Am Mittwoch, 25. Februar 2009 schrieb David Brownell: > On Monday 23 February 2009, Karsten Wiese wrote: > > ehci-hcd uses usb_get_urb() and usb_put_urb() in an unbalanced way > > How exactly are they unbalanced? The local *urb is NULL when usb_put_urb() is called after URB has been given back. > > causing > > isochronous URB's kref.counts incrementing once per usb_submit_urb() call. > > So long as they're decremented once on the giveback, no problem: > one increment, one decrement, all is fine. Actually the policy > was that each saved pointer gets a refcount, so each *ITD* (or > for fullspeed, SITD) gets a refcount. > > > > There is no need for ehci-hcd to deal with usb_get_urb() nor usb_put_urb() > > anyhow, > > That might now be true; usbcore has changed quite a lot since > that code was written. > > > > so patch removes their usages in ehci-hcd. > > It also removes NULL assignments that have no effect. > > Leave those NULL assignments please. They are not "no effect"; > they ensure that "old" pointers don't linger. Such things can > be useful invariants when tracking down bugs ... among other > things, it causes immediate trouble the instance anyone tries > to use a stale pointer. For me those NULL assignments make understanding the code harder. They make me wonder "why are they there?" as in xitd_urb_transaction an xITD is kmemset to zero before use. Thanks, Karsten -- 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
