On Tue, 23 May 2006, Franck Bui-Huu wrote:
> Franck Bui-Huu wrote:
> > /* wait for the completion of the URB */
> > - wait_for_completion(&urb_done);
> > - clear_bit(US_FLIDX_URB_ACTIVE, &us->flags);
> > + timeout = wait_for_completion_interruptible_timeout(
> > + &urb_done, timeout ? : MAX_SCHEDULE_TIMEOUT);
> >
> > - /* clean up the timeout timer */
> > - if (timeout > 0)
> > - del_timer_sync(&to_timer);
> > + if (test_and_clear_bit(US_FLIDX_URB_ACTIVE, &us->flags) && timeout <=
> > 0) {
>
> Thinking more about it, I'm wondering it the second condition (timeout <= 0)
> is really needed. Do you think so ?
The test is needed because you shouldn't print a debugging message or
call usb_unlink_urb when the URB completes normally. However you will
need to store the return value from
wait_for_completion_interruptible_timeout in a long instead of an int,
because MAX_SCHEDULE_TIMEOUT is defined as LONG_MAX.
On Tue, 23 May 2006, Oliver Neukum wrote:
> 1. You seem to have changed the semantics of the no timeout case. May you?
I don't see any changes in the semantics except for the _interruptible
part.
> 2. You don't handle signals.
Signals aren't too much of an issue. When this code runs in the context
of the usb-storage thread, all signals are masked. It might also run in
the context of modprobe, however. I don't know if a signal would need any
special handling; the fact that the URB was cancelled ought to be enough
to make the probe routine fail. It wouldn't hurt to check...
Alan Stern
-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
[email protected]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
