Merge tag 'for-usb-linus-2013-12-02' of git://git.kernel.org/pub/scm/linux/kernel/git/sarah/xhci into usb-linus

Sarah writes:

xhci: Regression fix for 3.13.

Hi Greg,

Here's one bug fix for 3.13.

usb-net added support for bulk scatter-gather in 3.12, and it triggered a
bug in the xHCI driver.  This bug causes xHCI hosts to send an unexpected
short transfer, which will cause the USB ethernet device to stop sending
packets.

The patch is marked for the 3.12 stable kernel.  It's a long standing bug,
but the usb-net drivers are the first to trigger it.  The only other
driver that does bulk scatter-gather (usb-storage) will not trigger this
bug.

I'm not sure what the effect of the no-op TRBs will be on various xHCI
host controllers, so I would only like to be conservative and only queue
it for 3.13 and 3.12 stable.

Please queue this for 3.13.

Sarah Sharp

1 files changed, 52 insertions, 2 deletions

diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 1e2f3f49584..53c2e296467 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -2973,8 +2973,58 @@ static int prepare_ring(struct xhci_hcd *xhci, struct xhci_ring *ep_ring,
 	}
 
 	while (1) {
-		if (room_on_ring(xhci, ep_ring, num_trbs))
-			break;
+		if (room_on_ring(xhci, ep_ring, num_trbs)) {
+			union xhci_trb *trb = ep_ring->enqueue;
+			unsigned int usable = ep_ring->enq_seg->trbs +
+					TRBS_PER_SEGMENT - 1 - trb;
+			u32 nop_cmd;
+
+			/*
+			 * Section 4.11.7.1 TD Fragments states that a link
+			 * TRB must only occur at the boundary between
+			 * data bursts (eg 512 bytes for 480M).
+			 * While it is possible to split a large fragment
+			 * we don't know the size yet.
+			 * Simplest solution is to fill the trb before the
+			 * LINK with nop commands.
+			 */
+			if (num_trbs == 1 || num_trbs <= usable || usable == 0)
+				break;
+
+			if (ep_ring->type != TYPE_BULK)
+				/*
+				 * While isoc transfers might have a buffer that
+				 * crosses a 64k boundary it is unlikely.
+				 * Since we can't add NOPs without generating
+				 * gaps in the traffic just hope it never
+				 * happens at the end of the ring.
+				 * This could be fixed by writing a LINK TRB
+				 * instead of the first NOP - however the
+				 * TRB_TYPE_LINK_LE32() calls would all need
+				 * changing to check the ring length.
+				 */
+				break;
+
+			if (num_trbs >= TRBS_PER_SEGMENT) {
+				xhci_err(xhci, "Too many fragments %d, max %d\n",
+						num_trbs, TRBS_PER_SEGMENT - 1);
+				return -ENOMEM;
+			}
+
+			nop_cmd = cpu_to_le32(TRB_TYPE(TRB_TR_NOOP) |
+					ep_ring->cycle_state);
+			ep_ring->num_trbs_free -= usable;
+			do {
+				trb->generic.field[0] = 0;
+				trb->generic.field[1] = 0;
+				trb->generic.field[2] = 0;
+				trb->generic.field[3] = nop_cmd;
+				trb++;
+			} while (--usable);
+			ep_ring->enqueue = trb;
+			if (room_on_ring(xhci, ep_ring, num_trbs))
+				break;
+		}
 
 		if (ep_ring == xhci->cmd_ring) {
 			xhci_err(xhci, "Do not support expand command ring\n");