[PATCH v2] usbip: Don't submit special requests twice

[PATCH v2] usbip: Don't submit special requests twice

[Simon Holesch]

Skip submitting URBs, when identical requests were already sent in
tweak_special_requests(). Instead call the completion handler directly
to return the result of the URB.

Even though submitting those requests twice should be harmless, there
are USB devices that react poorly to some duplicated requests.

One example is the ChipIdea controller implementation in U-Boot: The
second SET_CONFIURATION request makes U-Boot disable and re-enable all
endpoints. Re-enabling an endpoint in the ChipIdea controller, however,
was broken until U-Boot commit b272c8792502 ("usb: ci: Fix gadget
reinit").

Signed-off-by: Simon Holesch <simon@holesch.de>
---

Changes in v2:
- explain change in commit message

Thanks for the feedback!

 drivers/usb/usbip/stub_rx.c | 50 ++++++++++++++++++++++++-------------
 1 file changed, 32 insertions(+), 18 deletions(-)

diff --git a/drivers/usb/usbip/stub_rx.c b/drivers/usb/usbip/stub_rx.c
index fc01b31bbb87..f3fcc290ce5b 100644
--- a/drivers/usb/usbip/stub_rx.c
+++ b/drivers/usb/usbip/stub_rx.c
@@ -167,13 +167,13 @@ static int tweak_reset_device_cmd(struct urb *urb)
 /*
  * clear_halt, set_interface, and set_configuration require special tricks.
  */
-static void tweak_special_requests(struct urb *urb)
+static int tweak_special_requests(struct urb *urb)
 {
 	if (!urb || !urb->setup_packet)
-		return;
+		return 0;
 
 	if (usb_pipetype(urb->pipe) != PIPE_CONTROL)
-		return;
+		return 0;
 
 	if (is_clear_halt_cmd(urb))
 		/* tweak clear_halt */
@@ -189,8 +189,12 @@ static void tweak_special_requests(struct urb *urb)
 
 	else if (is_reset_device_cmd(urb))
 		tweak_reset_device_cmd(urb);
-	else
+	else {
 		usbip_dbg_stub_rx("no need to tweak\n");
+		return 0;
+	}
+
+	return 1;
 }
 
 /*
@@ -468,6 +472,7 @@ static void stub_recv_cmd_submit(struct stub_device *sdev,
 	int support_sg = 1;
 	int np = 0;
 	int ret, i;
+	int is_tweaked;
 
 	if (pipe == -1)
 		return;
@@ -580,8 +585,7 @@ static void stub_recv_cmd_submit(struct stub_device *sdev,
 		priv->urbs[i]->pipe = pipe;
 		priv->urbs[i]->complete = stub_complete;
 
-		/* no need to submit an intercepted request, but harmless? */
-		tweak_special_requests(priv->urbs[i]);
+		is_tweaked = tweak_special_requests(priv->urbs[i]);
 
 		masking_bogus_flags(priv->urbs[i]);
 	}
@@ -594,22 +598,32 @@ static void stub_recv_cmd_submit(struct stub_device *sdev,
 
 	/* urb is now ready to submit */
 	for (i = 0; i < priv->num_urbs; i++) {
-		ret = usb_submit_urb(priv->urbs[i], GFP_KERNEL);
+		if (!is_tweaked) {
+			ret = usb_submit_urb(priv->urbs[i], GFP_KERNEL);
 
-		if (ret == 0)
-			usbip_dbg_stub_rx("submit urb ok, seqnum %u\n",
-					pdu->base.seqnum);
-		else {
-			dev_err(&udev->dev, "submit_urb error, %d\n", ret);
-			usbip_dump_header(pdu);
-			usbip_dump_urb(priv->urbs[i]);
+			if (ret == 0)
+				usbip_dbg_stub_rx("submit urb ok, seqnum %u\n",
+						pdu->base.seqnum);
+			else {
+				dev_err(&udev->dev, "submit_urb error, %d\n", ret);
+				usbip_dump_header(pdu);
+				usbip_dump_urb(priv->urbs[i]);
 
+				/*
+				 * Pessimistic.
+				 * This connection will be discarded.
+				 */
+				usbip_event_add(ud, SDEV_EVENT_ERROR_SUBMIT);
+				break;
+			}
+		} else {
 			/*
-			 * Pessimistic.
-			 * This connection will be discarded.
+			 * An identical URB was already submitted in
+			 * tweak_special_requests(). Skip submitting this URB to not
+			 * duplicate the request.
 			 */
-			usbip_event_add(ud, SDEV_EVENT_ERROR_SUBMIT);
-			break;
+			priv->urbs[i]->status = 0;
+			stub_complete(priv->urbs[i]);
 		}
 	}
 
-- 
2.43.0

Re: [PATCH v2] usbip: Don't submit special requests twice

[Shuah Khan]

On 11/27/23 17:08, Simon Holesch wrote:
> Skip submitting URBs, when identical requests were already sent in
> tweak_special_requests(). Instead call the completion handler directly
> to return the result of the URB.
> 

Sounds reasonable. What happens when a tweak_* routine does nothing
or returns an error.

tweak_reset_device_cmd() returns 0 when it can't get the device lock
and after calling usb_reset_device().

tweak_set_interface_cmd() and tweak_clear_halt_cmd() return the value
the usb_* routine they call. Don't these cases be handled as well?

thanks,
-- Shuah