_eps,
+		gfp_t mem_flags);
+int xhci_address_device(struct usb_hcd *hcd, struct usb_device *udev);
+int xhci_update_device(struct usb_hcd *hcd, struct usb_device *udev);
+int xhci_set_usb2_hardware_lpm(struct usb_hcd *hcd,
+				struct usb_device *udev, int enable);
+int xhci_update_hub_device(struct usb_hcd *hcd, struct usb_device *hdev,
+			struct usb_tt *tt, gfp_t mem_flags);
+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_discover_or_reset_device(struct usb_hcd *hcd, struct usb_device *udev);
+int xhci_check_bandwidth(struct usb_hcd *hcd, struct usb_device *udev);
+void xhci_reset_bandwidth(struct usb_hcd *hcd, struct usb_device *udev);
+
+/* xHCI ring, segment, TRB, and TD functions */
+dma_addr_t xhci_trb_virt_to_dma(struct xhci_segment *seg, union xhci_trb *trb);
+struct xhci_segment *trb_in_td(struct xhci_segment *start_seg,
+		union xhci_trb *start_trb, union xhci_trb *end_trb,
+		dma_addr_t suspect_dma);
+int xhci_is_vendor_info_code(struct xhci_hcd *xhci, unsigned int trb_comp_code);
+void xhci_ring_cmd_db(struct xhci_hcd *xhci);
+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);
+int xhci_queue_vendor_command(struct xhci_hcd *xhci,
+		u32 field1, u32 field2, u32 field3, u32 field4);
+int xhci_queue_stop_endpoint(struct xhci_hcd *xhci, int slot_id,
+		unsigned int ep_index, int suspend);
+int xhci_queue_ctrl_tx(struct xhci_hcd *xhci, gfp_t mem_flags, struct urb *urb,
+		int slot_id, unsigned int ep_index);
+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_intr_tx(struct xhci_hcd *xhci, gfp_t mem_flags, struct urb *urb,
+		int slot_id, unsigned int ep_index);
+int xhci_queue_isoc_tx_prepare(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, bool command_must_succeed);
+int xhci_queue_evaluate_context(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
+		u32 slot_id, bool command_must_succeed);
+int xhci_queue_reset_ep(struct xhci_hcd *xhci, int slot_id,
+		unsigned int ep_index);
+int xhci_queue_reset_device(struct xhci_hcd *xhci, u32 slot_id);
+void xhci_find_new_dequeue_state(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		unsigned int stream_id, struct xhci_td *cur_td,
+		struct xhci_dequeue_state *state);
+void xhci_queue_new_dequeue_state(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		unsigned int stream_id,
+		struct xhci_dequeue_state *deq_state);
+void xhci_cleanup_stalled_ring(struct xhci_hcd *xhci,
+		struct usb_device *udev, unsigned int ep_index);
+void xhci_queue_config_ep_quirk(struct xhci_hcd *xhci,
+		unsigned int slot_id, unsigned int ep_index,
+		struct xhci_dequeue_state *deq_state);
+void xhci_stop_endpoint_command_watchdog(unsigned long arg);
+int xhci_cancel_cmd(struct xhci_hcd *xhci, struct xhci_command *command,
+		union xhci_trb *cmd_trb);
+void xhci_ring_ep_doorbell(struct xhci_hcd *xhci, unsigned int slot_id,
+		unsigned int ep_index, unsigned int stream_id);
+
+/* xHCI roothub code */
+void xhci_set_link_state(struct xhci_hcd *xhci, __le32 __iomem **port_array,
+				int port_id, u32 link_state);
+int xhci_enable_usb3_lpm_timeout(struct usb_hcd *hcd,
+			struct usb_device *udev, enum usb3_link_state state);
+int xhci_disable_usb3_lpm_timeout(struct usb_hcd *hcd,
+			struct usb_device *udev, enum usb3_link_state state);
+void xhci_test_and_clear_bit(struct xhci_hcd *xhci, __le32 __iomem **port_array,
+				int port_id, u32 port_bit);
+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);
+int xhci_find_raw_port_number(struct usb_hcd *hcd, int port1);
+
+#ifdef CONFIG_PM
+int xhci_bus_suspend(struct usb_hcd *hcd);
+int xhci_bus_resume(struct usb_hcd *hcd);
+#else
+#define	xhci_bus_suspend	NULL
+#define	xhci_bus_resume		NULL
+#endif	/* CONFIG_PM */
+
+u32 xhci_port_state_to_neutral(u32 state);
+int xhci_find_slot_id_by_port(struct usb_hcd *hcd, struct xhci_hcd *xhci,
+		u16 port);
+void xhci_ring_device(struct xhci_hcd *xhci, int slot_id);
+
+/* 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 */
diff --git a/include/linux/usb/ch11.h b/include/linux/usb/ch11.h
new file mode 100644
index 0000000..7692dc6
--- /dev/null
+++ b/include/linux/usb/ch11.h
@@ -0,0 +1,266 @@
+/*
+ * This file holds Hub protocol constants and data structures that are
+ * defined in chapter 11 (Hub Specification) of the USB 2.0 specification.
+ *
+ * It is used/shared between the USB core, the HCDs and couple of other USB
+ * drivers.
+ */
+
+#ifndef __LINUX_CH11_H
+#define __LINUX_CH11_H
+
+#include <linux/types.h>	/* __u8 etc */
+
+/*
+ * Hub request types
+ */
+
+#define USB_RT_HUB	(USB_TYPE_CLASS | USB_RECIP_DEVICE)
+#define USB_RT_PORT	(USB_TYPE_CLASS | USB_RECIP_OTHER)
+
+/*
+ * Hub class requests
+ * See USB 2.0 spec Table 11-16
+ */
+#define HUB_CLEAR_TT_BUFFER	8
+#define HUB_RESET_TT		9
+#define HUB_GET_TT_STATE	10
+#define HUB_STOP_TT		11
+
+/*
+ * Hub class additional requests defined by USB 3.0 spec
+ * See USB 3.0 spec Table 10-6
+ */
+#define HUB_SET_DEPTH		12
+#define HUB_GET_PORT_ERR_COUNT	13
+
+/*
+ * Hub Class feature numbers
+ * See USB 2.0 spec Table 11-17
+ */
+#define C_HUB_LOCAL_POWER	0
+#define C_HUB_OVER_CURRENT	1
+
+/*
+ * Port feature numbers
+ * See USB 2.0 spec Table 11-17
+ */
+#define USB_PORT_FEAT_CONNECTION	0
+#define USB_PORT_FEAT_ENABLE		1
+#define USB_PORT_FEAT_SUSPEND		2	/* L2 suspend */
+#define USB_PORT_FEAT_OVER_CURRENT	3
+#define USB_PORT_FEAT_RESET		4
+#define USB_PORT_FEAT_L1		5	/* L1 suspend */
+#define USB_PORT_FEAT_POWER		8
+#define USB_PORT_FEAT_LOWSPEED		9	/* Should never be used */
+#define USB_PORT_FEAT_C_CONNECTION	16
+#define USB_PORT_FEAT_C_ENABLE		17
+#define USB_PORT_FEAT_C_SUSPEND		18
+#define USB_PORT_FEAT_C_OVER_CURRENT	19
+#define USB_PORT_FEAT_C_RESET		20
+#define USB_PORT_FEAT_TEST              21
+#define USB_PORT_FEAT_INDICATOR         22
+#define USB_PORT_FEAT_C_PORT_L1         23
+
+/*
+ * Port feature selectors added by USB 3.0 spec.
+ * See USB 3.0 spec Table 10-7
+ */
+#define USB_PORT_FEAT_LINK_STATE		5
+#define USB_PORT_FEAT_U1_TIMEOUT		23
+#define USB_PORT_FEAT_U2_TIMEOUT		24
+#define USB_PORT_FEAT_C_PORT_LINK_STATE		25
+#define USB_PORT_FEAT_C_PORT_CONFIG_ERROR	26
+#define USB_PORT_FEAT_REMOTE_WAKE_MASK		27
+#define USB_PORT_FEAT_BH_PORT_RESET		28
+#define USB_PORT_FEAT_C_BH_PORT_RESET		29
+#define USB_PORT_FEAT_FORCE_LINKPM_ACCEPT	30
+
+#define USB_PORT_LPM_TIMEOUT(p)			(((p) & 0xff) << 8)
+
+/* USB 3.0 hub remote wake mask bits, see table 10-14 */
+#define USB_PORT_FEAT_REMOTE_WAKE_CONNECT	(1 << 8)
+#define USB_PORT_FEAT_REMOTE_WAKE_DISCONNECT	(1 << 9)
+#define USB_PORT_FEAT_REMOTE_WAKE_OVER_CURRENT	(1 << 10)
+
+/*
+ * Hub Status and Hub Change results
+ * See USB 2.0 spec Table 11-19 and Table 11-20
+ */
+struct usb_port_status {
+	__le16 wPortStatus;
+	__le16 wPortChange;
+} __attribute__ ((packed));
+
+/*
+ * wPortStatus bit field
+ * See USB 2.0 spec Table 11-21
+ */
+#define USB_PORT_STAT_CONNECTION	0x0001
+#define USB_PORT_STAT_ENABLE		0x0002
+#define USB_PORT_STAT_SUSPEND		0x0004
+#define USB_PORT_STAT_OVERCURRENT	0x0008
+#define USB_PORT_STAT_RESET		0x0010
+#define USB_PORT_STAT_L1		0x0020
+/* bits 6 to 7 are reserved */
+#define USB_PORT_STAT_POWER		0x0100
+#define USB_PORT_STAT_LOW_SPEED		0x0200
+#define USB_PORT_STAT_HIGH_SPEED        0x0400
+#define USB_PORT_STAT_TEST              0x0800
+#define USB_PORT_STAT_INDICATOR         0x1000
+/* bits 13 to 15 are reserved */
+
+/*
+ * Additions to wPortStatus bit field from USB 3.0
+ * See USB 3.0 spec Table 10-10
+ */
+#define USB_PORT_STAT_LINK_STATE	0x01e0
+#define USB_SS_PORT_STAT_POWER		0x0200
+#define USB_SS_PORT_STAT_SPEED		0x1c00
+#define USB_PORT_STAT_SPEED_5GBPS	0x0000
+/* Valid only if port is enabled */
+/* Bits that are the same from USB 2.0 */
+#define USB_SS_PORT_STAT_MASK (USB_PORT_STAT_CONNECTION |	    \
+				USB_PORT_STAT_ENABLE |	    \
+				USB_PORT_STAT_OVERCURRENT | \
+				USB_PORT_STAT_RESET)
+
+/*
+ * Definitions for PORT_LINK_STATE values
+ * (bits 5-8) in wPortStatus
+ */
+#define USB_SS_PORT_LS_U0		0x0000
+#define USB_SS_PORT_LS_U1		0x0020
+#define USB_SS_PORT_LS_U2		0x0040
+#define USB_SS_PORT_LS_U3		0x0060
+#define USB_SS_PORT_LS_SS_DISABLED	0x0080
+#define USB_SS_PORT_LS_RX_DETECT	0x00a0
+#define USB_SS_PORT_LS_SS_INACTIVE	0x00c0
+#define USB_SS_PORT_LS_POLLING		0x00e0
+#define USB_SS_PORT_LS_RECOVERY		0x0100
+#define USB_SS_PORT_LS_HOT_RESET	0x0120
+#define USB_SS_PORT_LS_COMP_MOD		0x0140
+#define USB_SS_PORT_LS_LOOPBACK		0x0160
+
+/*
+ * wPortChange bit field
+ * See USB 2.0 spec Table 11-22 and USB 2.0 LPM ECN Table-4.10
+ * Bits 0 to 5 shown, bits 6 to 15 are reserved
+ */
+#define USB_PORT_STAT_C_CONNECTION	0x0001
+#define USB_PORT_STAT_C_ENABLE		0x0002
+#define USB_PORT_STAT_C_SUSPEND		0x0004
+#define USB_PORT_STAT_C_OVERCURRENT	0x0008
+#define USB_PORT_STAT_C_RESET		0x0010
+#define USB_PORT_STAT_C_L1		0x0020
+/*
+ * USB 3.0 wPortChange bit fields
+ * See USB 3.0 spec Table 10-11
+ */
+#define USB_PORT_STAT_C_BH_RESET	0x0020
+#define USB_PORT_STAT_C_LINK_STATE	0x0040
+#define USB_PORT_STAT_C_CONFIG_ERROR	0x0080
+
+/*
+ * wHubCharacteristics (masks)
+ * See USB 2.0 spec Table 11-13, offset 3
+ */
+#define HUB_CHAR_LPSM		0x0003 /* Logical Power Switching Mode mask */
+#define HUB_CHAR_COMMON_LPSM	0x0000 /* All ports power control at once */
+#define HUB_CHAR_INDV_PORT_LPSM	0x0001 /* per-port power control */
+#define HUB_CHAR_NO_LPSM	0x0002 /* no power switching */
+
+#define HUB_CHAR_COMPOUND	0x0004 /* hub is part of a compound device */
+
+#define HUB_CHAR_OCPM		0x0018 /* Over-Current Protection Mode mask */
+#define HUB_CHAR_COMMON_OCPM	0x0000 /* All ports Over-Current reporting */
+#define HUB_CHAR_INDV_PORT_OCPM	0x0008 /* per-port Over-current reporting */
+#define HUB_CHAR_NO_OCPM	0x0010 /* No Over-current Protection support */
+
+#define HUB_CHAR_TTTT		0x0060 /* TT Think Time mask */
+#define HUB_CHAR_PORTIND	0x0080 /* per-port indicators (LEDs) */
+
+struct usb_hub_status {
+	__le16 wHubStatus;
+	__le16 wHubChange;
+} __attribute__ ((packed));
+
+/*
+ * Hub Status & Hub Change bit masks
+ * See USB 2.0 spec Table 11-19 and Table 11-20
+ * Bits 0 and 1 for wHubStatus and wHubChange
+ * Bits 2 to 15 are reserved for both
+ */
+#define HUB_STATUS_LOCAL_POWER	0x0001
+#define HUB_STATUS_OVERCURRENT	0x0002
+#define HUB_CHANGE_LOCAL_POWER	0x0001
+#define HUB_CHANGE_OVERCURRENT	0x0002
+
+
+/*
+ * Hub descriptor
+ * See USB 2.0 spec Table 11-13
+ */
+
+#define USB_DT_HUB			(USB_TYPE_CLASS | 0x09)
+#define USB_DT_SS_HUB			(USB_TYPE_CLASS | 0x0a)
+#define USB_DT_HUB_NONVAR_SIZE		7
+#define USB_DT_SS_HUB_SIZE              12
+
+/*
+ * Hub Device descriptor
+ * USB Hub class device protocols
+ */
+
+#define USB_HUB_PR_FS		0 /* Full speed hub */
+#define USB_HUB_PR_HS_NO_TT	0 /* Hi-speed hub without TT */
+#define USB_HUB_PR_HS_SINGLE_TT	1 /* Hi-speed hub with single TT */
+#define USB_HUB_PR_HS_MULTI_TT	2 /* Hi-speed hub with multiple TT */
+#define USB_HUB_PR_SS		3 /* Super speed hub */
+
+struct usb_hub_descriptor {
+	__u8  bDescLength;
+	__u8  bDescriptorType;
+	__u8  bNbrPorts;
+	__le16 wHubCharacteristics;
+	__u8  bPwrOn2PwrGood;
+	__u8  bHubContrCurrent;
+
+	/* 2.0 and 3.0 hubs differ here */
+	union {
+		struct {
+			/* add 1 bit for hub status change; round to bytes */
+			__u8  DeviceRemovable[(USB_MAXCHILDREN + 1 + 7) / 8];
+			__u8  PortPwrCtrlMask[(USB_MAXCHILDREN + 1 + 7) / 8];
+		}  __attribute__ ((packed)) hs;
+
+		struct {
+			__u8 bHubHdrDecLat;
+			__le16 wHubDelay;
+			__le16 DeviceRemovable;
+		}  __attribute__ ((packed)) ss;
+	} u;
+} __attribute__ ((packed));
+
+/* port indicator status selectors, tables 11-7 and 11-25 */
+#define HUB_LED_AUTO	0
+#define HUB_LED_AMBER	1
+#define HUB_LED_GREEN	2
+#define HUB_LED_OFF	3
+
+enum hub_led_mode {
+	INDICATOR_AUTO = 0,
+	INDICATOR_CYCLE,
+	/* software blinks for attention:  software, hardware, reserved */
+	INDICATOR_GREEN_BLINK, INDICATOR_GREEN_BLINK_OFF,
+	INDICATOR_AMBER_BLINK, INDICATOR_AMBER_BLINK_OFF,
+	INDICATOR_ALT_BLINK, INDICATOR_ALT_BLINK_OFF
+} __attribute__ ((packed));
+
+/* Transaction Translator Think Times, in bits */
+#define HUB_TTTT_8_BITS		0x00
+#define HUB_TTTT_16_BITS	0x20
+#define HUB_TTTT_24_BITS	0x40
+#define HUB_TTTT_32_BITS	0x60
+
+#endif /* __LINUX_CH11_H */
diff --git a/include/linux/usb/hcd.h b/include/linux/usb/hcd.h
new file mode 100644
index 0000000..f5f5c7d
--- /dev/null
+++ b/include/linux/usb/hcd.h
@@ -0,0 +1,672 @@
+/*
+ * Copyright (c) 2001-2002 by David Brownell
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation; either version 2 of the License, or (at your
+ * option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef __USB_CORE_HCD_H
+#define __USB_CORE_HCD_H
+
+#ifdef __KERNEL__
+
+#include <linux/rwsem.h>
+
+#define MAX_TOPO_LEVEL		6
+
+/* This file contains declarations of usbcore internals that are mostly
+ * used or exposed by Host Controller Drivers.
+ */
+
+/*
+ * USB Packet IDs (PIDs)
+ */
+#define USB_PID_EXT			0xf0	/* USB 2.0 LPM ECN */
+#define USB_PID_OUT			0xe1
+#define USB_PID_ACK			0xd2
+#define USB_PID_DATA0			0xc3
+#define USB_PID_PING			0xb4	/* USB 2.0 */
+#define USB_PID_SOF			0xa5
+#define USB_PID_NYET			0x96	/* USB 2.0 */
+#define USB_PID_DATA2			0x87	/* USB 2.0 */
+#define USB_PID_SPLIT			0x78	/* USB 2.0 */
+#define USB_PID_IN			0x69
+#define USB_PID_NAK			0x5a
+#define USB_PID_DATA1			0x4b
+#define USB_PID_PREAMBLE		0x3c	/* Token mode */
+#define USB_PID_ERR			0x3c	/* USB 2.0: handshake mode */
+#define USB_PID_SETUP			0x2d
+#define USB_PID_STALL			0x1e
+#define USB_PID_MDATA			0x0f	/* USB 2.0 */
+
+/*-------------------------------------------------------------------------*/
+
+/*
+ * USB Host Controller Driver (usb_hcd) framework
+ *
+ * Since "struct usb_bus" is so thin, you can't share much code in it.
+ * This framework is a layer over that, and should be more sharable.
+ *
+ * @authorized_default: Specifies if new devices are authorized to
+ *                      connect by default or they require explicit
+ *                      user space authorization; this bit is settable
+ *                      through /sys/class/usb_host/X/authorized_default.
+ *                      For the rest is RO, so we don't lock to r/w it.
+ */
+
+/*-------------------------------------------------------------------------*/
+
+struct usb_hcd {
+
+	/*
+	 * housekeeping
+	 */
+	struct usb_bus		self;		/* hcd is-a bus */
+	struct kref		kref;		/* reference counter */
+
+	const char		*product_desc;	/* product/vendor string */
+	int			speed;		/* Speed for this roothub.
+						 * May be different from
+						 * hcd->driver->flags & HCD_MASK
+						 */
+	char			irq_descr[24];	/* driver + bus # */
+
+	struct timer_list	rh_timer;	/* drives root-hub polling */
+	struct urb		*status_urb;	/* the current status urb */
+#ifdef CONFIG_PM_RUNTIME
+	struct work_struct	wakeup_work;	/* for remote wakeup */
+#endif
+
+	/*
+	 * hardware info/state
+	 */
+	const struct hc_driver	*driver;	/* hw-specific hooks */
+
+	/*
+	 * OTG and some Host controllers need software interaction with phys;
+	 * other external phys should be software-transparent
+	 */
+	struct usb_phy	*phy;
+
+	/* Flags that need to be manipulated atomically because they can
+	 * change while the host controller is running.  Always use
+	 * set_bit() or clear_bit() to change their values.
+	 */
+	unsigned long		flags;
+#define HCD_FLAG_HW_ACCESSIBLE		0	/* at full power */
+#define HCD_FLAG_POLL_RH		2	/* poll for rh status? */
+#define HCD_FLAG_POLL_PENDING		3	/* status has changed? */
+#define HCD_FLAG_WAKEUP_PENDING		4	/* root hub is resuming? */
+#define HCD_FLAG_RH_RUNNING		5	/* root hub is running? */
+#define HCD_FLAG_DEAD			6	/* controller has died? */
+
+	/* The flags can be tested using these macros; they are likely to
+	 * be slightly faster than test_bit().
+	 */
+#define HCD_HW_ACCESSIBLE(hcd)	((hcd)->flags & (1U << HCD_FLAG_HW_ACCESSIBLE))
+#define HCD_POLL_RH(hcd)	((hcd)->flags & (1U << HCD_FLAG_POLL_RH))
+#define HCD_POLL_PENDING(hcd)	((hcd)->flags & (1U << HCD_FLAG_POLL_PENDING))
+#define HCD_WAKEUP_PENDING(hcd)	((hcd)->flags & (1U << HCD_FLAG_WAKEUP_PENDING))
+#define HCD_RH_RUNNING(hcd)	((hcd)->flags & (1U << HCD_FLAG_RH_RUNNING))
+#define HCD_DEAD(hcd)		((hcd)->flags & (1U << HCD_FLAG_DEAD))
+
+	/* Flags that get set only during HCD registration or removal. */
+	unsigned		rh_registered:1;/* is root hub registered? */
+	unsigned		rh_pollable:1;	/* may we poll the root hub? */
+	unsigned		msix_enabled:1;	/* driver has MSI-X enabled? */
+
+	/* The next flag is a stopgap, to be removed when all the HCDs
+	 * support the new root-hub polling mechanism. */
+	unsigned		uses_new_polling:1;
+	unsigned		wireless:1;	/* Wireless USB HCD */
+	unsigned		authorized_default:1;
+	unsigned		has_tt:1;	/* Integrated TT in root hub */
+
+	unsigned int		irq;		/* irq allocated */
+	void __iomem		*regs;		/* device memory/io */
+	resource_size_t		rsrc_start;	/* memory/io resource start */
+	resource_size_t		rsrc_len;	/* memory/io resource length */
+	unsigned		power_budget;	/* in mA, 0 = no limit */
+
+	/* bandwidth_mutex should be taken before adding or removing
+	 * any new bus bandwidth constraints:
+	 *   1. Before adding a configuration for a new device.
+	 *   2. Before removing the configuration to put the device into
+	 *      the addressed state.
+	 *   3. Before selecting a different configuration.
+	 *   4. Before selecting an alternate interface setting.
+	 *
+	 * bandwidth_mutex should be dropped after a successful control message
+	 * to the device, or resetting the bandwidth after a failed attempt.
+	 */
+	struct mutex		*bandwidth_mutex;
+	struct usb_hcd		*shared_hcd;
+	struct usb_hcd		*primary_hcd;
+
+
+#define HCD_BUFFER_POOLS	4
+	struct dma_pool		*pool[HCD_BUFFER_POOLS];
+
+	int			state;
+#	define	__ACTIVE		0x01
+#	define	__SUSPEND		0x04
+#	define	__TRANSIENT		0x80
+
+#	define	HC_STATE_HALT		0
+#	define	HC_STATE_RUNNING	(__ACTIVE)
+#	define	HC_STATE_QUIESCING	(__SUSPEND|__TRANSIENT|__ACTIVE)
+#	define	HC_STATE_RESUMING	(__SUSPEND|__TRANSIENT)
+#	define	HC_STATE_SUSPENDED	(__SUSPEND)
+
+#define	HC_IS_RUNNING(state) ((state) & __ACTIVE)
+#define	HC_IS_SUSPENDED(state) ((state) & __SUSPEND)
+
+	/* more shared queuing code would be good; it should support
+	 * smarter scheduling, handle transaction translators, etc;
+	 * input size of periodic table to an interrupt scheduler.
+	 * (ohci 32, uhci 1024, ehci 256/512/1024).
+	 */
+
+	/* The HC driver's private data is stored at the end of
+	 * this structure.
+	 */
+	unsigned long hcd_priv[0]
+			__attribute__ ((aligned(sizeof(s64))));
+};
+
+/* 2.4 does this a bit differently ... */
+static inline struct usb_bus *hcd_to_bus(struct usb_hcd *hcd)
+{
+	return &hcd->self;
+}
+
+static inline struct usb_hcd *bus_to_hcd(struct usb_bus *bus)
+{
+	return container_of(bus, struct usb_hcd, self);
+}
+
+struct hcd_timeout {	/* timeouts we allocate */
+	struct list_head	timeout_list;
+	struct timer_list	timer;
+};
+
+/*-------------------------------------------------------------------------*/
+
+
+struct hc_driver {
+	const char	*description;	/* "ehci-hcd" etc */
+	const char	*product_desc;	/* product/vendor string */
+	size_t		hcd_priv_size;	/* size of private data */
+
+	/* irq handler */
+	irqreturn_t	(*irq) (struct usb_hcd *hcd);
+
+	int	flags;
+#define	HCD_MEMORY	0x0001		/* HC regs use memory (else I/O) */
+#define	HCD_LOCAL_MEM	0x0002		/* HC needs local memory */
+#define	HCD_SHARED	0x0004		/* Two (or more) usb_hcds share HW */
+#define	HCD_USB11	0x0010		/* USB 1.1 */
+#define	HCD_USB2	0x0020		/* USB 2.0 */
+#define	HCD_USB3	0x0040		/* USB 3.0 */
+#define	HCD_MASK	0x0070
+
+	/* called to init HCD and root hub */
+	int	(*reset) (struct usb_hcd *hcd);
+	int	(*start) (struct usb_hcd *hcd);
+
+	/* NOTE:  these suspend/resume calls relate to the HC as
+	 * a whole, not just the root hub; they're for PCI bus glue.
+	 */
+	/* called after suspending the hub, before entering D3 etc */
+	int	(*pci_suspend)(struct usb_hcd *hcd, bool do_wakeup);
+
+	/* called after entering D0 (etc), before resuming the hub */
+	int	(*pci_resume)(struct usb_hcd *hcd, bool hibernated);
+
+	/* cleanly make HCD stop writing memory and doing I/O */
+	void	(*stop) (struct usb_hcd *hcd);
+
+	/* shutdown HCD */
+	void	(*shutdown) (struct usb_hcd *hcd);
+
+	/* return current frame number */
+	int	(*get_frame_number) (struct usb_hcd *hcd);
+
+	/* manage i/o requests, device state */
+	int	(*urb_enqueue)(struct usb_hcd *hcd,
+				struct urb *urb, gfp_t mem_flags);
+	int	(*urb_dequeue)(struct usb_hcd *hcd,
+				struct urb *urb, int status);
+
+	/*
+	 * (optional) these hooks allow an HCD to override the default DMA
+	 * mapping and unmapping routines.  In general, they shouldn't be
+	 * necessary unless the host controller has special DMA requirements,
+	 * such as alignment contraints.  If these are not specified, the
+	 * general usb_hcd_(un)?map_urb_for_dma functions will be used instead
+	 * (and it may be a good idea to call these functions in your HCD
+	 * implementation)
+	 */
+	int	(*map_urb_for_dma)(struct usb_hcd *hcd, struct urb *urb,
+				   gfp_t mem_flags);
+	void    (*unmap_urb_for_dma)(struct usb_hcd *hcd, struct urb *urb);
+
+	/* hw synch, freeing endpoint resources that urb_dequeue can't */
+	void	(*endpoint_disable)(struct usb_hcd *hcd,
+			struct usb_host_endpoint *ep);
+
+	/* (optional) reset any endpoint state such as sequence number
+	   and current window */
+	void	(*endpoint_reset)(struct usb_hcd *hcd,
+			struct usb_host_endpoint *ep);
+
+	/* root hub support */
+	int	(*hub_status_data) (struct usb_hcd *hcd, char *buf);
+	int	(*hub_control) (struct usb_hcd *hcd,
+				u16 typeReq, u16 wValue, u16 wIndex,
+				char *buf, u16 wLength);
+	int	(*bus_suspend)(struct usb_hcd *);
+	int	(*bus_resume)(struct usb_hcd *);
+	int	(*start_port_reset)(struct usb_hcd *, unsigned port_num);
+
+		/* force handover of high-speed port to full-speed companion */
+	void	(*relinquish_port)(struct usb_hcd *, int);
+		/* has a port been handed over to a companion? */
+	int	(*port_handed_over)(struct usb_hcd *, int);
+
+		/* CLEAR_TT_BUFFER completion callback */
+	void	(*clear_tt_buffer_complete)(struct usb_hcd *,
+				struct usb_host_endpoint *);
+
+	/* xHCI specific functions */
+		/* Called by usb_alloc_dev to alloc HC device structures */
+	int	(*alloc_dev)(struct usb_hcd *, struct usb_device *);
+		/* Called by usb_disconnect to free HC device structures */
+	void	(*free_dev)(struct usb_hcd *, struct usb_device *);
+	/* Change a group of bulk endpoints to support multiple stream IDs */
+	int	(*alloc_streams)(struct usb_hcd *hcd, struct usb_device *udev,
+		struct usb_host_endpoint **eps, unsigned int num_eps,
+		unsigned int num_streams, gfp_t mem_flags);
+	/* Reverts a group of bulk endpoints back to not using stream IDs.
+	 * Can fail if we run out of memory.
+	 */
+	int	(*free_streams)(struct usb_hcd *hcd, struct usb_device *udev,
+		struct usb_host_endpoint **eps, unsigned int num_eps,
+		gfp_t mem_flags);
+
+	/* Bandwidth computation functions */
+	/* Note that add_endpoint() can only be called once per endpoint before
+	 * check_bandwidth() or reset_bandwidth() must be called.
+	 * drop_endpoint() can only be called once per endpoint also.
+	 * A call to xhci_drop_endpoint() followed by a call to
+	 * xhci_add_endpoint() will add the endpoint to the schedule with
+	 * possibly new parameters denoted by a different endpoint descriptor
+	 * in usb_host_endpoint.  A call to xhci_add_endpoint() followed by a
+	 * call to xhci_drop_endpoint() is not allowed.
+	 */
+		/* Allocate endpoint resources and add them to a new schedule */
+	int	(*add_endpoint)(struct usb_hcd *, struct usb_device *,
+				struct usb_host_endpoint *);
+		/* Drop an endpoint from a new schedule */
+	int	(*drop_endpoint)(struct usb_hcd *, struct usb_device *,
+				 struct usb_host_endpoint *);
+		/* Check that a new hardware configuration, set using
+		 * endpoint_enable and endpoint_disable, does not exceed bus
+		 * bandwidth.  This must be called before any set configuration
+		 * or set interface requests are sent to the device.
+		 */
+	int	(*check_bandwidth)(struct usb_hcd *, struct usb_device *);
+		/* Reset the device schedule to the last known good schedule,
+		 * which was set from a previous successful call to
+		 * check_bandwidth().  This reverts any add_endpoint() and
+		 * drop_endpoint() calls since that last successful call.
+		 * Used for when a check_bandwidth() call fails due to resource
+		 * or bandwidth constraints.
+		 */
+	void	(*reset_bandwidth)(struct usb_hcd *, struct usb_device *);
+		/* Returns the hardware-chosen device address */
+	int	(*address_device)(struct usb_hcd *, struct usb_device *udev);
+		/* Notifies the HCD after a hub descriptor is fetched.
+		 * Will block.
+		 */
+	int	(*update_hub_device)(struct usb_hcd *, struct usb_device *hdev,
+			struct usb_tt *tt, gfp_t mem_flags);
+	int	(*reset_device)(struct usb_hcd *, struct usb_device *);
+		/* Notifies the HCD after a device is connected and its
+		 * address is set
+		 */
+	int	(*update_device)(struct usb_hcd *, struct usb_device *);
+	int	(*set_usb2_hw_lpm)(struct usb_hcd *, struct usb_device *, int);
+	/* USB 3.0 Link Power Management */
+		/* Returns the USB3 hub-encoded value for the U1/U2 timeout. */
+	int	(*enable_usb3_lpm_timeout)(struct usb_hcd *,
+			struct usb_device *, enum usb3_link_state state);
+		/* The xHCI host controller can still fail the command to
+		 * disable the LPM timeouts, so this can return an error code.
+		 */
+	int	(*disable_usb3_lpm_timeout)(struct usb_hcd *,
+			struct usb_device *, enum usb3_link_state state);
+	int	(*find_raw_port_number)(struct usb_hcd *, int);
+};
+
+extern int usb_hcd_link_urb_to_ep(struct usb_hcd *hcd, struct urb *urb);
+extern int usb_hcd_check_unlink_urb(struct usb_hcd *hcd, struct urb *urb,
+		int status);
+extern void usb_hcd_unlink_urb_from_ep(struct usb_hcd *hcd, struct urb *urb);
+
+extern int usb_hcd_submit_urb(struct urb *urb, gfp_t mem_flags);
+extern int usb_hcd_unlink_urb(struct urb *urb, int status);
+extern void usb_hcd_giveback_urb(struct usb_hcd *hcd, struct urb *urb,
+		int status);
+extern int usb_hcd_map_urb_for_dma(struct usb_hcd *hcd, struct urb *urb,
+		gfp_t mem_flags);
+extern void usb_hcd_unmap_urb_setup_for_dma(struct usb_hcd *, struct urb *);
+extern void usb_hcd_unmap_urb_for_dma(struct usb_hcd *, struct urb *);
+extern void usb_hcd_flush_endpoint(struct usb_device *udev,
+		struct usb_host_endpoint *ep);
+extern void usb_hcd_disable_endpoint(struct usb_device *udev,
+		struct usb_host_endpoint *ep);
+extern void usb_hcd_reset_endpoint(struct usb_device *udev,
+		struct usb_host_endpoint *ep);
+extern void usb_hcd_synchronize_unlinks(struct usb_device *udev);
+extern int usb_hcd_alloc_bandwidth(struct usb_device *udev,
+		struct usb_host_config *new_config,
+		struct usb_host_interface *old_alt,
+		struct usb_host_interface *new_alt);
+extern int usb_hcd_get_frame_number(struct usb_device *udev);
+
+extern struct usb_hcd *usb_create_hcd(const struct hc_driver *driver,
+		struct device *dev, const char *bus_name);
+extern struct usb_hcd *usb_create_shared_hcd(const struct hc_driver *driver,
+		struct device *dev, const char *bus_name,
+		struct usb_hcd *shared_hcd);
+extern struct usb_hcd *usb_get_hcd(struct usb_hcd *hcd);
+extern void usb_put_hcd(struct usb_hcd *hcd);
+extern int usb_hcd_is_primary_hcd(struct usb_hcd *hcd);
+extern int usb_add_hcd(struct usb_hcd *hcd,
+		unsigned int irqnum, unsigned long irqflags);
+extern void usb_remove_hcd(struct usb_hcd *hcd);
+extern int usb_hcd_find_raw_port_number(struct usb_hcd *hcd, int port1);
+
+struct platform_device;
+extern void usb_hcd_platform_shutdown(struct platform_device *dev);
+
+#ifdef CONFIG_PCI
+struct pci_dev;
+struct pci_device_id;
+extern int usb_hcd_pci_probe(struct pci_dev *dev,
+				const struct pci_device_id *id);
+extern void usb_hcd_pci_remove(struct pci_dev *dev);
+extern void usb_hcd_pci_shutdown(struct pci_dev *dev);
+
+#ifdef CONFIG_PM_SLEEP
+extern const struct dev_pm_ops usb_hcd_pci_pm_ops;
+#endif
+#endif /* CONFIG_PCI */
+
+/* pci-ish (pdev null is ok) buffer alloc/mapping support */
+int hcd_buffer_create(struct usb_hcd *hcd);
+void hcd_buffer_destroy(struct usb_hcd *hcd);
+
+void *hcd_buffer_alloc(struct usb_bus *bus, size_t size,
+	gfp_t mem_flags, dma_addr_t *dma);
+void hcd_buffer_free(struct usb_bus *bus, size_t size,
+	void *addr, dma_addr_t dma);
+
+/* generic bus glue, needed for host controllers that don't use PCI */
+extern irqreturn_t usb_hcd_irq(int irq, void *__hcd);
+
+extern void usb_hc_died(struct usb_hcd *hcd);
+extern void usb_hcd_poll_rh_status(struct usb_hcd *hcd);
+extern void usb_wakeup_notification(struct usb_device *hdev,
+		unsigned int portnum);
+
+extern void usb_hcd_start_port_resume(struct usb_bus *bus, int portnum);
+extern void usb_hcd_end_port_resume(struct usb_bus *bus, int portnum);
+
+/* The D0/D1 toggle bits ... USE WITH CAUTION (they're almost hcd-internal) */
+#define usb_gettoggle(dev, ep, out) (((dev)->toggle[out] >> (ep)) & 1)
+#define	usb_dotoggle(dev, ep, out)  ((dev)->toggle[out] ^= (1 << (ep)))
+#define usb_settoggle(dev, ep, out, bit) \
+		((dev)->toggle[out] = ((dev)->toggle[out] & ~(1 << (ep))) | \
+		 ((bit) << (ep)))
+
+/* -------------------------------------------------------------------------- */
+
+/* Enumeration is only for the hub driver, or HCD virtual root hubs */
+extern struct usb_device *usb_alloc_dev(struct usb_device *parent,
+					struct usb_bus *, unsigned port);
+extern int usb_new_device(struct usb_device *dev);
+extern void usb_disconnect(struct usb_device **);
+
+extern int usb_get_configuration(struct usb_device *dev);
+extern void usb_destroy_configuration(struct usb_device *dev);
+
+/*-------------------------------------------------------------------------*/
+
+/*
+ * HCD Root Hub support
+ */
+
+#include <linux/usb/ch11.h>
+
+/*
+ * As of USB 2.0, full/low speed devices are segregated into trees.
+ * One type grows from USB 1.1 host controllers (OHCI, UHCI etc).
+ * The other type grows from high speed hubs when they connect to
+ * full/low speed devices using "Transaction Translators" (TTs).
+ *
+ * TTs should only be known to the hub driver, and high speed bus
+ * drivers (only EHCI for now).  They affect periodic scheduling and
+ * sometimes control/bulk error recovery.
+ */
+
+struct usb_device;
+
+struct usb_tt {
+	struct usb_device	*hub;	/* upstream highspeed hub */
+	int			multi;	/* true means one TT per port */
+	unsigned		think_time;	/* think time in ns */
+
+	/* for control/bulk error recovery (CLEAR_TT_BUFFER) */
+	spinlock_t		lock;
+	struct list_head	clear_list;	/* of usb_tt_clear */
+	struct work_struct	clear_work;
+};
+
+struct usb_tt_clear {
+	struct list_head	clear_list;
+	unsigned		tt;
+	u16			devinfo;
+	struct usb_hcd		*hcd;
+	struct usb_host_endpoint	*ep;
+};
+
+extern int usb_hub_clear_tt_buffer(struct urb *urb);
+extern void usb_ep0_reinit(struct usb_device *);
+
+/* (shifted) direction/type/recipient from the USB 2.0 spec, table 9.2 */
+#define DeviceRequest \
+	((USB_DIR_IN|USB_TYPE_STANDARD|USB_RECIP_DEVICE)<<8)
+#define DeviceOutRequest \
+	((USB_DIR_OUT|USB_TYPE_STANDARD|USB_RECIP_DEVICE)<<8)
+
+#define InterfaceRequest \
+	((USB_DIR_IN|USB_TYPE_STANDARD|USB_RECIP_INTERFACE)<<8)
+
+#define EndpointRequest \
+	((USB_DIR_IN|USB_TYPE_STANDARD|USB_RECIP_INTERFACE)<<8)
+#define EndpointOutRequest \
+	((USB_DIR_OUT|USB_TYPE_STANDARD|USB_RECIP_INTERFACE)<<8)
+
+/* class requests from the USB 2.0 hub spec, table 11-15 */
+/* GetBusState and SetHubDescriptor are optional, omitted */
+#define ClearHubFeature		(0x2000 | USB_REQ_CLEAR_FEATURE)
+#define ClearPortFeature	(0x2300 | USB_REQ_CLEAR_FEATURE)
+#define GetHubDescriptor	(0xa000 | USB_REQ_GET_DESCRIPTOR)
+#define GetHubStatus		(0xa000 | USB_REQ_GET_STATUS)
+#define GetPortStatus		(0xa300 | USB_REQ_GET_STATUS)
+#define SetHubFeature		(0x2000 | USB_REQ_SET_FEATURE)
+#define SetPortFeature		(0x2300 | USB_REQ_SET_FEATURE)
+
+
+/*-------------------------------------------------------------------------*/
+
+/* class requests from USB 3.0 hub spec, table 10-5 */
+#define SetHubDepth		(0x3000 | HUB_SET_DEPTH)
+#define GetPortErrorCount	(0x8000 | HUB_GET_PORT_ERR_COUNT)
+
+/*
+ * Generic bandwidth allocation constants/support
+ */
+#define FRAME_TIME_USECS	1000L
+#define BitTime(bytecount) (7 * 8 * bytecount / 6) /* with integer truncation */
+		/* Trying not to use worst-case bit-stuffing
+		 * of (7/6 * 8 * bytecount) = 9.33 * bytecount */
+		/* bytecount = data payload byte count */
+
+#define NS_TO_US(ns)	((ns + 500L) / 1000L)
+			/* convert & round nanoseconds to microseconds */
+
+
+/*
+ * Full/low speed bandwidth allocation constants/support.
+ */
+#define BW_HOST_DELAY	1000L		/* nanoseconds */
+#define BW_HUB_LS_SETUP	333L		/* nanoseconds */
+			/* 4 full-speed bit times (est.) */
+
+#define FRAME_TIME_BITS			12000L	/* frame = 1 millisecond */
+#define FRAME_TIME_MAX_BITS_ALLOC	(90L * FRAME_TIME_BITS / 100L)
+#define FRAME_TIME_MAX_USECS_ALLOC	(90L * FRAME_TIME_USECS / 100L)
+
+/*
+ * Ceiling [nano/micro]seconds (typical) for that many bytes at high speed
+ * ISO is a bit less, no ACK ... from USB 2.0 spec, 5.11.3 (and needed
+ * to preallocate bandwidth)
+ */
+#define USB2_HOST_DELAY	5	/* nsec, guess */
+#define HS_NSECS(bytes) (((55 * 8 * 2083) \
+	+ (2083UL * (3 + BitTime(bytes))))/1000 \
+	+ USB2_HOST_DELAY)
+#define HS_NSECS_ISO(bytes) (((38 * 8 * 2083) \
+	+ (2083UL * (3 + BitTime(bytes))))/1000 \
+	+ USB2_HOST_DELAY)
+#define HS_USECS(bytes)		NS_TO_US(HS_NSECS(bytes))
+#define HS_USECS_ISO(bytes)	NS_TO_US(HS_NSECS_ISO(bytes))
+
+extern long usb_calc_bus_time(int speed, int is_input,
+			int isoc, int bytecount);
+
+/*-------------------------------------------------------------------------*/
+
+extern void usb_set_device_state(struct usb_device *udev,
+		enum usb_device_state new_state);
+
+/*-------------------------------------------------------------------------*/
+
+/* exported only within usbcore */
+
+extern struct list_head usb_bus_list;
+extern struct mutex usb_bus_list_lock;
+extern wait_queue_head_t usb_kill_urb_queue;
+
+extern int usb_find_interface_driver(struct usb_device *dev,
+	struct usb_interface *interface);
+
+#define usb_endpoint_out(ep_dir)	(!((ep_dir) & USB_DIR_IN))
+
+#ifdef CONFIG_PM
+extern void usb_root_hub_lost_power(struct usb_device *rhdev);
+extern int hcd_bus_suspend(struct usb_device *rhdev, pm_message_t msg);
+extern int hcd_bus_resume(struct usb_device *rhdev, pm_message_t msg);
+#endif /* CONFIG_PM */
+
+#ifdef CONFIG_PM_RUNTIME
+extern void usb_hcd_resume_root_hub(struct usb_hcd *hcd);
+#else
+static inline void usb_hcd_resume_root_hub(struct usb_hcd *hcd)
+{
+	return;
+}
+#endif /* CONFIG_PM_RUNTIME */
+
+/*-------------------------------------------------------------------------*/
+
+#if defined(CONFIG_USB_MON) || defined(CONFIG_USB_MON_MODULE)
+
+struct usb_mon_operations {
+	void (*urb_submit)(struct usb_bus *bus, struct urb *urb);
+	void (*urb_submit_error)(struct usb_bus *bus, struct urb *urb, int err);
+	void (*urb_complete)(struct usb_bus *bus, struct urb *urb, int status);
+	/* void (*urb_unlink)(struct usb_bus *bus, struct urb *urb); */
+};
+
+extern struct usb_mon_operations *mon_ops;
+
+static inline void usbmon_urb_submit(struct usb_bus *bus, struct urb *urb)
+{
+	if (bus->monitored)
+		(*mon_ops->urb_submit)(bus, urb);
+}
+
+static inline void usbmon_urb_submit_error(struct usb_bus *bus, struct urb *urb,
+    int error)
+{
+	if (bus->monitored)
+		(*mon_ops->urb_submit_error)(bus, urb, error);
+}
+
+static inline void usbmon_urb_complete(struct usb_bus *bus, struct urb *urb,
+		int status)
+{
+	if (bus->monitored)
+		(*mon_ops->urb_complete)(bus, urb, status);
+}
+
+int usb_mon_register(struct usb_mon_operations *ops);
+void usb_mon_deregister(void);
+
+#else
+
+static inline void usbmon_urb_submit(struct usb_bus *bus, struct urb *urb) {}
+static inline void usbmon_urb_submit_error(struct usb_bus *bus, struct urb *urb,
+    int error) {}
+static inline void usbmon_urb_complete(struct usb_bus *bus, struct urb *urb,
+		int status) {}
+
+#endif /* CONFIG_USB_MON || CONFIG_USB_MON_MODULE */
+
+/*-------------------------------------------------------------------------*/
+
+/* random stuff */
+
+#define	RUN_CONTEXT (in_irq() ? "in_irq" \
+		: (in_interrupt() ? "in_interrupt" : "can sleep"))
+
+
+/* This rwsem is for use only by the hub driver and ehci-hcd.
+ * Nobody else should touch it.
+ */
+extern struct rw_semaphore ehci_cf_port_reset_rwsem;
+
+/* Keep track of which host controller drivers are loaded */
+#define USB_UHCI_LOADED		0
+#define USB_OHCI_LOADED		1
+#define USB_EHCI_LOADED		2
+extern unsigned long usb_hcds_loaded;
+
+#endif /* __KERNEL__ */
+
+#endif /* __USB_CORE_HCD_H */
-- 
1.7.9.5



More information about the U-Boot mailing list