USB: EHCI: split ehci_qh into hw and sw parts

The ehci_qh structure merged hw and sw together which is not good:
1. More and more items are being added into ehci_qh, the ehci_qh software
   part are unnecessary to be allocated in DMA qh_pool.
2. If HCD has local SRAM, the sw part will consume it too, and it won't
   bring any benefit.
3. For non-cache-coherence system, the entire ehci_qh is uncachable, actually
   we only need the hw part to be uncacheable. Spliting them will let the sw
   part to be cacheable.

Signed-off-by: Alek Du <alek.du@intel.com>
Cc: David Brownell <dbrownell@users.sourceforge.net>
CC: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

6 files changed, 127 insertions, 81 deletions

diff --git a/drivers/usb/host/ehci-dbg.c b/drivers/usb/host/ehci-dbg.c
index 1104caa0803..874d2000bf9 100644
--- a/drivers/usb/host/ehci-dbg.c
+++ b/drivers/usb/host/ehci-dbg.c
@@ -134,10 +134,11 @@ dbg_qtd (const char *label, struct ehci_hcd *ehci, struct ehci_qtd *qtd)
 static void __maybe_unused
 dbg_qh (const char *label, struct ehci_hcd *ehci, struct ehci_qh *qh)
 {
+	struct ehci_qh_hw *hw = qh->hw;
+
 	ehci_dbg (ehci, "%s qh %p n%08x info %x %x qtd %x\n", label,
-		qh, qh->hw_next, qh->hw_info1, qh->hw_info2,
-		qh->hw_current);
-	dbg_qtd ("overlay", ehci, (struct ehci_qtd *) &qh->hw_qtd_next);
+		qh, hw->hw_next, hw->hw_info1, hw->hw_info2, hw->hw_current);
+	dbg_qtd("overlay", ehci, (struct ehci_qtd *) &hw->hw_qtd_next);
 }
 
 static void __maybe_unused
@@ -400,31 +401,32 @@ static void qh_lines (
 	char			*next = *nextp;
 	char			mark;
 	__le32			list_end = EHCI_LIST_END(ehci);
+	struct ehci_qh_hw	*hw = qh->hw;
 
-	if (qh->hw_qtd_next == list_end)	/* NEC does this */
+	if (hw->hw_qtd_next == list_end)	/* NEC does this */
 		mark = '@';
 	else
-		mark = token_mark(ehci, qh->hw_token);
+		mark = token_mark(ehci, hw->hw_token);
 	if (mark == '/') {	/* qh_alt_next controls qh advance? */
-		if ((qh->hw_alt_next & QTD_MASK(ehci))
-				== ehci->async->hw_alt_next)
+		if ((hw->hw_alt_next & QTD_MASK(ehci))
+				== ehci->async->hw->hw_alt_next)
 			mark = '#';	/* blocked */
-		else if (qh->hw_alt_next == list_end)
+		else if (hw->hw_alt_next == list_end)
 			mark = '.';	/* use hw_qtd_next */
 		/* else alt_next points to some other qtd */
 	}
-	scratch = hc32_to_cpup(ehci, &qh->hw_info1);
-	hw_curr = (mark == '*') ? hc32_to_cpup(ehci, &qh->hw_current) : 0;
+	scratch = hc32_to_cpup(ehci, &hw->hw_info1);
+	hw_curr = (mark == '*') ? hc32_to_cpup(ehci, &hw->hw_current) : 0;
 	temp = scnprintf (next, size,
 			"qh/%p dev%d %cs ep%d %08x %08x (%08x%c %s nak%d)",
 			qh, scratch & 0x007f,
 			speed_char (scratch),
 			(scratch >> 8) & 0x000f,
-			scratch, hc32_to_cpup(ehci, &qh->hw_info2),
-			hc32_to_cpup(ehci, &qh->hw_token), mark,
-			(cpu_to_hc32(ehci, QTD_TOGGLE) & qh->hw_token)
+			scratch, hc32_to_cpup(ehci, &hw->hw_info2),
+			hc32_to_cpup(ehci, &hw->hw_token), mark,
+			(cpu_to_hc32(ehci, QTD_TOGGLE) & hw->hw_token)
 				? "data1" : "data0",
-			(hc32_to_cpup(ehci, &qh->hw_alt_next) >> 1) & 0x0f);
+			(hc32_to_cpup(ehci, &hw->hw_alt_next) >> 1) & 0x0f);
 	size -= temp;
 	next += temp;
 
@@ -435,10 +437,10 @@ static void qh_lines (
 		mark = ' ';
 		if (hw_curr == td->qtd_dma)
 			mark = '*';
-		else if (qh->hw_qtd_next == cpu_to_hc32(ehci, td->qtd_dma))
+		else if (hw->hw_qtd_next == cpu_to_hc32(ehci, td->qtd_dma))
 			mark = '+';
 		else if (QTD_LENGTH (scratch)) {
-			if (td->hw_alt_next == ehci->async->hw_alt_next)
+			if (td->hw_alt_next == ehci->async->hw->hw_alt_next)
 				mark = '#';
 			else if (td->hw_alt_next != list_end)
 				mark = '/';
@@ -550,12 +552,15 @@ static ssize_t fill_periodic_buffer(struct debug_buffer *buf)
 		next += temp;
 
 		do {
+			struct ehci_qh_hw *hw;
+
 			switch (hc32_to_cpu(ehci, tag)) {
 			case Q_TYPE_QH:
+				hw = p.qh->hw;
 				temp = scnprintf (next, size, " qh%d-%04x/%p",
 						p.qh->period,
 						hc32_to_cpup(ehci,
-								&p.qh->hw_info2)
+							&hw->hw_info2)
 							/* uframe masks */
 							& (QH_CMASK | QH_SMASK),
 						p.qh);
@@ -576,7 +581,7 @@ static ssize_t fill_periodic_buffer(struct debug_buffer *buf)
 				/* show more info the first time around */
 				if (temp == seen_count) {
 					u32	scratch = hc32_to_cpup(ehci,
-							&p.qh->hw_info1);
+							&hw->hw_info1);
 					struct ehci_qtd	*qtd;
 					char		*type = "";
 
@@ -609,7 +614,7 @@ static ssize_t fill_periodic_buffer(struct debug_buffer *buf)
 				} else
 					temp = 0;
 				if (p.qh) {
-					tag = Q_NEXT_TYPE(ehci, p.qh->hw_next);
+					tag = Q_NEXT_TYPE(ehci, hw->hw_next);
 					p = p.qh->qh_next;
 				}
 				break;
diff --git a/drivers/usb/host/ehci-hcd.c b/drivers/usb/host/ehci-hcd.c
index 2dc15f3ad14..7bee1638dfd 100644
--- a/drivers/usb/host/ehci-hcd.c
+++ b/drivers/usb/host/ehci-hcd.c
@@ -507,6 +507,7 @@ static int ehci_init(struct usb_hcd *hcd)
 	u32			temp;
 	int			retval;
 	u32			hcc_params;
+	struct ehci_qh_hw	*hw;
 
 	spin_lock_init(&ehci->lock);
 
@@ -550,12 +551,13 @@ static int ehci_init(struct usb_hcd *hcd)
 	 * from automatically advancing to the next td after short reads.
 	 */
 	ehci->async->qh_next.qh = NULL;
-	ehci->async->hw_next = QH_NEXT(ehci, ehci->async->qh_dma);
-	ehci->async->hw_info1 = cpu_to_hc32(ehci, QH_HEAD);
-	ehci->async->hw_token = cpu_to_hc32(ehci, QTD_STS_HALT);
-	ehci->async->hw_qtd_next = EHCI_LIST_END(ehci);
+	hw = ehci->async->hw;
+	hw->hw_next = QH_NEXT(ehci, ehci->async->qh_dma);
+	hw->hw_info1 = cpu_to_hc32(ehci, QH_HEAD);
+	hw->hw_token = cpu_to_hc32(ehci, QTD_STS_HALT);
+	hw->hw_qtd_next = EHCI_LIST_END(ehci);
 	ehci->async->qh_state = QH_STATE_LINKED;
-	ehci->async->hw_alt_next = QTD_NEXT(ehci, ehci->async->dummy->qtd_dma);
+	hw->hw_alt_next = QTD_NEXT(ehci, ehci->async->dummy->qtd_dma);
 
 	/* clear interrupt enables, set irq latency */
 	if (log2_irq_thresh < 0 || log2_irq_thresh > 6)
@@ -985,7 +987,7 @@ rescan:
 	/* endpoints can be iso streams.  for now, we don't
 	 * accelerate iso completions ... so spin a while.
 	 */
-	if (qh->hw_info1 == 0) {
+	if (qh->hw->hw_info1 == 0) {
 		ehci_vdbg (ehci, "iso delay\n");
 		goto idle_timeout;
 	}
diff --git a/drivers/usb/host/ehci-mem.c b/drivers/usb/host/ehci-mem.c
index 10d52919abb..aeda96e0af6 100644
--- a/drivers/usb/host/ehci-mem.c
+++ b/drivers/usb/host/ehci-mem.c
@@ -75,7 +75,8 @@ static void qh_destroy(struct ehci_qh *qh)
 	}
 	if (qh->dummy)
 		ehci_qtd_free (ehci, qh->dummy);
-	dma_pool_free (ehci->qh_pool, qh, qh->qh_dma);
+	dma_pool_free(ehci->qh_pool, qh->hw, qh->qh_dma);
+	kfree(qh);
 }
 
 static struct ehci_qh *ehci_qh_alloc (struct ehci_hcd *ehci, gfp_t flags)
@@ -83,12 +84,14 @@ static struct ehci_qh *ehci_qh_alloc (struct ehci_hcd *ehci, gfp_t flags)
 	struct ehci_qh		*qh;
 	dma_addr_t		dma;
 
-	qh = (struct ehci_qh *)
-		dma_pool_alloc (ehci->qh_pool, flags, &dma);
+	qh = kzalloc(sizeof *qh, GFP_ATOMIC);
 	if (!qh)
-		return qh;
-
-	memset (qh, 0, sizeof *qh);
+		goto done;
+	qh->hw = (struct ehci_qh_hw *)
+		dma_pool_alloc(ehci->qh_pool, flags, &dma);
+	if (!qh->hw)
+		goto fail;
+	memset(qh->hw, 0, sizeof *qh->hw);
 	qh->refcount = 1;
 	qh->ehci = ehci;
 	qh->qh_dma = dma;
@@ -99,10 +102,15 @@ static struct ehci_qh *ehci_qh_alloc (struct ehci_hcd *ehci, gfp_t flags)
 	qh->dummy = ehci_qtd_alloc (ehci, flags);
 	if (qh->dummy == NULL) {
 		ehci_dbg (ehci, "no dummy td\n");
-		dma_pool_free (ehci->qh_pool, qh, qh->qh_dma);
-		qh = NULL;
+		goto fail1;
 	}
+done:
 	return qh;
+fail1:
+	dma_pool_free(ehci->qh_pool, qh->hw, qh->qh_dma);
+fail:
+	kfree(qh);
+	return NULL;
 }
 
 /* to share a qh (cpu threads, or hc) */
@@ -180,7 +188,7 @@ static int ehci_mem_init (struct ehci_hcd *ehci, gfp_t flags)
 	/* QHs for control/bulk/intr transfers */
 	ehci->qh_pool = dma_pool_create ("ehci_qh",
 			ehci_to_hcd(ehci)->self.controller,
-			sizeof (struct ehci_qh),
+			sizeof(struct ehci_qh_hw),
 			32 /* byte alignment (for hw parts) */,
 			4096 /* can't cross 4K */);
 	if (!ehci->qh_pool) {
diff --git a/drivers/usb/host/ehci-q.c b/drivers/usb/host/ehci-q.c
index 7673554fa64..377ed530b92 100644
--- a/drivers/usb/host/ehci-q.c
+++ b/drivers/usb/host/ehci-q.c
@@ -87,31 +87,33 @@ qtd_fill(struct ehci_hcd *ehci, struct ehci_qtd *qtd, dma_addr_t buf,
 static inline void
 qh_update (struct ehci_hcd *ehci, struct ehci_qh *qh, struct ehci_qtd *qtd)
 {
+	struct ehci_qh_hw *hw = qh->hw;
+
 	/* writes to an active overlay are unsafe */
 	BUG_ON(qh->qh_state != QH_STATE_IDLE);
 
-	qh->hw_qtd_next = QTD_NEXT(ehci, qtd->qtd_dma);
-	qh->hw_alt_next = EHCI_LIST_END(ehci);
+	hw->hw_qtd_next = QTD_NEXT(ehci, qtd->qtd_dma);
+	hw->hw_alt_next = EHCI_LIST_END(ehci);
 
 	/* Except for control endpoints, we make hardware maintain data
 	 * toggle (like OHCI) ... here (re)initialize the toggle in the QH,
 	 * and set the pseudo-toggle in udev. Only usb_clear_halt() will
 	 * ever clear it.
 	 */
-	if (!(qh->hw_info1 & cpu_to_hc32(ehci, 1 << 14))) {
+	if (!(hw->hw_info1 & cpu_to_hc32(ehci, 1 << 14))) {
 		unsigned	is_out, epnum;
 
 		is_out = !(qtd->hw_token & cpu_to_hc32(ehci, 1 << 8));
-		epnum = (hc32_to_cpup(ehci, &qh->hw_info1) >> 8) & 0x0f;
+		epnum = (hc32_to_cpup(ehci, &hw->hw_info1) >> 8) & 0x0f;
 		if (unlikely (!usb_gettoggle (qh->dev, epnum, is_out))) {
-			qh->hw_token &= ~cpu_to_hc32(ehci, QTD_TOGGLE);
+			hw->hw_token &= ~cpu_to_hc32(ehci, QTD_TOGGLE);
 			usb_settoggle (qh->dev, epnum, is_out, 1);
 		}
 	}
 
 	/* HC must see latest qtd and qh data before we clear ACTIVE+HALT */
 	wmb ();
-	qh->hw_token &= cpu_to_hc32(ehci, QTD_TOGGLE | QTD_STS_PING);
+	hw->hw_token &= cpu_to_hc32(ehci, QTD_TOGGLE | QTD_STS_PING);
 }
 
 /* if it weren't for a common silicon quirk (writing the dummy into the qh
@@ -129,7 +131,7 @@ qh_refresh (struct ehci_hcd *ehci, struct ehci_qh *qh)
 		qtd = list_entry (qh->qtd_list.next,
 				struct ehci_qtd, qtd_list);
 		/* first qtd may already be partially processed */
-		if (cpu_to_hc32(ehci, qtd->qtd_dma) == qh->hw_current)
+		if (cpu_to_hc32(ehci, qtd->qtd_dma) == qh->hw->hw_current)
 			qtd = NULL;
 	}
 
@@ -260,7 +262,7 @@ __acquires(ehci->lock)
 		struct ehci_qh	*qh = (struct ehci_qh *) urb->hcpriv;
 
 		/* S-mask in a QH means it's an interrupt urb */
-		if ((qh->hw_info2 & cpu_to_hc32(ehci, QH_SMASK)) != 0) {
+		if ((qh->hw->hw_info2 & cpu_to_hc32(ehci, QH_SMASK)) != 0) {
 
 			/* ... update hc-wide periodic stats (for usbfs) */
 			ehci_to_hcd(ehci)->self.bandwidth_int_reqs--;
@@ -315,6 +317,7 @@ qh_completions (struct ehci_hcd *ehci, struct ehci_qh *qh)
 	unsigned		count = 0;
 	u8			state;
 	__le32			halt = HALT_BIT(ehci);
+	struct ehci_qh_hw	*hw = qh->hw;
 
 	if (unlikely (list_empty (&qh->qtd_list)))
 		return count;
@@ -392,7 +395,8 @@ qh_completions (struct ehci_hcd *ehci, struct ehci_qh *qh)
 					qtd->hw_token = cpu_to_hc32(ehci,
 							token);
 					wmb();
-					qh->hw_token = cpu_to_hc32(ehci, token);
+					hw->hw_token = cpu_to_hc32(ehci,
+							token);
 					goto retry_xacterr;
 				}
 				stopped = 1;
@@ -435,8 +439,8 @@ qh_completions (struct ehci_hcd *ehci, struct ehci_qh *qh)
 			/* qh unlinked; token in overlay may be most current */
 			if (state == QH_STATE_IDLE
 					&& cpu_to_hc32(ehci, qtd->qtd_dma)
-						== qh->hw_current) {
-				token = hc32_to_cpu(ehci, qh->hw_token);
+						== hw->hw_current) {
+				token = hc32_to_cpu(ehci, hw->hw_token);
 
 				/* An unlink may leave an incomplete
 				 * async transaction in the TT buffer.
@@ -449,9 +453,9 @@ qh_completions (struct ehci_hcd *ehci, struct ehci_qh *qh)
 			 * patch the qh later and so that completions can't
 			 * activate it while we "know" it's stopped.
 			 */
-			if ((halt & qh->hw_token) == 0) {
+			if ((halt & hw->hw_token) == 0) {
 halt:
-				qh->hw_token |= halt;
+				hw->hw_token |= halt;
 				wmb ();
 			}
 		}
@@ -510,7 +514,7 @@ halt:
 	 * it after fault cleanup, or recovering from silicon wrongly
 	 * overlaying the dummy qtd (which reduces DMA chatter).
 	 */
-	if (stopped != 0 || qh->hw_qtd_next == EHCI_LIST_END(ehci)) {
+	if (stopped != 0 || hw->hw_qtd_next == EHCI_LIST_END(ehci)) {
 		switch (state) {
 		case QH_STATE_IDLE:
 			qh_refresh(ehci, qh);
@@ -528,7 +532,7 @@ halt:
 			 * except maybe high bandwidth ...
 			 */
 			if ((cpu_to_hc32(ehci, QH_SMASK)
-					& qh->hw_info2) != 0) {
+					& hw->hw_info2) != 0) {
 				intr_deschedule (ehci, qh);
 				(void) qh_schedule (ehci, qh);
 			} else
@@ -649,7 +653,7 @@ qh_urb_transaction (
 		 * (this will usually be overridden later.)
 		 */
 		if (is_input)
-			qtd->hw_alt_next = ehci->async->hw_alt_next;
+			qtd->hw_alt_next = ehci->async->hw->hw_alt_next;
 
 		/* qh makes control packets use qtd toggle; maybe switch it */
 		if ((maxpacket & (this_qtd_len + (maxpacket - 1))) == 0)
@@ -744,6 +748,7 @@ qh_make (
 	int			is_input, type;
 	int			maxp = 0;
 	struct usb_tt		*tt = urb->dev->tt;
+	struct ehci_qh_hw	*hw;
 
 	if (!qh)
 		return qh;
@@ -890,8 +895,9 @@ done:
 
 	/* init as live, toggle clear, advance to dummy */
 	qh->qh_state = QH_STATE_IDLE;
-	qh->hw_info1 = cpu_to_hc32(ehci, info1);
-	qh->hw_info2 = cpu_to_hc32(ehci, info2);
+	hw = qh->hw;
+	hw->hw_info1 = cpu_to_hc32(ehci, info1);
+	hw->hw_info2 = cpu_to_hc32(ehci, info2);
 	usb_settoggle (urb->dev, usb_pipeendpoint (urb->pipe), !is_input, 1);
 	qh_refresh (ehci, qh);
 	return qh;
@@ -933,11 +939,11 @@ static void qh_link_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
 
 	/* splice right after start */
 	qh->qh_next = head->qh_next;
-	qh->hw_next = head->hw_next;
+	qh->hw->hw_next = head->hw->hw_next;
 	wmb ();
 
 	head->qh_next.qh = qh;
-	head->hw_next = dma;
+	head->hw->hw_next = dma;
 
 	qh_get(qh);
 	qh->xacterrs = 0;
@@ -984,7 +990,7 @@ static struct ehci_qh *qh_append_tds (
 
                         /* usb_reset_device() briefly reverts to address 0 */
                         if (usb_pipedevice (urb->pipe) == 0)
-                                qh->hw_info1 &= ~qh_addr_mask;
+				qh->hw->hw_info1 &= ~qh_addr_mask;
 		}
 
 		/* just one way to queue requests: swap with the dummy qtd.
@@ -1169,7 +1175,7 @@ static void start_unlink_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
 	while (prev->qh_next.qh != qh)
 		prev = prev->qh_next.qh;
 
-	prev->hw_next = qh->hw_next;
+	prev->hw->hw_next = qh->hw->hw_next;
 	prev->qh_next = qh->qh_next;
 	wmb ();
 
diff --git a/drivers/usb/host/ehci-sched.c b/drivers/usb/host/ehci-sched.c
index edd61ee9032..327437af212 100644
--- a/drivers/usb/host/ehci-sched.c
+++ b/drivers/usb/host/ehci-sched.c
@@ -60,6 +60,20 @@ periodic_next_shadow(struct ehci_hcd *ehci, union ehci_shadow *periodic,
 	}
 }
 
+static __hc32 *
+shadow_next_periodic(struct ehci_hcd *ehci, union ehci_shadow *periodic,
+		__hc32 tag)
+{
+	switch (hc32_to_cpu(ehci, tag)) {
+	/* our ehci_shadow.qh is actually software part */
+	case Q_TYPE_QH:
+		return &periodic->qh->hw->hw_next;
+	/* others are hw parts */
+	default:
+		return periodic->hw_next;
+	}
+}
+
 /* caller must hold ehci->lock */
 static void periodic_unlink (struct ehci_hcd *ehci, unsigned frame, void *ptr)
 {
@@ -71,7 +85,8 @@ static void periodic_unlink (struct ehci_hcd *ehci, unsigned frame, void *ptr)
 	while (here.ptr && here.ptr != ptr) {
 		prev_p = periodic_next_shadow(ehci, prev_p,
 				Q_NEXT_TYPE(ehci, *hw_p));
-		hw_p = here.hw_next;
+		hw_p = shadow_next_periodic(ehci, &here,
+				Q_NEXT_TYPE(ehci, *hw_p));
 		here = *prev_p;
 	}
 	/* an interrupt entry (at list end) could have been shared */
@@ -83,7 +98,7 @@ static void periodic_unlink (struct ehci_hcd *ehci, unsigned frame, void *ptr)
 	 */
 	*prev_p = *periodic_next_shadow(ehci, &here,
 			Q_NEXT_TYPE(ehci, *hw_p));
-	*hw_p = *here.hw_next;
+	*hw_p = *shadow_next_periodic(ehci, &here, Q_NEXT_TYPE(ehci, *hw_p));
 }
 
 /* how many of the uframe's 125 usecs are allocated? */
@@ -93,18 +108,20 @@ periodic_usecs (struct ehci_hcd *ehci, unsigned frame, unsigned uframe)
 	__hc32			*hw_p = &ehci->periodic [frame];
 	union ehci_shadow	*q = &ehci->pshadow [frame];
 	unsigned		usecs = 0;
+	struct ehci_qh_hw	*hw;
 
 	while (q->ptr) {
 		switch (hc32_to_cpu(ehci, Q_NEXT_TYPE(ehci, *hw_p))) {
 		case Q_TYPE_QH:
+			hw = q->qh->hw;
 			/* is it in the S-mask? */
-			if (q->qh->hw_info2 & cpu_to_hc32(ehci, 1 << uframe))
+			if (hw->hw_info2 & cpu_to_hc32(ehci, 1 << uframe))
 				usecs += q->qh->usecs;
 			/* ... or C-mask? */
-			if (q->qh->hw_info2 & cpu_to_hc32(ehci,
+			if (hw->hw_info2 & cpu_to_hc32(ehci,
 					1 << (8 + uframe)))
 				usecs += q->qh->c_usecs;
-			hw_p = &q->qh->hw_next;
+			hw_p = &hw->hw_next;
 			q = &q->qh->qh_next;
 			break;
 		// case Q_TYPE_FSTN:
@@ -237,10 +254,10 @@ periodic_tt_usecs (
 			continue;
 		case Q_TYPE_QH:
 			if (same_tt(dev, q->qh->dev)) {
-				uf = tt_start_uframe(ehci, q->qh->hw_info2);
+				uf = tt_start_uframe(ehci, q->qh->hw->hw_info2);
 				tt_usecs[uf] += q->qh->tt_usecs;
 			}
-			hw_p = &q->qh->hw_next;
+			hw_p = &q->qh->hw->hw_next;
 			q = &q->qh->qh_next;
 			continue;
 		case Q_TYPE_SITD:
@@ -375,6 +392,7 @@ static int tt_no_collision (
 	for (; frame < ehci->periodic_size; frame += period) {
 		union ehci_shadow	here;
 		__hc32			type;
+		struct ehci_qh_hw	*hw;
 
 		here = ehci->pshadow [frame];
 		type = Q_NEXT_TYPE(ehci, ehci->periodic [frame]);
@@ -385,17 +403,18 @@ static int tt_no_collision (
 				here = here.itd->itd_next;
 				continue;
 			case Q_TYPE_QH:
+				hw = here.qh->hw;
 				if (same_tt (dev, here.qh->dev)) {
 					u32		mask;
 
 					mask = hc32_to_cpu(ehci,
-							here.qh->hw_info2);
+							hw->hw_info2);
 					/* "knows" no gap is needed */
 					mask |= mask >> 8;
 					if (mask & uf_mask)
 						break;
 				}
-				type = Q_NEXT_TYPE(ehci, here.qh->hw_next);
+				type = Q_NEXT_TYPE(ehci, hw->hw_next);
 				here = here.qh->qh_next;
 				continue;
 			case Q_TYPE_SITD:
@@ -498,7 +517,8 @@ static int qh_link_periodic (struct ehci_hcd *ehci, struct ehci_qh *qh)
 
 	dev_dbg (&qh->dev->dev,
 		"link qh%d-%04x/%p start %d [%d/%d us]\n",
-		period, hc32_to_cpup(ehci, &qh->hw_info2) & (QH_CMASK | QH_SMASK),
+		period, hc32_to_cpup(ehci, &qh->hw->hw_info2)
+			& (QH_CMASK | QH_SMASK),
 		qh, qh->start, qh->usecs, qh->c_usecs);
 
 	/* high bandwidth, or otherwise every microframe */
@@ -517,7 +537,7 @@ static int qh_link_periodic (struct ehci_hcd *ehci, struct ehci_qh *qh)
 			if (type == cpu_to_hc32(ehci, Q_TYPE_QH))
 				break;
 			prev = periodic_next_shadow(ehci, prev, type);
-			hw_p = &here.qh->hw_next;
+			hw_p = shadow_next_periodic(ehci, &here, type);
 			here = *prev;
 		}
 
@@ -528,14 +548,14 @@ static int qh_link_periodic (struct ehci_hcd *ehci, struct ehci_qh *qh)
 			if (qh->period > here.qh->period)
 				break;
 			prev = &here.qh->qh_next;
-			hw_p = &here.qh->hw_next;
+			hw_p = &here.qh->hw->hw_next;
 			here = *prev;
 		}
 		/* link in this qh, unless some earlier pass did that */
 		if (qh != here.qh) {
 			qh->qh_next = here;
 			if (here.qh)
-				qh->hw_next = *hw_p;
+				qh->hw->hw_next = *hw_p;
 			wmb ();
 			prev->qh = qh;
 			*hw_p = QH_NEXT (ehci, qh->qh_dma);
@@ -581,7 +601,7 @@ static int qh_unlink_periodic(struct ehci_hcd *ehci, struct ehci_qh *qh)
 	dev_dbg (&qh->dev->dev,
 		"unlink qh%d-%04x/%p start %d [%d/%d us]\n",
 		qh->period,
-		hc32_to_cpup(ehci, &qh->hw_info2) & (QH_CMASK | QH_SMASK),
+		hc32_to_cpup(ehci, &qh->hw->hw_info2) & (QH_CMASK | QH_SMASK),
 		qh, qh->start, qh->usecs, qh->c_usecs);
 
 	/* qh->qh_next still "live" to HC */
@@ -596,6 +616,7 @@ static int qh_unlink_periodic(struct ehci_hcd *ehci, struct ehci_qh *qh)
 static void intr_deschedule (struct ehci_hcd *ehci, struct ehci_qh *qh)
 {
 	unsigned	wait;
+	struct ehci_qh_hw *hw = qh->hw;
 
 	qh_unlink_periodic (ehci, qh);
 
@@ -606,14 +627,14 @@ static void intr_deschedule (struct ehci_hcd *ehci, struct ehci_qh *qh)
 	 */
 	if (list_empty (&qh->qtd_list)
 			|| (cpu_to_hc32(ehci, QH_CMASK)
-					& qh->hw_info2) != 0)
+					& hw->hw_info2) != 0)
 		wait = 2;
 	else
 		wait = 55;	/* worst case: 3 * 1024 */
 
 	udelay (wait);
 	qh->qh_state = QH_STATE_IDLE;
-	qh->hw_next = EHCI_LIST_END(ehci);
+	hw->hw_next = EHCI_LIST_END(ehci);
 	wmb ();
 }
 
@@ -739,14 +760,15 @@ static int qh_schedule(struct ehci_hcd *ehci, struct ehci_qh *qh)
 	unsigned	uframe;
 	__hc32		c_mask;
 	unsigned	frame;		/* 0..(qh->period - 1), or NO_FRAME */
+	struct ehci_qh_hw	*hw = qh->hw;
 
 	qh_refresh(ehci, qh);
-	qh->hw_next = EHCI_LIST_END(ehci);
+	hw->hw_next = EHCI_LIST_END(ehci);
 	frame = qh->start;
 
 	/* reuse the previous schedule slots, if we can */
 	if (frame < qh->period) {
-		uframe = ffs(hc32_to_cpup(ehci, &qh->hw_info2) & QH_SMASK);
+		uframe = ffs(hc32_to_cpup(ehci, &hw->hw_info2) & QH_SMASK);
 		status = check_intr_schedule (ehci, frame, --uframe,
 				qh, &c_mask);
 	} else {
@@ -784,11 +806,11 @@ static int qh_schedule(struct ehci_hcd *ehci, struct ehci_qh *qh)
 		qh->start = frame;
 
 		/* reset S-frame and (maybe) C-frame masks */
-		qh->hw_info2 &= cpu_to_hc32(ehci, ~(QH_CMASK | QH_SMASK));
-		qh->hw_info2 |= qh->period
+		hw->hw_info2 &= cpu_to_hc32(ehci, ~(QH_CMASK | QH_SMASK));
+		hw->hw_info2 |= qh->period
 			? cpu_to_hc32(ehci, 1 << uframe)
 			: cpu_to_hc32(ehci, QH_SMASK);
-		qh->hw_info2 |= c_mask;
+		hw->hw_info2 |= c_mask;
 	} else
 		ehci_dbg (ehci, "reused qh %p schedule\n", qh);
 
@@ -2188,7 +2210,7 @@ restart:
 			case Q_TYPE_QH:
 				/* handle any completions */
 				temp.qh = qh_get (q.qh);
-				type = Q_NEXT_TYPE(ehci, q.qh->hw_next);
+				type = Q_NEXT_TYPE(ehci, q.qh->hw->hw_next);
 				q = q.qh->qh_next;
 				modified = qh_completions (ehci, temp.qh);
 				if (unlikely (list_empty (&temp.qh->qtd_list)))
diff --git a/drivers/usb/host/ehci.h b/drivers/usb/host/ehci.h
index acec1089407..fa12f20fbfe 100644
--- a/drivers/usb/host/ehci.h
+++ b/drivers/usb/host/ehci.h
@@ -299,8 +299,8 @@ union ehci_shadow {
  * These appear in both the async and (for interrupt) periodic schedules.
  */
 
-struct ehci_qh {
-	/* first part defined by EHCI spec */
+/* first part defined by EHCI spec */
+struct ehci_qh_hw {
 	__hc32			hw_next;	/* see EHCI 3.6.1 */
 	__hc32			hw_info1;       /* see EHCI 3.6.2 */
 #define	QH_HEAD		0x00008000
@@ -318,7 +318,10 @@ struct ehci_qh {
 	__hc32			hw_token;
 	__hc32			hw_buf [5];
 	__hc32			hw_buf_hi [5];
+} __attribute__ ((aligned(32)));
 
+struct ehci_qh {
+	struct ehci_qh_hw	*hw;
 	/* the rest is HCD-private */
 	dma_addr_t		qh_dma;		/* address of qh */
 	union ehci_shadow	qh_next;	/* ptr to qh; or periodic */
@@ -358,7 +361,7 @@ struct ehci_qh {
 
 	struct usb_device	*dev;		/* access to TT */
 	unsigned		clearing_tt:1;	/* Clear-TT-Buf in progress */
-} __attribute__ ((aligned (32)));
+};
 
 /*-------------------------------------------------------------------------*/