Thread (13 messages) 13 messages, 3 authors, 2017-09-19

Re: usb/gadget: stalls in dummy_timer

flat view

From: Andrey Konovalov <hidden>
Date: 2017-09-14 17:58:10
Also in: lkml

On Thu, Sep 14, 2017 at 7:49 PM, Alan Stern [off-list ref] wrote:
On Thu, 14 Sep 2017, Andrey Konovalov wrote:
quoted
Looked at this a little more.

dummy_timer() stucks in an infinite loop. It calls
usb_hcd_giveback_urb(), which in turn calls usbtouch_irq(), which
calls usb_submit_urb(), which calls dummy_urb_enqueue() and puts urb
back into dummy urb queue. dummy_timer() then does goto restart, finds
the urb and calls usb_hcd_giveback_urb() again. And this process goes
on again and again. It seems that something should either process the
urb and set urb->status or it should just expire.
There is some throttling code, but it applies only to bulk transfers.
Probably because the bandwidth limits for other types are slightly
different.  However, I don't think we need to worry about this level of
detail, since the driver makes a number of other approximations anyway.

Try the patch below; it should fix the problem.
Hi Alan,

Just tried your patch, my reproducer still hangs the kernel until all
memory is exhausted.

Thanks!
quoted hunk ↗ jump to hunk
Alan Stern



Index: usb-4.x/drivers/usb/gadget/udc/dummy_hcd.c
===================================================================
--- usb-4.x.orig/drivers/usb/gadget/udc/dummy_hcd.c
+++ usb-4.x/drivers/usb/gadget/udc/dummy_hcd.c
@@ -1781,7 +1781,6 @@ restart:
                struct dummy_request    *req;
                u8                      address;
                struct dummy_ep         *ep = NULL;
-               int                     type;
                int                     status = -EINPROGRESS;

                urb = urbp->urb;
@@ -1789,14 +1788,10 @@ restart:
                        goto return_urb;
                else if (dum_hcd->rh_state != DUMMY_RH_RUNNING)
                        continue;
-               type = usb_pipetype(urb->pipe);

-               /* used up this frame's non-periodic bandwidth?
-                * FIXME there's infinite bandwidth for control and
-                * periodic transfers ... unrealistic.
-                */
-               if (total <= 0 && type == PIPE_BULK)
-                       continue;
+               /* Used up this frame's bandwidth? */
+               if (total <= 0)
+                       break;

                /* find the gadget's ep for this request (if configured) */
                address = usb_pipeendpoint (urb->pipe);
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help