xref: /openbmc/linux/drivers/usb/host/xhci-ring.c (revision 7dd65feb)

/*
 * xHCI host controller driver
 *
 * Copyright (C) 2008 Intel Corp.
 *
 * Author: Sarah Sharp
 * Some code borrowed from the Linux EHCI driver.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
 * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 * for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software Foundation,
 * Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/*
 * Ring initialization rules:
 * 1. Each segment is initialized to zero, except for link TRBs.
 * 2. Ring cycle state = 0.  This represents Producer Cycle State (PCS) or
 *    Consumer Cycle State (CCS), depending on ring function.
 * 3. Enqueue pointer = dequeue pointer = address of first TRB in the segment.
 *
 * Ring behavior rules:
 * 1. A ring is empty if enqueue == dequeue.  This means there will always be at
 *    least one free TRB in the ring.  This is useful if you want to turn that
 *    into a link TRB and expand the ring.
 * 2. When incrementing an enqueue or dequeue pointer, if the next TRB is a
 *    link TRB, then load the pointer with the address in the link TRB.  If the
 *    link TRB had its toggle bit set, you may need to update the ring cycle
 *    state (see cycle bit rules).  You may have to do this multiple times
 *    until you reach a non-link TRB.
 * 3. A ring is full if enqueue++ (for the definition of increment above)
 *    equals the dequeue pointer.
 *
 * Cycle bit rules:
 * 1. When a consumer increments a dequeue pointer and encounters a toggle bit
 *    in a link TRB, it must toggle the ring cycle state.
 * 2. When a producer increments an enqueue pointer and encounters a toggle bit
 *    in a link TRB, it must toggle the ring cycle state.
 *
 * Producer rules:
 * 1. Check if ring is full before you enqueue.
 * 2. Write the ring cycle state to the cycle bit in the TRB you're enqueuing.
 *    Update enqueue pointer between each write (which may update the ring
 *    cycle state).
 * 3. Notify consumer.  If SW is producer, it rings the doorbell for command
 *    and endpoint rings.  If HC is the producer for the event ring,
 *    and it generates an interrupt according to interrupt modulation rules.
 *
 * Consumer rules:
 * 1. Check if TRB belongs to you.  If the cycle bit == your ring cycle state,
 *    the TRB is owned by the consumer.
 * 2. Update dequeue pointer (which may update the ring cycle state) and
 *    continue processing TRBs until you reach a TRB which is not owned by you.
 * 3. Notify the producer.  SW is the consumer for the event ring, and it
 *   updates event ring dequeue pointer.  HC is the consumer for the command and
 *   endpoint rings; it generates events on the event ring for these.
 */

#include <linux/scatterlist.h>
#include "xhci.h"

/*
 * Returns zero if the TRB isn't in this segment, otherwise it returns the DMA
 * address of the TRB.
 */
dma_addr_t xhci_trb_virt_to_dma(struct xhci_segment *seg,
		union xhci_trb *trb)
{
	unsigned long segment_offset;

	if (!seg || !trb || trb < seg->trbs)
		return 0;
	/* offset in TRBs */
	segment_offset = trb - seg->trbs;
	if (segment_offset > TRBS_PER_SEGMENT)
		return 0;
	return seg->dma + (segment_offset * sizeof(*trb));
}

/* Does this link TRB point to the first segment in a ring,
 * or was the previous TRB the last TRB on the last segment in the ERST?
 */
static inline bool last_trb_on_last_seg(struct xhci_hcd *xhci, struct xhci_ring *ring,
		struct xhci_segment *seg, union xhci_trb *trb)
{
	if (ring == xhci->event_ring)
		return (trb == &seg->trbs[TRBS_PER_SEGMENT]) &&
			(seg->next == xhci->event_ring->first_seg);
	else
		return trb->link.control & LINK_TOGGLE;
}

/* Is this TRB a link TRB or was the last TRB the last TRB in this event ring
 * segment?  I.e. would the updated event TRB pointer step off the end of the
 * event seg?
 */
static inline int last_trb(struct xhci_hcd *xhci, struct xhci_ring *ring,
		struct xhci_segment *seg, union xhci_trb *trb)
{
	if (ring == xhci->event_ring)
		return trb == &seg->trbs[TRBS_PER_SEGMENT];
	else
		return (trb->link.control & TRB_TYPE_BITMASK) == TRB_TYPE(TRB_LINK);
}

/* Updates trb to point to the next TRB in the ring, and updates seg if the next
 * TRB is in a new segment.  This does not skip over link TRBs, and it does not
 * effect the ring dequeue or enqueue pointers.
 */
static void next_trb(struct xhci_hcd *xhci,
		struct xhci_ring *ring,
		struct xhci_segment **seg,
		union xhci_trb **trb)
{
	if (last_trb(xhci, ring, *seg, *trb)) {
		*seg = (*seg)->next;
		*trb = ((*seg)->trbs);
	} else {
		*trb = (*trb)++;
	}
}

/*
 * See Cycle bit rules. SW is the consumer for the event ring only.
 * Don't make a ring full of link TRBs.  That would be dumb and this would loop.
 */
static void inc_deq(struct xhci_hcd *xhci, struct xhci_ring *ring, bool consumer)
{
	union xhci_trb *next = ++(ring->dequeue);
	unsigned long long addr;

	ring->deq_updates++;
	/* Update the dequeue pointer further if that was a link TRB or we're at
	 * the end of an event ring segment (which doesn't have link TRBS)
	 */
	while (last_trb(xhci, ring, ring->deq_seg, next)) {
		if (consumer && last_trb_on_last_seg(xhci, ring, ring->deq_seg, next)) {
			ring->cycle_state = (ring->cycle_state ? 0 : 1);
			if (!in_interrupt())
				xhci_dbg(xhci, "Toggle cycle state for ring %p = %i\n",
						ring,
						(unsigned int) ring->cycle_state);
		}
		ring->deq_seg = ring->deq_seg->next;
		ring->dequeue = ring->deq_seg->trbs;
		next = ring->dequeue;
	}
	addr = (unsigned long long) xhci_trb_virt_to_dma(ring->deq_seg, ring->dequeue);
	if (ring == xhci->event_ring)
		xhci_dbg(xhci, "Event ring deq = 0x%llx (DMA)\n", addr);
	else if (ring == xhci->cmd_ring)
		xhci_dbg(xhci, "Command ring deq = 0x%llx (DMA)\n", addr);
	else
		xhci_dbg(xhci, "Ring deq = 0x%llx (DMA)\n", addr);
}

/*
 * 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.
 */
static void inc_enq(struct xhci_hcd *xhci, struct xhci_ring *ring, bool consumer)
{
	u32 chain;
	union xhci_trb *next;
	unsigned long long addr;

	chain = ring->enqueue->generic.field[3] & TRB_CHAIN;
	next = ++(ring->enqueue);

	ring->enq_updates++;
	/* Update the dequeue pointer further if that was a link TRB or we're at
	 * the end of an event ring segment (which doesn't have link TRBS)
	 */
	while (last_trb(xhci, ring, ring->enq_seg, next)) {
		if (!consumer) {
			if (ring != xhci->event_ring) {
				/* If we're not dealing with 0.95 hardware,
				 * carry over the chain bit of the previous TRB
				 * (which may mean the chain bit is cleared).
				 */
				if (!xhci_link_trb_quirk(xhci)) {
					next->link.control &= ~TRB_CHAIN;
					next->link.control |= chain;
				}
				/* Give this link TRB to the hardware */
				wmb();
				if (next->link.control & TRB_CYCLE)
					next->link.control &= (u32) ~TRB_CYCLE;
				else
					next->link.control |= (u32) TRB_CYCLE;
			}
			/* Toggle the cycle bit after the last ring segment. */
			if (last_trb_on_last_seg(xhci, ring, ring->enq_seg, next)) {
				ring->cycle_state = (ring->cycle_state ? 0 : 1);
				if (!in_interrupt())
					xhci_dbg(xhci, "Toggle cycle state for ring %p = %i\n",
							ring,
							(unsigned int) ring->cycle_state);
			}
		}
		ring->enq_seg = ring->enq_seg->next;
		ring->enqueue = ring->enq_seg->trbs;
		next = ring->enqueue;
	}
	addr = (unsigned long long) xhci_trb_virt_to_dma(ring->enq_seg, ring->enqueue);
	if (ring == xhci->event_ring)
		xhci_dbg(xhci, "Event ring enq = 0x%llx (DMA)\n", addr);
	else if (ring == xhci->cmd_ring)
		xhci_dbg(xhci, "Command ring enq = 0x%llx (DMA)\n", addr);
	else
		xhci_dbg(xhci, "Ring enq = 0x%llx (DMA)\n", addr);
}

/*
 * Check to see if there's room to enqueue num_trbs on the ring.  See rules
 * above.
 * FIXME: this would be simpler and faster if we just kept track of the number
 * of free TRBs in a ring.
 */
static int room_on_ring(struct xhci_hcd *xhci, struct xhci_ring *ring,
		unsigned int num_trbs)
{
	int i;
	union xhci_trb *enq = ring->enqueue;
	struct xhci_segment *enq_seg = ring->enq_seg;

	/* Check if ring is empty */
	if (enq == ring->dequeue)
		return 1;
	/* Make sure there's an extra empty TRB available */
	for (i = 0; i <= num_trbs; ++i) {
		if (enq == ring->dequeue)
			return 0;
		enq++;
		while (last_trb(xhci, ring, enq_seg, enq)) {
			enq_seg = enq_seg->next;
			enq = enq_seg->trbs;
		}
	}
	return 1;
}

void xhci_set_hc_event_deq(struct xhci_hcd *xhci)
{
	u64 temp;
	dma_addr_t deq;

	deq = xhci_trb_virt_to_dma(xhci->event_ring->deq_seg,
			xhci->event_ring->dequeue);
	if (deq == 0 && !in_interrupt())
		xhci_warn(xhci, "WARN something wrong with SW event ring "
				"dequeue ptr.\n");
	/* Update HC event ring dequeue pointer */
	temp = xhci_read_64(xhci, &xhci->ir_set->erst_dequeue);
	temp &= ERST_PTR_MASK;
	/* Don't clear the EHB bit (which is RW1C) because
	 * there might be more events to service.
	 */
	temp &= ~ERST_EHB;
	xhci_dbg(xhci, "// Write event ring dequeue pointer, preserving EHB bit\n");
	xhci_write_64(xhci, ((u64) deq & (u64) ~ERST_PTR_MASK) | temp,
			&xhci->ir_set->erst_dequeue);
}

/* Ring the host controller doorbell after placing a command on the ring */
void xhci_ring_cmd_db(struct xhci_hcd *xhci)
{
	u32 temp;

	xhci_dbg(xhci, "// Ding dong!\n");
	temp = xhci_readl(xhci, &xhci->dba->doorbell[0]) & DB_MASK;
	xhci_writel(xhci, temp | DB_TARGET_HOST, &xhci->dba->doorbell[0]);
	/* Flush PCI posted writes */
	xhci_readl(xhci, &xhci->dba->doorbell[0]);
}

static void ring_ep_doorbell(struct xhci_hcd *xhci,
		unsigned int slot_id,
		unsigned int ep_index)
{
	struct xhci_virt_ep *ep;
	unsigned int ep_state;
	u32 field;
	__u32 __iomem *db_addr = &xhci->dba->doorbell[slot_id];

	ep = &xhci->devs[slot_id]->eps[ep_index];
	ep_state = ep->ep_state;
	/* Don't ring the doorbell for this endpoint if there are pending
	 * cancellations because the we don't want to interrupt processing.
	 */
	if (!(ep_state & EP_HALT_PENDING) && !(ep_state & SET_DEQ_PENDING)
			&& !(ep_state & EP_HALTED)) {
		field = xhci_readl(xhci, db_addr) & DB_MASK;
		xhci_writel(xhci, field | EPI_TO_DB(ep_index), db_addr);
		/* Flush PCI posted writes - FIXME Matthew Wilcox says this
		 * isn't time-critical and we shouldn't make the CPU wait for
		 * the flush.
		 */
		xhci_readl(xhci, db_addr);
	}
}

/*
 * Find the segment that trb is in.  Start searching in start_seg.
 * If we must move past a segment that has a link TRB with a toggle cycle state
 * bit set, then we will toggle the value pointed at by cycle_state.
 */
static struct xhci_segment *find_trb_seg(
		struct xhci_segment *start_seg,
		union xhci_trb	*trb, int *cycle_state)
{
	struct xhci_segment *cur_seg = start_seg;
	struct xhci_generic_trb *generic_trb;

	while (cur_seg->trbs > trb ||
			&cur_seg->trbs[TRBS_PER_SEGMENT - 1] < trb) {
		generic_trb = &cur_seg->trbs[TRBS_PER_SEGMENT - 1].generic;
		if (TRB_TYPE(generic_trb->field[3]) == TRB_LINK &&
				(generic_trb->field[3] & LINK_TOGGLE))
			*cycle_state = ~(*cycle_state) & 0x1;
		cur_seg = cur_seg->next;
		if (cur_seg == start_seg)
			/* Looped over the entire list.  Oops! */
			return 0;
	}
	return cur_seg;
}

/*
 * Move the xHC's endpoint ring dequeue pointer past cur_td.
 * Record the new state of the xHC's endpoint ring dequeue segment,
 * dequeue pointer, and new consumer cycle state in state.
 * Update our internal representation of the ring's dequeue pointer.
 *
 * We do this in three jumps:
 *  - First we update our new ring state to be the same as when the xHC stopped.
 *  - Then we traverse the ring to find the segment that contains
 *    the last TRB in the TD.  We toggle the xHC's new cycle state when we pass
 *    any link TRBs with the toggle cycle bit set.
 *  - Finally we move the dequeue state one TRB further, toggling the cycle bit
 *    if we've moved it past a link TRB with the toggle cycle bit set.
 */
void xhci_find_new_dequeue_state(struct xhci_hcd *xhci,
		unsigned int slot_id, unsigned int ep_index,
		struct xhci_td *cur_td, struct xhci_dequeue_state *state)
{
	struct xhci_virt_device *dev = xhci->devs[slot_id];
	struct xhci_ring *ep_ring = dev->eps[ep_index].ring;
	struct xhci_generic_trb *trb;
	struct xhci_ep_ctx *ep_ctx;
	dma_addr_t addr;

	state->new_cycle_state = 0;
	xhci_dbg(xhci, "Finding segment containing stopped TRB.\n");
	state->new_deq_seg = find_trb_seg(cur_td->start_seg,
			dev->eps[ep_index].stopped_trb,
			&state->new_cycle_state);
	if (!state->new_deq_seg)
		BUG();
	/* Dig out the cycle state saved by the xHC during the stop ep cmd */
	xhci_dbg(xhci, "Finding endpoint context\n");
	ep_ctx = xhci_get_ep_ctx(xhci, dev->out_ctx, ep_index);
	state->new_cycle_state = 0x1 & ep_ctx->deq;

	state->new_deq_ptr = cur_td->last_trb;
	xhci_dbg(xhci, "Finding segment containing last TRB in TD.\n");
	state->new_deq_seg = find_trb_seg(state->new_deq_seg,
			state->new_deq_ptr,
			&state->new_cycle_state);
	if (!state->new_deq_seg)
		BUG();

	trb = &state->new_deq_ptr->generic;
	if (TRB_TYPE(trb->field[3]) == TRB_LINK &&
				(trb->field[3] & LINK_TOGGLE))
		state->new_cycle_state = ~(state->new_cycle_state) & 0x1;
	next_trb(xhci, ep_ring, &state->new_deq_seg, &state->new_deq_ptr);

	/* Don't update the ring cycle state for the producer (us). */
	xhci_dbg(xhci, "New dequeue segment = %p (virtual)\n",
			state->new_deq_seg);
	addr = xhci_trb_virt_to_dma(state->new_deq_seg, state->new_deq_ptr);
	xhci_dbg(xhci, "New dequeue pointer = 0x%llx (DMA)\n",
			(unsigned long long) addr);
	xhci_dbg(xhci, "Setting dequeue pointer in internal ring state.\n");
	ep_ring->dequeue = state->new_deq_ptr;
	ep_ring->deq_seg = state->new_deq_seg;
}

static void td_to_noop(struct xhci_hcd *xhci, struct xhci_ring *ep_ring,
		struct xhci_td *cur_td)
{
	struct xhci_segment *cur_seg;
	union xhci_trb *cur_trb;

	for (cur_seg = cur_td->start_seg, cur_trb = cur_td->first_trb;
			true;
			next_trb(xhci, ep_ring, &cur_seg, &cur_trb)) {
		if ((cur_trb->generic.field[3] & TRB_TYPE_BITMASK) ==
				TRB_TYPE(TRB_LINK)) {
			/* Unchain any chained Link TRBs, but
			 * leave the pointers intact.
			 */
			cur_trb->generic.field[3] &= ~TRB_CHAIN;
			xhci_dbg(xhci, "Cancel (unchain) link TRB\n");
			xhci_dbg(xhci, "Address = %p (0x%llx dma); "
					"in seg %p (0x%llx dma)\n",
					cur_trb,
					(unsigned long long)xhci_trb_virt_to_dma(cur_seg, cur_trb),
					cur_seg,
					(unsigned long long)cur_seg->dma);
		} else {
			cur_trb->generic.field[0] = 0;
			cur_trb->generic.field[1] = 0;
			cur_trb->generic.field[2] = 0;
			/* Preserve only the cycle bit of this TRB */
			cur_trb->generic.field[3] &= TRB_CYCLE;
			cur_trb->generic.field[3] |= TRB_TYPE(TRB_TR_NOOP);
			xhci_dbg(xhci, "Cancel TRB %p (0x%llx dma) "
					"in seg %p (0x%llx dma)\n",
					cur_trb,
					(unsigned long long)xhci_trb_virt_to_dma(cur_seg, cur_trb),
					cur_seg,
					(unsigned long long)cur_seg->dma);
		}
		if (cur_trb == cur_td->last_trb)
			break;
	}
}

static int queue_set_tr_deq(struct xhci_hcd *xhci, int slot_id,
		unsigned int ep_index, struct xhci_segment *deq_seg,
		union xhci_trb *deq_ptr, u32 cycle_state);

void xhci_queue_new_dequeue_state(struct xhci_hcd *xhci,
		unsigned int slot_id, unsigned int ep_index,
		struct xhci_dequeue_state *deq_state)
{
	struct xhci_virt_ep *ep = &xhci->devs[slot_id]->eps[ep_index];

	xhci_dbg(xhci, "Set TR Deq Ptr cmd, new deq seg = %p (0x%llx dma), "
			"new deq ptr = %p (0x%llx dma), new cycle = %u\n",
			deq_state->new_deq_seg,
			(unsigned long long)deq_state->new_deq_seg->dma,
			deq_state->new_deq_ptr,
			(unsigned long long)xhci_trb_virt_to_dma(deq_state->new_deq_seg, deq_state->new_deq_ptr),
			deq_state->new_cycle_state);
	queue_set_tr_deq(xhci, slot_id, ep_index,
			deq_state->new_deq_seg,
			deq_state->new_deq_ptr,
			(u32) deq_state->new_cycle_state);
	/* Stop the TD queueing code from ringing the doorbell until
	 * this command completes.  The HC won't set the dequeue pointer
	 * if the ring is running, and ringing the doorbell starts the
	 * ring running.
	 */
	ep->ep_state |= SET_DEQ_PENDING;
}

static inline void xhci_stop_watchdog_timer_in_irq(struct xhci_hcd *xhci,
		struct xhci_virt_ep *ep)
{
	ep->ep_state &= ~EP_HALT_PENDING;
	/* Can't del_timer_sync in interrupt, so we attempt to cancel.  If the
	 * timer is running on another CPU, we don't decrement stop_cmds_pending
	 * (since we didn't successfully stop the watchdog timer).
	 */
	if (del_timer(&ep->stop_cmd_timer))
		ep->stop_cmds_pending--;
}

/* Must be called with xhci->lock held in interrupt context */
static void xhci_giveback_urb_in_irq(struct xhci_hcd *xhci,
		struct xhci_td *cur_td, int status, char *adjective)
{
	struct usb_hcd *hcd = xhci_to_hcd(xhci);

	cur_td->urb->hcpriv = NULL;
	usb_hcd_unlink_urb_from_ep(hcd, cur_td->urb);
	xhci_dbg(xhci, "Giveback %s URB %p\n", adjective, cur_td->urb);

	spin_unlock(&xhci->lock);
	usb_hcd_giveback_urb(hcd, cur_td->urb, status);
	kfree(cur_td);
	spin_lock(&xhci->lock);
	xhci_dbg(xhci, "%s URB given back\n", adjective);
}

/*
 * When we get a command completion for a Stop Endpoint Command, we need to
 * unlink any cancelled TDs from the ring.  There are two ways to do that:
 *
 *  1. If the HW was in the middle of processing the TD that needs to be
 *     cancelled, then we must move the ring's dequeue pointer past the last TRB
 *     in the TD with a Set Dequeue Pointer Command.
 *  2. Otherwise, we turn all the TRBs in the TD into No-op TRBs (with the chain
 *     bit cleared) so that the HW will skip over them.
 */
static void handle_stopped_endpoint(struct xhci_hcd *xhci,
		union xhci_trb *trb)
{
	unsigned int slot_id;
	unsigned int ep_index;
	struct xhci_ring *ep_ring;
	struct xhci_virt_ep *ep;
	struct list_head *entry;
	struct xhci_td *cur_td = 0;
	struct xhci_td *last_unlinked_td;

	struct xhci_dequeue_state deq_state;

	memset(&deq_state, 0, sizeof(deq_state));
	slot_id = TRB_TO_SLOT_ID(trb->generic.field[3]);
	ep_index = TRB_TO_EP_INDEX(trb->generic.field[3]);
	ep = &xhci->devs[slot_id]->eps[ep_index];
	ep_ring = ep->ring;

	if (list_empty(&ep->cancelled_td_list)) {
		xhci_stop_watchdog_timer_in_irq(xhci, ep);
		ring_ep_doorbell(xhci, slot_id, ep_index);
		return;
	}

	/* Fix up the ep ring first, so HW stops executing cancelled TDs.
	 * We have the xHCI lock, so nothing can modify this list until we drop
	 * it.  We're also in the event handler, so we can't get re-interrupted
	 * if another Stop Endpoint command completes
	 */
	list_for_each(entry, &ep->cancelled_td_list) {
		cur_td = list_entry(entry, struct xhci_td, cancelled_td_list);
		xhci_dbg(xhci, "Cancelling TD starting at %p, 0x%llx (dma).\n",
				cur_td->first_trb,
				(unsigned long long)xhci_trb_virt_to_dma(cur_td->start_seg, cur_td->first_trb));
		/*
		 * If we stopped on the TD we need to cancel, then we have to
		 * move the xHC endpoint ring dequeue pointer past this TD.
		 */
		if (cur_td == ep->stopped_td)
			xhci_find_new_dequeue_state(xhci, slot_id, ep_index, cur_td,
					&deq_state);
		else
			td_to_noop(xhci, ep_ring, cur_td);
		/*
		 * The event handler won't see a completion for this TD anymore,
		 * so remove it from the endpoint ring's TD list.  Keep it in
		 * the cancelled TD list for URB completion later.
		 */
		list_del(&cur_td->td_list);
	}
	last_unlinked_td = cur_td;
	xhci_stop_watchdog_timer_in_irq(xhci, ep);

	/* If necessary, queue a Set Transfer Ring Dequeue Pointer command */
	if (deq_state.new_deq_ptr && deq_state.new_deq_seg) {
		xhci_queue_new_dequeue_state(xhci,
				slot_id, ep_index, &deq_state);
		xhci_ring_cmd_db(xhci);
	} else {
		/* Otherwise just ring the doorbell to restart the ring */
		ring_ep_doorbell(xhci, slot_id, ep_index);
	}

	/*
	 * Drop the lock and complete the URBs in the cancelled TD list.
	 * New TDs to be cancelled might be added to the end of the list before
	 * we can complete all the URBs for the TDs we already unlinked.
	 * So stop when we've completed the URB for the last TD we unlinked.
	 */
	do {
		cur_td = list_entry(ep->cancelled_td_list.next,
				struct xhci_td, cancelled_td_list);
		list_del(&cur_td->cancelled_td_list);

		/* Clean up the cancelled URB */
		/* Doesn't matter what we pass for status, since the core will
		 * just overwrite it (because the URB has been unlinked).
		 */
		xhci_giveback_urb_in_irq(xhci, cur_td, 0, "cancelled");

		/* Stop processing the cancelled list if the watchdog timer is
		 * running.
		 */
		if (xhci->xhc_state & XHCI_STATE_DYING)
			return;
	} while (cur_td != last_unlinked_td);

	/* Return to the event handler with xhci->lock re-acquired */
}

/* Watchdog timer function for when a stop endpoint command fails to complete.
 * In this case, we assume the host controller is broken or dying or dead.  The
 * host may still be completing some other events, so we have to be careful to
 * let the event ring handler and the URB dequeueing/enqueueing functions know
 * through xhci->state.
 *
 * The timer may also fire if the host takes a very long time to respond to the
 * command, and the stop endpoint command completion handler cannot delete the
 * timer before the timer function is called.  Another endpoint cancellation may
 * sneak in before the timer function can grab the lock, and that may queue
 * another stop endpoint command and add the timer back.  So we cannot use a
 * simple flag to say whether there is a pending stop endpoint command for a
 * particular endpoint.
 *
 * Instead we use a combination of that flag and a counter for the number of
 * pending stop endpoint commands.  If the timer is the tail end of the last
 * stop endpoint command, and the endpoint's command is still pending, we assume
 * the host is dying.
 */
void xhci_stop_endpoint_command_watchdog(unsigned long arg)
{
	struct xhci_hcd *xhci;
	struct xhci_virt_ep *ep;
	struct xhci_virt_ep *temp_ep;
	struct xhci_ring *ring;
	struct xhci_td *cur_td;
	int ret, i, j;

	ep = (struct xhci_virt_ep *) arg;
	xhci = ep->xhci;

	spin_lock(&xhci->lock);

	ep->stop_cmds_pending--;
	if (xhci->xhc_state & XHCI_STATE_DYING) {
		xhci_dbg(xhci, "Stop EP timer ran, but another timer marked "
				"xHCI as DYING, exiting.\n");
		spin_unlock(&xhci->lock);
		return;
	}
	if (!(ep->stop_cmds_pending == 0 && (ep->ep_state & EP_HALT_PENDING))) {
		xhci_dbg(xhci, "Stop EP timer ran, but no command pending, "
				"exiting.\n");
		spin_unlock(&xhci->lock);
		return;
	}

	xhci_warn(xhci, "xHCI host not responding to stop endpoint command.\n");
	xhci_warn(xhci, "Assuming host is dying, halting host.\n");
	/* Oops, HC is dead or dying or at least not responding to the stop
	 * endpoint command.
	 */
	xhci->xhc_state |= XHCI_STATE_DYING;
	/* Disable interrupts from the host controller and start halting it */
	xhci_quiesce(xhci);
	spin_unlock(&xhci->lock);

	ret = xhci_halt(xhci);

	spin_lock(&xhci->lock);
	if (ret < 0) {
		/* This is bad; the host is not responding to commands and it's
		 * not allowing itself to be halted.  At least interrupts are
		 * disabled, so we can set HC_STATE_HALT and notify the
		 * USB core.  But if we call usb_hc_died(), it will attempt to
		 * disconnect all device drivers under this host.  Those
		 * disconnect() methods will wait for all URBs to be unlinked,
		 * so we must complete them.
		 */
		xhci_warn(xhci, "Non-responsive xHCI host is not halting.\n");
		xhci_warn(xhci, "Completing active URBs anyway.\n");
		/* We could turn all TDs on the rings to no-ops.  This won't
		 * help if the host has cached part of the ring, and is slow if
		 * we want to preserve the cycle bit.  Skip it and hope the host
		 * doesn't touch the memory.
		 */
	}
	for (i = 0; i < MAX_HC_SLOTS; i++) {
		if (!xhci->devs[i])
			continue;
		for (j = 0; j < 31; j++) {
			temp_ep = &xhci->devs[i]->eps[j];
			ring = temp_ep->ring;
			if (!ring)
				continue;
			xhci_dbg(xhci, "Killing URBs for slot ID %u, "
					"ep index %u\n", i, j);
			while (!list_empty(&ring->td_list)) {
				cur_td = list_first_entry(&ring->td_list,
						struct xhci_td,
						td_list);
				list_del(&cur_td->td_list);
				if (!list_empty(&cur_td->cancelled_td_list))
					list_del(&cur_td->cancelled_td_list);
				xhci_giveback_urb_in_irq(xhci, cur_td,
						-ESHUTDOWN, "killed");
			}
			while (!list_empty(&temp_ep->cancelled_td_list)) {
				cur_td = list_first_entry(
						&temp_ep->cancelled_td_list,
						struct xhci_td,
						cancelled_td_list);
				list_del(&cur_td->cancelled_td_list);
				xhci_giveback_urb_in_irq(xhci, cur_td,
						-ESHUTDOWN, "killed");
			}
		}
	}
	spin_unlock(&xhci->lock);
	xhci_to_hcd(xhci)->state = HC_STATE_HALT;
	xhci_dbg(xhci, "Calling usb_hc_died()\n");
	usb_hc_died(xhci_to_hcd(xhci));
	xhci_dbg(xhci, "xHCI host controller is dead.\n");
}

/*
 * When we get a completion for a Set Transfer Ring Dequeue Pointer command,
 * we need to clear the set deq pending flag in the endpoint ring state, so that
 * the TD queueing code can ring the doorbell again.  We also need to ring the
 * endpoint doorbell to restart the ring, but only if there aren't more
 * cancellations pending.
 */
static void handle_set_deq_completion(struct xhci_hcd *xhci,
		struct xhci_event_cmd *event,
		union xhci_trb *trb)
{
	unsigned int slot_id;
	unsigned int ep_index;
	struct xhci_ring *ep_ring;
	struct xhci_virt_device *dev;
	struct xhci_ep_ctx *ep_ctx;
	struct xhci_slot_ctx *slot_ctx;

	slot_id = TRB_TO_SLOT_ID(trb->generic.field[3]);
	ep_index = TRB_TO_EP_INDEX(trb->generic.field[3]);
	dev = xhci->devs[slot_id];
	ep_ring = dev->eps[ep_index].ring;
	ep_ctx = xhci_get_ep_ctx(xhci, dev->out_ctx, ep_index);
	slot_ctx = xhci_get_slot_ctx(xhci, dev->out_ctx);

	if (GET_COMP_CODE(event->status) != COMP_SUCCESS) {
		unsigned int ep_state;
		unsigned int slot_state;

		switch (GET_COMP_CODE(event->status)) {
		case COMP_TRB_ERR:
			xhci_warn(xhci, "WARN Set TR Deq Ptr cmd invalid because "
					"of stream ID configuration\n");
			break;
		case COMP_CTX_STATE:
			xhci_warn(xhci, "WARN Set TR Deq Ptr cmd failed due "
					"to incorrect slot or ep state.\n");
			ep_state = ep_ctx->ep_info;
			ep_state &= EP_STATE_MASK;
			slot_state = slot_ctx->dev_state;
			slot_state = GET_SLOT_STATE(slot_state);
			xhci_dbg(xhci, "Slot state = %u, EP state = %u\n",
					slot_state, ep_state);
			break;
		case COMP_EBADSLT:
			xhci_warn(xhci, "WARN Set TR Deq Ptr cmd failed because "
					"slot %u was not enabled.\n", slot_id);
			break;
		default:
			xhci_warn(xhci, "WARN Set TR Deq Ptr cmd with unknown "
					"completion code of %u.\n",
					GET_COMP_CODE(event->status));
			break;
		}
		/* OK what do we do now?  The endpoint state is hosed, and we
		 * should never get to this point if the synchronization between
		 * queueing, and endpoint state are correct.  This might happen
		 * if the device gets disconnected after we've finished
		 * cancelling URBs, which might not be an error...
		 */
	} else {
		xhci_dbg(xhci, "Successful Set TR Deq Ptr cmd, deq = @%08llx\n",
				ep_ctx->deq);
	}

	dev->eps[ep_index].ep_state &= ~SET_DEQ_PENDING;
	ring_ep_doorbell(xhci, slot_id, ep_index);
}

static void handle_reset_ep_completion(struct xhci_hcd *xhci,
		struct xhci_event_cmd *event,
		union xhci_trb *trb)
{
	int slot_id;
	unsigned int ep_index;
	struct xhci_ring *ep_ring;

	slot_id = TRB_TO_SLOT_ID(trb->generic.field[3]);
	ep_index = TRB_TO_EP_INDEX(trb->generic.field[3]);
	ep_ring = xhci->devs[slot_id]->eps[ep_index].ring;
	/* This command will only fail if the endpoint wasn't halted,
	 * but we don't care.
	 */
	xhci_dbg(xhci, "Ignoring reset ep completion code of %u\n",
			(unsigned int) GET_COMP_CODE(event->status));

	/* HW with the reset endpoint quirk needs to have a configure endpoint
	 * command complete before the endpoint can be used.  Queue that here
	 * because the HW can't handle two commands being queued in a row.
	 */
	if (xhci->quirks & XHCI_RESET_EP_QUIRK) {
		xhci_dbg(xhci, "Queueing configure endpoint command\n");
		xhci_queue_configure_endpoint(xhci,
				xhci->devs[slot_id]->in_ctx->dma, slot_id,
				false);
		xhci_ring_cmd_db(xhci);
	} else {
		/* Clear our internal halted state and restart the ring */
		xhci->devs[slot_id]->eps[ep_index].ep_state &= ~EP_HALTED;
		ring_ep_doorbell(xhci, slot_id, ep_index);
	}
}

/* Check to see if a command in the device's command queue matches this one.
 * Signal the completion or free the command, and return 1.  Return 0 if the
 * completed command isn't at the head of the command list.
 */
static int handle_cmd_in_cmd_wait_list(struct xhci_hcd *xhci,
		struct xhci_virt_device *virt_dev,
		struct xhci_event_cmd *event)
{
	struct xhci_command *command;

	if (list_empty(&virt_dev->cmd_list))
		return 0;

	command = list_entry(virt_dev->cmd_list.next,
			struct xhci_command, cmd_list);
	if (xhci->cmd_ring->dequeue != command->command_trb)
		return 0;

	command->status =
		GET_COMP_CODE(event->status);
	list_del(&command->cmd_list);
	if (command->completion)
		complete(command->completion);
	else
		xhci_free_command(xhci, command);
	return 1;
}

static void handle_cmd_completion(struct xhci_hcd *xhci,
		struct xhci_event_cmd *event)
{
	int slot_id = TRB_TO_SLOT_ID(event->flags);
	u64 cmd_dma;
	dma_addr_t cmd_dequeue_dma;
	struct xhci_input_control_ctx *ctrl_ctx;
	struct xhci_virt_device *virt_dev;
	unsigned int ep_index;
	struct xhci_ring *ep_ring;
	unsigned int ep_state;

	cmd_dma = event->cmd_trb;
	cmd_dequeue_dma = xhci_trb_virt_to_dma(xhci->cmd_ring->deq_seg,
			xhci->cmd_ring->dequeue);
	/* Is the command ring deq ptr out of sync with the deq seg ptr? */
	if (cmd_dequeue_dma == 0) {
		xhci->error_bitmask |= 1 << 4;
		return;
	}
	/* Does the DMA address match our internal dequeue pointer address? */
	if (cmd_dma != (u64) cmd_dequeue_dma) {
		xhci->error_bitmask |= 1 << 5;
		return;
	}
	switch (xhci->cmd_ring->dequeue->generic.field[3] & TRB_TYPE_BITMASK) {
	case TRB_TYPE(TRB_ENABLE_SLOT):
		if (GET_COMP_CODE(event->status) == COMP_SUCCESS)
			xhci->slot_id = slot_id;
		else
			xhci->slot_id = 0;
		complete(&xhci->addr_dev);
		break;
	case TRB_TYPE(TRB_DISABLE_SLOT):
		if (xhci->devs[slot_id])
			xhci_free_virt_device(xhci, slot_id);
		break;
	case TRB_TYPE(TRB_CONFIG_EP):
		virt_dev = xhci->devs[slot_id];
		if (handle_cmd_in_cmd_wait_list(xhci, virt_dev, event))
			break;
		/*
		 * Configure endpoint commands can come from the USB core
		 * configuration or alt setting changes, or because the HW
		 * needed an extra configure endpoint command after a reset
		 * endpoint command.  In the latter case, the xHCI driver is
		 * not waiting on the configure endpoint command.
		 */
		ctrl_ctx = xhci_get_input_control_ctx(xhci,
				virt_dev->in_ctx);
		/* Input ctx add_flags are the endpoint index plus one */
		ep_index = xhci_last_valid_endpoint(ctrl_ctx->add_flags) - 1;
		/* A usb_set_interface() call directly after clearing a halted
		 * condition may race on this quirky hardware.
		 * Not worth worrying about, since this is prototype hardware.
		 */
		if (xhci->quirks & XHCI_RESET_EP_QUIRK &&
				ep_index != (unsigned int) -1 &&
				ctrl_ctx->add_flags - SLOT_FLAG ==
					ctrl_ctx->drop_flags) {
			ep_ring = xhci->devs[slot_id]->eps[ep_index].ring;
			ep_state = xhci->devs[slot_id]->eps[ep_index].ep_state;
			if (!(ep_state & EP_HALTED))
				goto bandwidth_change;
			xhci_dbg(xhci, "Completed config ep cmd - "
					"last ep index = %d, state = %d\n",
					ep_index, ep_state);
			/* Clear our internal halted state and restart ring */
			xhci->devs[slot_id]->eps[ep_index].ep_state &=
				~EP_HALTED;
			ring_ep_doorbell(xhci, slot_id, ep_index);
			break;
		}
bandwidth_change:
		xhci_dbg(xhci, "Completed config ep cmd\n");
		xhci->devs[slot_id]->cmd_status =
			GET_COMP_CODE(event->status);
		complete(&xhci->devs[slot_id]->cmd_completion);
		break;
	case TRB_TYPE(TRB_EVAL_CONTEXT):
		virt_dev = xhci->devs[slot_id];
		if (handle_cmd_in_cmd_wait_list(xhci, virt_dev, event))
			break;
		xhci->devs[slot_id]->cmd_status = GET_COMP_CODE(event->status);
		complete(&xhci->devs[slot_id]->cmd_completion);
		break;
	case TRB_TYPE(TRB_ADDR_DEV):
		xhci->devs[slot_id]->cmd_status = GET_COMP_CODE(event->status);
		complete(&xhci->addr_dev);
		break;
	case TRB_TYPE(TRB_STOP_RING):
		handle_stopped_endpoint(xhci, xhci->cmd_ring->dequeue);
		break;
	case TRB_TYPE(TRB_SET_DEQ):
		handle_set_deq_completion(xhci, event, xhci->cmd_ring->dequeue);
		break;
	case TRB_TYPE(TRB_CMD_NOOP):
		++xhci->noops_handled;
		break;
	case TRB_TYPE(TRB_RESET_EP):
		handle_reset_ep_completion(xhci, event, xhci->cmd_ring->dequeue);
		break;
	default:
		/* Skip over unknown commands on the event ring */
		xhci->error_bitmask |= 1 << 6;
		break;
	}
	inc_deq(xhci, xhci->cmd_ring, false);
}

static void handle_port_status(struct xhci_hcd *xhci,
		union xhci_trb *event)
{
	u32 port_id;

	/* Port status change events always have a successful completion code */
	if (GET_COMP_CODE(event->generic.field[2]) != COMP_SUCCESS) {
		xhci_warn(xhci, "WARN: xHC returned failed port status event\n");
		xhci->error_bitmask |= 1 << 8;
	}
	/* FIXME: core doesn't care about all port link state changes yet */
	port_id = GET_PORT_ID(event->generic.field[0]);
	xhci_dbg(xhci, "Port Status Change Event for port %d\n", port_id);

	/* Update event ring dequeue pointer before dropping the lock */
	inc_deq(xhci, xhci->event_ring, true);
	xhci_set_hc_event_deq(xhci);

	spin_unlock(&xhci->lock);
	/* Pass this up to the core */
	usb_hcd_poll_rh_status(xhci_to_hcd(xhci));
	spin_lock(&xhci->lock);
}

/*
 * This TD is defined by the TRBs starting at start_trb in start_seg and ending
 * at end_trb, which may be in another segment.  If the suspect DMA address is a
 * TRB in this TD, this function returns that TRB's segment.  Otherwise it
 * returns 0.
 */
struct xhci_segment *trb_in_td(struct xhci_segment *start_seg,
		union xhci_trb	*start_trb,
		union xhci_trb	*end_trb,
		dma_addr_t	suspect_dma)
{
	dma_addr_t start_dma;
	dma_addr_t end_seg_dma;
	dma_addr_t end_trb_dma;
	struct xhci_segment *cur_seg;

	start_dma = xhci_trb_virt_to_dma(start_seg, start_trb);
	cur_seg = start_seg;

	do {
		if (start_dma == 0)
			return 0;
		/* We may get an event for a Link TRB in the middle of a TD */
		end_seg_dma = xhci_trb_virt_to_dma(cur_seg,
				&cur_seg->trbs[TRBS_PER_SEGMENT - 1]);
		/* If the end TRB isn't in this segment, this is set to 0 */
		end_trb_dma = xhci_trb_virt_to_dma(cur_seg, end_trb);

		if (end_trb_dma > 0) {
			/* The end TRB is in this segment, so suspect should be here */
			if (start_dma <= end_trb_dma) {
				if (suspect_dma >= start_dma && suspect_dma <= end_trb_dma)
					return cur_seg;
			} else {
				/* Case for one segment with
				 * a TD wrapped around to the top
				 */
				if ((suspect_dma >= start_dma &&
							suspect_dma <= end_seg_dma) ||
						(suspect_dma >= cur_seg->dma &&
						 suspect_dma <= end_trb_dma))
					return cur_seg;
			}
			return 0;
		} else {
			/* Might still be somewhere in this segment */
			if (suspect_dma >= start_dma && suspect_dma <= end_seg_dma)
				return cur_seg;
		}
		cur_seg = cur_seg->next;
		start_dma = xhci_trb_virt_to_dma(cur_seg, &cur_seg->trbs[0]);
	} while (cur_seg != start_seg);

	return 0;
}

static void xhci_cleanup_halted_endpoint(struct xhci_hcd *xhci,
		unsigned int slot_id, unsigned int ep_index,
		struct xhci_td *td, union xhci_trb *event_trb)
{
	struct xhci_virt_ep *ep = &xhci->devs[slot_id]->eps[ep_index];
	ep->ep_state |= EP_HALTED;
	ep->stopped_td = td;
	ep->stopped_trb = event_trb;
	xhci_queue_reset_ep(xhci, slot_id, ep_index);
	xhci_cleanup_stalled_ring(xhci, td->urb->dev, ep_index);
	xhci_ring_cmd_db(xhci);
}

/* Check if an error has halted the endpoint ring.  The class driver will
 * cleanup the halt for a non-default control endpoint if we indicate a stall.
 * However, a babble and other errors also halt the endpoint ring, and the class
 * driver won't clear the halt in that case, so we need to issue a Set Transfer
 * Ring Dequeue Pointer command manually.
 */
static int xhci_requires_manual_halt_cleanup(struct xhci_hcd *xhci,
		struct xhci_ep_ctx *ep_ctx,
		unsigned int trb_comp_code)
{
	/* TRB completion codes that may require a manual halt cleanup */
	if (trb_comp_code == COMP_TX_ERR ||
			trb_comp_code == COMP_BABBLE ||
			trb_comp_code == COMP_SPLIT_ERR)
		/* The 0.96 spec says a babbling control endpoint
		 * is not halted. The 0.96 spec says it is.  Some HW
		 * claims to be 0.95 compliant, but it halts the control
		 * endpoint anyway.  Check if a babble halted the
		 * endpoint.
		 */
		if ((ep_ctx->ep_info & EP_STATE_MASK) == EP_STATE_HALTED)
			return 1;

	return 0;
}

/*
 * If this function returns an error condition, it means it got a Transfer
 * event with a corrupted Slot ID, Endpoint ID, or TRB DMA address.
 * At this point, the host controller is probably hosed and should be reset.
 */
static int handle_tx_event(struct xhci_hcd *xhci,
		struct xhci_transfer_event *event)
{
	struct xhci_virt_device *xdev;
	struct xhci_virt_ep *ep;
	struct xhci_ring *ep_ring;
	unsigned int slot_id;
	int ep_index;
	struct xhci_td *td = 0;
	dma_addr_t event_dma;
	struct xhci_segment *event_seg;
	union xhci_trb *event_trb;
	struct urb *urb = 0;
	int status = -EINPROGRESS;
	struct xhci_ep_ctx *ep_ctx;
	u32 trb_comp_code;

	xhci_dbg(xhci, "In %s\n", __func__);
	slot_id = TRB_TO_SLOT_ID(event->flags);
	xdev = xhci->devs[slot_id];
	if (!xdev) {
		xhci_err(xhci, "ERROR Transfer event pointed to bad slot\n");
		return -ENODEV;
	}

	/* Endpoint ID is 1 based, our index is zero based */
	ep_index = TRB_TO_EP_ID(event->flags) - 1;
	xhci_dbg(xhci, "%s - ep index = %d\n", __func__, ep_index);
	ep = &xdev->eps[ep_index];
	ep_ring = ep->ring;
	ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
	if (!ep_ring || (ep_ctx->ep_info & EP_STATE_MASK) == EP_STATE_DISABLED) {
		xhci_err(xhci, "ERROR Transfer event pointed to disabled endpoint\n");
		return -ENODEV;
	}

	event_dma = event->buffer;
	/* This TRB should be in the TD at the head of this ring's TD list */
	xhci_dbg(xhci, "%s - checking for list empty\n", __func__);
	if (list_empty(&ep_ring->td_list)) {
		xhci_warn(xhci, "WARN Event TRB for slot %d ep %d with no TDs queued?\n",
				TRB_TO_SLOT_ID(event->flags), ep_index);
		xhci_dbg(xhci, "Event TRB with TRB type ID %u\n",
				(unsigned int) (event->flags & TRB_TYPE_BITMASK)>>10);
		xhci_print_trb_offsets(xhci, (union xhci_trb *) event);
		urb = NULL;
		goto cleanup;
	}
	xhci_dbg(xhci, "%s - getting list entry\n", __func__);
	td = list_entry(ep_ring->td_list.next, struct xhci_td, td_list);

	/* Is this a TRB in the currently executing TD? */
	xhci_dbg(xhci, "%s - looking for TD\n", __func__);
	event_seg = trb_in_td(ep_ring->deq_seg, ep_ring->dequeue,
			td->last_trb, event_dma);
	xhci_dbg(xhci, "%s - found event_seg = %p\n", __func__, event_seg);
	if (!event_seg) {
		/* HC is busted, give up! */
		xhci_err(xhci, "ERROR Transfer event TRB DMA ptr not part of current TD\n");
		return -ESHUTDOWN;
	}
	event_trb = &event_seg->trbs[(event_dma - event_seg->dma) / sizeof(*event_trb)];
	xhci_dbg(xhci, "Event TRB with TRB type ID %u\n",
			(unsigned int) (event->flags & TRB_TYPE_BITMASK)>>10);
	xhci_dbg(xhci, "Offset 0x00 (buffer lo) = 0x%x\n",
			lower_32_bits(event->buffer));
	xhci_dbg(xhci, "Offset 0x04 (buffer hi) = 0x%x\n",
			upper_32_bits(event->buffer));
	xhci_dbg(xhci, "Offset 0x08 (transfer length) = 0x%x\n",
			(unsigned int) event->transfer_len);
	xhci_dbg(xhci, "Offset 0x0C (flags) = 0x%x\n",
			(unsigned int) event->flags);

	/* Look for common error cases */
	trb_comp_code = GET_COMP_CODE(event->transfer_len);
	switch (trb_comp_code) {
	/* Skip codes that require special handling depending on
	 * transfer type
	 */
	case COMP_SUCCESS:
	case COMP_SHORT_TX:
		break;
	case COMP_STOP:
		xhci_dbg(xhci, "Stopped on Transfer TRB\n");
		break;
	case COMP_STOP_INVAL:
		xhci_dbg(xhci, "Stopped on No-op or Link TRB\n");
		break;
	case COMP_STALL:
		xhci_warn(xhci, "WARN: Stalled endpoint\n");
		ep->ep_state |= EP_HALTED;
		status = -EPIPE;
		break;
	case COMP_TRB_ERR:
		xhci_warn(xhci, "WARN: TRB error on endpoint\n");
		status = -EILSEQ;
		break;
	case COMP_SPLIT_ERR:
	case COMP_TX_ERR:
		xhci_warn(xhci, "WARN: transfer error on endpoint\n");
		status = -EPROTO;
		break;
	case COMP_BABBLE:
		xhci_warn(xhci, "WARN: babble error on endpoint\n");
		status = -EOVERFLOW;
		break;
	case COMP_DB_ERR:
		xhci_warn(xhci, "WARN: HC couldn't access mem fast enough\n");
		status = -ENOSR;
		break;
	default:
		if (trb_comp_code >= 224 && trb_comp_code <= 255) {
			/* Vendor defined "informational" completion code,
			 * treat as not-an-error.
			 */
			xhci_dbg(xhci, "Vendor defined info completion code %u\n",
					trb_comp_code);
			xhci_dbg(xhci, "Treating code as success.\n");
			status = 0;
			break;
		}
		xhci_warn(xhci, "ERROR Unknown event condition, HC probably busted\n");
		urb = NULL;
		goto cleanup;
	}
	/* Now update the urb's actual_length and give back to the core */
	/* Was this a control transfer? */
	if (usb_endpoint_xfer_control(&td->urb->ep->desc)) {
		xhci_debug_trb(xhci, xhci->event_ring->dequeue);
		switch (trb_comp_code) {
		case COMP_SUCCESS:
			if (event_trb == ep_ring->dequeue) {
				xhci_warn(xhci, "WARN: Success on ctrl setup TRB without IOC set??\n");
				status = -ESHUTDOWN;
			} else if (event_trb != td->last_trb) {
				xhci_warn(xhci, "WARN: Success on ctrl data TRB without IOC set??\n");
				status = -ESHUTDOWN;
			} else {
				xhci_dbg(xhci, "Successful control transfer!\n");
				status = 0;
			}
			break;
		case COMP_SHORT_TX:
			xhci_warn(xhci, "WARN: short transfer on control ep\n");
			if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
				status = -EREMOTEIO;
			else
				status = 0;
			break;

		default:
			if (!xhci_requires_manual_halt_cleanup(xhci,
						ep_ctx, trb_comp_code))
				break;
			xhci_dbg(xhci, "TRB error code %u, "
					"halted endpoint index = %u\n",
					trb_comp_code, ep_index);
			/* else fall through */
		case COMP_STALL:
			/* Did we transfer part of the data (middle) phase? */
			if (event_trb != ep_ring->dequeue &&
					event_trb != td->last_trb)
				td->urb->actual_length =
					td->urb->transfer_buffer_length
					- TRB_LEN(event->transfer_len);
			else
				td->urb->actual_length = 0;

			xhci_cleanup_halted_endpoint(xhci,
					slot_id, ep_index, td, event_trb);
			goto td_cleanup;
		}
		/*
		 * Did we transfer any data, despite the errors that might have
		 * happened?  I.e. did we get past the setup stage?
		 */
		if (event_trb != ep_ring->dequeue) {
			/* The event was for the status stage */
			if (event_trb == td->last_trb) {
				if (td->urb->actual_length != 0) {
					/* Don't overwrite a previously set error code */
					if ((status == -EINPROGRESS ||
								status == 0) &&
							(td->urb->transfer_flags
							 & URB_SHORT_NOT_OK))
						/* Did we already see a short data stage? */
						status = -EREMOTEIO;
				} else {
					td->urb->actual_length =
						td->urb->transfer_buffer_length;
				}
			} else {
			/* Maybe the event was for the data stage? */
				if (trb_comp_code != COMP_STOP_INVAL) {
					/* We didn't stop on a link TRB in the middle */
					td->urb->actual_length =
						td->urb->transfer_buffer_length -
						TRB_LEN(event->transfer_len);
					xhci_dbg(xhci, "Waiting for status stage event\n");
					urb = NULL;
					goto cleanup;
				}
			}
		}
	} else {
		switch (trb_comp_code) {
		case COMP_SUCCESS:
			/* Double check that the HW transferred everything. */
			if (event_trb != td->last_trb) {
				xhci_warn(xhci, "WARN Successful completion "
						"on short TX\n");
				if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
					status = -EREMOTEIO;
				else
					status = 0;
			} else {
				if (usb_endpoint_xfer_bulk(&td->urb->ep->desc))
					xhci_dbg(xhci, "Successful bulk "
							"transfer!\n");
				else
					xhci_dbg(xhci, "Successful interrupt "
							"transfer!\n");
				status = 0;
			}
			break;
		case COMP_SHORT_TX:
			if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
				status = -EREMOTEIO;
			else
				status = 0;
			break;
		default:
			/* Others already handled above */
			break;
		}
		dev_dbg(&td->urb->dev->dev,
				"ep %#x - asked for %d bytes, "
				"%d bytes untransferred\n",
				td->urb->ep->desc.bEndpointAddress,
				td->urb->transfer_buffer_length,
				TRB_LEN(event->transfer_len));
		/* Fast path - was this the last TRB in the TD for this URB? */
		if (event_trb == td->last_trb) {
			if (TRB_LEN(event->transfer_len) != 0) {
				td->urb->actual_length =
					td->urb->transfer_buffer_length -
					TRB_LEN(event->transfer_len);
				if (td->urb->transfer_buffer_length <
						td->urb->actual_length) {
					xhci_warn(xhci, "HC gave bad length "
							"of %d bytes left\n",
							TRB_LEN(event->transfer_len));
					td->urb->actual_length = 0;
					if (td->urb->transfer_flags &
							URB_SHORT_NOT_OK)
						status = -EREMOTEIO;
					else
						status = 0;
				}
				/* Don't overwrite a previously set error code */
				if (status == -EINPROGRESS) {
					if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
						status = -EREMOTEIO;
					else
						status = 0;
				}
			} else {
				td->urb->actual_length = td->urb->transfer_buffer_length;
				/* Ignore a short packet completion if the
				 * untransferred length was zero.
				 */
				if (status == -EREMOTEIO)
					status = 0;
			}
		} else {
			/* Slow path - walk the list, starting from the dequeue
			 * pointer, to get the actual length transferred.
			 */
			union xhci_trb *cur_trb;
			struct xhci_segment *cur_seg;

			td->urb->actual_length = 0;
			for (cur_trb = ep_ring->dequeue, cur_seg = ep_ring->deq_seg;
					cur_trb != event_trb;
					next_trb(xhci, ep_ring, &cur_seg, &cur_trb)) {
				if (TRB_TYPE(cur_trb->generic.field[3]) != TRB_TR_NOOP &&
						TRB_TYPE(cur_trb->generic.field[3]) != TRB_LINK)
					td->urb->actual_length +=
						TRB_LEN(cur_trb->generic.field[2]);
			}
			/* If the ring didn't stop on a Link or No-op TRB, add
			 * in the actual bytes transferred from the Normal TRB
			 */
			if (trb_comp_code != COMP_STOP_INVAL)
				td->urb->actual_length +=
					TRB_LEN(cur_trb->generic.field[2]) -
					TRB_LEN(event->transfer_len);
		}
	}
	if (trb_comp_code == COMP_STOP_INVAL ||
			trb_comp_code == COMP_STOP) {
		/* The Endpoint Stop Command completion will take care of any
		 * stopped TDs.  A stopped TD may be restarted, so don't update
		 * the ring dequeue pointer or take this TD off any lists yet.
		 */
		ep->stopped_td = td;
		ep->stopped_trb = event_trb;
	} else {
		if (trb_comp_code == COMP_STALL) {
			/* The transfer is completed from the driver's
			 * perspective, but we need to issue a set dequeue
			 * command for this stalled endpoint to move the dequeue
			 * pointer past the TD.  We can't do that here because
			 * the halt condition must be cleared first.  Let the
			 * USB class driver clear the stall later.
			 */
			ep->stopped_td = td;
			ep->stopped_trb = event_trb;
		} else if (xhci_requires_manual_halt_cleanup(xhci,
					ep_ctx, trb_comp_code)) {
			/* Other types of errors halt the endpoint, but the
			 * class driver doesn't call usb_reset_endpoint() unless
			 * the error is -EPIPE.  Clear the halted status in the
			 * xHCI hardware manually.
			 */
			xhci_cleanup_halted_endpoint(xhci,
					slot_id, ep_index, td, event_trb);
		} else {
			/* Update ring dequeue pointer */
			while (ep_ring->dequeue != td->last_trb)
				inc_deq(xhci, ep_ring, false);
			inc_deq(xhci, ep_ring, false);
		}

td_cleanup:
		/* Clean up the endpoint's TD list */
		urb = td->urb;
		/* Do one last check of the actual transfer length.
		 * If the host controller said we transferred more data than
		 * the buffer length, urb->actual_length will be a very big
		 * number (since it's unsigned).  Play it safe and say we didn't
		 * transfer anything.
		 */
		if (urb->actual_length > urb->transfer_buffer_length) {
			xhci_warn(xhci, "URB transfer length is wrong, "
					"xHC issue? req. len = %u, "
					"act. len = %u\n",
					urb->transfer_buffer_length,
					urb->actual_length);
			urb->actual_length = 0;
			if (td->urb->transfer_flags & URB_SHORT_NOT_OK)
				status = -EREMOTEIO;
			else
				status = 0;
		}
		list_del(&td->td_list);
		/* Was this TD slated to be cancelled but completed anyway? */
		if (!list_empty(&td->cancelled_td_list))
			list_del(&td->cancelled_td_list);

		/* Leave the TD around for the reset endpoint function to use
		 * (but only if it's not a control endpoint, since we already
		 * queued the Set TR dequeue pointer command for stalled
		 * control endpoints).
		 */
		if (usb_endpoint_xfer_control(&urb->ep->desc) ||
			(trb_comp_code != COMP_STALL &&
				trb_comp_code != COMP_BABBLE)) {
			kfree(td);
		}
		urb->hcpriv = NULL;
	}
cleanup:
	inc_deq(xhci, xhci->event_ring, true);
	xhci_set_hc_event_deq(xhci);

	/* FIXME for multi-TD URBs (who have buffers bigger than 64MB) */
	if (urb) {
		usb_hcd_unlink_urb_from_ep(xhci_to_hcd(xhci), urb);
		xhci_dbg(xhci, "Giveback URB %p, len = %d, status = %d\n",
				urb, urb->actual_length, status);
		spin_unlock(&xhci->lock);
		usb_hcd_giveback_urb(xhci_to_hcd(xhci), urb, status);
		spin_lock(&xhci->lock);
	}
	return 0;
}

/*
 * This function handles all OS-owned events on the event ring.  It may drop
 * xhci->lock between event processing (e.g. to pass up port status changes).
 */
void xhci_handle_event(struct xhci_hcd *xhci)
{
	union xhci_trb *event;
	int update_ptrs = 1;
	int ret;

	xhci_dbg(xhci, "In %s\n", __func__);
	if (!xhci->event_ring || !xhci->event_ring->dequeue) {
		xhci->error_bitmask |= 1 << 1;
		return;
	}

	event = xhci->event_ring->dequeue;
	/* Does the HC or OS own the TRB? */
	if ((event->event_cmd.flags & TRB_CYCLE) !=
			xhci->event_ring->cycle_state) {
		xhci->error_bitmask |= 1 << 2;
		return;
	}
	xhci_dbg(xhci, "%s - OS owns TRB\n", __func__);

	/* FIXME: Handle more event types. */
	switch ((event->event_cmd.flags & TRB_TYPE_BITMASK)) {
	case TRB_TYPE(TRB_COMPLETION):
		xhci_dbg(xhci, "%s - calling handle_cmd_completion\n", __func__);
		handle_cmd_completion(xhci, &event->event_cmd);
		xhci_dbg(xhci, "%s - returned from handle_cmd_completion\n", __func__);
		break;
	case TRB_TYPE(TRB_PORT_STATUS):
		xhci_dbg(xhci, "%s - calling handle_port_status\n", __func__);
		handle_port_status(xhci, event);
		xhci_dbg(xhci, "%s - returned from handle_port_status\n", __func__);
		update_ptrs = 0;
		break;
	case TRB_TYPE(TRB_TRANSFER):
		xhci_dbg(xhci, "%s - calling handle_tx_event\n", __func__);
		ret = handle_tx_event(xhci, &event->trans_event);
		xhci_dbg(xhci, "%s - returned from handle_tx_event\n", __func__);
		if (ret < 0)
			xhci->error_bitmask |= 1 << 9;
		else
			update_ptrs = 0;
		break;
	default:
		xhci->error_bitmask |= 1 << 3;
	}
	/* Any of the above functions may drop and re-acquire the lock, so check
	 * to make sure a watchdog timer didn't mark the host as non-responsive.
	 */
	if (xhci->xhc_state & XHCI_STATE_DYING) {
		xhci_dbg(xhci, "xHCI host dying, returning from "
				"event handler.\n");
		return;
	}

	if (update_ptrs) {
		/* Update SW and HC event ring dequeue pointer */
		inc_deq(xhci, xhci->event_ring, true);
		xhci_set_hc_event_deq(xhci);
	}
	/* Are there more items on the event ring? */
	xhci_handle_event(xhci);
}

/****		Endpoint Ring Operations	****/

/*
 * Generic function for queueing a TRB on a ring.
 * The caller must have checked to make sure there's room on the ring.
 */
static void queue_trb(struct xhci_hcd *xhci, struct xhci_ring *ring,
		bool consumer,
		u32 field1, u32 field2, u32 field3, u32 field4)
{
	struct xhci_generic_trb *trb;

	trb = &ring->enqueue->generic;
	trb->field[0] = field1;
	trb->field[1] = field2;
	trb->field[2] = field3;
	trb->field[3] = field4;
	inc_enq(xhci, ring, consumer);
}

/*
 * Does various checks on the endpoint ring, and makes it ready to queue num_trbs.
 * FIXME allocate segments if the ring is full.
 */
static int prepare_ring(struct xhci_hcd *xhci, struct xhci_ring *ep_ring,
		u32 ep_state, unsigned int num_trbs, gfp_t mem_flags)
{
	/* Make sure the endpoint has been added to xHC schedule */
	xhci_dbg(xhci, "Endpoint state = 0x%x\n", ep_state);
	switch (ep_state) {
	case EP_STATE_DISABLED:
		/*
		 * USB core changed config/interfaces without notifying us,
		 * or hardware is reporting the wrong state.
		 */
		xhci_warn(xhci, "WARN urb submitted to disabled ep\n");
		return -ENOENT;
	case EP_STATE_ERROR:
		xhci_warn(xhci, "WARN waiting for error on ep to be cleared\n");
		/* FIXME event handling code for error needs to clear it */
		/* XXX not sure if this should be -ENOENT or not */
		return -EINVAL;
	case EP_STATE_HALTED:
		xhci_dbg(xhci, "WARN halted endpoint, queueing URB anyway.\n");
	case EP_STATE_STOPPED:
	case EP_STATE_RUNNING:
		break;
	default:
		xhci_err(xhci, "ERROR unknown endpoint state for ep\n");
		/*
		 * FIXME issue Configure Endpoint command to try to get the HC
		 * back into a known state.
		 */
		return -EINVAL;
	}
	if (!room_on_ring(xhci, ep_ring, num_trbs)) {
		/* FIXME allocate more room */
		xhci_err(xhci, "ERROR no room on ep ring\n");
		return -ENOMEM;
	}
	return 0;
}

static int prepare_transfer(struct xhci_hcd *xhci,
		struct xhci_virt_device *xdev,
		unsigned int ep_index,
		unsigned int num_trbs,
		struct urb *urb,
		struct xhci_td **td,
		gfp_t mem_flags)
{
	int ret;
	struct xhci_ep_ctx *ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
	ret = prepare_ring(xhci, xdev->eps[ep_index].ring,
			ep_ctx->ep_info & EP_STATE_MASK,
			num_trbs, mem_flags);
	if (ret)
		return ret;
	*td = kzalloc(sizeof(struct xhci_td), mem_flags);
	if (!*td)
		return -ENOMEM;
	INIT_LIST_HEAD(&(*td)->td_list);
	INIT_LIST_HEAD(&(*td)->cancelled_td_list);

	ret = usb_hcd_link_urb_to_ep(xhci_to_hcd(xhci), urb);
	if (unlikely(ret)) {
		kfree(*td);
		return ret;
	}

	(*td)->urb = urb;
	urb->hcpriv = (void *) (*td);
	/* Add this TD to the tail of the endpoint ring's TD list */
	list_add_tail(&(*td)->td_list, &xdev->eps[ep_index].ring->td_list);
	(*td)->start_seg = xdev->eps[ep_index].ring->enq_seg;
	(*td)->first_trb = xdev->eps[ep_index].ring->enqueue;

	return 0;
}

static unsigned int count_sg_trbs_needed(struct xhci_hcd *xhci, struct urb *urb)
{
	int num_sgs, num_trbs, running_total, temp, i;
	struct scatterlist *sg;

	sg = NULL;
	num_sgs = urb->num_sgs;
	temp = urb->transfer_buffer_length;

	xhci_dbg(xhci, "count sg list trbs: \n");
	num_trbs = 0;
	for_each_sg(urb->sg->sg, sg, num_sgs, i) {
		unsigned int previous_total_trbs = num_trbs;
		unsigned int len = sg_dma_len(sg);

		/* Scatter gather list entries may cross 64KB boundaries */
		running_total = TRB_MAX_BUFF_SIZE -
			(sg_dma_address(sg) & ((1 << TRB_MAX_BUFF_SHIFT) - 1));
		if (running_total != 0)
			num_trbs++;

		/* How many more 64KB chunks to transfer, how many more TRBs? */
		while (running_total < sg_dma_len(sg)) {
			num_trbs++;
			running_total += TRB_MAX_BUFF_SIZE;
		}
		xhci_dbg(xhci, " sg #%d: dma = %#llx, len = %#x (%d), num_trbs = %d\n",
				i, (unsigned long long)sg_dma_address(sg),
				len, len, num_trbs - previous_total_trbs);

		len = min_t(int, len, temp);
		temp -= len;
		if (temp == 0)
			break;
	}
	xhci_dbg(xhci, "\n");
	if (!in_interrupt())
		dev_dbg(&urb->dev->dev, "ep %#x - urb len = %d, sglist used, num_trbs = %d\n",
				urb->ep->desc.bEndpointAddress,
				urb->transfer_buffer_length,
				num_trbs);
	return num_trbs;
}

static void check_trb_math(struct urb *urb, int num_trbs, int running_total)
{
	if (num_trbs != 0)
		dev_dbg(&urb->dev->dev, "%s - ep %#x - Miscalculated number of "
				"TRBs, %d left\n", __func__,
				urb->ep->desc.bEndpointAddress, num_trbs);
	if (running_total != urb->transfer_buffer_length)
		dev_dbg(&urb->dev->dev, "%s - ep %#x - Miscalculated tx length, "
				"queued %#x (%d), asked for %#x (%d)\n",
				__func__,
				urb->ep->desc.bEndpointAddress,
				running_total, running_total,
				urb->transfer_buffer_length,
				urb->transfer_buffer_length);
}

static void giveback_first_trb(struct xhci_hcd *xhci, int slot_id,
		unsigned int ep_index, int start_cycle,
		struct xhci_generic_trb *start_trb, struct xhci_td *td)
{
	/*
	 * Pass all the TRBs to the hardware at once and make sure this write
	 * isn't reordered.
	 */
	wmb();
	start_trb->field[3] |= start_cycle;
	ring_ep_doorbell(xhci, slot_id, ep_index);
}

/*
 * xHCI uses normal TRBs for both bulk and interrupt.  When the interrupt
 * endpoint is to be serviced, the xHC will consume (at most) one TD.  A TD
 * (comprised of sg list entries) can take several service intervals to
 * transmit.
 */
int xhci_queue_intr_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
		struct urb *urb, int slot_id, unsigned int ep_index)
{
	struct xhci_ep_ctx *ep_ctx = xhci_get_ep_ctx(xhci,
			xhci->devs[slot_id]->out_ctx, ep_index);
	int xhci_interval;
	int ep_interval;

	xhci_interval = EP_INTERVAL_TO_UFRAMES(ep_ctx->ep_info);
	ep_interval = urb->interval;
	/* Convert to microframes */
	if (urb->dev->speed == USB_SPEED_LOW ||
			urb->dev->speed == USB_SPEED_FULL)
		ep_interval *= 8;
	/* FIXME change this to a warning and a suggestion to use the new API
	 * to set the polling interval (once the API is added).
	 */
	if (xhci_interval != ep_interval) {
		if (!printk_ratelimit())
			dev_dbg(&urb->dev->dev, "Driver uses different interval"
					" (%d microframe%s) than xHCI "
					"(%d microframe%s)\n",
					ep_interval,
					ep_interval == 1 ? "" : "s",
					xhci_interval,
					xhci_interval == 1 ? "" : "s");
		urb->interval = xhci_interval;
		/* Convert back to frames for LS/FS devices */
		if (urb->dev->speed == USB_SPEED_LOW ||
				urb->dev->speed == USB_SPEED_FULL)
			urb->interval /= 8;
	}
	return xhci_queue_bulk_tx(xhci, GFP_ATOMIC, urb, slot_id, ep_index);
}

/*
 * The TD size is the number of bytes remaining in the TD (including this TRB),
 * right shifted by 10.
 * It must fit in bits 21:17, so it can't be bigger than 31.
 */
static u32 xhci_td_remainder(unsigned int remainder)
{
	u32 max = (1 << (21 - 17 + 1)) - 1;

	if ((remainder >> 10) >= max)
		return max << 17;
	else
		return (remainder >> 10) << 17;
}

static int queue_bulk_sg_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
		struct urb *urb, int slot_id, unsigned int ep_index)
{
	struct xhci_ring *ep_ring;
	unsigned int num_trbs;
	struct xhci_td *td;
	struct scatterlist *sg;
	int num_sgs;
	int trb_buff_len, this_sg_len, running_total;
	bool first_trb;
	u64 addr;

	struct xhci_generic_trb *start_trb;
	int start_cycle;

	ep_ring = xhci->devs[slot_id]->eps[ep_index].ring;
	num_trbs = count_sg_trbs_needed(xhci, urb);
	num_sgs = urb->num_sgs;

	trb_buff_len = prepare_transfer(xhci, xhci->devs[slot_id],
			ep_index, num_trbs, urb, &td, mem_flags);
	if (trb_buff_len < 0)
		return trb_buff_len;
	/*
	 * Don't give the first TRB to the hardware (by toggling the cycle bit)
	 * until we've finished creating all the other TRBs.  The ring's cycle
	 * state may change as we enqueue the other TRBs, so save it too.
	 */
	start_trb = &ep_ring->enqueue->generic;
	start_cycle = ep_ring->cycle_state;

	running_total = 0;
	/*
	 * How much data is in the first TRB?
	 *
	 * There are three forces at work for TRB buffer pointers and lengths:
	 * 1. We don't want to walk off the end of this sg-list entry buffer.
	 * 2. The transfer length that the driver requested may be smaller than
	 *    the amount of memory allocated for this scatter-gather list.
	 * 3. TRBs buffers can't cross 64KB boundaries.
	 */
	sg = urb->sg->sg;
	addr = (u64) sg_dma_address(sg);
	this_sg_len = sg_dma_len(sg);
	trb_buff_len = TRB_MAX_BUFF_SIZE -
		(addr & ((1 << TRB_MAX_BUFF_SHIFT) - 1));
	trb_buff_len = min_t(int, trb_buff_len, this_sg_len);
	if (trb_buff_len > urb->transfer_buffer_length)
		trb_buff_len = urb->transfer_buffer_length;
	xhci_dbg(xhci, "First length to xfer from 1st sglist entry = %u\n",
			trb_buff_len);

	first_trb = true;
	/* Queue the first TRB, even if it's zero-length */
	do {
		u32 field = 0;
		u32 length_field = 0;
		u32 remainder = 0;

		/* Don't change the cycle bit of the first TRB until later */
		if (first_trb)
			first_trb = false;
		else
			field |= ep_ring->cycle_state;

		/* Chain all the TRBs together; clear the chain bit in the last
		 * TRB to indicate it's the last TRB in the chain.
		 */
		if (num_trbs > 1) {
			field |= TRB_CHAIN;
		} else {
			/* FIXME - add check for ZERO_PACKET flag before this */
			td->last_trb = ep_ring->enqueue;
			field |= TRB_IOC;
		}
		xhci_dbg(xhci, " sg entry: dma = %#x, len = %#x (%d), "
				"64KB boundary at %#x, end dma = %#x\n",
				(unsigned int) addr, trb_buff_len, trb_buff_len,
				(unsigned int) (addr + TRB_MAX_BUFF_SIZE) & ~(TRB_MAX_BUFF_SIZE - 1),
				(unsigned int) addr + trb_buff_len);
		if (TRB_MAX_BUFF_SIZE -
				(addr & ((1 << TRB_MAX_BUFF_SHIFT) - 1)) < trb_buff_len) {
			xhci_warn(xhci, "WARN: sg dma xfer crosses 64KB boundaries!\n");
			xhci_dbg(xhci, "Next boundary at %#x, end dma = %#x\n",
					(unsigned int) (addr + TRB_MAX_BUFF_SIZE) & ~(TRB_MAX_BUFF_SIZE - 1),
					(unsigned int) addr + trb_buff_len);
		}
		remainder = xhci_td_remainder(urb->transfer_buffer_length -
				running_total) ;
		length_field = TRB_LEN(trb_buff_len) |
			remainder |
			TRB_INTR_TARGET(0);
		queue_trb(xhci, ep_ring, false,
				lower_32_bits(addr),
				upper_32_bits(addr),
				length_field,
				/* We always want to know if the TRB was short,
				 * or we won't get an event when it completes.
				 * (Unless we use event data TRBs, which are a
				 * waste of space and HC resources.)
				 */
				field | TRB_ISP | TRB_TYPE(TRB_NORMAL));
		--num_trbs;
		running_total += trb_buff_len;

		/* Calculate length for next transfer --
		 * Are we done queueing all the TRBs for this sg entry?
		 */
		this_sg_len -= trb_buff_len;
		if (this_sg_len == 0) {
			--num_sgs;
			if (num_sgs == 0)
				break;
			sg = sg_next(sg);
			addr = (u64) sg_dma_address(sg);
			this_sg_len = sg_dma_len(sg);
		} else {
			addr += trb_buff_len;
		}

		trb_buff_len = TRB_MAX_BUFF_SIZE -
			(addr & ((1 << TRB_MAX_BUFF_SHIFT) - 1));
		trb_buff_len = min_t(int, trb_buff_len, this_sg_len);
		if (running_total + trb_buff_len > urb->transfer_buffer_length)
			trb_buff_len =
				urb->transfer_buffer_length - running_total;
	} while (running_total < urb->transfer_buffer_length);

	check_trb_math(urb, num_trbs, running_total);
	giveback_first_trb(xhci, slot_id, ep_index, start_cycle, start_trb, td);
	return 0;
}

/* This is very similar to what ehci-q.c qtd_fill() does */
int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
		struct urb *urb, int slot_id, unsigned int ep_index)
{
	struct xhci_ring *ep_ring;
	struct xhci_td *td;
	int num_trbs;
	struct xhci_generic_trb *start_trb;
	bool first_trb;
	int start_cycle;
	u32 field, length_field;

	int running_total, trb_buff_len, ret;
	u64 addr;

	if (urb->sg)
		return queue_bulk_sg_tx(xhci, mem_flags, urb, slot_id, ep_index);

	ep_ring = xhci->devs[slot_id]->eps[ep_index].ring;

	num_trbs = 0;
	/* How much data is (potentially) left before the 64KB boundary? */
	running_total = TRB_MAX_BUFF_SIZE -
		(urb->transfer_dma & ((1 << TRB_MAX_BUFF_SHIFT) - 1));

	/* If there's some data on this 64KB chunk, or we have to send a
	 * zero-length transfer, we need at least one TRB
	 */
	if (running_total != 0 || urb->transfer_buffer_length == 0)
		num_trbs++;
	/* How many more 64KB chunks to transfer, how many more TRBs? */
	while (running_total < urb->transfer_buffer_length) {
		num_trbs++;
		running_total += TRB_MAX_BUFF_SIZE;
	}
	/* FIXME: this doesn't deal with URB_ZERO_PACKET - need one more */

	if (!in_interrupt())
		dev_dbg(&urb->dev->dev, "ep %#x - urb len = %#x (%d), addr = %#llx, num_trbs = %d\n",
				urb->ep->desc.bEndpointAddress,
				urb->transfer_buffer_length,
				urb->transfer_buffer_length,
				(unsigned long long)urb->transfer_dma,
				num_trbs);

	ret = prepare_transfer(xhci, xhci->devs[slot_id], ep_index,
			num_trbs, urb, &td, mem_flags);
	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;

	running_total = 0;
	/* How much data is in the first TRB? */
	addr = (u64) urb->transfer_dma;
	trb_buff_len = TRB_MAX_BUFF_SIZE -
		(urb->transfer_dma & ((1 << TRB_MAX_BUFF_SHIFT) - 1));
	if (urb->transfer_buffer_length < trb_buff_len)
		trb_buff_len = urb->transfer_buffer_length;

	first_trb = true;

	/* Queue the first TRB, even if it's zero-length */
	do {
		u32 remainder = 0;
		field = 0;

		/* Don't change the cycle bit of the first TRB until later */
		if (first_trb)
			first_trb = false;
		else
			field |= ep_ring->cycle_state;

		/* Chain all the TRBs together; clear the chain bit in the last
		 * TRB to indicate it's the last TRB in the chain.
		 */
		if (num_trbs > 1) {
			field |= TRB_CHAIN;
		} else {
			/* FIXME - add check for ZERO_PACKET flag before this */
			td->last_trb = ep_ring->enqueue;
			field |= TRB_IOC;
		}
		remainder = xhci_td_remainder(urb->transfer_buffer_length -
				running_total);
		length_field = TRB_LEN(trb_buff_len) |
			remainder |
			TRB_INTR_TARGET(0);
		queue_trb(xhci, ep_ring, false,
				lower_32_bits(addr),
				upper_32_bits(addr),
				length_field,
				/* We always want to know if the TRB was short,
				 * or we won't get an event when it completes.
				 * (Unless we use event data TRBs, which are a
				 * waste of space and HC resources.)
				 */
				field | TRB_ISP | TRB_TYPE(TRB_NORMAL));
		--num_trbs;
		running_total += trb_buff_len;

		/* Calculate length for next transfer */
		addr += trb_buff_len;
		trb_buff_len = urb->transfer_buffer_length - running_total;
		if (trb_buff_len > TRB_MAX_BUFF_SIZE)
			trb_buff_len = TRB_MAX_BUFF_SIZE;
	} while (running_total < urb->transfer_buffer_length);

	check_trb_math(urb, num_trbs, running_total);
	giveback_first_trb(xhci, slot_id, ep_index, start_cycle, start_trb, td);
	return 0;
}

/* Caller must have locked xhci->lock */
int xhci_queue_ctrl_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
		struct urb *urb, int slot_id, unsigned int ep_index)
{
	struct xhci_ring *ep_ring;
	int num_trbs;
	int ret;
	struct usb_ctrlrequest *setup;
	struct xhci_generic_trb *start_trb;
	int start_cycle;
	u32 field, length_field;
	struct xhci_td *td;

	ep_ring = xhci->devs[slot_id]->eps[ep_index].ring;

	/*
	 * Need to copy setup packet into setup TRB, so we can't use the setup
	 * DMA address.
	 */
	if (!urb->setup_packet)
		return -EINVAL;

	if (!in_interrupt())
		xhci_dbg(xhci, "Queueing ctrl tx for slot id %d, ep %d\n",
				slot_id, 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 (urb->transfer_buffer_length > 0)
		num_trbs++;
	ret = prepare_transfer(xhci, xhci->devs[slot_id], ep_index, num_trbs,
			urb, &td, mem_flags);
	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;

	/* Queue setup TRB - see section 6.4.1.2.1 */
	/* FIXME better way to translate setup_packet into two u32 fields? */
	setup = (struct usb_ctrlrequest *) urb->setup_packet;
	queue_trb(xhci, ep_ring, false,
			/* FIXME endianness is probably going to bite my ass here. */
			setup->bRequestType | setup->bRequest << 8 | setup->wValue << 16,
			setup->wIndex | setup->wLength << 16,
			TRB_LEN(8) | TRB_INTR_TARGET(0),
			/* Immediate data in pointer */
			TRB_IDT | TRB_TYPE(TRB_SETUP));

	/* If there's data, queue data TRBs */
	field = 0;
	length_field = TRB_LEN(urb->transfer_buffer_length) |
		xhci_td_remainder(urb->transfer_buffer_length) |
		TRB_INTR_TARGET(0);
	if (urb->transfer_buffer_length > 0) {
		if (setup->bRequestType & USB_DIR_IN)
			field |= TRB_DIR_IN;
		queue_trb(xhci, ep_ring, false,
				lower_32_bits(urb->transfer_dma),
				upper_32_bits(urb->transfer_dma),
				length_field,
				/* Event on short tx */
				field | TRB_ISP | TRB_TYPE(TRB_DATA) | ep_ring->cycle_state);
	}

	/* Save the DMA address of the last TRB in the TD */
	td->last_trb = ep_ring->enqueue;

	/* Queue status TRB - see Table 7 and sections 4.11.2.2 and 6.4.1.2.3 */
	/* If the device sent data, the status stage is an OUT transfer */
	if (urb->transfer_buffer_length > 0 && setup->bRequestType & USB_DIR_IN)
		field = 0;
	else
		field = TRB_DIR_IN;
	queue_trb(xhci, ep_ring, false,
			0,
			0,
			TRB_INTR_TARGET(0),
			/* Event on completion */
			field | TRB_IOC | TRB_TYPE(TRB_STATUS) | ep_ring->cycle_state);

	giveback_first_trb(xhci, slot_id, ep_index, start_cycle, start_trb, td);
	return 0;
}

/****		Command Ring Operations		****/

/* Generic function for queueing a command TRB on the command ring.
 * Check to make sure there's room on the command ring for one command TRB.
 * Also check that there's room reserved for commands that must not fail.
 * If this is a command that must not fail, meaning command_must_succeed = TRUE,
 * then only check for the number of reserved spots.
 * Don't decrement xhci->cmd_ring_reserved_trbs after we've queued the TRB
 * because the command event handler may want to resubmit a failed command.
 */
static int queue_command(struct xhci_hcd *xhci, u32 field1, u32 field2,
		u32 field3, u32 field4, bool command_must_succeed)
{
	int reserved_trbs = xhci->cmd_ring_reserved_trbs;
	if (!command_must_succeed)
		reserved_trbs++;

	if (!room_on_ring(xhci, xhci->cmd_ring, reserved_trbs)) {
		if (!in_interrupt())
			xhci_err(xhci, "ERR: No room for command on command ring\n");
		if (command_must_succeed)
			xhci_err(xhci, "ERR: Reserved TRB counting for "
					"unfailable commands failed.\n");
		return -ENOMEM;
	}
	queue_trb(xhci, xhci->cmd_ring, false, field1, field2, field3,
			field4 | xhci->cmd_ring->cycle_state);
	return 0;
}

/* Queue a no-op command on the command ring */
static int queue_cmd_noop(struct xhci_hcd *xhci)
{
	return queue_command(xhci, 0, 0, 0, TRB_TYPE(TRB_CMD_NOOP), false);
}

/*
 * Place a no-op command on the command ring to test the command and
 * event ring.
 */
void *xhci_setup_one_noop(struct xhci_hcd *xhci)
{
	if (queue_cmd_noop(xhci) < 0)
		return NULL;
	xhci->noops_submitted++;
	return xhci_ring_cmd_db;
}

/* Queue a slot enable or disable request on the command ring */
int xhci_queue_slot_control(struct xhci_hcd *xhci, u32 trb_type, u32 slot_id)
{
	return queue_command(xhci, 0, 0, 0,
			TRB_TYPE(trb_type) | SLOT_ID_FOR_TRB(slot_id), false);
}

/* Queue an address device command TRB */
int xhci_queue_address_device(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
		u32 slot_id)
{
	return queue_command(xhci, lower_32_bits(in_ctx_ptr),
			upper_32_bits(in_ctx_ptr), 0,
			TRB_TYPE(TRB_ADDR_DEV) | SLOT_ID_FOR_TRB(slot_id),
			false);
}

/* Queue a configure endpoint command TRB */
int xhci_queue_configure_endpoint(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
		u32 slot_id, bool command_must_succeed)
{
	return queue_command(xhci, lower_32_bits(in_ctx_ptr),
			upper_32_bits(in_ctx_ptr), 0,
			TRB_TYPE(TRB_CONFIG_EP) | SLOT_ID_FOR_TRB(slot_id),
			command_must_succeed);
}

/* Queue an evaluate context command TRB */
int xhci_queue_evaluate_context(struct xhci_hcd *xhci, dma_addr_t in_ctx_ptr,
		u32 slot_id)
{
	return queue_command(xhci, lower_32_bits(in_ctx_ptr),
			upper_32_bits(in_ctx_ptr), 0,
			TRB_TYPE(TRB_EVAL_CONTEXT) | SLOT_ID_FOR_TRB(slot_id),
			false);
}

int xhci_queue_stop_endpoint(struct xhci_hcd *xhci, int slot_id,
		unsigned int ep_index)
{
	u32 trb_slot_id = SLOT_ID_FOR_TRB(slot_id);
	u32 trb_ep_index = EP_ID_FOR_TRB(ep_index);
	u32 type = TRB_TYPE(TRB_STOP_RING);

	return queue_command(xhci, 0, 0, 0,
			trb_slot_id | trb_ep_index | type, false);
}

/* Set Transfer Ring Dequeue Pointer command.
 * This should not be used for endpoints that have streams enabled.
 */
static int queue_set_tr_deq(struct xhci_hcd *xhci, int slot_id,
		unsigned int ep_index, struct xhci_segment *deq_seg,
		union xhci_trb *deq_ptr, u32 cycle_state)
{
	dma_addr_t addr;
	u32 trb_slot_id = SLOT_ID_FOR_TRB(slot_id);
	u32 trb_ep_index = EP_ID_FOR_TRB(ep_index);
	u32 type = TRB_TYPE(TRB_SET_DEQ);

	addr = xhci_trb_virt_to_dma(deq_seg, deq_ptr);
	if (addr == 0) {
		xhci_warn(xhci, "WARN Cannot submit Set TR Deq Ptr\n");
		xhci_warn(xhci, "WARN deq seg = %p, deq pt = %p\n",
				deq_seg, deq_ptr);
		return 0;
	}
	return queue_command(xhci, lower_32_bits(addr) | cycle_state,
			upper_32_bits(addr), 0,
			trb_slot_id | trb_ep_index | type, false);
}

int xhci_queue_reset_ep(struct xhci_hcd *xhci, int slot_id,
		unsigned int ep_index)
{
	u32 trb_slot_id = SLOT_ID_FOR_TRB(slot_id);
	u32 trb_ep_index = EP_ID_FOR_TRB(ep_index);
	u32 type = TRB_TYPE(TRB_RESET_EP);

	return queue_command(xhci, 0, 0, 0, trb_slot_id | trb_ep_index | type,
			false);
}