[PATCH] USB ohci-hcd, don't force SLAB_ATOMIC allocations

This is a minor cleanup to let per-request memory allocations block,
when the caller allows (it provided the bitmask).  The driver used
to work that way until something like 2.4.3; an update (a few months
back) to how the "dma_addr_t" hashes to a "struct ohci_td *" lets us
simplify things again.  Another benfit:  it blocks irqs for less time
on the submit path.  (The ehci driver already acts this way.)

drivers/usb/host/ohci-hcd.c

@@ -203,28 +203,28 @@ static int ohci_urb_enqueue (
 		return -ENOMEM;
 	memset (urb_priv, 0, sizeof (urb_priv_t) + size * sizeof (struct td *));
 	
-	spin_lock_irqsave (&ohci->lock, flags);
-
-	/* don't submit to a dead HC */
-	if (ohci->disabled || ohci->sleeping) {
-		retval = -ENODEV;
-		goto fail;
-	}
-
 	/* fill the private part of the URB */
 	urb_priv->length = size;
 	urb_priv->ed = ed;	
 
-	/* allocate the TDs (updating hash chains) */
+	/* allocate the TDs (deferring hash chain updates) */
 	for (i = 0; i < size; i++) {
-		urb_priv->td [i] = td_alloc (ohci, SLAB_ATOMIC);
+		urb_priv->td [i] = td_alloc (ohci, mem_flags);
 		if (!urb_priv->td [i]) {
 			urb_priv->length = i;
-			retval = -ENOMEM;
-			goto fail;
+			urb_free_priv (ohci, urb_priv);
+			return -ENOMEM;
 		}
 	}	
 
+	spin_lock_irqsave (&ohci->lock, flags);
+
+	/* don't submit to a dead HC */
+	if (ohci->disabled || ohci->sleeping) {
+		retval = -ENODEV;
+		goto fail;
+	}
+
 	/* schedule the ed if needed */
 	if (ed->state == ED_IDLE) {
 		retval = ed_schedule (ohci, ed);

drivers/usb/host/ohci-mem.c

@@ -97,17 +97,11 @@ td_alloc (struct ohci_hcd *hc, int mem_flags)
 
 	td = pci_pool_alloc (hc->td_cache, mem_flags, &dma);
 	if (td) {
-		int	hash;
-
 		/* in case hc fetches it, make it look dead */
 		memset (td, 0, sizeof *td);
 		td->hwNextTD = cpu_to_le32 (dma);
 		td->td_dma = dma;
-
-		/* hash it for later reverse mapping */
-		hash = TD_HASH_FUNC (dma);
-		td->td_hash = hc->td_hash [hash];
-		hc->td_hash [hash] = td;
+		/* hashed in td_fill */
 	}
 	return td;
 }

drivers/usb/host/ohci-q.c

@@ -463,13 +463,14 @@ static void start_urb_unlink (struct ohci_hcd *ohci, struct ed *ed)
 /* enqueue next TD for this URB (OHCI spec 5.2.8.2) */
 
 static void
-td_fill (unsigned int info,
+td_fill (struct ohci_hcd *ohci, u32 info,
 	dma_addr_t data, int len,
 	struct urb *urb, int index)
 {
 	struct td		*td, *td_pt;
 	struct urb_priv		*urb_priv = urb->hcpriv;
 	int			is_iso = info & TD_ISO;
+	int			hash;
 
 	// ASSERT (index < urb_priv->length);
 
@@ -516,11 +517,16 @@ td_fill (unsigned int info,
 		td->hwBE = 0;
 	td->hwNextTD = cpu_to_le32 (td_pt->td_dma);
 
-	/* HC might read the TD right after we link it ... */
-	wmb ();
-
 	/* append to queue */
 	list_add_tail (&td->td_list, &td->ed->td_list);
+
+	/* hash it for later reverse mapping */
+	hash = TD_HASH_FUNC (td->td_dma);
+	td->td_hash = ohci->td_hash [hash];
+	ohci->td_hash [hash] = td;
+
+	/* HC might read the TD (or cachelines) right away ... */
+	wmb ();
 	td->ed->hwTailP = td->hwNextTD;
 }
 
@@ -578,7 +584,7 @@ static void td_submit_urb (
 			: TD_T_TOGGLE | TD_CC | TD_DP_IN;
 		/* TDs _could_ transfer up to 8K each */
 		while (data_len > 4096) {
-			td_fill (info, data, 4096, urb, cnt);
+			td_fill (ohci, info, data, 4096, urb, cnt);
 			data += 4096;
 			data_len -= 4096;
 			cnt++;
@@ -586,11 +592,11 @@ static void td_submit_urb (
 		/* maybe avoid ED halt on final TD short read */
 		if (!(urb->transfer_flags & URB_SHORT_NOT_OK))
 			info |= TD_R;
-		td_fill (info, data, data_len, urb, cnt);
+		td_fill (ohci, info, data, data_len, urb, cnt);
 		cnt++;
 		if ((urb->transfer_flags & URB_ZERO_PACKET)
 				&& cnt < urb_priv->length) {
-			td_fill (info, 0, 0, urb, cnt);
+			td_fill (ohci, info, 0, 0, urb, cnt);
 			cnt++;
 		}
 		/* maybe kickstart bulk list */
@@ -605,17 +611,17 @@ static void td_submit_urb (
 	 */
 	case PIPE_CONTROL:
 		info = TD_CC | TD_DP_SETUP | TD_T_DATA0;
-		td_fill (info, urb->setup_dma, 8, urb, cnt++);
+		td_fill (ohci, info, urb->setup_dma, 8, urb, cnt++);
 		if (data_len > 0) {
 			info = TD_CC | TD_R | TD_T_DATA1;
 			info |= is_out ? TD_DP_OUT : TD_DP_IN;
 			/* NOTE:  mishandles transfers >8K, some >4K */
-			td_fill (info, data, data_len, urb, cnt++);
+			td_fill (ohci, info, data, data_len, urb, cnt++);
 		}
 		info = is_out
 			? TD_CC | TD_DP_IN | TD_T_DATA1
 			: TD_CC | TD_DP_OUT | TD_T_DATA1;
-		td_fill (info, data, 0, urb, cnt++);
+		td_fill (ohci, info, data, 0, urb, cnt++);
 		/* maybe kickstart control list */
 		wmb ();
 		writel (OHCI_CLF, &ohci->regs->cmdstatus);
@@ -634,7 +640,7 @@ static void td_submit_urb (
 			// a 2^16 iso range, vs other HCs max of 2^10)
 			frame += cnt * urb->interval;
 			frame &= 0xffff;
-			td_fill (TD_CC | TD_ISO | frame,
+			td_fill (ohci, TD_CC | TD_ISO | frame,
 				data + urb->iso_frame_desc [cnt].offset,
 				urb->iso_frame_desc [cnt].length, urb, cnt);
 		}