qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
* [Qemu-devel] usb-musb: calls usb_packet_complete() on packets with no owner
@ 2011-06-13 11:39 Peter Maydell
  2011-06-14  7:16 ` Gerd Hoffmann
  0 siblings, 1 reply; 2+ messages in thread
From: Peter Maydell @ 2011-06-13 11:39 UTC (permalink / raw)
  To: QEMU Developers, Gerd Hoffmann

The function usb_packet_complete() in hw/usb.c says:
/* Notify the controller that an async packet is complete.  This should only
   be called for packets previously deferred by returning USB_RET_ASYNC from
   handle_packet. */

and this is checked by means of an assert(p->owner != NULL). However,
hw/usb-musb.c's musb_packet() function does this:

    if (s->port.dev)
        ret = usb_handle_packet(s->port.dev, &ep->packey[dir].p);
    else
        ret = USB_RET_NODEV;

    if (ret == USB_RET_ASYNC) {
        ep->status[dir] = len;
        return;
    }

    ep->status[dir] = ret;
    usb_packet_complete(s->port.dev, &ep->packey[dir].p);

which will call usb_packet_complete() on packets which did not
return USB_RET_ASYNC from usb_handle_packet, and so trips this
assert.

Any suggestions about what the right way to fix this is?
(I'm a bit confused about the comment that usb_packet_complete
is to 'notify the controller' -- usb-musb is the controller...)

(I do have a repro case but it's in qemu-linaro because it
uses the omap3 port; I can't seem to persuade my n800 image
to actually use the usb-otg.)

thanks
-- PMM

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [Qemu-devel] usb-musb: calls usb_packet_complete() on packets with no owner
  2011-06-13 11:39 [Qemu-devel] usb-musb: calls usb_packet_complete() on packets with no owner Peter Maydell
@ 2011-06-14  7:16 ` Gerd Hoffmann
  0 siblings, 0 replies; 2+ messages in thread
From: Gerd Hoffmann @ 2011-06-14  7:16 UTC (permalink / raw)
  To: Peter Maydell; +Cc: QEMU Developers

   Hi,

>      usb_packet_complete(s->port.dev,&ep->packey[dir].p);
>
> which will call usb_packet_complete() on packets which did not
> return USB_RET_ASYNC from usb_handle_packet, and so trips this
> assert.

> Any suggestions about what the right way to fix this is?
> (I'm a bit confused about the comment that usb_packet_complete
> is to 'notify the controller' -- usb-musb is the controller...)

The usual work flow is that the usb device (usb-msd.c does this for 
example) kicks the I/O, then returns USB_RET_ASYNC, and when the I/O is 
completed it calls usb_packet_complete().

In case of the musb controller this will call musb_schedule_cb() which 
is hooked into musb_port_ops->complete.  So simply calling 
musb_schedule_cb() directly for non-async packets should work I think.

cheers,
   Gerd

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2011-06-14  7:16 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-06-13 11:39 [Qemu-devel] usb-musb: calls usb_packet_complete() on packets with no owner Peter Maydell
2011-06-14  7:16 ` Gerd Hoffmann

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).