Commit a6835090 authored Jan 26, 2016 by Mathias Nyman Committed by Greg Kroah-Hartman Feb 03, 2016

Revert "xhci: don't finish a TD if we get a short-transfer event mid TD"



This reverts commit e210c422 ("xhci: don't finish a TD if we get a
short transfer event mid TD")

Turns out that most host controllers do not follow the xHCI specs and never
send the second event for the last TRB in the TD if there was a short event
mid-TD.

Returning the URB directly after the first short-transfer event is far
better than never returning the URB. (class drivers usually timeout
after 30sec). For the hosts that do send the second event we will go
back to treating it as misplaced event and print an error message for it.

The origial patch was sent to stable kernels and needs to be reverted from
there as well

Cc: stable@vger.kernel.org
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parent 36f90b0a

drivers/usb/host/xhci-ring.c

+0 −10

Original line number	Diff line number	Diff line
		@@ -2193,10 +2193,6 @@ static int process_bulk_intr_td(struct xhci_hcd xhci, struct xhci_td td,
		}
		/* Fast path - was this the last TRB in the TD for this URB? */
		} else if (event_trb == td->last_trb) {
		if (td->urb_length_set && trb_comp_code == COMP_SHORT_TX)
		return finish_td(xhci, td, event_trb, event, ep,
		status, false);

		if (EVENT_TRB_LEN(le32_to_cpu(event->transfer_len)) != 0) {
		td->urb->actual_length =
		td->urb->transfer_buffer_length -
		@@ -2248,12 +2244,6 @@ static int process_bulk_intr_td(struct xhci_hcd xhci, struct xhci_td td,
		td->urb->actual_length +=
		TRB_LEN(le32_to_cpu(cur_trb->generic.field[2])) -
		EVENT_TRB_LEN(le32_to_cpu(event->transfer_len));

		if (trb_comp_code == COMP_SHORT_TX) {
		xhci_dbg(xhci, "mid bulk/intr SP, wait for last TRB event\n");
		td->urb_length_set = true;
		return 0;
		}
		}

		return finish_td(xhci, td, event_trb, event, ep, status, false);

Admin message