Hi,
I think I see what is going on there. I believe this commit is a
regression after r365239, that the XHCI driver only resets the data
toggle when a STALL PID or USB error is received! But that is not the
case for the BULK OUT endpoint, in your traces, because USB mass storage
has two endpoints.
https://svnweb.freebsd.org/changeset/base/365239
Can you try the attached patch instead of reverting that commit and see
if the XHCI now sends a DATA0, as expected?
Bjoern and Kristof, can you also test this patch?
--HPS
diff --git a/sys/dev/usb/controller/xhci.c b/sys/dev/usb/controller/xhci.c
index f95996b7ab3..eb88140cf55 100644
--- a/sys/dev/usb/controller/xhci.c
+++ b/sys/dev/usb/controller/xhci.c
@@ -3825,6 +3825,13 @@ xhci_configure_reset_endpoint(struct usb_xfer *xfer)
err = xhci_cmd_stop_ep(sc, 0, epno, index);
if (err != 0)
DPRINTF("Could not stop endpoint %u\n", epno);
+ /*
+ * Need to reset the data toggle, because stop
+ * endpoint doesn't do that:
+ */
+ err = xhci_cmd_reset_ep(sc, 0, epno, index);
+ if (err != 0)
+ DPRINTF("Could not reset endpoint %u\n", epno);
break;
}
