For the MOST packet channel there are two dedicated USB endpoints. But
internally the hardware has actually one channel for data forwarding from
and to MOST. To have the hardware clean up its state machine correctly in
case of an error, both USB pipes need to be reset. This patch triggers the
host to also clear the OUT endpoint's halt condition in case an IN endpoint
has signaled to be stalled.
Signed-off-by: Christian Gromm <christian.gromm@xxxxxxxxxxxxx>
---
drivers/staging/most/usb/usb.c | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/drivers/staging/most/usb/usb.c b/drivers/staging/most/usb/usb.c
index acd0a36..49f5495 100644
--- a/drivers/staging/most/usb/usb.c
+++ b/drivers/staging/most/usb/usb.c
@@ -815,6 +815,21 @@ static void wq_clear_halt(struct work_struct *wq_obj)
if (usb_clear_halt(mdev->usb_device, pipe))
dev_warn(&mdev->usb_device->dev, "Failed to reset endpoint.\n");
+ /* If the functional Stall condition has been set on an
+ * asynchronous rx channel, we need to clear the tx channel
+ * too, since the hardware runs its clean-up sequence on both
+ * channels, as they are physically one on the network.
+ *
+ * The USB interface that exposes the asynchronous channels
+ * contains always two endpoints, and two only.
+ */
+ if (mdev->conf[channel].data_type == MOST_CH_ASYNC &&
+ mdev->conf[channel].direction == MOST_CH_RX) {
+ int peer = 1 - channel;
+ int snd_pipe = usb_sndbulkpipe(mdev->usb_device,
+ mdev->ep_address[peer]);
+ usb_clear_halt(mdev->usb_device, snd_pipe);
+ }
mdev->is_channel_healthy[channel] = true;
most_resume_enqueue(&mdev->iface, channel);
mutex_unlock(&mdev->io_mutex);
--
2.7.4
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
