In case we unload a gadget driver while any of
os_desc_req or req are still pending, we need
to make sure to dequeue them.
By using our setup_pending and os_desc_pending
flags we achieve that in a way that doesn't
cause any regressions because we won't dequeue
a request which was already completed.
The original idea came from Li Jun's commit
f2267089ea17fa97b796b1b4247e3f8957655df3
(usb: gadget: composite: dequeue cdev->req
before free it in composite_dev_cleanup) which,
unfortunately, caused two regressions (kfree()
being called before usb_ep_dequeue() and calling
usb_ep_dequeue() when the request was already
completed). That commit also didn't take care
of os_desc_req which can fall into the same
situation so we must care for that one too.
Note that in order to make code slightly easier
to read, we introduce composite_ep_queue() which
hides details about how to fiddle with our pending
flags.
Signed-off-by: Felipe Balbi <balbi@xxxxxx>
---
drivers/usb/gadget/composite.c | 51 +++++++++++++++++++++++++++++++++++++++---
1 file changed, 48 insertions(+), 3 deletions(-)
diff --git a/drivers/usb/gadget/composite.c b/drivers/usb/gadget/composite.c
index 09602dc..e071d58 100644
--- a/drivers/usb/gadget/composite.c
+++ b/drivers/usb/gadget/composite.c
@@ -1246,10 +1246,49 @@ EXPORT_SYMBOL_GPL(usb_string_ids_n);
static void composite_setup_complete(struct usb_ep *ep, struct usb_request *req)
{
+ struct usb_composite_dev *cdev;
+
if (req->status || req->actual != req->length)
DBG((struct usb_composite_dev *) ep->driver_data,
"setup complete --> %d, %d/%d\n",
req->status, req->actual, req->length);
+
+ /*
+ * REVIST The same ep0 requests are shared with function drivers
+ * so they don't have to maintain the same ->complete() stubs.
+ *
+ * Because of that, we need to check for the validity of ->context
+ * here, even though we know we've set it to something useful.
+ */
+ if (!req->context)
+ return;
+
+ cdev = req->context;
+
+ if (cdev->req == req)
+ cdev->setup_pending = false;
+ else if (cdev->os_desc_req == req)
+ cdev->os_desc_pending = false;
+ else
+ WARN(1, "unknown request %p\n", req);
+}
+
+static int composite_ep0_queue(struct usb_composite_dev *cdev,
+ struct usb_request *req, gfp_t gfp_flags)
+{
+ int ret;
+
+ ret = usb_ep_queue(cdev->gadget->ep0, req, gfp_flags);
+ if (ret == 0) {
+ if (cdev->req == req)
+ cdev->setup_pending = true;
+ else if (cdev->os_desc_req == req)
+ cdev->os_desc_pending = true;
+ else
+ WARN(1, "unknown request %p\n", req);
+ }
+
+ return ret;
}
static int count_ext_compat(struct usb_configuration *c)
@@ -1690,7 +1729,7 @@ unknown:
req->length = value;
req->context = cdev;
req->zero = value < w_length;
- value = usb_ep_queue(gadget->ep0, req, GFP_ATOMIC);
+ value = composite_ep0_queue(cdev, req, GFP_ATOMIC);
if (value < 0) {
DBG(cdev, "ep_queue --> %d\n", value);
req->status = 0;
@@ -1762,7 +1801,7 @@ unknown:
req->length = value;
req->context = cdev;
req->zero = value < w_length;
- value = usb_ep_queue(gadget->ep0, req, GFP_ATOMIC);
+ value = composite_ep0_queue(cdev, req, GFP_ATOMIC);
if (value < 0) {
DBG(cdev, "ep_queue --> %d\n", value);
req->status = 0;
@@ -1957,10 +1996,16 @@ void composite_dev_cleanup(struct usb_composite_dev *cdev)
kfree(uc);
}
if (cdev->os_desc_req) {
+ if (cdev->os_desc_pending)
+ usb_ep_dequeue(cdev->gadget->ep0, cdev->os_desc_req);
+
kfree(cdev->os_desc_req->buf);
usb_ep_free_request(cdev->gadget->ep0, cdev->os_desc_req);
}
if (cdev->req) {
+ if (cdev->setup_pending)
+ usb_ep_dequeue(cdev->gadget->ep0, cdev->req);
+
kfree(cdev->req->buf);
usb_ep_free_request(cdev->gadget->ep0, cdev->req);
}
@@ -2165,7 +2210,7 @@ void usb_composite_setup_continue(struct usb_composite_dev *cdev)
DBG(cdev, "%s: Completing delayed status\n", __func__);
req->length = 0;
req->context = cdev;
- value = usb_ep_queue(cdev->gadget->ep0, req, GFP_ATOMIC);
+ value = composite_ep0_queue(cdev, req, GFP_ATOMIC);
if (value < 0) {
DBG(cdev, "ep_queue --> %d\n", value);
req->status = 0;
--
2.1.0.GIT
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
