On Mon, May 17, 2004 at 12:41:42PM -0400, Alan Stern wrote:
> +void usb_kill_urb(struct urb *urb)
> +{
> + if (!(urb->dev && urb->dev->bus && urb->dev->bus->op))
> + return;
> + urb->reject = 1;
> + urb->dev->bus->op->unlink_urb(urb, -ENOENT);
> + wait_event(usb_kill_urb_queue, atomic_read(&urb->kref.refcount) <= 1);
> + urb->reject = 0;
> }
Heh, no, you can't just sit and spin on the refcount, because that will
not work if I do:
usb_free_urb(urb);
usb_kill_urb(urb);
And that last urb->reject = 0 will oops if I do that :)
(and yes, some drivers do that already today with a call to
usb_submit_urb() followed by a call to usb_free_urb().)
Or what if I do:
usb_get_urb(urb);
usb_get_urb(urb);
usb_get_urb(urb);
usb_kill_urb(urb);
which will cause this function to deadlock.
You need to grab a reference to the urb at the beginning of the call,
and decrement it at the end of it. But in the middle there, you can't
count on a paticular refcount value, so you will have to trigger off of
something else.
thanks,
greg k-h
-------------------------------------------------------
This SF.Net email is sponsored by: SourceForge.net Broadband
Sign-up now for SourceForge Broadband and get the fastest
6.0/768 connection for only $19.95/mo for the first 3 months!
http://ads.osdn.com/?ad_id=2562&alloc_id=6184&op=click
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
