diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
index 64d036804715b6554f46f24c39256b39221d975c..44eeaa016f1b8adac25a4d84facecc39c4c98e18 100644
--- a/drivers/usb/host/xhci-mem.c
+++ b/drivers/usb/host/xhci-mem.c
@@ -1390,6 +1390,22 @@ struct xhci_command *xhci_alloc_command(struct xhci_hcd *xhci,
 	return command;
 }
 
+void xhci_urb_free_priv(struct xhci_hcd *xhci, struct urb_priv *urb_priv)
+{
+	int last;
+
+	if (!urb_priv)
+		return;
+
+	last = urb_priv->length - 1;
+	if (last >= 0) {
+		int	i;
+		for (i = 0; i <= last; i++)
+			kfree(urb_priv->td[i]);
+	}
+	kfree(urb_priv);
+}
+
 void xhci_free_command(struct xhci_hcd *xhci,
 		struct xhci_command *command)
 {
diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 4c3501003b8eee21c2c85ec6fa5d1aa90b403aaa..fa8c93559133274846db7eefde2f454ce6b45330 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -578,16 +578,24 @@ 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);
+	struct urb	*urb;
+	struct urb_priv	*urb_priv;
 
-	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);
+	urb = cur_td->urb;
+	urb_priv = urb->hcpriv;
+	urb_priv->td_cnt++;
 
-	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);
+	/* Only giveback urb when this is the last td in urb */
+	if (urb_priv->td_cnt == urb_priv->length) {
+		usb_hcd_unlink_urb_from_ep(hcd, urb);
+		xhci_dbg(xhci, "Giveback %s URB %p\n", adjective, urb);
+
+		spin_unlock(&xhci->lock);
+		usb_hcd_giveback_urb(hcd, urb, status);
+		xhci_urb_free_priv(xhci, urb_priv);
+		spin_lock(&xhci->lock);
+		xhci_dbg(xhci, "%s URB given back\n", adjective);
+	}
 }
 
 /*
@@ -1272,6 +1280,7 @@ static int finish_td(struct xhci_hcd *xhci, struct xhci_td *td,
 	struct urb *urb = NULL;
 	struct xhci_ep_ctx *ep_ctx;
 	int ret = 0;
+	struct urb_priv	*urb_priv;
 	u32 trb_comp_code;
 
 	slot_id = TRB_TO_SLOT_ID(event->flags);
@@ -1325,6 +1334,7 @@ static int finish_td(struct xhci_hcd *xhci, struct xhci_td *td,
 td_cleanup:
 		/* Clean up the endpoint's TD list */
 		urb = td->urb;
+		urb_priv = urb->hcpriv;
 
 		/* Do one last check of the actual transfer length.
 		 * If the host controller said we transferred more data than
@@ -1349,7 +1359,10 @@ static int finish_td(struct xhci_hcd *xhci, struct xhci_td *td,
 		if (!list_empty(&td->cancelled_td_list))
 			list_del(&td->cancelled_td_list);
 
-		ret = 1;
+		urb_priv->td_cnt++;
+		/* Giveback the urb when all the tds are completed */
+		if (urb_priv->td_cnt == urb_priv->length)
+			ret = 1;
 	}
 
 	return ret;
@@ -1588,6 +1601,7 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 	union xhci_trb *event_trb;
 	struct urb *urb = NULL;
 	int status = -EINPROGRESS;
+	struct urb_priv *urb_priv;
 	struct xhci_ep_ctx *ep_ctx;
 	u32 trb_comp_code;
 	int ret = 0;
@@ -1770,6 +1784,7 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 
 		if (ret) {
 			urb = td->urb;
+			urb_priv = urb->hcpriv;
 			/* Leave the TD around for the reset endpoint function
 			 * to use(but only if it's not a control endpoint,
 			 * since we already queued the Set TR dequeue pointer
@@ -1778,7 +1793,7 @@ static int handle_tx_event(struct xhci_hcd *xhci,
 			if (usb_endpoint_xfer_control(&urb->ep->desc) ||
 				(trb_comp_code != COMP_STALL &&
 					trb_comp_code != COMP_BABBLE))
-				kfree(td);
+				xhci_urb_free_priv(xhci, urb_priv);
 
 			usb_hcd_unlink_urb_from_ep(xhci_to_hcd(xhci), urb);
 			xhci_dbg(xhci, "Giveback URB %p, len = %d, "
@@ -1979,10 +1994,12 @@ static int prepare_transfer(struct xhci_hcd *xhci,
 		unsigned int stream_id,
 		unsigned int num_trbs,
 		struct urb *urb,
-		struct xhci_td **td,
+		unsigned int td_index,
 		gfp_t mem_flags)
 {
 	int ret;
+	struct urb_priv *urb_priv;
+	struct xhci_td	*td;
 	struct xhci_ring *ep_ring;
 	struct xhci_ep_ctx *ep_ctx = xhci_get_ep_ctx(xhci, xdev->out_ctx, ep_index);
 
@@ -1998,24 +2015,29 @@ static int prepare_transfer(struct xhci_hcd *xhci,
 			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;
+	urb_priv = urb->hcpriv;
+	td = urb_priv->td[td_index];
+
+	INIT_LIST_HEAD(&td->td_list);
+	INIT_LIST_HEAD(&td->cancelled_td_list);
+
+	if (td_index == 0) {
+		ret = usb_hcd_link_urb_to_ep(xhci_to_hcd(xhci), urb);
+		if (unlikely(ret)) {
+			xhci_urb_free_priv(xhci, urb_priv);
+			urb->hcpriv = NULL;
+			return ret;
+		}
 	}
 
-	(*td)->urb = urb;
-	urb->hcpriv = (void *) (*td);
+	td->urb = urb;
 	/* Add this TD to the tail of the endpoint ring's TD list */
-	list_add_tail(&(*td)->td_list, &ep_ring->td_list);
-	(*td)->start_seg = ep_ring->enq_seg;
-	(*td)->first_trb = ep_ring->enqueue;
+	list_add_tail(&td->td_list, &ep_ring->td_list);
+	td->start_seg = ep_ring->enq_seg;
+	td->first_trb = ep_ring->enqueue;
+
+	urb_priv->td[td_index] = td;
 
 	return 0;
 }
@@ -2154,6 +2176,7 @@ static int queue_bulk_sg_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 {
 	struct xhci_ring *ep_ring;
 	unsigned int num_trbs;
+	struct urb_priv *urb_priv;
 	struct xhci_td *td;
 	struct scatterlist *sg;
 	int num_sgs;
@@ -2174,9 +2197,13 @@ static int queue_bulk_sg_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 
 	trb_buff_len = prepare_transfer(xhci, xhci->devs[slot_id],
 			ep_index, urb->stream_id,
-			num_trbs, urb, &td, mem_flags);
+			num_trbs, urb, 0, mem_flags);
 	if (trb_buff_len < 0)
 		return trb_buff_len;
+
+	urb_priv = urb->hcpriv;
+	td = urb_priv->td[0];
+
 	/*
 	 * Don't give the first TRB to the hardware (by toggling the cycle bit)
 	 * until we've finished creating all the other TRBs.  The ring's cycle
@@ -2297,6 +2324,7 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 		struct urb *urb, int slot_id, unsigned int ep_index)
 {
 	struct xhci_ring *ep_ring;
+	struct urb_priv *urb_priv;
 	struct xhci_td *td;
 	int num_trbs;
 	struct xhci_generic_trb *start_trb;
@@ -2342,10 +2370,13 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 
 	ret = prepare_transfer(xhci, xhci->devs[slot_id],
 			ep_index, urb->stream_id,
-			num_trbs, urb, &td, mem_flags);
+			num_trbs, urb, 0, mem_flags);
 	if (ret < 0)
 		return ret;
 
+	urb_priv = urb->hcpriv;
+	td = urb_priv->td[0];
+
 	/*
 	 * Don't give the first TRB to the hardware (by toggling the cycle bit)
 	 * until we've finished creating all the other TRBs.  The ring's cycle
@@ -2431,6 +2462,7 @@ int xhci_queue_ctrl_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 	struct xhci_generic_trb *start_trb;
 	int start_cycle;
 	u32 field, length_field;
+	struct urb_priv *urb_priv;
 	struct xhci_td *td;
 
 	ep_ring = xhci_urb_to_transfer_ring(xhci, urb);
@@ -2458,10 +2490,13 @@ int xhci_queue_ctrl_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 		num_trbs++;
 	ret = prepare_transfer(xhci, xhci->devs[slot_id],
 			ep_index, urb->stream_id,
-			num_trbs, urb, &td, mem_flags);
+			num_trbs, urb, 0, mem_flags);
 	if (ret < 0)
 		return ret;
 
+	urb_priv = urb->hcpriv;
+	td = urb_priv->td[0];
+
 	/*
 	 * Don't give the first TRB to the hardware (by toggling the cycle bit)
 	 * until we've finished creating all the other TRBs.  The ring's cycle
diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
index 3106d22ae053fe7feeaf6cbe0a66635df7e994f2..295a0a2063a6ea875455698ba1f6fa37d7640885 100644
--- a/drivers/usb/host/xhci.c
+++ b/drivers/usb/host/xhci.c
@@ -804,7 +804,8 @@ int xhci_urb_enqueue(struct usb_hcd *hcd, struct urb *urb, gfp_t mem_flags)
 	unsigned long flags;
 	int ret = 0;
 	unsigned int slot_id, ep_index;
-
+	struct urb_priv	*urb_priv;
+	int size, i;
 
 	if (!urb || xhci_check_args(hcd, urb->dev, urb->ep, true, __func__) <= 0)
 		return -EINVAL;
@@ -824,6 +825,30 @@ int xhci_urb_enqueue(struct usb_hcd *hcd, struct urb *urb, gfp_t mem_flags)
 		ret = -ESHUTDOWN;
 		goto exit;
 	}
+
+	if (usb_endpoint_xfer_isoc(&urb->ep->desc))
+		size = urb->number_of_packets;
+	else
+		size = 1;
+
+	urb_priv = kzalloc(sizeof(struct urb_priv) +
+				  size * sizeof(struct xhci_td *), mem_flags);
+	if (!urb_priv)
+		return -ENOMEM;
+
+	for (i = 0; i < size; i++) {
+		urb_priv->td[i] = kzalloc(sizeof(struct xhci_td), mem_flags);
+		if (!urb_priv->td[i]) {
+			urb_priv->length = i;
+			xhci_urb_free_priv(xhci, urb_priv);
+			return -ENOMEM;
+		}
+	}
+
+	urb_priv->length = size;
+	urb_priv->td_cnt = 0;
+	urb->hcpriv = urb_priv;
+
 	if (usb_endpoint_xfer_control(&urb->ep->desc)) {
 		/* Check to see if the max packet size for the default control
 		 * endpoint changed during FS device enumeration
@@ -877,6 +902,8 @@ int xhci_urb_enqueue(struct usb_hcd *hcd, struct urb *urb, gfp_t mem_flags)
 exit:
 	return ret;
 dying:
+	xhci_urb_free_priv(xhci, urb_priv);
+	urb->hcpriv = NULL;
 	xhci_dbg(xhci, "Ep 0x%x: URB %p submitted for "
 			"non-responsive xHCI host.\n",
 			urb->ep->desc.bEndpointAddress, urb);
@@ -918,9 +945,10 @@ int xhci_urb_enqueue(struct usb_hcd *hcd, struct urb *urb, gfp_t mem_flags)
 int xhci_urb_dequeue(struct usb_hcd *hcd, struct urb *urb, int status)
 {
 	unsigned long flags;
-	int ret;
+	int ret, i;
 	u32 temp;
 	struct xhci_hcd *xhci;
+	struct urb_priv	*urb_priv;
 	struct xhci_td *td;
 	unsigned int ep_index;
 	struct xhci_ring *ep_ring;
@@ -935,12 +963,12 @@ int xhci_urb_dequeue(struct usb_hcd *hcd, struct urb *urb, int status)
 	temp = xhci_readl(xhci, &xhci->op_regs->status);
 	if (temp == 0xffffffff) {
 		xhci_dbg(xhci, "HW died, freeing TD.\n");
-		td = (struct xhci_td *) urb->hcpriv;
+		urb_priv = urb->hcpriv;
 
 		usb_hcd_unlink_urb_from_ep(hcd, urb);
 		spin_unlock_irqrestore(&xhci->lock, flags);
 		usb_hcd_giveback_urb(xhci_to_hcd(xhci), urb, -ESHUTDOWN);
-		kfree(td);
+		xhci_urb_free_priv(xhci, urb_priv);
 		return ret;
 	}
 	if (xhci->xhc_state & XHCI_STATE_DYING) {
@@ -968,9 +996,14 @@ int xhci_urb_dequeue(struct usb_hcd *hcd, struct urb *urb, int status)
 
 	xhci_dbg(xhci, "Endpoint ring:\n");
 	xhci_debug_ring(xhci, ep_ring);
-	td = (struct xhci_td *) urb->hcpriv;
 
-	list_add_tail(&td->cancelled_td_list, &ep->cancelled_td_list);
+	urb_priv = urb->hcpriv;
+
+	for (i = urb_priv->td_cnt; i < urb_priv->length; i++) {
+		td = urb_priv->td[i];
+		list_add_tail(&td->cancelled_td_list, &ep->cancelled_td_list);
+	}
+
 	/* Queue a stop endpoint command, but only if this is
 	 * the first cancellation to be handled.
 	 */
diff --git a/drivers/usb/host/xhci.h b/drivers/usb/host/xhci.h
index f4dfb26a65aaea2ec52b3e3a695a4bbf5d4db93f..ebf62082950bb24dd862d507a5c1ea594d43ce5e 100644
--- a/drivers/usb/host/xhci.h
+++ b/drivers/usb/host/xhci.h
@@ -1090,6 +1090,12 @@ struct xhci_scratchpad {
 	dma_addr_t *sp_dma_buffers;
 };
 
+struct urb_priv {
+	int	length;
+	int	td_cnt;
+	struct	xhci_td	*td[0];
+};
+
 /*
  * Each segment table entry is 4*32bits long.  1K seems like an ok size:
  * (1K bytes * 8bytes/bit) / (4*32 bits) = 64 segment entries in the table,
@@ -1347,6 +1353,7 @@ struct xhci_ring *xhci_stream_id_to_ring(
 struct xhci_command *xhci_alloc_command(struct xhci_hcd *xhci,
 		bool allocate_in_ctx, bool allocate_completion,
 		gfp_t mem_flags);
+void xhci_urb_free_priv(struct xhci_hcd *xhci, struct urb_priv *urb_priv);
 void xhci_free_command(struct xhci_hcd *xhci,
 		struct xhci_command *command);