USB: xHCI: Add stack support for xHCI

This adds stack layer for eXtensible Host Controller Interface
which facilitates use of USB 3.0 in host mode.

Adapting xHCI host controller driver in linux-kernel
by Sarah Sharp to needs in u-boot.

Initial porting from Linux kernel version 3.4, with following
top commit history of drivers/usb/host/xhci* :
cf84055 xHCI: Cleanup isoc transfer ring when TD length mismatch found

This adds the basic xHCI host controller driver with bare minimum
features:
- Control/Bulk transfer support has been added with required
  infrastructure for necessary xHC data structures.
- Stream protocol hasn't been supported yet.
- No support for quirky devices has been added.

Signed-off-by: Vikas C Sajjan <vikas.sajjan@samsung.com>
Signed-off-by: Julius Werner <jwerner@chromium.org>
Signed-off-by: Vivek Gautam <gautam.vivek@samsung.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Minkyu Kang <mk7.kang@samsung.com>
Cc: Dan Murphy <dmurphy@ti.com>
Cc: Marek Vasut <marex@denx.de>

5 files changed, 3947 insertions, 0 deletions

diff --git a/drivers/usb/host/Makefile b/drivers/usb/host/Makefile
index ff6c80e66f..6bd6c861fc 100644
--- a/drivers/usb/host/Makefile
+++ b/drivers/usb/host/Makefile
@@ -42,6 +42,9 @@ COBJS-$(CONFIG_USB_EHCI_SPEAR) += ehci-spear.o
 COBJS-$(CONFIG_USB_EHCI_TEGRA) += ehci-tegra.o
 COBJS-$(CONFIG_USB_EHCI_VCT) += ehci-vct.o
 
+# xhci
+COBJS-$(CONFIG_USB_XHCI) += xhci.o xhci-mem.o xhci-ring.o
+
 COBJS	:= $(COBJS-y)
 SRCS	:= $(COBJS:.o=.c)
 OBJS	:= $(addprefix $(obj),$(COBJS))
diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
new file mode 100644
index 0000000000..89908e8a80
--- /dev/null
+++ b/drivers/usb/host/xhci-mem.c
@@ -0,0 +1,720 @@
+/*
+ * USB HOST XHCI Controller stack
+ *
+ * Based on xHCI host controller driver in linux-kernel
+ * by Sarah Sharp.
+ *
+ * Copyright (C) 2008 Intel Corp.
+ * Author: Sarah Sharp
+ *
+ * Copyright (C) 2013 Samsung Electronics Co.Ltd
+ * Authors: Vivek Gautam <gautam.vivek@samsung.com>
+ *	    Vikas Sajjan <vikas.sajjan@samsung.com>
+ *
+ * SPDX-License-Identifier:	GPL-2.0+
+ */
+
+#include <common.h>
+#include <asm/byteorder.h>
+#include <usb.h>
+#include <malloc.h>
+#include <asm/cache.h>
+#include <asm-generic/errno.h>
+
+#include "xhci.h"
+
+#define CACHELINE_SIZE		CONFIG_SYS_CACHELINE_SIZE
+/**
+ * flushes the address passed till the length
+ *
+ * @param addr	pointer to memory region to be flushed
+ * @param len	the length of the cache line to be flushed
+ * @return none
+ */
+void xhci_flush_cache(uint32_t addr, u32 len)
+{
+	BUG_ON((void *)addr == NULL || len == 0);
+
+	flush_dcache_range(addr & ~(CACHELINE_SIZE - 1),
+				ALIGN(addr + len, CACHELINE_SIZE));
+}
+
+/**
+ * invalidates the address passed till the length
+ *
+ * @param addr	pointer to memory region to be invalidates
+ * @param len	the length of the cache line to be invalidated
+ * @return none
+ */
+void xhci_inval_cache(uint32_t addr, u32 len)
+{
+	BUG_ON((void *)addr == NULL || len == 0);
+
+	invalidate_dcache_range(addr & ~(CACHELINE_SIZE - 1),
+				ALIGN(addr + len, CACHELINE_SIZE));
+}
+
+
+/**
+ * frees the "segment" pointer passed
+ *
+ * @param ptr	pointer to "segement" to be freed
+ * @return none
+ */
+static void xhci_segment_free(struct xhci_segment *seg)
+{
+	free(seg->trbs);
+	seg->trbs = NULL;
+
+	free(seg);
+}
+
+/**
+ * frees the "ring" pointer passed
+ *
+ * @param ptr	pointer to "ring" to be freed
+ * @return none
+ */
+static void xhci_ring_free(struct xhci_ring *ring)
+{
+	struct xhci_segment *seg;
+	struct xhci_segment *first_seg;
+
+	BUG_ON(!ring);
+
+	first_seg = ring->first_seg;
+	seg = first_seg->next;
+	while (seg != first_seg) {
+		struct xhci_segment *next = seg->next;
+		xhci_segment_free(seg);
+		seg = next;
+	}
+	xhci_segment_free(first_seg);
+
+	free(ring);
+}
+
+/**
+ * frees the "xhci_container_ctx" pointer passed
+ *
+ * @param ptr	pointer to "xhci_container_ctx" to be freed
+ * @return none
+ */
+static void xhci_free_container_ctx(struct xhci_container_ctx *ctx)
+{
+	free(ctx->bytes);
+	free(ctx);
+}
+
+/**
+ * frees the virtual devices for "xhci_ctrl" pointer passed
+ *
+ * @param ptr	pointer to "xhci_ctrl" whose virtual devices are to be freed
+ * @return none
+ */
+static void xhci_free_virt_devices(struct xhci_ctrl *ctrl)
+{
+	int i;
+	int slot_id;
+	struct xhci_virt_device *virt_dev;
+
+	/*
+	 * refactored here to loop through all virt_dev
+	 * Slot ID 0 is reserved
+	 */
+	for (slot_id = 0; slot_id < MAX_HC_SLOTS; slot_id++) {
+		virt_dev = ctrl->devs[slot_id];
+		if (!virt_dev)
+			continue;
+
+		ctrl->dcbaa->dev_context_ptrs[slot_id] = 0;
+
+		for (i = 0; i < 31; ++i)
+			if (virt_dev->eps[i].ring)
+				xhci_ring_free(virt_dev->eps[i].ring);
+
+		if (virt_dev->in_ctx)
+			xhci_free_container_ctx(virt_dev->in_ctx);
+		if (virt_dev->out_ctx)
+			xhci_free_container_ctx(virt_dev->out_ctx);
+
+		free(virt_dev);
+		/* make sure we are pointing to NULL */
+		ctrl->devs[slot_id] = NULL;
+	}
+}
+
+/**
+ * frees all the memory allocated
+ *
+ * @param ptr	pointer to "xhci_ctrl" to be cleaned up
+ * @return none
+ */
+void xhci_cleanup(struct xhci_ctrl *ctrl)
+{
+	xhci_ring_free(ctrl->event_ring);
+	xhci_ring_free(ctrl->cmd_ring);
+	xhci_free_virt_devices(ctrl);
+	free(ctrl->erst.entries);
+	free(ctrl->dcbaa);
+	memset(ctrl, '\0', sizeof(struct xhci_ctrl));
+}
+
+/**
+ * Malloc the aligned memory
+ *
+ * @param size	size of memory to be allocated
+ * @return allocates the memory and returns the aligned pointer
+ */
+static void *xhci_malloc(unsigned int size)
+{
+	void *ptr;
+	size_t cacheline_size = max(XHCI_ALIGNMENT, CACHELINE_SIZE);
+
+	ptr = memalign(cacheline_size, ALIGN(size, cacheline_size));
+	BUG_ON(!ptr);
+	memset(ptr, '\0', size);
+
+	xhci_flush_cache((uint32_t)ptr, size);
+
+	return ptr;
+}
+
+/**
+ * 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
+ * address of the next segment.  The caller needs to set any Link TRB
+ * related flags, such as End TRB, Toggle Cycle, and no snoop.
+ *
+ * @param prev	pointer to the previous segment
+ * @param next	pointer to the next segment
+ * @param link_trbs	flag to indicate whether to link the trbs or NOT
+ * @return none
+ */
+static void xhci_link_segments(struct xhci_segment *prev,
+				struct xhci_segment *next, bool link_trbs)
+{
+	u32 val;
+	u64 val_64 = 0;
+
+	if (!prev || !next)
+		return;
+	prev->next = next;
+	if (link_trbs) {
+		val_64 = (uintptr_t)next->trbs;
+		prev->trbs[TRBS_PER_SEGMENT-1].link.segment_ptr = val_64;
+
+		/*
+		 * 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_LINK << TRB_TYPE_SHIFT);
+
+		prev->trbs[TRBS_PER_SEGMENT-1].link.control = cpu_to_le32(val);
+	}
+}
+
+/**
+ * Initialises the Ring's enqueue,dequeue,enq_seg pointers
+ *
+ * @param ring	pointer to the RING to be intialised
+ * @return none
+ */
+static void xhci_initialize_ring_info(struct xhci_ring *ring)
+{
+	/*
+	 * 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.
+	 */
+	ring->cycle_state = 1;
+}
+
+/**
+ * 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'"
+ *
+ * @param	none
+ * @return pointer to the newly allocated SEGMENT
+ */
+static struct xhci_segment *xhci_segment_alloc(void)
+{
+	struct xhci_segment *seg;
+
+	seg = (struct xhci_segment *)malloc(sizeof(struct xhci_segment));
+	BUG_ON(!seg);
+
+	seg->trbs = (union xhci_trb *)xhci_malloc(SEGMENT_SIZE);
+
+	seg->next = NULL;
+
+	return seg;
+}
+
+/**
+ * Create a new ring with zero or more segments.
+ * TODO: current code only uses one-time-allocated single-segment rings
+ * of 1KB anyway, so we might as well get rid of all the segment and
+ * linking code (and maybe increase the size a bit, e.g. 4KB).
+ *
+ *
+ * Link each segment together into a ring.
+ * Set the end flag and the cycle toggle bit on the last segment.
+ * See section 4.9.2 and figures 15 and 16 of XHCI spec rev1.0.
+ *
+ * @param num_segs	number of segments in the ring
+ * @param link_trbs	flag to indicate whether to link the trbs or NOT
+ * @return pointer to the newly created RING
+ */
+struct xhci_ring *xhci_ring_alloc(unsigned int num_segs, bool link_trbs)
+{
+	struct xhci_ring *ring;
+	struct xhci_segment *prev;
+
+	ring = (struct xhci_ring *)malloc(sizeof(struct xhci_ring));
+	BUG_ON(!ring);
+
+	if (num_segs == 0)
+		return ring;
+
+	ring->first_seg = xhci_segment_alloc();
+	BUG_ON(!ring->first_seg);
+
+	num_segs--;
+
+	prev = ring->first_seg;
+	while (num_segs > 0) {
+		struct xhci_segment *next;
+
+		next = xhci_segment_alloc();
+		BUG_ON(!next);
+
+		xhci_link_segments(prev, next, link_trbs);
+
+		prev = next;
+		num_segs--;
+	}
+	xhci_link_segments(prev, ring->first_seg, link_trbs);
+	if (link_trbs) {
+		/* See section 4.9.2.1 and 6.4.4.1 */
+		prev->trbs[TRBS_PER_SEGMENT-1].link.control |=
+					cpu_to_le32(LINK_TOGGLE);
+	}
+	xhci_initialize_ring_info(ring);
+
+	return ring;
+}
+
+/**
+ * Allocates the Container context
+ *
+ * @param ctrl	Host controller data structure
+ * @param type type of XHCI Container Context
+ * @return NULL if failed else pointer to the context on success
+ */
+static struct xhci_container_ctx
+		*xhci_alloc_container_ctx(struct xhci_ctrl *ctrl, int type)
+{
+	struct xhci_container_ctx *ctx;
+
+	ctx = (struct xhci_container_ctx *)
+		malloc(sizeof(struct xhci_container_ctx));
+	BUG_ON(!ctx);
+
+	BUG_ON((type != XHCI_CTX_TYPE_DEVICE) && (type != XHCI_CTX_TYPE_INPUT));
+	ctx->type = type;
+	ctx->size = (MAX_EP_CTX_NUM + 1) *
+			CTX_SIZE(readl(&ctrl->hccr->cr_hccparams));
+	if (type == XHCI_CTX_TYPE_INPUT)
+		ctx->size += CTX_SIZE(readl(&ctrl->hccr->cr_hccparams));
+
+	ctx->bytes = (u8 *)xhci_malloc(ctx->size);
+
+	return ctx;
+}
+
+/**
+ * Allocating virtual device
+ *
+ * @param udev	pointer to USB deivce structure
+ * @return 0 on success else -1 on failure
+ */
+int xhci_alloc_virt_device(struct usb_device *udev)
+{
+	u64 byte_64 = 0;
+	unsigned int slot_id = udev->slot_id;
+	struct xhci_virt_device *virt_dev;
+	struct xhci_ctrl *ctrl = udev->controller;
+
+	/* Slot ID 0 is reserved */
+	if (ctrl->devs[slot_id]) {
+		printf("Virt dev for slot[%d] already allocated\n", slot_id);
+		return -EEXIST;
+	}
+
+	ctrl->devs[slot_id] = (struct xhci_virt_device *)
+					malloc(sizeof(struct xhci_virt_device));
+
+	if (!ctrl->devs[slot_id]) {
+		puts("Failed to allocate virtual device\n");
+		return -ENOMEM;
+	}
+
+	memset(ctrl->devs[slot_id], 0, sizeof(struct xhci_virt_device));
+	virt_dev = ctrl->devs[slot_id];
+
+	/* Allocate the (output) device context that will be used in the HC. */
+	virt_dev->out_ctx = xhci_alloc_container_ctx(ctrl,
+					XHCI_CTX_TYPE_DEVICE);
+	if (!virt_dev->out_ctx) {
+		puts("Failed to allocate out context for virt dev\n");
+		return -ENOMEM;
+	}
+
+	/* Allocate the (input) device context for address device command */
+	virt_dev->in_ctx = xhci_alloc_container_ctx(ctrl,
+					XHCI_CTX_TYPE_INPUT);
+	if (!virt_dev->in_ctx) {
+		puts("Failed to allocate in context for virt dev\n");
+		return -ENOMEM;
+	}
+
+	/* Allocate endpoint 0 ring */
+	virt_dev->eps[0].ring = xhci_ring_alloc(1, true);
+
+	byte_64 = (uintptr_t)(virt_dev->out_ctx->bytes);
+
+	/* Point to output device context in dcbaa. */
+	ctrl->dcbaa->dev_context_ptrs[slot_id] = byte_64;
+
+	xhci_flush_cache((uint32_t)&ctrl->dcbaa->dev_context_ptrs[slot_id],
+							sizeof(__le64));
+	return 0;
+}
+
+/**
+ * Allocates the necessary data structures
+ * for XHCI host controller
+ *
+ * @param ctrl	Host controller data structure
+ * @param hccr	pointer to HOST Controller Control Registers
+ * @param hcor	pointer to HOST Controller Operational Registers
+ * @return 0 if successful else -1 on failure
+ */
+int xhci_mem_init(struct xhci_ctrl *ctrl, struct xhci_hccr *hccr,
+					struct xhci_hcor *hcor)
+{
+	uint64_t val_64;
+	uint64_t trb_64;
+	uint32_t val;
+	unsigned long deq;
+	int i;
+	struct xhci_segment *seg;
+
+	/* DCBAA initialization */
+	ctrl->dcbaa = (struct xhci_device_context_array *)
+			xhci_malloc(sizeof(struct xhci_device_context_array));
+	if (ctrl->dcbaa == NULL) {
+		puts("unable to allocate DCBA\n");
+		return -ENOMEM;
+	}
+
+	val_64 = (uintptr_t)ctrl->dcbaa;
+	/* Set the pointer in DCBAA register */
+	xhci_writeq(&hcor->or_dcbaap, val_64);
+
+	/* Command ring control pointer register initialization */
+	ctrl->cmd_ring = xhci_ring_alloc(1, true);
+
+	/* Set the address in the Command Ring Control register */
+	trb_64 = (uintptr_t)ctrl->cmd_ring->first_seg->trbs;
+	val_64 = xhci_readq(&hcor->or_crcr);
+	val_64 = (val_64 & (u64) CMD_RING_RSVD_BITS) |
+		(trb_64 & (u64) ~CMD_RING_RSVD_BITS) |
+		ctrl->cmd_ring->cycle_state;
+	xhci_writeq(&hcor->or_crcr, val_64);
+
+	/* write the address of db register */
+	val = xhci_readl(&hccr->cr_dboff);
+	val &= DBOFF_MASK;
+	ctrl->dba = (struct xhci_doorbell_array *)((char *)hccr + val);
+
+	/* write the address of runtime register */
+	val = xhci_readl(&hccr->cr_rtsoff);
+	val &= RTSOFF_MASK;
+	ctrl->run_regs = (struct xhci_run_regs *)((char *)hccr + val);
+
+	/* writting the address of ir_set structure */
+	ctrl->ir_set = &ctrl->run_regs->ir_set[0];
+
+	/* Event ring does not maintain link TRB */
+	ctrl->event_ring = xhci_ring_alloc(ERST_NUM_SEGS, false);
+	ctrl->erst.entries = (struct xhci_erst_entry *)
+		xhci_malloc(sizeof(struct xhci_erst_entry) * ERST_NUM_SEGS);
+
+	ctrl->erst.num_entries = ERST_NUM_SEGS;
+
+	for (val = 0, seg = ctrl->event_ring->first_seg;
+			val < ERST_NUM_SEGS;
+			val++) {
+		trb_64 = 0;
+		trb_64 = (uintptr_t)seg->trbs;
+		struct xhci_erst_entry *entry = &ctrl->erst.entries[val];
+		xhci_writeq(&entry->seg_addr, trb_64);
+		entry->seg_size = cpu_to_le32(TRBS_PER_SEGMENT);
+		entry->rsvd = 0;
+		seg = seg->next;
+	}
+	xhci_flush_cache((uint32_t)ctrl->erst.entries,
+			ERST_NUM_SEGS * sizeof(struct xhci_erst_entry));
+
+	deq = (unsigned long)ctrl->event_ring->dequeue;
+
+	/* Update HC event ring dequeue pointer */
+	xhci_writeq(&ctrl->ir_set->erst_dequeue,
+				(u64)deq & (u64)~ERST_PTR_MASK);
+
+	/* set ERST count with the number of entries in the segment table */
+	val = xhci_readl(&ctrl->ir_set->erst_size);
+	val &= ERST_SIZE_MASK;
+	val |= ERST_NUM_SEGS;
+	xhci_writel(&ctrl->ir_set->erst_size, val);
+
+	/* this is the event ring segment table pointer */
+	val_64 = xhci_readq(&ctrl->ir_set->erst_base);
+	val_64 &= ERST_PTR_MASK;
+	val_64 |= ((u32)(ctrl->erst.entries) & ~ERST_PTR_MASK);
+
+	xhci_writeq(&ctrl->ir_set->erst_base, val_64);
+
+	/* initializing the virtual devices to NULL */
+	for (i = 0; i < MAX_HC_SLOTS; ++i)
+		ctrl->devs[i] = NULL;
+
+	/*
+	 * Just Zero'ing this register completely,
+	 * or some spurious Device Notification Events
+	 * might screw things here.
+	 */
+	xhci_writel(&hcor->or_dnctrl, 0x0);
+
+	return 0;
+}
+
+/**
+ * Give the input control context for the passed container context
+ *
+ * @param ctx	pointer to the context
+ * @return pointer to the Input control context data
+ */
+struct xhci_input_control_ctx
+		*xhci_get_input_control_ctx(struct xhci_container_ctx *ctx)
+{
+	BUG_ON(ctx->type != XHCI_CTX_TYPE_INPUT);
+	return (struct xhci_input_control_ctx *)ctx->bytes;
+}
+
+/**
+ * Give the slot context for the passed container context
+ *
+ * @param ctrl	Host controller data structure
+ * @param ctx	pointer to the context
+ * @return pointer to the slot control context data
+ */
+struct xhci_slot_ctx *xhci_get_slot_ctx(struct xhci_ctrl *ctrl,
+				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(readl(&ctrl->hccr->cr_hccparams)));
+}
+
+/**
+ * Gets the EP context from based on the ep_index
+ *
+ * @param ctrl	Host controller data structure
+ * @param ctx	context container
+ * @param ep_index	index of the endpoint
+ * @return pointer to the End point context
+ */
+struct xhci_ep_ctx *xhci_get_ep_ctx(struct xhci_ctrl *ctrl,
+				    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(readl(&ctrl->hccr->cr_hccparams))));
+}
+
+/**
+ * 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.
+ *
+ * @param ctrl	Host controller data structure
+ * @param in_ctx contains the input context
+ * @param out_ctx contains the input context
+ * @param ep_index index of the end point
+ * @return none
+ */
+void xhci_endpoint_copy(struct xhci_ctrl *ctrl,
+			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(ctrl, out_ctx, ep_index);
+	in_ep_ctx = xhci_get_ep_ctx(ctrl, 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.
+ *
+ * @param ctrl	Host controller data structure
+ * @param in_ctx contains the inpout context
+ * @param out_ctx contains the inpout context
+ * @return none
+ */
+void xhci_slot_copy(struct xhci_ctrl *ctrl, 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(ctrl, in_ctx);
+	out_slot_ctx = xhci_get_slot_ctx(ctrl, 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;
+}
+
+/**
+ * Setup an xHCI virtual device for a Set Address command
+ *
+ * @param udev pointer to the Device Data Structure
+ * @return returns negative value on failure else 0 on success
+ */
+void xhci_setup_addressable_virt_dev(struct usb_device *udev)
+{
+	struct usb_device *hop = udev;
+	struct xhci_virt_device *virt_dev;
+	struct xhci_ep_ctx *ep0_ctx;
+	struct xhci_slot_ctx *slot_ctx;
+	u32 port_num = 0;
+	u64 trb_64 = 0;
+	struct xhci_ctrl *ctrl = udev->controller;
+
+	virt_dev = ctrl->devs[udev->slot_id];
+
+	BUG_ON(!virt_dev);
+
+	/* Extract the EP0 and Slot Ctrl */
+	ep0_ctx = xhci_get_ep_ctx(ctrl, virt_dev->in_ctx, 0);
+	slot_ctx = xhci_get_slot_ctx(ctrl, virt_dev->in_ctx);
+
+	/* Only the control endpoint is valid - one endpoint context */
+	slot_ctx->dev_info |= cpu_to_le32(LAST_CTX(1) | 0);
+
+	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;
+	default:
+		/* Speed was set earlier, this shouldn't happen. */
+		BUG();
+	}
+
+	/* Extract the root hub port number */
+	if (hop->parent)
+		while (hop->parent->parent)
+			hop = hop->parent;
+	port_num = hop->portnr;
+	debug("port_num = %d\n", port_num);
+
+	slot_ctx->dev_info2 |=
+			cpu_to_le32(((port_num & ROOT_HUB_PORT_MASK) <<
+				ROOT_HUB_PORT_SHIFT));
+
+	/* Step 4 - ring already allocated */
+	/* Step 5 */
+	ep0_ctx->ep_info2 = cpu_to_le32(CTRL_EP << EP_TYPE_SHIFT);
+	debug("SPEED = %d\n", udev->speed);
+
+	switch (udev->speed) {
+	case USB_SPEED_SUPER:
+		ep0_ctx->ep_info2 |= cpu_to_le32(((512 & MAX_PACKET_MASK) <<
+					MAX_PACKET_SHIFT));
+		debug("Setting Packet size = 512bytes\n");
+		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(((64 & MAX_PACKET_MASK) <<
+					MAX_PACKET_SHIFT));
+		debug("Setting Packet size = 64bytes\n");
+		break;
+	case USB_SPEED_LOW:
+		ep0_ctx->ep_info2 |= cpu_to_le32(((8 & MAX_PACKET_MASK) <<
+					MAX_PACKET_SHIFT));
+		debug("Setting Packet size = 8bytes\n");
+		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(((0 & MAX_BURST_MASK) << MAX_BURST_SHIFT) |
+			((3 & ERROR_COUNT_MASK) << ERROR_COUNT_SHIFT));
+
+	trb_64 = (uintptr_t)virt_dev->eps[0].ring->first_seg->trbs;
+	ep0_ctx->deq = cpu_to_le64(trb_64 | virt_dev->eps[0].ring->cycle_state);
+
+	/* Steps 7 and 8 were done in xhci_alloc_virt_device() */
+
+	xhci_flush_cache((uint32_t)ep0_ctx, sizeof(struct xhci_ep_ctx));
+	xhci_flush_cache((uint32_t)slot_ctx, sizeof(struct xhci_slot_ctx));
+}
diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
new file mode 100644
index 0000000000..19c3ec6211
--- /dev/null
+++ b/drivers/usb/host/xhci-ring.c
@@ -0,0 +1,939 @@
+/*
+ * USB HOST XHCI Controller stack
+ *
+ * Based on xHCI host controller driver in linux-kernel
+ * by Sarah Sharp.
+ *
+ * Copyright (C) 2008 Intel Corp.
+ * Author: Sarah Sharp
+ *
+ * Copyright (C) 2013 Samsung Electronics Co.Ltd
+ * Authors: Vivek Gautam <gautam.vivek@samsung.com>
+ *	    Vikas Sajjan <vikas.sajjan@samsung.com>
+ *
+ * SPDX-License-Identifier:	GPL-2.0+
+ */
+
+#include <common.h>
+#include <asm/byteorder.h>
+#include <usb.h>
+#include <asm/unaligned.h>
+#include <asm-generic/errno.h>
+
+#include "xhci.h"
+
+/**
+ * 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 ?
+ *
+ * @param ctrl	Host controller data structure
+ * @param ring	pointer to the ring
+ * @param seg	poniter to the segment to which TRB belongs
+ * @param trb	poniter to the ring trb
+ * @return 1 if this TRB a link TRB else 0
+ */
+static int last_trb(struct xhci_ctrl *ctrl, struct xhci_ring *ring,
+			struct xhci_segment *seg, union xhci_trb *trb)
+{
+	if (ring == ctrl->event_ring)
+		return trb == &seg->trbs[TRBS_PER_SEGMENT];
+	else
+		return TRB_TYPE_LINK_LE32(trb->link.control);
+}
+
+/**
+ * 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?
+ *
+ * @param ctrl	Host controller data structure
+ * @param ring	pointer to the ring
+ * @param seg	poniter to the segment to which TRB belongs
+ * @param trb	poniter to the ring trb
+ * @return 1 if this TRB is the last TRB on the last segment else 0
+ */
+static bool last_trb_on_last_seg(struct xhci_ctrl *ctrl,
+				 struct xhci_ring *ring,
+				 struct xhci_segment *seg,
+				 union xhci_trb *trb)
+{
+	if (ring == ctrl->event_ring)
+		return ((trb == &seg->trbs[TRBS_PER_SEGMENT]) &&
+			(seg->next == ring->first_seg));
+	else
+		return le32_to_cpu(trb->link.control) & LINK_TOGGLE;
+}
+
+/**
+ * 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.
+ *
+ * @param ctrl	Host controller data structure
+ * @param ring	pointer to the ring
+ * @param more_trbs_coming	flag to indicate whether more trbs
+ *				are expected or NOT.
+ *				Will you enqueue more TRBs before calling
+ *				prepare_ring()?
+ * @return none
+ */
+static void inc_enq(struct xhci_ctrl *ctrl, struct xhci_ring *ring,
+						bool more_trbs_coming)
+{
+	u32 chain;
+	union xhci_trb *next;
+
+	chain = le32_to_cpu(ring->enqueue->generic.field[3]) & TRB_CHAIN;
+	next = ++(ring->enqueue);
+
+	/*
+	 * 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(ctrl, ring, ring->enq_seg, next)) {
+		if (ring != ctrl->event_ring) {
+			/*
+			 * 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).
+			 */
+			next->link.control &= cpu_to_le32(~TRB_CHAIN);
+			next->link.control |= cpu_to_le32(chain);
+
+			next->link.control ^= cpu_to_le32(TRB_CYCLE);
+			xhci_flush_cache((uint32_t)next,
+						sizeof(union xhci_trb));
+		}
+		/* Toggle the cycle bit after the last ring segment. */
+		if (last_trb_on_last_seg(ctrl, 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;
+	}
+}
+
+/**
+ * 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.
+ *
+ * @param ctrl	Host controller data structure
+ * @param ring	Ring whose Dequeue TRB pointer needs to be incremented.
+ * return none
+ */
+static void inc_deq(struct xhci_ctrl *ctrl, struct xhci_ring *ring)
+{
+	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(ctrl, ring, ring->deq_seg, ring->dequeue)) {
+			if (ring == ctrl->event_ring &&
+					last_trb_on_last_seg(ctrl, 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(ctrl, ring, ring->deq_seg, ring->dequeue));
+}
+
+/**
+ * Generic function for queueing a TRB on a ring.
+ * The caller must have checked to make sure there's room on the ring.
+ *
+ * @param	more_trbs_coming:   Will you enqueue more TRBs before calling
+ *				prepare_ring()?
+ * @param ctrl	Host controller data structure
+ * @param ring	pointer to the ring
+ * @param more_trbs_coming	flag to indicate whether more trbs
+ * @param trb_fields	pointer to trb field array containing TRB contents
+ * @return pointer to the enqueued trb
+ */
+static struct xhci_generic_trb *queue_trb(struct xhci_ctrl *ctrl,
+					  struct xhci_ring *ring,
+					  bool more_trbs_coming,
+					  unsigned int *trb_fields)
+{
+	struct xhci_generic_trb *trb;
+	int i;
+
+	trb = &ring->enqueue->generic;
+
+	for (i = 0; i < 4; i++)
+		trb->field[i] = cpu_to_le32(trb_fields[i]);
+
+	xhci_flush_cache((uint32_t)trb, sizeof(struct xhci_generic_trb));
+
+	inc_enq(ctrl, ring, more_trbs_coming);
+
+	return trb;
+}
+
+/**
+ * Does various checks on the endpoint ring, and makes it ready
+ * to queue num_trbs.
+ *
+ * @param ctrl		Host controller data structure
+ * @param ep_ring	pointer to the EP Transfer Ring
+ * @param ep_state	State of the End Point
+ * @return error code in case of invalid ep_state, 0 on success
+ */
+static int prepare_ring(struct xhci_ctrl *ctrl, struct xhci_ring *ep_ring,
+							u32 ep_state)
+{
+	union xhci_trb *next = ep_ring->enqueue;
+
+	/* 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.
+		 */
+		puts("WARN urb submitted to disabled ep\n");
+		return -ENOENT;
+	case EP_STATE_ERROR:
+		puts("WARN waiting for error on ep to be cleared\n");
+		return -EINVAL;
+	case EP_STATE_HALTED:
+		puts("WARN halted endpoint, queueing URB anyway.\n");
+	case EP_STATE_STOPPED:
+	case EP_STATE_RUNNING:
+		debug("EP STATE RUNNING.\n");
+		break;
+	default:
+		puts("ERROR unknown endpoint state for ep\n");
+		return -EINVAL;
+	}
+
+	while (last_trb(ctrl, ep_ring, ep_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.
+		 */
+		next->link.control &= cpu_to_le32(~TRB_CHAIN);
+
+		next->link.control ^= cpu_to_le32(TRB_CYCLE);
+
+		xhci_flush_cache((uint32_t)next, sizeof(union xhci_trb));
+
+		/* Toggle the cycle bit after the last ring segment. */
+		if (last_trb_on_last_seg(ctrl, ep_ring,
+					ep_ring->enq_seg, next))
+			ep_ring->cycle_state = (ep_ring->cycle_state ? 0 : 1);
+		ep_ring->enq_seg = ep_ring->enq_seg->next;
+		ep_ring->enqueue = ep_ring->enq_seg->trbs;
+		next = ep_ring->enqueue;
+	}
+
+	return 0;
+}
+
+/**
+ * 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.
+ *
+ * @param ctrl		Host controller data structure
+ * @param ptr		Pointer address to write in the first two fields (opt.)
+ * @param slot_id	Slot ID to encode in the flags field (opt.)
+ * @param ep_index	Endpoint index to encode in the flags field (opt.)
+ * @param cmd		Command type to enqueue
+ * @return none
+ */
+void xhci_queue_command(struct xhci_ctrl *ctrl, u8 *ptr, u32 slot_id,
+			u32 ep_index, trb_type cmd)
+{
+	u32 fields[4];
+	u64 val_64 = (uintptr_t)ptr;
+
+	BUG_ON(prepare_ring(ctrl, ctrl->cmd_ring, EP_STATE_RUNNING));
+
+	fields[0] = lower_32_bits(val_64);
+	fields[1] = upper_32_bits(val_64);
+	fields[2] = 0;
+	fields[3] = TRB_TYPE(cmd) | EP_ID_FOR_TRB(ep_index) |
+		    SLOT_ID_FOR_TRB(slot_id) | ctrl->cmd_ring->cycle_state;
+
+	queue_trb(ctrl, ctrl->cmd_ring, false, fields);
+
+	/* Ring the command ring doorbell */
+	xhci_writel(&ctrl->dba->doorbell[0], DB_VALUE_HOST);
+}
+
+/**
+ * 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.
+ *
+ * @param remainder	remaining packets to be sent
+ * @return remainder if remainder is less than max else max
+ */
+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;
+}
+
+/**
+ * Finds out the remanining packets to be sent
+ *
+ * @param running_total	total size sent so far
+ * @param trb_buff_len	length of the TRB Buffer
+ * @param total_packet_count	total packet count
+ * @param maxpacketsize		max packet size of current pipe
+ * @param num_trbs_left		number of TRBs left to be processed
+ * @return 0 if running_total or trb_buff_len is 0, else remainder
+ */
+static u32 xhci_v1_0_td_remainder(int running_total,
+				int trb_buff_len,
+				unsigned int total_packet_count,
+				int maxpacketsize,
+				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) / maxpacketsize;
+
+	if ((total_packet_count - packets_transferred) > 31)
+		return 31 << 17;
+	return (total_packet_count - packets_transferred) << 17;
+}
+
+/**
+ * Ring the doorbell of the End Point
+ *
+ * @param udev		pointer to the USB device structure
+ * @param ep_index	index of the endpoint
+ * @param start_cycle	cycle flag of the first TRB
+ * @param start_trb	pionter to the first TRB
+ * @return none
+ */
+static void giveback_first_trb(struct usb_device *udev, int ep_index,
+				int start_cycle,
+				struct xhci_generic_trb *start_trb)
+{
+	struct xhci_ctrl *ctrl = udev->controller;
+
+	/*
+	 * Pass all the TRBs to the hardware at once and make sure this write
+	 * isn't reordered.
+	 */
+	if (start_cycle)
+		start_trb->field[3] |= cpu_to_le32(start_cycle);
+	else
+		start_trb->field[3] &= cpu_to_le32(~TRB_CYCLE);
+
+	xhci_flush_cache((uint32_t)start_trb, sizeof(struct xhci_generic_trb));
+
+	/* Ringing EP doorbell here */
+	xhci_writel(&ctrl->dba->doorbell[udev->slot_id],
+				DB_VALUE(ep_index, 0));
+
+	return;
+}
+
+/**** POLLING mechanism for XHCI ****/
+
+/**
+ * Finalizes a handled event TRB by advancing our dequeue pointer and giving
+ * the TRB back to the hardware for recycling. Must call this exactly once at
+ * the end of each event handler, and not touch the TRB again afterwards.
+ *
+ * @param ctrl	Host controller data structure
+ * @return none
+ */
+void xhci_acknowledge_event(struct xhci_ctrl *ctrl)
+{
+	/* Advance our dequeue pointer to the next event */
+	inc_deq(ctrl, ctrl->event_ring);
+
+	/* Inform the hardware */
+	xhci_writeq(&ctrl->ir_set->erst_dequeue,
+		(uintptr_t)ctrl->event_ring->dequeue | ERST_EHB);
+}
+
+/**
+ * Checks if there is a new event to handle on the event ring.
+ *
+ * @param ctrl	Host controller data structure
+ * @return 0 if failure else 1 on success
+ */
+static int event_ready(struct xhci_ctrl *ctrl)
+{
+	union xhci_trb *event;
+
+	xhci_inval_cache((uint32_t)ctrl->event_ring->dequeue,
+					sizeof(union xhci_trb));
+
+	event = ctrl->event_ring->dequeue;
+
+	/* Does the HC or OS own the TRB? */
+	if ((le32_to_cpu(event->event_cmd.flags) & TRB_CYCLE) !=
+		ctrl->event_ring->cycle_state)
+		return 0;
+
+	return 1;
+}
+
+/**
+ * Waits for a specific type of event and returns it. Discards unexpected
+ * events. Caller *must* call xhci_acknowledge_event() after it is finished
+ * processing the event, and must not access the returned pointer afterwards.
+ *
+ * @param ctrl		Host controller data structure
+ * @param expected	TRB type expected from Event TRB
+ * @return pointer to event trb
+ */
+union xhci_trb *xhci_wait_for_event(struct xhci_ctrl *ctrl, trb_type expected)
+{
+	trb_type type;
+	unsigned long ts = get_timer(0);
+
+	do {
+		union xhci_trb *event = ctrl->event_ring->dequeue;
+
+		if (!event_ready(ctrl))
+			continue;
+
+		type = TRB_FIELD_TO_TYPE(le32_to_cpu(event->event_cmd.flags));
+		if (type == expected)
+			return event;
+
+		if (type == TRB_PORT_STATUS)
+		/* TODO: remove this once enumeration has been reworked */
+			/*
+			 * Port status change events always have a
+			 * successful completion code
+			 */
+			BUG_ON(GET_COMP_CODE(
+				le32_to_cpu(event->generic.field[2])) !=
+								COMP_SUCCESS);
+		else
+			printf("Unexpected XHCI event TRB, skipping... "
+				"(%08x %08x %08x %08x)\n",
+				le32_to_cpu(event->generic.field[0]),
+				le32_to_cpu(event->generic.field[1]),
+				le32_to_cpu(event->generic.field[2]),
+				le32_to_cpu(event->generic.field[3]));
+
+		xhci_acknowledge_event(ctrl);
+	} while (get_timer(ts) < XHCI_TIMEOUT);
+
+	if (expected == TRB_TRANSFER)
+		return NULL;
+
+	printf("XHCI timeout on event type %d... cannot recover.\n", expected);
+	BUG();
+}
+
+/*
+ * Stops transfer processing for an endpoint and throws away all unprocessed
+ * TRBs by setting the xHC's dequeue pointer to our enqueue pointer. The next
+ * xhci_bulk_tx/xhci_ctrl_tx on this enpoint will add new transfers there and
+ * ring the doorbell, causing this endpoint to start working again.
+ * (Careful: This will BUG() when there was no transfer in progress. Shouldn't
+ * happen in practice for current uses and is too complicated to fix right now.)
+ */
+static void abort_td(struct usb_device *udev, int ep_index)
+{
+	struct xhci_ctrl *ctrl = udev->controller;
+	struct xhci_ring *ring =  ctrl->devs[udev->slot_id]->eps[ep_index].ring;
+	union xhci_trb *event;
+	u32 field;
+
+	xhci_queue_command(ctrl, NULL, udev->slot_id, ep_index, TRB_STOP_RING);
+
+	event = xhci_wait_for_event(ctrl, TRB_TRANSFER);
+	field = le32_to_cpu(event->trans_event.flags);
+	BUG_ON(TRB_TO_SLOT_ID(field) != udev->slot_id);
+	BUG_ON(TRB_TO_EP_INDEX(field) != ep_index);
+	BUG_ON(GET_COMP_CODE(le32_to_cpu(event->trans_event.transfer_len
+		!= COMP_STOP)));
+	xhci_acknowledge_event(ctrl);
+
+	event = xhci_wait_for_event(ctrl, TRB_COMPLETION);
+	BUG_ON(TRB_TO_SLOT_ID(le32_to_cpu(event->event_cmd.flags))
+		!= udev->slot_id || GET_COMP_CODE(le32_to_cpu(
+		event->event_cmd.status)) != COMP_SUCCESS);
+	xhci_acknowledge_event(ctrl);
+
+	xhci_queue_command(ctrl, (void *)((uintptr_t)ring->enqueue |
+		ring->cycle_state), udev->slot_id, ep_index, TRB_SET_DEQ);
+	event = xhci_wait_for_event(ctrl, TRB_COMPLETION);
+	BUG_ON(TRB_TO_SLOT_ID(le32_to_cpu(event->event_cmd.flags))
+		!= udev->slot_id || GET_COMP_CODE(le32_to_cpu(
+		event->event_cmd.status)) != COMP_SUCCESS);
+	xhci_acknowledge_event(ctrl);
+}
+
+static void record_transfer_result(struct usb_device *udev,
+				   union xhci_trb *event, int length)
+{
+	udev->act_len = min(length, length -
+		EVENT_TRB_LEN(le32_to_cpu(event->trans_event.transfer_len)));
+
+	switch (GET_COMP_CODE(le32_to_cpu(event->trans_event.transfer_len))) {
+	case COMP_SUCCESS:
+		BUG_ON(udev->act_len != length);
+		/* fallthrough */
+	case COMP_SHORT_TX:
+		udev->status = 0;
+		break;
+	case COMP_STALL:
+		udev->status = USB_ST_STALLED;
+		break;
+	case COMP_DB_ERR:
+	case COMP_TRB_ERR:
+		udev->status = USB_ST_BUF_ERR;
+		break;
+	case COMP_BABBLE:
+		udev->status = USB_ST_BABBLE_DET;
+		break;
+	default:
+		udev->status = 0x80;  /* USB_ST_TOO_LAZY_TO_MAKE_A_NEW_MACRO */
+	}
+}
+
+/**** Bulk and Control transfer methods ****/
+/**
+ * Queues up the BULK Request
+ *
+ * @param udev		pointer to the USB device structure
+ * @param pipe		contains the DIR_IN or OUT , devnum
+ * @param length	length of the buffer
+ * @param buffer	buffer to be read/written based on the request
+ * @return returns 0 if successful else -1 on failure
+ */
+int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
+			int length, void *buffer)
+{
+	int num_trbs = 0;
+	struct xhci_generic_trb *start_trb;
+	bool first_trb = 0;
+	int start_cycle;
+	u32 field = 0;
+	u32 length_field = 0;
+	struct xhci_ctrl *ctrl = udev->controller;
+	int slot_id = udev->slot_id;
+	int ep_index;
+	struct xhci_virt_device *virt_dev;
+	struct xhci_ep_ctx *ep_ctx;
+	struct xhci_ring *ring;		/* EP transfer ring */
+	union xhci_trb *event;
+
+	int running_total, trb_buff_len;
+	unsigned int total_packet_count;
+	int maxpacketsize;
+	u64 addr;
+	int ret;
+	u32 trb_fields[4];
+	u64 val_64 = (uintptr_t)buffer;
+
+	debug("dev=%p, pipe=%lx, buffer=%p, length=%d\n",
+		udev, pipe, buffer, length);
+
+	ep_index = usb_pipe_ep_index(pipe);
+	virt_dev = ctrl->devs[slot_id];
+
+	xhci_inval_cache((uint32_t)virt_dev->out_ctx->bytes,
+					virt_dev->out_ctx->size);
+
+	ep_ctx = xhci_get_ep_ctx(ctrl, virt_dev->out_ctx, ep_index);
+
+	ring = virt_dev->eps[ep_index].ring;
+	/*
+	 * How much data is (potentially) left before the 64KB boundary?
+	 * XHCI Spec puts restriction( TABLE 49 and 6.4.1 section of XHCI Spec)
+	 * that the buffer should not span 64KB boundary. if so
+	 * we send request in more than 1 TRB by chaining them.
+	 */
+	running_total = TRB_MAX_BUFF_SIZE -
+			(lower_32_bits(val_64) & (TRB_MAX_BUFF_SIZE - 1));
+	trb_buff_len = running_total;
+	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 || length == 0)
+		num_trbs++;
+
+	/* How many more 64KB chunks to transfer, how many more TRBs? */
+	while (running_total < length) {
+		num_trbs++;
+		running_total += TRB_MAX_BUFF_SIZE;
+	}
+
+	/*
+	 * XXX: Calling routine prepare_ring() called in place of
+	 * prepare_trasfer() as there in 'Linux' since we are not
+	 * maintaining multiple TDs/transfer at the same time.
+	 */
+	ret = prepare_ring(ctrl, ring,
+			   le32_to_cpu(ep_ctx->ep_info) & EP_STATE_MASK);
+	if (ret < 0)
+		return ret;
+
+	/*
+	 * 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 = &ring->enqueue->generic;
+	start_cycle = ring->cycle_state;
+
+	running_total = 0;
+	maxpacketsize = usb_maxpacket(udev, pipe);
+
+	total_packet_count = DIV_ROUND_UP(length, maxpacketsize);
+
+	/* How much data is in the first TRB? */
+	/*
+	 * How much data is (potentially) left before the 64KB boundary?
+	 * XHCI Spec puts restriction( TABLE 49 and 6.4.1 section of XHCI Spec)
+	 * that the buffer should not span 64KB boundary. if so
+	 * we send request in more than 1 TRB by chaining them.
+	 */
+	addr = val_64;
+
+	if (trb_buff_len > length)
+		trb_buff_len = length;
+
+	first_trb = true;
+
+	/* flush the buffer before use */
+	xhci_flush_cache((uint32_t)buffer, length);
+
+	/* 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 |= TRB_CYCLE;
+		} else {
+			field |= 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
+			field |= TRB_IOC;
+
+		/* Only set interrupt on short packet for IN endpoints */
+		if (usb_pipein(pipe))
+			field |= TRB_ISP;
+
+		/* Set the TRB length, TD size, and interrupter fields. */
+		if (HC_VERSION(xhci_readl(&ctrl->hccr->cr_capbase)) < 0x100)
+			remainder = xhci_td_remainder(length - running_total);
+		else
+			remainder = xhci_v1_0_td_remainder(running_total,
+							   trb_buff_len,
+							   total_packet_count,
+							   maxpacketsize,
+							   num_trbs - 1);
+
+		length_field = ((trb_buff_len & TRB_LEN_MASK) |
+				remainder |
+				((0 & TRB_INTR_TARGET_MASK) <<
+				TRB_INTR_TARGET_SHIFT));
+
+		trb_fields[0] = lower_32_bits(addr);
+		trb_fields[1] = upper_32_bits(addr);
+		trb_fields[2] = length_field;
+		trb_fields[3] = field | (TRB_NORMAL << TRB_TYPE_SHIFT);
+
+		queue_trb(ctrl, ring, (num_trbs > 1), trb_fields);
+
+		--num_trbs;
+
+		running_total += trb_buff_len;
+
+		/* Calculate length for next transfer */
+		addr += trb_buff_len;
+		trb_buff_len = min((length - running_total), TRB_MAX_BUFF_SIZE);
+	} while (running_total < length);
+
+	giveback_first_trb(udev, ep_index, start_cycle, start_trb);
+
+	event = xhci_wait_for_event(ctrl, TRB_TRANSFER);
+	if (!event) {
+		debug("XHCI bulk transfer timed out, aborting...\n");
+		abort_td(udev, ep_index);
+		udev->status = USB_ST_NAK_REC;  /* closest thing to a timeout */
+		udev->act_len = 0;
+		return -ETIMEDOUT;
+	}
+	field = le32_to_cpu(event->trans_event.flags);
+
+	BUG_ON(TRB_TO_SLOT_ID(field) != slot_id);
+	BUG_ON(TRB_TO_EP_INDEX(field) != ep_index);
+	BUG_ON(*(void **)(uintptr_t)le64_to_cpu(event->trans_event.buffer) -
+		buffer > (size_t)length);
+
+	record_transfer_result(udev, event, length);
+	xhci_acknowledge_event(ctrl);
+	xhci_inval_cache((uint32_t)buffer, length);
+
+	return (udev->status != USB_ST_NOT_PROC) ? 0 : -1;
+}
+
+/**
+ * Queues up the Control Transfer Request
+ *
+ * @param udev	pointer to the USB device structure
+ * @param pipe		contains the DIR_IN or OUT , devnum
+ * @param req		request type
+ * @param length	length of the buffer
+ * @param buffer	buffer to be read/written based on the request
+ * @return returns 0 if successful else error code on failure
+ */
+int xhci_ctrl_tx(struct usb_device *udev, unsigned long pipe,
+			struct devrequest *req,	int length,
+			void *buffer)
+{
+	int ret;
+	int start_cycle;
+	int num_trbs;
+	u32 field;
+	u32 length_field;
+	u64 buf_64 = 0;
+	struct xhci_generic_trb *start_trb;
+	struct xhci_ctrl *ctrl = udev->controller;
+	int slot_id = udev->slot_id;
+	int ep_index;
+	u32 trb_fields[4];
+	struct xhci_virt_device *virt_dev = ctrl->devs[slot_id];
+	struct xhci_ring *ep_ring;
+	union xhci_trb *event;
+
+	debug("req=%u (%#x), type=%u (%#x), value=%u (%#x), index=%u\n",
+		req->request, req->request,
+		req->requesttype, req->requesttype,
+		le16_to_cpu(req->value), le16_to_cpu(req->value),
+		le16_to_cpu(req->index));
+
+	ep_index = usb_pipe_ep_index(pipe);
+
+	ep_ring = virt_dev->eps[ep_index].ring;
+
+	/*
+	 * Check to see if the max packet size for the default control
+	 * endpoint changed during FS device enumeration
+	 */
+	if (udev->speed == USB_SPEED_FULL) {
+		ret = xhci_check_maxpacket(udev);
+		if (ret < 0)
+			return ret;
+	}
+
+	xhci_inval_cache((uint32_t)virt_dev->out_ctx->bytes,
+				virt_dev->out_ctx->size);
+
+	struct xhci_ep_ctx *ep_ctx = NULL;
+	ep_ctx = xhci_get_ep_ctx(ctrl, virt_dev->out_ctx, ep_index);
+
+	/* 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 (length > 0)
+		num_trbs++;
+	/*
+	 * XXX: Calling routine prepare_ring() called in place of
+	 * prepare_trasfer() as there in 'Linux' since we are not
+	 * maintaining multiple TDs/transfer at the same time.
+	 */
+	ret = prepare_ring(ctrl, ep_ring,
+				le32_to_cpu(ep_ctx->ep_info) & EP_STATE_MASK);
+
+	if (ret < 0)
+		return ret;
+
+	/*
+	 * 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;
+
+	debug("start_trb %p, start_cycle %d\n", start_trb, start_cycle);
+
+	/* Queue setup TRB - see section 6.4.1.2.1 */
+	/* FIXME better way to translate setup_packet into two u32 fields? */
+	field = 0;
+	field |= TRB_IDT | (TRB_SETUP << TRB_TYPE_SHIFT);
+	if (start_cycle == 0)
+		field |= 0x1;
+
+	/* xHCI 1.0 6.4.1.2.1: Transfer Type field */
+	if (HC_VERSION(xhci_readl(&ctrl->hccr->cr_capbase)) == 0x100) {
+		if (length > 0) {
+			if (req->requesttype & USB_DIR_IN)
+				field |= (TRB_DATA_IN << TRB_TX_TYPE_SHIFT);
+			else
+				field |= (TRB_DATA_OUT << TRB_TX_TYPE_SHIFT);
+		}
+	}
+
+	debug("req->requesttype = %d, req->request = %d,"
+		"le16_to_cpu(req->value) = %d,"
+		"le16_to_cpu(req->index) = %d,"
+		"le16_to_cpu(req->length) = %d\n",
+		req->requesttype, req->request, le16_to_cpu(req->value),
+		le16_to_cpu(req->index), le16_to_cpu(req->length));
+
+	trb_fields[0] = req->requesttype | req->request << 8 |
+				le16_to_cpu(req->value) << 16;
+	trb_fields[1] = le16_to_cpu(req->index) |
+			le16_to_cpu(req->length) << 16;
+	/* TRB_LEN | (TRB_INTR_TARGET) */
+	trb_fields[2] = (8 | ((0 & TRB_INTR_TARGET_MASK) <<
+			TRB_INTR_TARGET_SHIFT));
+	/* Immediate data in pointer */
+	trb_fields[3] = field;
+	queue_trb(ctrl, ep_ring, true, trb_fields);
+
+	/* Re-initializing field to zero */
+	field = 0;
+	/* If there's data, queue data TRBs */
+	/* Only set interrupt on short packet for IN endpoints */
+	if (usb_pipein(pipe))
+		field = TRB_ISP | (TRB_DATA << TRB_TYPE_SHIFT);
+	else
+		field = (TRB_DATA << TRB_TYPE_SHIFT);
+
+	length_field = (length & TRB_LEN_MASK) | xhci_td_remainder(length) |
+			((0 & TRB_INTR_TARGET_MASK) << TRB_INTR_TARGET_SHIFT);
+	debug("length_field = %d, length = %d,"
+		"xhci_td_remainder(length) = %d , TRB_INTR_TARGET(0) = %d\n",
+		length_field, (length & TRB_LEN_MASK),
+		xhci_td_remainder(length), 0);
+
+	if (length > 0) {
+		if (req->requesttype & USB_DIR_IN)
+			field |= TRB_DIR_IN;
+		buf_64 = (uintptr_t)buffer;
+
+		trb_fields[0] = lower_32_bits(buf_64);
+		trb_fields[1] = upper_32_bits(buf_64);
+		trb_fields[2] = length_field;
+		trb_fields[3] = field | ep_ring->cycle_state;
+
+		xhci_flush_cache((uint32_t)buffer, length);
+		queue_trb(ctrl, ep_ring, true, trb_fields);
+	}
+
+	/*
+	 * 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 */
+	field = 0;
+	if (length > 0 && req->requesttype & USB_DIR_IN)
+		field = 0;
+	else
+		field = TRB_DIR_IN;
+
+	trb_fields[0] = 0;
+	trb_fields[1] = 0;
+	trb_fields[2] = ((0 & TRB_INTR_TARGET_MASK) << TRB_INTR_TARGET_SHIFT);
+		/* Event on completion */
+	trb_fields[3] = field | TRB_IOC |
+			(TRB_STATUS << TRB_TYPE_SHIFT) |
+			ep_ring->cycle_state;
+
+	queue_trb(ctrl, ep_ring, false, trb_fields);
+
+	giveback_first_trb(udev, ep_index, start_cycle, start_trb);
+
+	event = xhci_wait_for_event(ctrl, TRB_TRANSFER);
+	if (!event)
+		goto abort;
+	field = le32_to_cpu(event->trans_event.flags);
+
+	BUG_ON(TRB_TO_SLOT_ID(field) != slot_id);
+	BUG_ON(TRB_TO_EP_INDEX(field) != ep_index);
+
+	record_transfer_result(udev, event, length);
+	xhci_acknowledge_event(ctrl);
+
+	/* Invalidate buffer to make it available to usb-core */
+	if (length > 0)
+		xhci_inval_cache((uint32_t)buffer, length);
+
+	if (GET_COMP_CODE(le32_to_cpu(event->trans_event.transfer_len))
+			== COMP_SHORT_TX) {
+		/* Short data stage, clear up additional status stage event */
+		event = xhci_wait_for_event(ctrl, TRB_TRANSFER);
+		if (!event)
+			goto abort;
+		BUG_ON(TRB_TO_SLOT_ID(field) != slot_id);
+		BUG_ON(TRB_TO_EP_INDEX(field) != ep_index);
+		xhci_acknowledge_event(ctrl);
+	}
+
+	return (udev->status != USB_ST_NOT_PROC) ? 0 : -1;
+
+abort:
+	debug("XHCI control transfer timed out, aborting...\n");
+	abort_td(udev, ep_index);
+	udev->status = USB_ST_NAK_REC;
+	udev->act_len = 0;
+	return -ETIMEDOUT;
+}
diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
new file mode 100644
index 0000000000..57fa4ff9a7
--- /dev/null
+++ b/drivers/usb/host/xhci.c
@@ -0,0 +1,1030 @@
+/*
+ * USB HOST XHCI Controller stack
+ *
+ * Based on xHCI host controller driver in linux-kernel
+ * by Sarah Sharp.
+ *
+ * Copyright (C) 2008 Intel Corp.
+ * Author: Sarah Sharp
+ *
+ * Copyright (C) 2013 Samsung Electronics Co.Ltd
+ * Authors: Vivek Gautam <gautam.vivek@samsung.com>
+ *	    Vikas Sajjan <vikas.sajjan@samsung.com>
+ *
+ * SPDX-License-Identifier:	GPL-2.0+
+ */
+
+/**
+ * This file gives the xhci stack for usb3.0 looking into
+ * xhci specification Rev1.0 (5/21/10).
+ * The quirk devices support hasn't been given yet.
+ */
+
+#include <common.h>
+#include <asm/byteorder.h>
+#include <usb.h>
+#include <malloc.h>
+#include <watchdog.h>
+#include <asm/cache.h>
+#include <asm/unaligned.h>
+#include <asm-generic/errno.h>
+#include "xhci.h"
+
+#ifndef CONFIG_USB_MAX_CONTROLLER_COUNT
+#define CONFIG_USB_MAX_CONTROLLER_COUNT 1
+#endif
+
+static struct descriptor {
+	struct usb_hub_descriptor hub;
+	struct usb_device_descriptor device;
+	struct usb_config_descriptor config;
+	struct usb_interface_descriptor interface;
+	struct usb_endpoint_descriptor endpoint;
+	struct usb_ss_ep_comp_descriptor ep_companion;
+} __attribute__ ((packed)) descriptor = {
+	{
+		0xc,		/* bDescLength */
+		0x2a,		/* bDescriptorType: hub descriptor */
+		2,		/* bNrPorts -- runtime modified */
+		cpu_to_le16(0x8), /* wHubCharacteristics */
+		10,		/* bPwrOn2PwrGood */
+		0,		/* bHubCntrCurrent */
+		{},		/* Device removable */
+		{}		/* at most 7 ports! XXX */
+	},
+	{
+		0x12,		/* bLength */
+		1,		/* bDescriptorType: UDESC_DEVICE */
+		cpu_to_le16(0x0300), /* bcdUSB: v3.0 */
+		9,		/* bDeviceClass: UDCLASS_HUB */
+		0,		/* bDeviceSubClass: UDSUBCLASS_HUB */
+		3,		/* bDeviceProtocol: UDPROTO_SSHUBSTT */
+		9,		/* bMaxPacketSize: 512 bytes  2^9 */
+		0x0000,		/* idVendor */
+		0x0000,		/* idProduct */
+		cpu_to_le16(0x0100), /* bcdDevice */
+		1,		/* iManufacturer */
+		2,		/* iProduct */
+		0,		/* iSerialNumber */
+		1		/* bNumConfigurations: 1 */
+	},
+	{
+		0x9,
+		2,		/* bDescriptorType: UDESC_CONFIG */
+		cpu_to_le16(0x1f), /* includes SS endpoint descriptor */
+		1,		/* bNumInterface */
+		1,		/* bConfigurationValue */
+		0,		/* iConfiguration */
+		0x40,		/* bmAttributes: UC_SELF_POWER */
+		0		/* bMaxPower */
+	},
+	{
+		0x9,		/* bLength */
+		4,		/* bDescriptorType: UDESC_INTERFACE */
+		0,		/* bInterfaceNumber */
+		0,		/* bAlternateSetting */
+		1,		/* bNumEndpoints */
+		9,		/* bInterfaceClass: UICLASS_HUB */
+		0,		/* bInterfaceSubClass: UISUBCLASS_HUB */
+		0,		/* bInterfaceProtocol: UIPROTO_HSHUBSTT */
+		0		/* iInterface */
+	},
+	{
+		0x7,		/* bLength */
+		5,		/* bDescriptorType: UDESC_ENDPOINT */
+		0x81,		/* bEndpointAddress: IN endpoint 1 */
+		3,		/* bmAttributes: UE_INTERRUPT */
+		8,		/* wMaxPacketSize */
+		255		/* bInterval */
+	},
+	{
+		0x06,		/* ss_bLength */
+		0x30,		/* ss_bDescriptorType: SS EP Companion */
+		0x00,		/* ss_bMaxBurst: allows 1 TX between ACKs */
+		/* ss_bmAttributes: 1 packet per service interval */
+		0x00,
+		/* ss_wBytesPerInterval: 15 bits for max 15 ports */
+		cpu_to_le16(0x02),
+	},
+};
+
+static struct xhci_ctrl xhcic[CONFIG_USB_MAX_CONTROLLER_COUNT];
+
+/**
+ * Waits for as per specified amount of time
+ * for the "result" to match with "done"
+ *
+ * @param ptr	pointer to the register to be read
+ * @param mask	mask for the value read
+ * @param done	value to be campared with result
+ * @param usec	time to wait till
+ * @return 0 if handshake is success else < 0 on failure
+ */
+static int handshake(uint32_t volatile *ptr, uint32_t mask,
+					uint32_t done, int usec)
+{
+	uint32_t result;
+
+	do {
+		result = xhci_readl(ptr);
+		if (result == ~(uint32_t)0)
+			return -ENODEV;
+		result &= mask;
+		if (result == done)
+			return 0;
+		usec--;
+		udelay(1);
+	} while (usec > 0);
+
+	return -ETIMEDOUT;
+}
+
+/**
+ * Set the run bit and wait for the host to be running.
+ *
+ * @param hcor	pointer to host controller operation registers
+ * @return status of the Handshake
+ */
+static int xhci_start(struct xhci_hcor *hcor)
+{
+	u32 temp;
+	int ret;
+
+	puts("Starting the controller\n");
+	temp = xhci_readl(&hcor->or_usbcmd);
+	temp |= (CMD_RUN);
+	xhci_writel(&hcor->or_usbcmd, temp);
+
+	/*
+	 * Wait for the HCHalted Status bit to be 0 to indicate the host is
+	 * running.
+	 */
+	ret = handshake(&hcor->or_usbsts, STS_HALT, 0, XHCI_MAX_HALT_USEC);
+	if (ret)
+		debug("Host took too long to start, "
+				"waited %u microseconds.\n",
+				XHCI_MAX_HALT_USEC);
+	return ret;
+}
+
+/**
+ * Resets the XHCI Controller
+ *
+ * @param hcor	pointer to host controller operation registers
+ * @return -EBUSY if XHCI Controller is not halted else status of handshake
+ */
+int xhci_reset(struct xhci_hcor *hcor)
+{
+	u32 cmd;
+	u32 state;
+	int ret;
+
+	/* Halting the Host first */
+	debug("// Halt the HC\n");
+	state = xhci_readl(&hcor->or_usbsts) & STS_HALT;
+	if (!state) {
+		cmd = xhci_readl(&hcor->or_usbcmd);
+		cmd &= ~CMD_RUN;
+		xhci_writel(&hcor->or_usbcmd, cmd);
+	}
+
+	ret = handshake(&hcor->or_usbsts,
+			STS_HALT, STS_HALT, XHCI_MAX_HALT_USEC);
+	if (ret) {
+		printf("Host not halted after %u microseconds.\n",
+				XHCI_MAX_HALT_USEC);
+		return -EBUSY;
+	}
+
+	debug("// Reset the HC\n");
+	cmd = xhci_readl(&hcor->or_usbcmd);
+	cmd |= CMD_RESET;
+	xhci_writel(&hcor->or_usbcmd, cmd);
+
+	ret = handshake(&hcor->or_usbcmd, CMD_RESET, 0, XHCI_MAX_RESET_USEC);
+	if (ret)
+		return ret;
+
+	/*
+	 * xHCI cannot write to any doorbells or operational registers other
+	 * than status until the "Controller Not Ready" flag is cleared.
+	 */
+	return handshake(&hcor->or_usbsts, STS_CNR, 0, XHCI_MAX_RESET_USEC);
+}
+
+/**
+ * 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)
+ *
+ * @param desc	USB enpdoint Descriptor
+ * @return index of the Endpoint
+ */
+static unsigned int xhci_get_ep_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) ? 0 : 1));
+
+	return index;
+}
+
+/**
+ * Issue a configure endpoint command or evaluate context command
+ * and wait for it to finish.
+ *
+ * @param udev	pointer to the Device Data Structure
+ * @param ctx_change	flag to indicate the Context has changed or NOT
+ * @return 0 on success, -1 on failure
+ */
+static int xhci_configure_endpoints(struct usb_device *udev, bool ctx_change)
+{
+	struct xhci_container_ctx *in_ctx;
+	struct xhci_virt_device *virt_dev;
+	struct xhci_ctrl *ctrl = udev->controller;
+	union xhci_trb *event;
+
+	virt_dev = ctrl->devs[udev->slot_id];
+	in_ctx = virt_dev->in_ctx;
+
+	xhci_flush_cache((uint32_t)in_ctx->bytes, in_ctx->size);
+	xhci_queue_command(ctrl, in_ctx->bytes, udev->slot_id, 0,
+			   ctx_change ? TRB_EVAL_CONTEXT : TRB_CONFIG_EP);
+	event = xhci_wait_for_event(ctrl, TRB_COMPLETION);
+	BUG_ON(TRB_TO_SLOT_ID(le32_to_cpu(event->event_cmd.flags))
+		!= udev->slot_id);
+
+	switch (GET_COMP_CODE(le32_to_cpu(event->event_cmd.status))) {
+	case COMP_SUCCESS:
+		debug("Successful %s command\n",
+			ctx_change ? "Evaluate Context" : "Configure Endpoint");
+		break;
+	default:
+		printf("ERROR: %s command returned completion code %d.\n",
+			ctx_change ? "Evaluate Context" : "Configure Endpoint",
+			GET_COMP_CODE(le32_to_cpu(event->event_cmd.status)));
+		return -EINVAL;
+	}
+
+	xhci_acknowledge_event(ctrl);
+
+	return 0;
+}
+
+/**
+ * Configure the endpoint, programming the device contexts.
+ *
+ * @param udev	pointer to the USB device structure
+ * @return returns the status of the xhci_configure_endpoints
+ */
+static int xhci_set_configuration(struct usb_device *udev)
+{
+	struct xhci_container_ctx *in_ctx;
+	struct xhci_container_ctx *out_ctx;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	struct xhci_slot_ctx *slot_ctx;
+	struct xhci_ep_ctx *ep_ctx[MAX_EP_CTX_NUM];
+	int cur_ep;
+	int max_ep_flag = 0;
+	int ep_index;
+	unsigned int dir;
+	unsigned int ep_type;
+	struct xhci_ctrl *ctrl = udev->controller;
+	int num_of_ep;
+	int ep_flag = 0;
+	u64 trb_64 = 0;
+	int slot_id = udev->slot_id;
+	struct xhci_virt_device *virt_dev = ctrl->devs[slot_id];
+	struct usb_interface *ifdesc;
+
+	out_ctx = virt_dev->out_ctx;
+	in_ctx = virt_dev->in_ctx;
+
+	num_of_ep = udev->config.if_desc[0].no_of_ep;
+	ifdesc = &udev->config.if_desc[0];
+
+	ctrl_ctx = xhci_get_input_control_ctx(in_ctx);
+	/* Zero the input context control */
+	ctrl_ctx->add_flags = 0;
+	ctrl_ctx->drop_flags = 0;
+
+	/* EP_FLAG gives values 1 & 4 for EP1OUT and EP2IN */
+	for (cur_ep = 0; cur_ep < num_of_ep; cur_ep++) {
+		ep_flag = xhci_get_ep_index(&ifdesc->ep_desc[cur_ep]);
+		ctrl_ctx->add_flags |= cpu_to_le32(1 << (ep_flag + 1));
+		if (max_ep_flag < ep_flag)
+			max_ep_flag = ep_flag;
+	}
+
+	xhci_inval_cache((uint32_t)out_ctx->bytes, out_ctx->size);
+
+	/* slot context */
+	xhci_slot_copy(ctrl, in_ctx, out_ctx);
+	slot_ctx = xhci_get_slot_ctx(ctrl, in_ctx);
+	slot_ctx->dev_info &= ~(LAST_CTX_MASK);
+	slot_ctx->dev_info |= cpu_to_le32(LAST_CTX(max_ep_flag + 1) | 0);
+
+	xhci_endpoint_copy(ctrl, in_ctx, out_ctx, 0);
+
+	/* filling up ep contexts */
+	for (cur_ep = 0; cur_ep < num_of_ep; cur_ep++) {
+		struct usb_endpoint_descriptor *endpt_desc = NULL;
+
+		endpt_desc = &ifdesc->ep_desc[cur_ep];
+		trb_64 = 0;
+
+		ep_index = xhci_get_ep_index(endpt_desc);
+		ep_ctx[ep_index] = xhci_get_ep_ctx(ctrl, in_ctx, ep_index);
+
+		/* Allocate the ep rings */
+		virt_dev->eps[ep_index].ring = xhci_ring_alloc(1, true);
+		if (!virt_dev->eps[ep_index].ring)
+			return -ENOMEM;
+
+		/*NOTE: ep_desc[0] actually represents EP1 and so on */
+		dir = (((endpt_desc->bEndpointAddress) & (0x80)) >> 7);
+		ep_type = (((endpt_desc->bmAttributes) & (0x3)) | (dir << 2));
+		ep_ctx[ep_index]->ep_info2 =
+			cpu_to_le32(ep_type << EP_TYPE_SHIFT);
+		ep_ctx[ep_index]->ep_info2 |=
+			cpu_to_le32(MAX_PACKET
+			(get_unaligned(&endpt_desc->wMaxPacketSize)));
+
+		ep_ctx[ep_index]->ep_info2 |=
+			cpu_to_le32(((0 & MAX_BURST_MASK) << MAX_BURST_SHIFT) |
+			((3 & ERROR_COUNT_MASK) << ERROR_COUNT_SHIFT));
+
+		trb_64 = (uintptr_t)
+				virt_dev->eps[ep_index].ring->enqueue;
+		ep_ctx[ep_index]->deq = cpu_to_le64(trb_64 |
+				virt_dev->eps[ep_index].ring->cycle_state);
+	}
+
+	return xhci_configure_endpoints(udev, false);
+}
+
+/**
+ * Issue an Address Device command (which will issue a SetAddress request to
+ * the device).
+ *
+ * @param udev pointer to the Device Data Structure
+ * @return 0 if successful else error code on failure
+ */
+static int xhci_address_device(struct usb_device *udev)
+{
+	int ret = 0;
+	struct xhci_ctrl *ctrl = udev->controller;
+	struct xhci_slot_ctx *slot_ctx;
+	struct xhci_input_control_ctx *ctrl_ctx;
+	struct xhci_virt_device *virt_dev;
+	int slot_id = udev->slot_id;
+	union xhci_trb *event;
+
+	virt_dev = ctrl->devs[slot_id];
+
+	/*
+	 * This is the first Set Address since device plug-in
+	 * so setting up the slot context.
+	 */
+	debug("Setting up addressable devices\n");
+	xhci_setup_addressable_virt_dev(udev);
+
+	ctrl_ctx = xhci_get_input_control_ctx(virt_dev->in_ctx);
+	ctrl_ctx->add_flags = cpu_to_le32(SLOT_FLAG | EP0_FLAG);
+	ctrl_ctx->drop_flags = 0;
+
+	xhci_queue_command(ctrl, (void *)ctrl_ctx, slot_id, 0, TRB_ADDR_DEV);
+	event = xhci_wait_for_event(ctrl, TRB_COMPLETION);
+	BUG_ON(TRB_TO_SLOT_ID(le32_to_cpu(event->event_cmd.flags)) != slot_id);
+
+	switch (GET_COMP_CODE(le32_to_cpu(event->event_cmd.status))) {
+	case COMP_CTX_STATE:
+	case COMP_EBADSLT:
+		printf("Setup ERROR: address device command for slot %d.\n",
+								slot_id);
+		ret = -EINVAL;
+		break;
+	case COMP_TX_ERR:
+		puts("Device not responding to set address.\n");
+		ret = -EPROTO;
+		break;
+	case COMP_DEV_ERR:
+		puts("ERROR: Incompatible device"
+					"for address device command.\n");
+		ret = -ENODEV;
+		break;
+	case COMP_SUCCESS:
+		debug("Successful Address Device command\n");
+		udev->status = 0;
+		break;
+	default:
+		printf("ERROR: unexpected command completion code 0x%x.\n",
+			GET_COMP_CODE(le32_to_cpu(event->event_cmd.status)));
+		ret = -EINVAL;
+		break;
+	}
+
+	xhci_acknowledge_event(ctrl);
+
+	if (ret < 0)
+		/*
+		 * TODO: Unsuccessful Address Device command shall leave the
+		 * slot in default state. So, issue Disable Slot command now.
+		 */
+		return ret;
+
+	xhci_inval_cache((uint32_t)virt_dev->out_ctx->bytes,
+				virt_dev->out_ctx->size);
+	slot_ctx = xhci_get_slot_ctx(ctrl, virt_dev->out_ctx);
+
+	debug("xHC internal address is: %d\n",
+		le32_to_cpu(slot_ctx->dev_state) & DEV_ADDR_MASK);
+
+	return 0;
+}
+
+/**
+ * Issue Enable slot command to the controller to allocate
+ * device slot and assign the slot id. It fails if the xHC
+ * ran out of device slots, the Enable Slot command timed out,
+ * or allocating memory failed.
+ *
+ * @param udev	pointer to the Device Data Structure
+ * @return Returns 0 on succes else return error code on failure
+ */
+int usb_alloc_device(struct usb_device *udev)
+{
+	union xhci_trb *event;
+	struct xhci_ctrl *ctrl = udev->controller;
+	int ret;
+
+	/*
+	 * Root hub will be first device to be initailized.
+	 * If this device is root-hub, don't do any xHC related
+	 * stuff.
+	 */
+	if (ctrl->rootdev == 0) {
+		udev->speed = USB_SPEED_SUPER;
+		return 0;
+	}
+
+	xhci_queue_command(ctrl, NULL, 0, 0, TRB_ENABLE_SLOT);
+	event = xhci_wait_for_event(ctrl, TRB_COMPLETION);
+	BUG_ON(GET_COMP_CODE(le32_to_cpu(event->event_cmd.status))
+		!= COMP_SUCCESS);
+
+	udev->slot_id = TRB_TO_SLOT_ID(le32_to_cpu(event->event_cmd.flags));
+
+	xhci_acknowledge_event(ctrl);
+
+	ret = xhci_alloc_virt_device(udev);
+	if (ret < 0) {
+		/*
+		 * TODO: Unsuccessful Address Device command shall leave
+		 * the slot in default. So, issue Disable Slot command now.
+		 */
+		puts("Could not allocate xHCI USB device data structures\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+/*
+ * 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.
+ *
+ * @param udev	pointer to the Device Data Structure
+ * @return returns the status of the xhci_configure_endpoints
+ */
+int xhci_check_maxpacket(struct usb_device *udev)
+{
+	struct xhci_ctrl *ctrl = udev->controller;
+	unsigned int slot_id = udev->slot_id;
+	int ep_index = 0;	/* control endpoint */
+	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;
+	struct usb_interface *ifdesc;
+
+	ifdesc = &udev->config.if_desc[0];
+
+	out_ctx = ctrl->devs[slot_id]->out_ctx;
+	xhci_inval_cache((uint32_t)out_ctx->bytes, out_ctx->size);
+
+	ep_ctx = xhci_get_ep_ctx(ctrl, out_ctx, ep_index);
+	hw_max_packet_size = MAX_PACKET_DECODED(le32_to_cpu(ep_ctx->ep_info2));
+	max_packet_size = usb_endpoint_maxp(&ifdesc->ep_desc[0]);
+	if (hw_max_packet_size != max_packet_size) {
+		debug("Max Packet Size for ep 0 changed.\n");
+		debug("Max packet size in usb_device = %d\n", max_packet_size);
+		debug("Max packet size in xHCI HW = %d\n", hw_max_packet_size);
+		debug("Issuing evaluate context command.\n");
+
+		/* Set up the modified control endpoint 0 */
+		xhci_endpoint_copy(ctrl, ctrl->devs[slot_id]->in_ctx,
+				ctrl->devs[slot_id]->out_ctx, ep_index);
+		in_ctx = ctrl->devs[slot_id]->in_ctx;
+		ep_ctx = xhci_get_ep_ctx(ctrl, 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(in_ctx);
+		ctrl_ctx->add_flags = cpu_to_le32(EP0_FLAG);
+		ctrl_ctx->drop_flags = 0;
+
+		ret = xhci_configure_endpoints(udev, true);
+	}
+	return ret;
+}
+
+/**
+ * Clears the Change bits of the Port Status Register
+ *
+ * @param wValue	request value
+ * @param wIndex	request index
+ * @param addr		address of posrt status register
+ * @param port_status	state of port status register
+ * @return none
+ */
+static void xhci_clear_port_change_bit(u16 wValue,
+		u16 wIndex, volatile uint32_t *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_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;
+	default:
+		/* Should never happen */
+		return;
+	}
+
+	/* Change bits are all write 1 to clear */
+	xhci_writel(addr, port_status | status);
+
+	port_status = xhci_readl(addr);
+	debug("clear port %s change, actual port %d status  = 0x%x\n",
+			port_change_bit, wIndex, port_status);
+}
+
+/**
+ * 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.
+ *
+ * @param state	state of the Port Status and Control Regsiter
+ * @return a value that would result in the port being in the
+ *	   same state, if the value was written to the port
+ *	   status control register.
+ */
+static u32 xhci_port_state_to_neutral(u32 state)
+{
+	/* Save read-only status and port state */
+	return (state & XHCI_PORT_RO) | (state & XHCI_PORT_RWS);
+}
+
+/**
+ * Submits the Requests to the XHCI Host Controller
+ *
+ * @param udev pointer to the USB device structure
+ * @param pipe contains the DIR_IN or OUT , devnum
+ * @param buffer buffer to be read/written based on the request
+ * @return returns 0 if successful else -1 on failure
+ */
+static int xhci_submit_root(struct usb_device *udev, unsigned long pipe,
+			void *buffer, struct devrequest *req)
+{
+	uint8_t tmpbuf[4];
+	u16 typeReq;
+	void *srcptr = NULL;
+	int len, srclen;
+	uint32_t reg;
+	volatile uint32_t *status_reg;
+	struct xhci_ctrl *ctrl = udev->controller;
+	struct xhci_hcor *hcor = ctrl->hcor;
+
+	if (((req->requesttype & USB_RT_PORT) &&
+	     le16_to_cpu(req->index)) > CONFIG_SYS_USB_XHCI_MAX_ROOT_PORTS) {
+		printf("The request port(%d) is not configured\n",
+			le16_to_cpu(req->index) - 1);
+		return -EINVAL;
+	}
+
+	status_reg = (volatile uint32_t *)
+		     (&hcor->portregs[le16_to_cpu(req->index) - 1].or_portsc);
+	srclen = 0;
+
+	typeReq = req->request | req->requesttype << 8;
+
+	switch (typeReq) {
+	case DeviceRequest | USB_REQ_GET_DESCRIPTOR:
+		switch (le16_to_cpu(req->value) >> 8) {
+		case USB_DT_DEVICE:
+			debug("USB_DT_DEVICE request\n");
+			srcptr = &descriptor.device;
+			srclen = 0x12;
+			break;
+		case USB_DT_CONFIG:
+			debug("USB_DT_CONFIG config\n");
+			srcptr = &descriptor.config;
+			srclen = 0x19;
+			break;
+		case USB_DT_STRING:
+			debug("USB_DT_STRING config\n");
+			switch (le16_to_cpu(req->value) & 0xff) {
+			case 0:	/* Language */
+				srcptr = "\4\3\11\4";
+				srclen = 4;
+				break;
+			case 1:	/* Vendor String  */
+				srcptr = "\16\3u\0-\0b\0o\0o\0t\0";
+				srclen = 14;
+				break;
+			case 2:	/* Product Name */
+				srcptr = "\52\3X\0H\0C\0I\0 "
+					 "\0H\0o\0s\0t\0 "
+					 "\0C\0o\0n\0t\0r\0o\0l\0l\0e\0r\0";
+				srclen = 42;
+				break;
+			default:
+				printf("unknown value DT_STRING %x\n",
+					le16_to_cpu(req->value));
+				goto unknown;
+			}
+			break;
+		default:
+			printf("unknown value %x\n", le16_to_cpu(req->value));
+			goto unknown;
+		}
+		break;
+	case USB_REQ_GET_DESCRIPTOR | ((USB_DIR_IN | USB_RT_HUB) << 8):
+		switch (le16_to_cpu(req->value) >> 8) {
+		case USB_DT_HUB:
+			debug("USB_DT_HUB config\n");
+			srcptr = &descriptor.hub;
+			srclen = 0x8;
+			break;
+		default:
+			printf("unknown value %x\n", le16_to_cpu(req->value));
+			goto unknown;
+		}
+		break;
+	case USB_REQ_SET_ADDRESS | (USB_RECIP_DEVICE << 8):
+		debug("USB_REQ_SET_ADDRESS\n");
+		ctrl->rootdev = le16_to_cpu(req->value);
+		break;
+	case DeviceOutRequest | USB_REQ_SET_CONFIGURATION:
+		/* Do nothing */
+		break;
+	case USB_REQ_GET_STATUS | ((USB_DIR_IN | USB_RT_HUB) << 8):
+		tmpbuf[0] = 1;	/* USB_STATUS_SELFPOWERED */
+		tmpbuf[1] = 0;
+		srcptr = tmpbuf;
+		srclen = 2;
+		break;
+	case USB_REQ_GET_STATUS | ((USB_RT_PORT | USB_DIR_IN) << 8):
+		memset(tmpbuf, 0, 4);
+		reg = xhci_readl(status_reg);
+		if (reg & PORT_CONNECT) {
+			tmpbuf[0] |= USB_PORT_STAT_CONNECTION;
+			switch (reg & DEV_SPEED_MASK) {
+			case XDEV_FS:
+				debug("SPEED = FULLSPEED\n");
+				break;
+			case XDEV_LS:
+				debug("SPEED = LOWSPEED\n");
+				tmpbuf[1] |= USB_PORT_STAT_LOW_SPEED >> 8;
+				break;
+			case XDEV_HS:
+				debug("SPEED = HIGHSPEED\n");
+				tmpbuf[1] |= USB_PORT_STAT_HIGH_SPEED >> 8;
+				break;
+			case XDEV_SS:
+				debug("SPEED = SUPERSPEED\n");
+				tmpbuf[1] |= USB_PORT_STAT_SUPER_SPEED >> 8;
+				break;
+			}
+		}
+		if (reg & PORT_PE)
+			tmpbuf[0] |= USB_PORT_STAT_ENABLE;
+		if ((reg & PORT_PLS_MASK) == XDEV_U3)
+			tmpbuf[0] |= USB_PORT_STAT_SUSPEND;
+		if (reg & PORT_OC)
+			tmpbuf[0] |= USB_PORT_STAT_OVERCURRENT;
+		if (reg & PORT_RESET)
+			tmpbuf[0] |= USB_PORT_STAT_RESET;
+		if (reg & PORT_POWER)
+			/*
+			 * XXX: This Port power bit (for USB 3.0 hub)
+			 * we are faking in USB 2.0 hub port status;
+			 * since there's a change in bit positions in
+			 * two:
+			 * USB 2.0 port status PP is at position[8]
+			 * USB 3.0 port status PP is at position[9]
+			 * So, we are still keeping it at position [8]
+			 */
+			tmpbuf[1] |= USB_PORT_STAT_POWER >> 8;
+		if (reg & PORT_CSC)
+			tmpbuf[2] |= USB_PORT_STAT_C_CONNECTION;
+		if (reg & PORT_PEC)
+			tmpbuf[2] |= USB_PORT_STAT_C_ENABLE;
+		if (reg & PORT_OCC)
+			tmpbuf[2] |= USB_PORT_STAT_C_OVERCURRENT;
+		if (reg & PORT_RC)
+			tmpbuf[2] |= USB_PORT_STAT_C_RESET;
+
+		srcptr = tmpbuf;
+		srclen = 4;
+		break;
+	case USB_REQ_SET_FEATURE | ((USB_DIR_OUT | USB_RT_PORT) << 8):
+		reg = xhci_readl(status_reg);
+		reg = xhci_port_state_to_neutral(reg);
+		switch (le16_to_cpu(req->value)) {
+		case USB_PORT_FEAT_ENABLE:
+			reg |= PORT_PE;
+			xhci_writel(status_reg, reg);
+			break;
+		case USB_PORT_FEAT_POWER:
+			reg |= PORT_POWER;
+			xhci_writel(status_reg, reg);
+			break;
+		case USB_PORT_FEAT_RESET:
+			reg |= PORT_RESET;
+			xhci_writel(status_reg, reg);
+			break;
+		default:
+			printf("unknown feature %x\n", le16_to_cpu(req->value));
+			goto unknown;
+		}
+		break;
+	case USB_REQ_CLEAR_FEATURE | ((USB_DIR_OUT | USB_RT_PORT) << 8):
+		reg = xhci_readl(status_reg);
+		reg = xhci_port_state_to_neutral(reg);
+		switch (le16_to_cpu(req->value)) {
+		case USB_PORT_FEAT_ENABLE:
+			reg &= ~PORT_PE;
+			break;
+		case USB_PORT_FEAT_POWER:
+			reg &= ~PORT_POWER;
+			break;
+		case USB_PORT_FEAT_C_RESET:
+		case USB_PORT_FEAT_C_CONNECTION:
+		case USB_PORT_FEAT_C_OVER_CURRENT:
+		case USB_PORT_FEAT_C_ENABLE:
+			xhci_clear_port_change_bit((le16_to_cpu(req->value)),
+							le16_to_cpu(req->index),
+							status_reg, reg);
+			break;
+		default:
+			printf("unknown feature %x\n", le16_to_cpu(req->value));
+			goto unknown;
+		}
+		xhci_writel(status_reg, reg);
+		break;
+	default:
+		puts("Unknown request\n");
+		goto unknown;
+	}
+
+	debug("scrlen = %d\n req->length = %d\n",
+		srclen, le16_to_cpu(req->length));
+
+	len = min(srclen, le16_to_cpu(req->length));
+
+	if (srcptr != NULL && len > 0)
+		memcpy(buffer, srcptr, len);
+	else
+		debug("Len is 0\n");
+
+	udev->act_len = len;
+	udev->status = 0;
+
+	return 0;
+
+unknown:
+	udev->act_len = 0;
+	udev->status = USB_ST_STALLED;
+
+	return -ENODEV;
+}
+
+/**
+ * Submits the INT request to XHCI Host cotroller
+ *
+ * @param udev	pointer to the USB device
+ * @param pipe		contains the DIR_IN or OUT , devnum
+ * @param buffer	buffer to be read/written based on the request
+ * @param length	length of the buffer
+ * @param interval	interval of the interrupt
+ * @return 0
+ */
+int
+submit_int_msg(struct usb_device *udev, unsigned long pipe, void *buffer,
+						int length, int interval)
+{
+	/*
+	 * TODO: Not addressing any interrupt type transfer requests
+	 * Add support for it later.
+	 */
+	return -EINVAL;
+}
+
+/**
+ * submit the BULK type of request to the USB Device
+ *
+ * @param udev	pointer to the USB device
+ * @param pipe		contains the DIR_IN or OUT , devnum
+ * @param buffer	buffer to be read/written based on the request
+ * @param length	length of the buffer
+ * @return returns 0 if successful else -1 on failure
+ */
+int
+submit_bulk_msg(struct usb_device *udev, unsigned long pipe, void *buffer,
+								int length)
+{
+	if (usb_pipetype(pipe) != PIPE_BULK) {
+		printf("non-bulk pipe (type=%lu)", usb_pipetype(pipe));
+		return -EINVAL;
+	}
+
+	return xhci_bulk_tx(udev, pipe, length, buffer);
+}
+
+/**
+ * submit the control type of request to the Root hub/Device based on the devnum
+ *
+ * @param udev	pointer to the USB device
+ * @param pipe		contains the DIR_IN or OUT , devnum
+ * @param buffer	buffer to be read/written based on the request
+ * @param length	length of the buffer
+ * @param setup		Request type
+ * @return returns 0 if successful else -1 on failure
+ */
+int
+submit_control_msg(struct usb_device *udev, unsigned long pipe, void *buffer,
+					int length, struct devrequest *setup)
+{
+	struct xhci_ctrl *ctrl = udev->controller;
+	int ret = 0;
+
+	if (usb_pipetype(pipe) != PIPE_CONTROL) {
+		printf("non-control pipe (type=%lu)", usb_pipetype(pipe));
+		return -EINVAL;
+	}
+
+	if (usb_pipedevice(pipe) == ctrl->rootdev)
+		return xhci_submit_root(udev, pipe, buffer, setup);
+
+	if (setup->request == USB_REQ_SET_ADDRESS)
+		return xhci_address_device(udev);
+
+	if (setup->request == USB_REQ_SET_CONFIGURATION) {
+		ret = xhci_set_configuration(udev);
+		if (ret) {
+			puts("Failed to configure xHCI endpoint\n");
+			return ret;
+		}
+	}
+
+	return xhci_ctrl_tx(udev, pipe, setup, length, buffer);
+}
+
+/**
+ * Intialises the XHCI host controller
+ * and allocates the necessary data structures
+ *
+ * @param index	index to the host controller data structure
+ * @return pointer to the intialised controller
+ */
+int usb_lowlevel_init(int index, void **controller)
+{
+	uint32_t val;
+	uint32_t val2;
+	uint32_t reg;
+	struct xhci_hccr *hccr;
+	struct xhci_hcor *hcor;
+	struct xhci_ctrl *ctrl;
+
+	if (xhci_hcd_init(index, &hccr, (struct xhci_hcor **)&hcor) != 0)
+		return -ENODEV;
+
+	if (xhci_reset(hcor) != 0)
+		return -ENODEV;
+
+	ctrl = &xhcic[index];
+
+	ctrl->hccr = hccr;
+	ctrl->hcor = hcor;
+
+	/*
+	 * Program the Number of Device Slots Enabled field in the CONFIG
+	 * register with the max value of slots the HC can handle.
+	 */
+	val = (xhci_readl(&hccr->cr_hcsparams1) & HCS_SLOTS_MASK);
+	val2 = xhci_readl(&hcor->or_config);
+	val |= (val2 & ~HCS_SLOTS_MASK);
+	xhci_writel(&hcor->or_config, val);
+
+	/* initializing xhci data structures */
+	if (xhci_mem_init(ctrl, hccr, hcor) < 0)
+		return -ENOMEM;
+
+	reg = xhci_readl(&hccr->cr_hcsparams1);
+	descriptor.hub.bNbrPorts = ((reg & HCS_MAX_PORTS_MASK) >>
+						HCS_MAX_PORTS_SHIFT);
+	printf("Register %x NbrPorts %d\n", reg, descriptor.hub.bNbrPorts);
+
+	/* Port Indicators */
+	reg = xhci_readl(&hccr->cr_hccparams);
+	if (HCS_INDICATOR(reg))
+		put_unaligned(get_unaligned(&descriptor.hub.wHubCharacteristics)
+				| 0x80, &descriptor.hub.wHubCharacteristics);
+
+	/* Port Power Control */
+	if (HCC_PPC(reg))
+		put_unaligned(get_unaligned(&descriptor.hub.wHubCharacteristics)
+				| 0x01, &descriptor.hub.wHubCharacteristics);
+
+	if (xhci_start(hcor)) {
+		xhci_reset(hcor);
+		return -ENODEV;
+	}
+
+	/* Zero'ing IRQ control register and IRQ pending register */
+	xhci_writel(&ctrl->ir_set->irq_control, 0x0);
+	xhci_writel(&ctrl->ir_set->irq_pending, 0x0);
+
+	reg = HC_VERSION(xhci_readl(&hccr->cr_capbase));
+	printf("USB XHCI %x.%02x\n", reg >> 8, reg & 0xff);
+
+	*controller = &xhcic[index];
+
+	return 0;
+}
+
+/**
+ * Stops the XHCI host controller
+ * and cleans up all the related data structures
+ *
+ * @param index	index to the host controller data structure
+ * @return none
+ */
+int usb_lowlevel_stop(int index)
+{
+	struct xhci_ctrl *ctrl = (xhcic + index);
+	u32 temp;
+
+	xhci_reset(ctrl->hcor);
+
+	debug("// Disabling event ring interrupts\n");
+	temp = xhci_readl(&ctrl->hcor->or_usbsts);
+	xhci_writel(&ctrl->hcor->or_usbsts, temp & ~STS_EINT);
+	temp = xhci_readl(&ctrl->ir_set->irq_pending);
+	xhci_writel(&ctrl->ir_set->irq_pending, ER_IRQ_DISABLE(temp));
+
+	xhci_hcd_stop(index);
+
+	xhci_cleanup(ctrl);
+
+	return 0;
+}
diff --git a/drivers/usb/host/xhci.h b/drivers/usb/host/xhci.h
new file mode 100644
index 0000000000..ceb1573d86
--- /dev/null
+++ b/drivers/usb/host/xhci.h
@@ -0,0 +1,1255 @@
+/*
+ * USB HOST XHCI Controller
+ *
+ * Based on xHCI host controller driver in linux-kernel
+ * by Sarah Sharp.
+ *
+ * Copyright (C) 2008 Intel Corp.
+ * Author: Sarah Sharp
+ *
+ * Copyright (C) 2013 Samsung Electronics Co.Ltd
+ * Authors: Vivek Gautam <gautam.vivek@samsung.com>
+ *	    Vikas Sajjan <vikas.sajjan@samsung.com>
+ *
+ * SPDX-License-Identifier:	GPL-2.0+
+ */
+
+#ifndef HOST_XHCI_H_
+#define HOST_XHCI_H_
+
+#include <asm/cache.h>
+#include <asm/io.h>
+#include <linux/list.h>
+
+#define upper_32_bits(n) (u32)((n) >> 32)
+#define lower_32_bits(n) (u32)(n)
+
+#define MAX_EP_CTX_NUM		31
+#define XHCI_ALIGNMENT		64
+/* Generic timeout for XHCI events */
+#define XHCI_TIMEOUT		5000
+/* 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
+
+/* Up to 16 ms to halt an HC */
+#define XHCI_MAX_HALT_USEC	(16*1000)
+
+#define XHCI_MAX_RESET_USEC	(250*1000)
+
+/*
+ * 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))
+
+/*
+ * XHCI Register Space.
+ */
+struct xhci_hccr {
+	uint32_t cr_capbase;
+	uint32_t cr_hcsparams1;
+	uint32_t cr_hcsparams2;
+	uint32_t cr_hcsparams3;
+	uint32_t cr_hccparams;
+	uint32_t cr_dboff;
+	uint32_t cr_rtsoff;
+
+/* 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_SHIFT	24
+#define HCS_MAX_PORTS_MASK	(0x7f << HCS_MAX_PORTS_SHIFT)
+#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)
+
+};
+
+struct xhci_hcor_port_regs {
+	volatile uint32_t or_portsc;
+	volatile uint32_t or_portpmsc;
+	volatile uint32_t or_portli;
+	volatile uint32_t reserved_3;
+};
+
+struct xhci_hcor {
+	volatile uint32_t or_usbcmd;
+	volatile uint32_t or_usbsts;
+	volatile uint32_t or_pagesize;
+	volatile uint32_t reserved_0[2];
+	volatile uint32_t or_dnctrl;
+	volatile uint64_t or_crcr;
+	volatile uint32_t reserved_1[4];
+	volatile uint64_t or_dcbaap;
+	volatile uint32_t or_config;
+	volatile uint32_t reserved_2[241];
+	struct xhci_hcor_port_regs portregs[CONFIG_SYS_USB_XHCI_MAX_ROOT_PORTS];
+
+	uint32_t reserved_4[CONFIG_SYS_USB_XHCI_MAX_ROOT_PORTS * 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). */
+
+/* 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)
+/* bit 24 reserved */
+/* 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)
+/* Inactivity timer value for transitions into U2 */
+#define PORT_U2_TIMEOUT(p)	(((p) & 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 {
+	volatile __le32	irq_pending;
+	volatile __le32	irq_control;
+	volatile __le32	erst_size;
+	volatile __le32	rsvd;
+	volatile __le64	erst_base;
+	volatile __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 {
+	volatile __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;
+};
+
+/**
+ * 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 ROOT_HUB_PORT_MASK		(0xff)
+#define ROOT_HUB_PORT_SHIFT		(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 ERROR_COUNT_SHIFT	(1)
+#define ERROR_COUNT_MASK	(0x3)
+#define CTX_TO_EP_TYPE(p)	(((p) >> 3) & 0x7)
+#define EP_TYPE(p)		((p) << 3)
+#define EP_TYPE_SHIFT		(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 MAX_BURST_MASK		(0xff)
+#define MAX_BURST_SHIFT		(8)
+#define CTX_TO_MAX_BURST(p)	(((p) >> 8) & 0xff)
+#define MAX_PACKET(p)		(((p)&0xffff) << 16)
+#define MAX_PACKET_MASK		(0xffff)
+#define MAX_PACKET_DECODED(p)	(((p) >> 16) & 0xffff)
+#define MAX_PACKET_SHIFT	(16)
+
+/* 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 {
+	volatile __le32	drop_flags;
+	volatile __le32	add_flags;
+	__le32	rsvd2[6];
+};
+
+
+/**
+ * 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];
+};
+/* 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 */
+	volatile __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	COMP_CODE_SHIFT		(24)
+#define GET_COMP_CODE(p)	(((p) & COMP_CODE_MASK) >> 24)
+
+typedef enum {
+	COMP_SUCCESS = 1,
+	/* Data Buffer Error */
+	COMP_DB_ERR, /* 2 */
+	/* Babble Detected Error */
+	COMP_BABBLE, /* 3 */
+	/* USB Transaction Error */
+	COMP_TX_ERR, /* 4 */
+	/* TRB Error - some TRB field is invalid */
+	COMP_TRB_ERR, /* 5 */
+	/* Stall Error - USB device is stalled */
+	COMP_STALL, /* 6 */
+	/* Resource Error - HC doesn't have memory for that device configuration */
+	COMP_ENOMEM, /* 7 */
+	/* Bandwidth Error - not enough room in schedule for this dev config */
+	COMP_BW_ERR, /* 8 */
+	/* No Slots Available Error - HC ran out of device slots */
+	COMP_ENOSLOTS, /* 9 */
+	/* Invalid Stream Type Error */
+	COMP_STREAM_ERR, /* 10 */
+	/* Slot Not Enabled Error - doorbell rung for disabled device slot */
+	COMP_EBADSLT, /* 11 */
+	/* Endpoint Not Enabled Error */
+	COMP_EBADEP,/* 12 */
+	/* Short Packet */
+	COMP_SHORT_TX, /* 13 */
+	/* Ring Underrun - doorbell rung for an empty isoc OUT ep ring */
+	COMP_UNDERRUN, /* 14 */
+	/* Ring Overrun - isoc IN ep ring is empty when ep is scheduled to RX */
+	COMP_OVERRUN, /* 15 */
+	/* Virtual Function Event Ring Full Error */
+	COMP_VF_FULL, /* 16 */
+	/* Parameter Error - Context parameter is invalid */
+	COMP_EINVAL, /* 17 */
+	/* Bandwidth Overrun Error - isoc ep exceeded its allocated bandwidth */
+	COMP_BW_OVER,/* 18 */
+	/* Context State Error - illegal context state transition requested */
+	COMP_CTX_STATE,/* 19 */
+	/* No Ping Response Error - HC didn't get PING_RESPONSE in time to TX */
+	COMP_PING_ERR,/* 20 */
+	/* Event Ring is full */
+	COMP_ER_FULL,/* 21 */
+	/* Incompatible Device Error */
+	COMP_DEV_ERR,/* 22 */
+	/* Missed Service Error - HC couldn't service an isoc ep within interval */
+	COMP_MISSED_INT,/* 23 */
+	/* Successfully stopped command ring */
+	COMP_CMD_STOP, /* 24 */
+	/* Successfully aborted current command and stopped command ring */
+	COMP_CMD_ABORT, /* 25 */
+	/* Stopped - transfer was terminated by a stop endpoint command */
+	COMP_STOP,/* 26 */
+	/* Same as COMP_EP_STOPPED, but the transferred length in the event
+	 * is invalid */
+	COMP_STOP_INVAL, /* 27*/
+	/* Control Abort Error - Debug Capability - control pipe aborted */
+	COMP_DBG_ABORT, /* 28 */
+	/* Max Exit Latency Too Large Error */
+	COMP_MEL_ERR,/* 29 */
+	/* TRB type 30 reserved */
+	/* Isoc Buffer Overrun - an isoc IN ep sent more data than could fit in TD */
+	COMP_BUFF_OVER = 31,
+	/* Event Lost Error - xHC has an "internal event overrun condition" */
+	COMP_ISSUES, /* 32 */
+	/* Undefined Error - reported when other error codes don't apply */
+	COMP_UNKNOWN, /* 33 */
+	/* Invalid Stream ID Error */
+	COMP_STRID_ERR, /* 34 */
+	/* Secondary Bandwidth Error - may be returned by a Configure Endpoint cmd */
+	COMP_2ND_BW_ERR, /* 35 */
+	/* Split Transaction Error */
+	COMP_SPLIT_ERR /* 36 */
+
+} xhci_comp_code;
+
+struct xhci_link_trb {
+	/* 64-bit segment pointer*/
+	volatile __le64 segment_ptr;
+	volatile __le32 intr_target;
+	volatile __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 */
+	volatile __le64 cmd_trb;
+	volatile __le32 status;
+	volatile __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	TRB_TO_SLOT_ID_SHIFT		(24)
+#define	TRB_TO_SLOT_ID_MASK		(0xff << TRB_TO_SLOT_ID_SHIFT)
+#define	SLOT_ID_FOR_TRB(p)		(((p) & 0xff) << 24)
+#define	SLOT_ID_FOR_TRB_MASK		(0xff)
+#define	SLOT_ID_FOR_TRB_SHIFT		(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)
+#define	PORT_ID_SHIFT			(24)
+#define	PORT_ID_MASK			(0xff << PORT_ID_SHIFT)
+
+/* Normal TRB fields */
+/* transfer_len bitmasks - bits 0:16 */
+#define	TRB_LEN(p)			((p) & 0x1ffff)
+#define	TRB_LEN_MASK			(0x1ffff)
+/* Interrupter Target - which MSI-X vector to target the completion event at */
+#define	TRB_INTR_TARGET_SHIFT		(22)
+#define	TRB_INTR_TARGET_MASK		(0x3ff)
+#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_TX_TYPE_SHIFT	(16)
+#define	TRB_DATA_OUT		2
+#define	TRB_DATA_IN		3
+
+/* Isochronous TRB specific fields */
+#define TRB_SIA			(1 << 31)
+
+struct xhci_generic_trb {
+	volatile __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_TYPE_SHIFT		(10)
+#define TRB_FIELD_TO_TYPE(p)	(((p) & TRB_TYPE_BITMASK) >> 10)
+
+/* TRB type IDs */
+typedef enum {
+	/* bulk, interrupt, isoc scatter/gather, and control data stage */
+	TRB_NORMAL = 1,
+	/* setup stage for control transfers */
+	TRB_SETUP, /* 2 */
+	/* data stage for control transfers */
+	TRB_DATA, /* 3 */
+	/* status stage for control transfers */
+	TRB_STATUS, /* 4 */
+	/* isoc transfers */
+	TRB_ISOC, /* 5 */
+	/* TRB for linking ring segments */
+	TRB_LINK, /* 6 */
+	/* TRB for EVENT DATA */
+	TRB_EVENT_DATA, /* 7 */
+	/* Transfer Ring No-op (not for the command ring) */
+	TRB_TR_NOOP, /* 8 */
+	/* Command TRBs */
+	/* Enable Slot Command */
+	TRB_ENABLE_SLOT, /* 9 */
+	/* Disable Slot Command */
+	TRB_DISABLE_SLOT, /* 10 */
+	/* Address Device Command */
+	TRB_ADDR_DEV, /* 11 */
+	/* Configure Endpoint Command */
+	TRB_CONFIG_EP, /* 12 */
+	/* Evaluate Context Command */
+	TRB_EVAL_CONTEXT, /* 13 */
+	/* Reset Endpoint Command */
+	TRB_RESET_EP, /* 14 */
+	/* Stop Transfer Ring Command */
+	TRB_STOP_RING, /* 15 */
+	/* Set Transfer Ring Dequeue Pointer Command */
+	TRB_SET_DEQ, /* 16 */
+	/* Reset Device Command */
+	TRB_RESET_DEV, /* 17 */
+	/* Force Event Command (opt) */
+	TRB_FORCE_EVENT, /* 18 */
+	/* Negotiate Bandwidth Command (opt) */
+	TRB_NEG_BANDWIDTH, /* 19 */
+	/* Set Latency Tolerance Value Command (opt) */
+	TRB_SET_LT, /* 20 */
+	/* Get port bandwidth Command */
+	TRB_GET_BW, /* 21 */
+	/* Force Header Command - generate a transaction or link management packet */
+	TRB_FORCE_HEADER, /* 22 */
+	/* No-op Command - not for transfer rings */
+	TRB_CMD_NOOP, /* 23 */
+	/* TRB IDs 24-31 reserved */
+	/* Event TRBS */
+	/* Transfer Event */
+	TRB_TRANSFER = 32,
+	/* Command Completion Event */
+	TRB_COMPLETION, /* 33 */
+	/* Port Status Change Event */
+	TRB_PORT_STATUS, /* 34 */
+	/* Bandwidth Request Event (opt) */
+	TRB_BANDWIDTH_EVENT, /* 35 */
+	/* Doorbell Event (opt) */
+	TRB_DOORBELL, /* 36 */
+	/* Host Controller Event */
+	TRB_HC_EVENT, /* 37 */
+	/* Device Notification Event - device sent function wake notification */
+	TRB_DEV_NOTE, /* 38 */
+	/* MFINDEX Wrap Event - microframe counter wrapped */
+	TRB_MFINDEX_WRAP, /* 39 */
+	/* TRB IDs 40-47 reserved, 48-63 is vendor-defined */
+	/* Nec vendor-specific command completion event. */
+	TRB_NEC_CMD_COMP = 48, /* 48 */
+	/* Get NEC firmware revision. */
+	TRB_NEC_GET_FW, /* 49 */
+} trb_type;
+
+#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)))
+
+/*
+ * 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 SEGMENT_SIZE		(TRBS_PER_SEGMENT*16)
+/* SEGMENT_SHIFT should be log2(SEGMENT_SIZE).
+ * Change this if you change TRBS_PER_SEGMENT!
+ */
+#define SEGMENT_SHIFT		10
+/* 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;
+};
+
+struct xhci_ring {
+	struct xhci_segment	*first_seg;
+	union  xhci_trb		*enqueue;
+	struct xhci_segment	*enq_seg;
+	union  xhci_trb		*dequeue;
+	struct xhci_segment	*deq_seg;
+	/*
+	 * 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.
+	 */
+	volatile u32		cycle_state;
+	unsigned int		num_segs;
+};
+
+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;
+	/* Num entries the ERST can contain */
+	unsigned int		erst_size;
+};
+
+/*
+ * 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	3
+/* Initial number of event segment rings allocated */
+#define	ERST_ENTRIES	3
+/* Initial allocated size of the ERST, in number of entries */
+#define	ERST_SIZE	64
+/* 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 xhci_virt_ep {
+	struct xhci_ring		*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)
+};
+
+#define CTX_SIZE(_hcc) (HCC_64BYTE_CONTEXT(_hcc) ? 64 : 32)
+
+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 */
+#define	XHCI_MAX_RINGS_CACHED	31
+	struct xhci_virt_ep		eps[31];
+};
+
+/* TODO: copied from ehci.h - can be refactored? */
+/* xHCI spec says all registers are little endian */
+static inline unsigned int xhci_readl(uint32_t volatile *regs)
+{
+	return readl(regs);
+}
+
+static inline void xhci_writel(uint32_t volatile *regs, const unsigned int val)
+{
+	writel(val, regs);
+}
+
+/*
+ * Registers should always be accessed with double word or quad word accesses.
+ * Some xHCI implementations may support 64-bit address pointers.  Registers
+ * with 64-bit address pointers should be written to with dword accesses by
+ * writing the low dword first (ptr[0]), then the high dword (ptr[1]) second.
+ * xHCI implementations that do not support 64-bit address pointers will ignore
+ * the high dword, and write order is irrelevant.
+ */
+static inline u64 xhci_readq(__le64 volatile *regs)
+{
+	__u32 *ptr = (__u32 *)regs;
+	u64 val_lo = readl(ptr);
+	u64 val_hi = readl(ptr + 1);
+	return val_lo + (val_hi << 32);
+}
+
+static inline void xhci_writeq(__le64 volatile *regs, const u64 val)
+{
+	__u32 *ptr = (__u32 *)regs;
+	u32 val_lo = lower_32_bits(val);
+	/* FIXME */
+	u32 val_hi = 0;
+	writel(val_lo, ptr);
+	writel(val_hi, ptr + 1);
+}
+
+int xhci_hcd_init(int index, struct xhci_hccr **ret_hccr,
+					struct xhci_hcor **ret_hcor);
+void xhci_hcd_stop(int index);
+
+
+/*************************************************************
+	EXTENDED CAPABILITY DEFINITIONS
+*************************************************************/
+/* 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 1:2, 5:12, and 17:19 need to be preserved; bits 21:28 should be zero */
+#define	XHCI_LEGACY_DISABLE_SMI		((0x3 << 1) + (0xff << 5) + (0x7 << 17))
+
+/* 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)
+
+struct xhci_ctrl {
+	struct xhci_hccr *hccr;	/* R/O registers, not need for volatile */
+	struct xhci_hcor *hcor;
+	struct xhci_doorbell_array *dba;
+	struct xhci_run_regs *run_regs;
+	struct xhci_device_context_array *dcbaa		\
+			__attribute__ ((aligned(ARCH_DMA_MINALIGN)));
+	struct xhci_ring *event_ring;
+	struct xhci_ring *cmd_ring;
+	struct xhci_ring *transfer_ring;
+	struct xhci_segment *seg;
+	struct xhci_intr_reg *ir_set;
+	struct xhci_erst erst;
+	struct xhci_erst_entry entry[ERST_NUM_SEGS];
+	struct xhci_virt_device *devs[MAX_HC_SLOTS];
+	int rootdev;
+};
+
+unsigned long trb_addr(struct xhci_segment *seg, union xhci_trb *trb);
+struct xhci_input_control_ctx
+		*xhci_get_input_control_ctx(struct xhci_container_ctx *ctx);
+struct xhci_slot_ctx *xhci_get_slot_ctx(struct xhci_ctrl *ctrl,
+					struct xhci_container_ctx *ctx);
+struct xhci_ep_ctx *xhci_get_ep_ctx(struct xhci_ctrl *ctrl,
+				    struct xhci_container_ctx *ctx,
+				    unsigned int ep_index);
+void xhci_endpoint_copy(struct xhci_ctrl *ctrl,
+			struct xhci_container_ctx *in_ctx,
+			struct xhci_container_ctx *out_ctx,
+			unsigned int ep_index);
+void xhci_slot_copy(struct xhci_ctrl *ctrl,
+		    struct xhci_container_ctx *in_ctx,
+		    struct xhci_container_ctx *out_ctx);
+void xhci_setup_addressable_virt_dev(struct usb_device *udev);
+void xhci_queue_command(struct xhci_ctrl *ctrl, u8 *ptr,
+			u32 slot_id, u32 ep_index, trb_type cmd);
+void xhci_acknowledge_event(struct xhci_ctrl *ctrl);
+union xhci_trb *xhci_wait_for_event(struct xhci_ctrl *ctrl, trb_type expected);
+int xhci_bulk_tx(struct usb_device *udev, unsigned long pipe,
+		 int length, void *buffer);
+int xhci_ctrl_tx(struct usb_device *udev, unsigned long pipe,
+		 struct devrequest *req, int length, void *buffer);
+int xhci_check_maxpacket(struct usb_device *udev);
+void xhci_flush_cache(uint32_t addr, u32 type_len);
+void xhci_inval_cache(uint32_t addr, u32 type_len);
+void xhci_cleanup(struct xhci_ctrl *ctrl);
+struct xhci_ring *xhci_ring_alloc(unsigned int num_segs, bool link_trbs);
+int xhci_alloc_virt_device(struct usb_device *udev);
+int xhci_mem_init(struct xhci_ctrl *ctrl, struct xhci_hccr *hccr,
+		  struct xhci_hcor *hcor);
+
+#endif /* HOST_XHCI_H_ */